Structmodel
Uploaded by
Rodrigo AcuñaStructmodel
Uploaded by
Rodrigo AcuñaSTRUCTURAL ECONOMETRIC MODELLING
METHODOLOGY AND TOOLS WITH
APPLICATIONS UNDER EVIEWS
JEAN-LOUIS BRILLET.
CONTENTS
1 FOREWORD 14
INTRODUCTION 15
THE EXAMPLE: A VERY BASIC MODEL 18
1 19
2 CHAPTER 1: NOTATIONS AND DEFINITIONS 19
1.1 THE MODEL AS A SET OF EQUATIONS 19
1.2 THE ELEMENTS IN A MODEL 19
1.2.1 VARIABLES: ENDOGENOUS AND EXOGENOUS 19
1.2.2 EQUATIONS: BEHAVIORAL AND IDENTITIES 21
1.2.3 PARAMETERS 24
1.2.4 THE RANDOM TERM 25
1.2.5 RESIDUALS VERSUS ERRORS 26
1.2.6 FORMULATIONS 27
1.2.7 THE TIME DIMENSION 27
1.2.8 LINEARITY 32
1.2.9 OTHER PROPERTIES 36
1.2.10 CONSTRAINTS THE MODEL MUST MEET 40
1.2.11 NORMALIZATION AND IDENTIFICATION 47
1.2.12 CONCLUSION 51
2 CHAPTER 2: MODEL APPLICATIONS 52
2.1 OPERATIONAL DIAGNOSES 52
2.1.1 DIFFERENT TYPES OF DIAGNOSES: SCENARIOS AND SHOCKS 52
1
2.1.2 ADVANTAGES OF MODELS 52
2.1.3 A CERTAIN REASSESSING 53
2.2 THEORETICAL MODELS 54
2.3 QUANTIFIED SMALL MODELS 55
2.3.1 WITH SCIENTIFIC PURPOSE 55
2.3.2 WITH AN EDUCATIONAL PURPOSE 55
3 CHAPTER 3: MODEL TYPES 56
3.1 THE FIELD 56
3.2 THE SIZE 56
3.2.1 QUASI-ACCOUNTING MODELS 57
3.2.2 DETERMINANTS OF THE SIZE 57
3.3 THE HORIZON 58
3.3.1 FOR FORECASTING 58
3.3.2 FOR MODEL ANALYSIS 58
3.3.3 A CLASSIFICATION 58
3.4 THE PERIODICITY 58
3.5 OTHER MODELS 59
4 CHAPTER 4: GENERAL ELEMENTS 60
4.1 THE STAGES IN THE PROCESS 60
4.1.1 PREPARING THE MODEL 60
4.1.2 ESTIMATION 60
4.1.3 SOLVING AND TESTING OVER THE PAST. 60
4.1.4 SOLVING AND TESTING OVER THE FUTURE 61
4.1.5 USING THE MODEL FOR FORECASTS AND POLICY STUDIES 61
4.2 HOW TO ORGANIZE THE DEVELOPMENT OF THE MODEL 61
5 CHAPTER 5: PREPARING THE MODEL 64
2
5.1 PREPARING THE MODEL: THE FRAMEWORK 64
5.1.1 PRINT VIEW 68
5.2 PREPARING THE MODEL: SPECIFIC DATA ISSUES 69
5.2.1 TYPES OF DATA 69
5.2.2 THE ACCESS TO DATA 70
5.2.3 THE MODE OF TRANSMISSION 70
5.2.4 PREPARING THE DATA FOR THE TRANSFER 78
5.2.5 THE PRELIMINARY PROCESSING OF SERIES 79
5.2.6 UPDATES 82
5.2.7 SUPPRESSIONS 82
5.2.8 THE DOCUMENTATION 83
5.2.9 CONSEQUENCES ON WORK ORGANIZATION 85
5.2.10 THE PRACTICAL OPTIONS 85
5.3 ACCESSING INTERNATIONAL DATA BASES 86
5.3.1 THE WORLD BANK 87
5.3.2 THE ORIGINAL SERIES 88
5.3.3 CREATING A SINGLE COUNTRY WORKFILE 89
5.3.4 WORKING ON SEVERAL (OR ALL) COUNTRIES) 89
5.3.5 THE ALTERNATE SOURCES 94
5.3.6 BACK TO OUR EXAMPLE 104
5.3.7 USING LOOPS AND GROUPS IN EVIEWS 140
5.3.8 COMPARING WORKFILES : THE WFCOMPARE COMMAND 144
6 CHAPTER 6 THE ESTIMATION OF EQUATIONS 146
6.1 THE PROCESS OF ESTIMATION 146
6.2 SPECIFIC ISSUES 147
6.2.1 THE R2 OR R-SQUARED 147
3
6.2.2 THE CONSTANT TERM 150
6.3 APPLICATIONS: OUR MODEL 152
6.3.1 THE CHANGE IN INVENTORIES 153
6.3.2 INVESTMENT: THE NECESSITY TO ESTABLISH A CONSISTENT THEORETICAL EQUATION PRIOR TO ESTIMATION 163
6.3.3 EMPLOYMENT: STATIONARITY, ERROR CORRECTION MODELS, BREAKPOINT TEST. 169
6.3.4 EXPORTS: AUTOREGRESSIVE PROCESS, COINTEGRATION, LONG-TERM STABILITY. 183
6.3.5 IMPORTS: GOING FURTHER ON COINTEGRATION AND LONG-TERM STABILITY. 197
6.3.6 BACK TO THE RESIDUAL CHECK 210
6.3.7 THE PRESENT MODEL 212
7 CHAPTER 7: TESTING THE MODEL THROUGH SIMULATIONS OVER THE PAST 214
7.1 THE SOLUTION 215
7.1.1 GAUSS-SEIDEL 215
7.1.2 RITZ-JORDAN 216
7.1.3 NEWTON AND ITS VARIANTS 216
7.1.4 BROYDEN’S METHOD 220
7.1.5 ITERATIONS AND TEST OF CONVERGENCE 221
7.1.6 STUDY OF THE CONVERGENCE 224
7.1.7 SOLVING THE MODEL: BASIC EVIEWS TECHNIQUES 247
7.2 A FIRST VALIDATION 266
7.2.1 EX POST SIMULATIONS 266
7.2.2 EX-POST FORECASTS 273
7.2.3 SOLVING THE MODEL: SCENARIOS 273
7.2.4 OVERRIDING THROUGH MENUS 277
7.2.5 OUR EXAMPLE 278
7.2.6 ANALYTIC SHOCKS 280
7.2.7 OUR EXAMPLE 282
4
8 CHAPTER 8: TESTING THE MODEL OVER THE FUTURE 286
8.1 MAKING THE MODEL CONVERGE IN THE LONG RUN 286
8.1.1 THE ASSUMPTIONS 287
8.1.2 ADAPTING THE FORMULATIONS 288
8.1.3 IMPROVING THE CHANCE (AND SPEED) OF CONVERGENCE 290
8.1.4 SOLVING PARTIAL MODELS 291
8.1.5 CHECKING THE EXISTENCE OF A LONG-TERM SOLUTION 291
8.2 CONVERGENCE PROBLEMS IN THE SHORT RUN 293
8.3 CONVERGENCE PROBLEMS IN THE MEDIUM RUN 293
8.4 TESTING THE RESULTS 294
8.4.1 OF SIMULATIONS 294
8.4.2 OF SHOCKS 294
8.5 EVIEWS FEATURES 295
8.5.1 PRODUCING SIMULATIONS 295
8.5.2 PRODUCING A BASE SOLUTION 296
8.5.3 PRODUCING SHOCKS 297
8.5.4 CHANGING MODEL SPECIFICATIONS 298
8.5.5 UPDATING THE MODEL SPECIFICATIONS 300
8.5.6 MODEL PROTECTION 302
8.6 SOME RECENT USEFUL FEATURES TO BETTER MANAGE SIMULATIONS 302
8.6.1 THE GRAPH OF DEPENDENCIES 302
8.6.2 INTERPRETING THE LIST OF VARIABLES 303
8.6.3 THE “PRINT”DISPLAY 304
8.6.4 EQUATION SEARCH 305
8.7 RATIONAL EXPECTATIONS 305
8.7.1 THEFRAMEWORK 305
5
8.7.2 CONSEQUENCES FOR MODEL SIMULATIONS 306
8.7.3 TECHNICAL ELEMENTS 307
8.7.4 OUR EXAMPLE 307
8.7.5 THE TEST 308
8.7.6 THE RESULTS 308
8.8 STOCHASTIC SIMULATIONS 311
8.8.1 PURELY STATISTICAL ERRORS 311
8.8.2 FORMULATION AND ASSUMPTION ERRORS 311
8.8.3 CONSIDERING THE ERRORS 311
8.8.4 THE INTEREST OF THE TECHNIQUE 313
8.8.5 BACK TO OUR EXAMPLE 314
8.8.6 THE RESULTS 314
8.9 GOING FURTHER: STUDYING MODEL PROPERTIES 317
8.9.1 EIGENVALUE ANALYSIS 317
8.9.2 THE CASE OF ERROR CORRECTION MODELS: A SIMPLE EXAMPLE 328
8.10 MISCELLANEOUS MODELLING FUNCTIONS 331
8.10.1 MODEL DEPENDENCIES 331
8.10.2 DEBUGGING TOOLS FOR EVIEWS PROGRAMS. 331
8.10.3 PROGRAM DEPENDENCY TRACKING 333
9 CHAPTER 9: USING MODELS 333
9.1 OPERATIONAL DIAGNOSES 333
9.1.1 SCENARIOS AND THEIR DIFFERENT TYPES 333
9.1.2 MANAGING ACTUAL FORECASTS: MANAGING THE RESIDUALS 335
9.1.3 MANAGING ACTUAL FORECASTS: THE TARGETING OF SIMULATIONS 336
9.1.4 DIFFERENT TYPES OF SHOCKS 337
9.1.5 THE FORECAST: TECHNICAL ASPECTS 339
6
9.1.6 CHANGING MODEL SPECIFICATIONS 340
9.1.7 OPTIMAL CONTROL 342
9.2 TEACHING WITH MODELS 342
9.3 PRESENTATION OF RESULTS 343
9.3.1 GENERAL ISSUES 343
9.3.2 TABLES 343
9.3.3 GRAPHS 344
10 CHAPTER 10: APPLYING THE ABOVE PRINCIPLES TO OPERATIONAL CASES 349
10.1 THE ACCOUNTING FRAMEWORK 349
10.1.1 THE AGENTS: A FIRST DEFINITION 350
10.1.2 THE OPERATIONS 350
10.1.3 THE INTEGRATED ECONOMIC ACCOUNTS 351
10.1.4 SECTORS, BRANCHES AND PRODUCTS, 351
10.1.5 AGENTS SUBDIVISIONS 352
10.1.6 A MULTI COUNTRY MODEL 352
10.2 A SINGLE COUNTRY, SINGLE PRODUCT MODEL 352
10.2.1 THE ECONOMIC ASPECTS 352
10.2.2 THE EVIEWS PROGRAMS 370
10.3 A SINGLE COUNTRY, MULTI PRODUCT MODEL 464
10.3.1 THE MAIN REASONS FOR PRODUCT DECOMPOSITION 464
10.3.2 INTRODUCING INTERMEDIATE CONSUMPTION 467
10.3.3 SPECIFIC SECTORAL ISSUES 468
10.3.4 THE DATA 469
10.3.5 CONSEQUENCES FOR ESTIMATIONS 470
10.3.6 THE PRODUCTION FUNCTION 471
10.3.7 UNEMPLOYMENT 472
7
10.3.8 CHANGE IN INVENTORIES 472
10.3.9 HOUSEHOLD CONSUMPTION 472
10.3.10 EXTERNAL TRADE 472
10.3.11 WAGES 473
10.3.12 PRICES 473
10.4 A MULTI COUNTRY, SINGLE PRODUCT MODEL 484
10.4.1 FIRST ISSUE: THE MODELS 484
10.4.2 SECOND ISSUE: INTER-COUNTRY TRADE 485
10.4.3 A CONSISTENT METHOD: MACSIM 486
10.4.4 THE EVIEWS PROGRAM 499
10.5 A REGIONAL MODEL 499
10.6 A MULTI COUNTRY, MULTI PRODUCT MODEL 500
10.6.1 THE EVIEWS PROGRAM 500
11 CHAPTER 10 : QUASI ACCOUNTING MODELS 501
11.1 THE ORIGINALITY OF THE APPROACH 501
11.1.1 THE PRODUCT DETAIL 501
11.1.2 THE OPERATIONS DETAIL 502
11.2 THE LEVEL OF SIMPLIFICATION 503
11.3 THE GOOD POINTS 504
11.4 TWO EXAMPLES 504
11.4.1 WORKING WITH BOTH TYPES 511
12 CHAPTER 11 : STOCK-FLOW CONSISTENT MODELS 512
12.1 THE GENERAL APPROCH 512
12.2 THE ACCOUNTING FRAMEWORK 514
12.2.1 THE AGENTS 514
12.2.2 THE OPERATIONS 514
8
12.3 A PROPOSAL 515
12.3.1 THE FRAMEWORK FOR THE NEW PRODUCTION BLOCK 515
12.3.2 THE FINANCIAL BLOCK 517
12.3.3 THE ADDITIONAL CHANGES LINKED TO THE FINANCIAL ACCOUNT. 517
12.4 EVOLVING TOWARDS A FULL MACROECONOMIC MODEL 518
12.4.1 A FULL PROPOSAL 519
12.4.2 THE CAUSALITIES 520
12.4.3 THE CAUSALITIES ( KEYS) 520
12.4.4 EXPLICITING THE BEHAVIORS 521
12.5 THE SIMULATIONS 527
THE SHOCKS 528
12.5.2 CONCLUSION ON MODEL PROPERTIES 533
13 SOLVING A MODEL FOR TARGET VALUES 536
13.1 THE METHOD USING A PROGRAM 536
13.1.1 INTRODUCTION 536
13.1.2 THE PROBLEM 536
13.1.3 THE LINEAR CASE 537
13.1.4 THE NON-LINEAR CASE 538
13.1.5 A FEW QUESTIONINGS 539
13.1.6 A FEW IDEAS ON CHOSING THE INSTRUMENTS 540
13.1.7 A TEST ON A VERY SMALL MODEL. 540
13.1.8 A TEST ON A LARGER MODEL 549
13.2 USING THE EVIEWS INTEGRATED COMMAND 550
13.2.1 AN EXAMPLE BASED ON OUR SMALL MODEL 551
13.2.2 USING A MENU 552
13.2.3 WITH TWO TARGETS 553
9
13.3 RESPECIFICATION OF ENDOGENOUS AND EXOGENOUS VARIABLES 555
13.3.1 THE FIRST EXAMPLE 556
13.3.2 A MORE COMPLEX EXAMPLE 557
13.3.3 A LESS LOGICAL ONE 557
13.4 THE CONSEQUENCES FOR THE SOLVING PROCESS 559
14 A LIST OF USEFUL SERIES FOR PRODUCING A SINGLE COUNTRY, SINGLE PRODUCT MODEL 559
14.1 THE PERIODICITY 560
14.2 GENERAL ELEMENTS 560
14.3 THE SERIES 560
14.3.1 THE SUPPLY - DEMAND VARIABLES AT CONSTANT PRICES (WHOLE ECONOMY) 560
14.3.2 THE SAME ELEMENTS AT CURRENT PRICES 561
14.3.3 THE SAME ELEMENTS DECOMPOSED INTO PRODUCTS 561
14.4 PRODUCTIVE CAPACITY 561
14.5 EMPLOYMENT 561
14.6 PRICE DEFLATORS 561
14.7 HOUSEHOLDS ACCOUNT 562
14.8 FIRMS ACCOUNT (ALL TYPES OF FIRMS) 562
14.9 REST OF THE WORLD 562
14.10 GOVERNMENT ACCOUNT (ELEMENTS NOT DESCRIBED EARLIER) 563
14.11 FINANCIAL DOMAIN 563
15 A LIST OF USEFUL SERIES FOR A MODEL WITH SEVERAL SECTORS 564
15.1 THE PERIODICITY 564
15.2 GENERAL ELEMENTS 564
15.3 THE SERIES 564
15.3.1 THE SUPPLY - DEMAND VARIABLES AT CONSTANT PRICES (WHOLE ECONOMY) 564
15.3.2 THE SAME ELEMENTS AT CURRENT PRICES 565
10
15.4 PRODUCTIVE CAPACITY 565
15.5 EMPLOYMENT 565
15.6 PRICE DEFLATORS 566
15.7 HOUSEHOLDS ACCOUNT 566
15.8 FIRMS ACCOUNT (ALL TYPES OF FIRMS) 566
15.9 REST OF THE WORLD 567
15.10 GOVERNMENT ACCOUNT (ELEMENTS NOT DESCRIBED EARLIER) 567
15.11 FINANCIAL DOMAIN 567
16 RECENT ADDITIONS TO SESSION MANAGEMENT 568
16.1 SNAPSHOTS 568
16.2 CLOUD MANAGEMENT 568
16.2.1 GOOGLE DRIVE 568
17 USING EVIEWS MENUS TO BUILD AND MANAGE AN ECONOMETRIC MODEL 569
17.1 GENERAL MENU FEATURES 569
17.1.1 COMMAND CAPTURE 569
17.1.2 WINDOW MANAGEMENT AND DOCKING 569
17.1.3 SNAPSHOTS 570
17.1.4 LOGGING 570
17.2 A VERY SMALL MODEL 570
17.3 THE DATA 571
17.4 CREATING THE WORKFILE 571
17.4.1 PROGRAMMING 573
17.5 READING THE DATA 573
11
577
17.5.1 PROGRAMMING 577
17.6 CREATING THE PAGE 577
17.6.1 PROGRAMMING 578
17.7 LINKING THE DATA 578
17.7.1 PROGRAMMING 579
17.8 DISPLAYING THE DATA 580
17.9 CREATING THE MODEL 582
17.9.1 PROGRAMMING 586
17.10 CHECKING THE MODEL 586
17.11 ESTIMATING CONSUMPTION 587
17.11.1 A VERY SIMPLE FORMULA 587
17.11.2 IMPROVING THE FORMULA 589
17.12 SOLVING THE MODEL 595
17.12.1 PROGRAMMING 601
17.13 A SHOCK 601
12
602
17.13.1 PROGRAMMING 605
17.14 FORECASTING 606
17.14.1 CREATING A PAGE 606
17.14.2 THE ASSUMPTIONS 608
17.14.3 BASIC SIMULATION 609
17.14.4 A SHOCK 611
17.15 A FULL PROGRAM 617
17.15.1 A RESIDUAL IN THE ESTIMATED EQUATION 617
17.15.2 A RESIDUAL CHECK 618
17.15.3 COMPUTING BASEFUTURE VALUES FOR THE ENDOGENOUS 618
17.15.4 A BASELINE SOLUTION USING A SCENARIO 618
17.15.5 PRODUCING A SHOCK FROM SCENARIO 618
17.15.6 COMPUTING DIFFERENCES USING A LOOP 619
17.15.7 THE PROGRAM 619
18 INDEX 629
13
1 FOREWORD
The purpose of this book is a little special.
First, of course, by its subject: structural econometric modelling no longer looks so fashionable, having lost ground to
Computable General Equilibrium models and in particular their Dynamic Stochastic versions.
We will contend that while this might be true in the academic field (you just have to look at the program of congresses
and symposiums) there is still a lot of room for structural models. Indeed, many institutions are still using them and
even building new ones, both in developed and developing countries. We shall try to show that this position is quite
justified, and that for a large part of the modelling applications, in particular the analysis and interpretation of
macroeconomic interactions, the call for structural models remains a good strategy, arguably the best one.
But we shall not stop at proving the usefulness of these models. For the people we have convinced, or which were so
already, we will provide a set of tools facilitating all the tasks in the modelling process. Starting from elementary
elements, it will lead by stages the user to a level at which he should be able to build, manage and use his
professional, operational model.
This means this book will, as its title says, focus essentially on applied and even technical features, which does not
mean it will be so simplistic.
After a necessary description of the field, we shall use the largest part of the book to show the reader how to build his
own model, from general strategies to technical details. For this we shall rely on a specific example, presented at the
beginning, and which we will follow through all the steps of model development. When the situation becomes more
complex (with the addition of product and international dimensions), we shall still find this model at the core of the
cases.
This model will also be present in the background, when we address new directions, which we think are quite
compatible with our approach: Quasi-Accounting, and Stock-Flow Consistent models.
Our examples will be based on EViews package, the most popular modeling product presently available This will allow
us to be more helpful to EViews users, concentrating on its practice (including some tricks).
Finally, just as important if not more so, we shall provide a set of files allowing readers to practice modelling (either
alone or as part of a course). And for more advanced users, we shall give access to files allowing to produce
operational (if small) models, which they can adapt to their own ideas, with the tedious tasks: producing the data,
defining the accounting framework and organizing simulations over the future, being already prepared.
All these elements are provided for free, and downloadable on the EViews site, at the address:
https://www.eviews.com/StructModel/structmodel.html
This version of the book takes into account the features of the last version of EViews; EViews 13. However, most of
the text is valid for earlier versions. The main differences come from improvements in the user-friendliness.
14
INTRODUCTION
Since an early date in the twentieth century, economists have tried to produce mathematical tools which, applied to a
given practical problem, formalized a given economic theory to produce a reliable numerical picture. The most natural
application is of course to forecast the future, and indeed this goal was present from the first. But one can also
consider learning the consequences of an unforeseen event, or measuring the efficiency of a change in the present
policy, or even improving the understanding of a set of mechanisms too complex to be grasped by the human mind.
In the beginning (let us say since the 1930s) the field was occupied by the “structural” models. They start from a given
economic framework, defining the behaviors of the individual agents according to some globally consistent economic
theory. They use the available data to associate to these behaviors reliable formulas, which are linked by identities
guaranteeing the consistency of the whole set. These models can be placed halfway between the two above
categories: they do rely on statistics, and also on theory. To accept a formula, it must respect both types of criteria.
The use of this kind of models, which occupied the whole field at the beginning, is now mostly restricted to policy
analysis and medium-term forecasting. For the latter, they show huge advantages: the full theoretical formulations
provide a clear and understandable picture, including the measurement of individual influences. They allow also to
introduce stability constraints leading to identified long-term equilibriums, and to separate this equilibrium from the
dynamic fluctuations which lead to it.
In the last decades, new kinds of models have emerged, which share the present market.
• The “VAR” models. They try to give the most reliable image of the near future, using a complex estimated
structure of lagged elements, based essentially on the statistical quality, although economic theory can be
introduced, mostly through constraints on the specifications. The main use of this tool is to produce short-term
assessments.
• The Quasi-Accounting models, which rely on very basic behaviors, most of the time calibrated. This allows to
treat cases where data is available for extremely limited sample periods, or where the fine detail (generally in
products) forbids to apply econometrics with a good chance of global success.
• Stock-Flow Consistent models, which answer two criticisms addressed to structural models: producing incomplete
and formally unbalanced models, and not taking enough into account the stocks, in particular of financial assets.
By detailing these assets by agent and category, SFCs allow to consider sophisticated financial behaviors,
sometimes at the expense of the “real” properties.
• And last (but not the least) Computable General Equilibrium models. They use a detailed structure with a priori
formulations and calibrated coefficients to solve a generally local problem, through the application of one or
several optimizing behaviors. The issues typically addressed are optimizing resource allocations, or describing the
consequences of trade agreements. The mechanisms described contain generally little dynamics.
This is no longer true for the Dynamic Stochastic General Equilibrium models, which dominate the current field. They
include dynamic behaviors and take into account the uncertainty in economic evolutions. Compared to the traditional
models (see later) they formalize explicitly the optimizing equilibria, based on the aggregated behavior of individual
agents. This means that they allow agents to adapt their behavior to changes is the rules governing the behaviors of
others, including the State, in principle escaping the Lucas critique. As the model does not rely on traditional
estimated equations, calibration is required for most parameters.
Compared to CGEs and DSGEs, optimization behaviors are present (as we shall see later) and introduced in the
estimated equations. But they are frozen there, in a state associated with a period, and the behavior of other agents
at the time. If these conditions do not change, the statistical validation is an important advantage. But sensitivity to
shocks is flawed, in a way which is difficult to measure.
A very good (and objective) description of the issue can be found in:
15
http://en.wikipedia.org/wiki/Dynamic_stochastic_general_equilibrium
http://en.wikipedia.org/wiki/Macroeconomic_model#Empirical_forecasting_models
It seems to us that the main criterion in the choice between DSGEs and traditional structural models lie in the tradeoff
between statistical validation and adaptability of behaviors.
In the last years, popularity of structural econometric modelling seems to have stabilized. A personal hint for this (if
not an actual proof) is the sustained demand for participation in structural modelling projects, observed on the sites of
companies devoted to international cooperation.
Another issue is that being the first tool produced (in the thirties of the last century) it was applied immediately to the
ambitious task of producing reliable forecasts. The complexity of the economy, and the presence of many random
shocks makes this completely unrealistic (and this is even more true today). During the golden years of structural
modelling, when economy was growing at a regular (and high) rate, forecasting was as easy as riding a tame horse on
a straight path: anybody could do it. But when the horse turned into a wild one, the quality of the rider showed, and it
did not stay in the saddle too long. Failing to succeed in a task too difficult for any tool (including VAR and CGE models,
which do not have to forecast the medium-term), gave discredit to structural models and all their uses, including
policy analysis and even the understanding and interpretation of complex economic mechanisms, applications for
which neither VAR nor CGE can compete in our opinion.
Also, the role of financial issues has grown, which the initial structural models were not well equipped to address. But
Stock-Flow Consistent versions can be an answer to this problem.
Anyway, even with limited ambitions, producing a sound econometric structural model is not a simple task. Even a
professional economist, having an excellent knowledge of both economic theory (but not necessarily a complete and
consistent picture) and econometric techniques (but not necessarily of their practical application) will find it quite
difficult producing a reliable and operational econometric model.
The purpose of this book is to shorten the learning process, in several ways.
After a global presentation of economic models:
• Notations, definitions, mathematical characteristics (dynamics, linearity, continuity, identifiability…).
• Applications: economic theory, forecast, education.
• Classification of existing models.
We shall describe how to organize the sequence of model building tasks, from data production and framework
specification to actual operational studies.
For each task, we shall give all the necessary elements of methodology.
We shall present the main economic options available, with some theoretical explanations.
All these explanations will be based on a practical example, the production of a very small model of the French
economy. The size will not forbid us to address most of the problems encountered in the process.
The methods, techniques and solutions proposed will be based on the EViews software. This will allow us to present
some useful features and tricks, and to provide a sequence of complete programs, which the user can modify at will,
but not necessarily too heavily, as all the models of this type share a number of common elements. The main issue is
of course the estimation process, each case leading generally to an original version of each behavioral equation.
16
A set of documented programs is available on demand, following the above principles
• For the small example,
• For a more detailed product, a model for a single country, not far from an operational version.
These programs will allow to:
• Import the original data
• Build the model framework
• Transform the data to conform to the elements in the model.
• Estimate a set of equations, starting with standard behaviors, possibly updated.
• Check the technical and theoretical consistency of the resulting model.
• Produce forecasts and policy studies.
In each case, we shall present programs which actually work. An econometric solution will be found, reliable both in
statistical and economic terms. And the properties of the models will be rather satisfying, with a long-term solution
and reasonable dynamics leading to it.
Finally, we shall address the more complex problems: multi-sector and multi-country models (and both options
combined). The specific issues will be described, and a framework for a three-product model will be provided,
following the same lines as the previous example.
The goal of this book is therefore both limited and ambitious. Without getting into theoretically complex features, it
should give readers all the elements required to construct their own model. Being relieved of the more technical (and
tedious) tasks, they will be allowed to concentrate on the more intelligent (and interesting) ones.
Readers must be aware they will find here neither a full description of econometric and statistical methods, nor a
course in economic theory. We shall give basic elements on these fields, and rather focus on their links with the
modelling process itself. For more detailed information, one can refer to the list of references provided at the end of
the volume.
Concerning Quasi-Accounting and even more Stock-Flow Consistent models, for which our experience is much more
limited, we will be even less directive.
17
THE EXAMPLE: A VERY BASIC MODEL
To present the elements and the framework of a structural econometric model, we shall use a specific example, which
we shall address permanently during our presentation. In spite of its limited size, we think it remains quite
representative of the class of models we are considering in this manual.
At the start of any model building process, one has to specify in a broad manner the logic of his model, and the
behaviors he wants his model to describe. No equation needs to be established at this time. We shall place ourselves
in this situation.
In our example, an economist has decided to build a very simple model of the French economy. As our tests will be
based on actual data, a country had to be chosen, but the principles apply to any medium sized industrialized country.
Our model includes the following elements.
• Based on their production expectations and the productivity of factors, firms invest and hire workers to adapt
their productive capacity. However, they exert some caution in this process, as they do not want to be stuck
with unused elements.
• The levels reached in practice define potential production.
• Firms also build up inventories.
• Households obtain wages, based on total employment (including civil servants) but also a share of Gross
Domestic Product. They consume part of this revenue.
• Final demand is defined as the sum of its components: consumption, productive investment, housing
investment, the change in inventories, and government demand.
• Imports are a share of local demand («domestic demand»). But the less capacities remain available, the more
an increase in demand will call for imports.
• Exports follow world demand, but producers are limited by available capacities, and their priority is satisfying
local demand.
• Supply is equal to demand.
• Productive capital grows with investment, but is subject to depreciation.
The above framework looks rather straightforward, and certainly simplistic. Obviously, it lacks many elements, such as
prices, financial concepts, and taxes. This will be addressed as later developments.
Let us no go further for the time being. One can observe that if we have not built a single equation yet, a few are
already implicit from the above text.
18
1
2 CHAPTER 1: NOTATIONS AND DEFINITION S
Before we start presenting the process of model building, we must define the concepts we shall use. They will be
based on individual examples taken from our (future) model.
1.1 THE MODEL AS A SET OF EQUATIONS
In a general way, a model will be defined as a set of fully defined formulas describing the links between a set of
concepts.
Formally, a model can be written as the vector function of variables.
𝑓(. . . . ) = 0
We shall address in turn:
• The nature of elements appearing in the function.
• The nature of the functions themselves.
1.2 THE ELEMENTS IN A MODEL
1.2.1 VARIABLES: ENDOGENOUS AND EXOGENOUS
Obviously, a model will be used to measure economic concepts, depending on other elements.
Two variable types will appear in a model:
• Endogenous variables, or results, whose value will be obtained by solving the system of equations,
• Exogenous variables, or assumptions, whose value is known from outside considerations, and which
obviously condition the solution.
If the model is solved over past periods, this value should be known. But in forecasting operations, it will have to be
chosen by the model builder (or user).
For the system to be solved, the number of endogenous variables must correspond to the number of equations.
Our formulation becomes:
𝒇(𝒙, 𝒚) = 𝟎
19
with
• x vector of exogenous variables
• y vector of endogenous variable (with the same dimension as f).
For instance, in our model:
• Imports will be endogenous, as they depend on local demand. Exports too, depending on world demand.
• World demand will be exogenous, as we are building a model for a single country, and we are going to
neglect the impact of local variables on the world economy. Of course, this impact exists, as France is (still) an
important country, and its growth has some influence on the world economy. But the relatively limited
improvement can only be obtained at the very high cost of building a world model. This simplification would
be less acceptable for a model of the USA, or China, or the European Union as a whole (we shall address this
issue in detail later).
Technically, one can dispute the fact that exports are endogenous. As we make them depend only on an exogenous
world demand, they are de facto predetermined, apart from an unforecastable error. But they have to be considered
endogenous. Our model describes the local economy, and one of its features is the exports of local firms, allowed by
the external assumption on foreign demand, but following a local behavior.
As to Government demand, models of the present type will keep it also exogenous, but for different reasons:
• The main goal of this model is to show its user (which can be the State, or a State advising agency, an
independent economist playing the role of the State, or even a student answering a test on applied
economics) the consequences of its decisions. So these decisions must be left free, and not forced on him.
• The behavior of the State is almost impossible to formalize, as it has few targets (mostly growth, inflation,
unemployment, budget and trade balances) and a much larger number of available instruments1. If their base
values are more or less fixed, it can deviate from them arbitrarily, without too much delay. To achieve the
same goal, past French governments have used different global approaches, calling for different panels of
individual instruments.2
• The State alone has enough individual power to influence significantly the national economy.
Each of the two exogenous elements is characteristic of a broader category:
• Variables considered as external to the modeled area, on which economic agents considered by the model
have no or little influence. In addition to the world environment for a national model, this can mean
population3, or meteorological conditions4, or the area available for farming. The theoretical framework of
the model can also suppose exogenous structural elements, such as the real interest rate, the evolution of
factor productivity, or the depreciation rate of capital.
1
Not here, but in the general case.
2
For instance, to decrease unemployment, a government can increase demand or reduce firms’ taxes, and the tax
instrument can be social security contributions, or subsidies
3
In long-term models growth might affect the death and birthed rates thus population.
4
Which can depend on growth (glasshouse effect).
20
• Variables controlled by an agent, but whose decision process the model does not describe. Even if it was
formally possible, the model builder wants to master their value, to measure their consequences on the
economic balance. These will be referred to as decision variables or «instruments".
Changing the assumptions on these two types of variables, therefore, will relate to questions of very different spirit:
• What happens if perhaps...? (the price of oil increases abruptly).
• What happens if I (the State), decide...? (to decrease VAT rate5).
The second type of question can be inverted: what decision do I have to take to obtain this particular result? (By how
much should I decrease the rate of employers’ social contributions to create 1000 jobs?). This means that the status
(exogenous/endogenous) of some variables is changed: the answer calls for specific techniques or solving a
transformed model. We will deal with this later.
Sometimes the two approaches can also be combined; by considering first the consequences of an evolution of
uncontrolled elements, and then supposing a reaction of the State, for instance a change in policy that would return
the situation to normal. For instance, the State could use its own tools to compensate losses in external trade due to a
drop in world demand.
From a model to another, the field described can change, but also the separation between endogenous and
exogenous. The real interest rate can change its nature depending on the endogeneity of the financial sector,
technical progress can be assumed as a trend or depend on growth, and the level of population can depend on
revenue.
1.2.2 EQUATIONS: BEHAVIORAL AND IDENTITIES
1.2.2.1 Behaviors
The first role of the model is to describe “behaviors”: the model builder, following most of the time an existing
economic theory, will establish a functional form describing the behavior of a given agent, and will use econometrics
to choose a precise formulation, with estimated parameters.
In describing consumption, one might suppose that its share in household income is determined by
• The level of income (a higher income will make consumption less attractive or necessary, compared to
savings6).
• Recent variations of income (consumers take time in adapting their habits to their new status).
• The evolution of unemployment: if it grows, the prospect of losing a job will lead households to increase
reserves.
• Inflation: it defines the contribution required to maintain the purchasing power of financial savings.
Once identified, all these elements will be united in a formula, or rather a set of possible formulas (households can
consider present inflation, or the average over the last year; the increase in unemployment can use its level or
percentage change). These formulas will be confronted with the available data, to find a specification statistically
acceptable on the whole, each element participating significantly in the explanation, and presenting coefficient values
5
Provided the EU commission will allow it.
6
Let us recall that investment in housing is considered as savings.
21
consistent with economic theory. Once parameters are estimated, each element of the resulting formulation will
contribute to the logical behavior of the associated agent.
But the process is not always so straightforward. Two other cases can be considered.
• The behavior can be formalized, but not directly as estimation-ready formulas. A framework has first to be
formalized, then processed through analytical transformations possibly including derivations and
maximizations, leading finally to the equation (or set of equations) to estimate. This will be the case for our
Cobb-Douglas production function (page 105) for which we compute the combination of labor and capital
which maximize profits for a given production level, according to a set of formulas obtained outside the
model. Or for the definition of the wage rate as the result of negotiations between workers unions and firm
managers, based on their respective negotiating power.
• Often the model builder will not be able to formulate precisely the equation, but will consider a set of
potential explanatory elements, waiting for econometric diagnoses to make a final choice between
formulations (generally linear). For instance, the exchange rate might depend on the comparison of local and
foreign inflation, and on the trade balance.
In any case, even if the exact intensity of influences is unknown to the model builder 7, economic theory generally
defines an interval of validity, and especially a sign. Whatever the significance of the statistical explanation, it will be
rejected if its sign does not correspond to theory. In the example above, the increase of labor demand must generate
gains in the purchasing power of the wage rate.
Anyway, the less the theoretical value of the estimated coefficient is known, the more care must be applied to the
numerical properties of the model, at simulation time.
The formulation of these theoretical equations often makes use of specific operators, allowing alternative
calculations: Boolean variables, operators for maximum and minimum. For instance, in disequilibrium models, the
theoretical equation can include a constraint. We can consider also the case of a function of production with
complementary factors, where the level of each factor determines an individual constraint:
𝐶𝐴𝑃 = 𝑚𝑖𝑛( 𝑝𝑙. 𝐿, 𝑝𝑘. 𝐾)
with CAP production capacity, L employment, K capital, and pl, pk the associated productivities
1.2.2.2 Identities
A model composed only of behavioral equations cannot generally be used as such. Additional equations will be
needed, this time describing undisputable and exact relationships.
Several cases can be identified, which can apply simultaneously:
• Some concepts are linked by an accounting formula, and we need to ensure their numerical coherence. For
example, once the model has defined household revenue, it cannot estimate savings and consumption
7
Otherwise he would not have estimated it.
22
separately as the sum of the two is known8. A single element will be estimated: it can be savings,
consumption, the savings ratio or the consumption ratio, and the other elements will follow, using an
identity.
• Some concepts are linked by a causal sequence of elements, and some elements in the chain are not defined
by behaviors. For example, if we estimate firms’ employment and household consumption, we must
formalize household revenue (as a sum including wages) to make job creation improve consumption. And in
our example, defining final demand (as a sum of its components) ensures that imports will follow
consumption.
Of course, one can consider eliminating these identities by replacing each element they compute by the
corresponding formula. This is not always technically possible, but in any case it would:
o Lead to overly complex formulations, difficult to interpret and slower to compute.
o Discard potentially interesting information.
In addition, one will be led to introduce:
• Intermediate variables simplifying formulations (and speeding up computations). Even if the growth rate of
the real wage rate, which uses a slightly complex expression, was not considered interesting as an economic
quantity, it will be useful to define it, if it appears as an explanatory element in many equations.
• Purely descriptive elements: the ratio of Government Balance to GDP is a crucial element in evaluating the
financial health of the state (and one of the « Maastricht » criteria for entering the European Monetary
Union).
• Finally, economic theory is not always absent from this type of equation: the supply – demand equilibrium
has to be enforced:
Q (supply from local producers) + M (foreign supply to the country) = FD (demand from local agents) + X (foreign
demand to the country).
And the choice of the variable which balances it has a strong theoretical impact on model properties.
o If exports and imports come from behaviors, and demand from the sum of its components, we need to
compute Q as:
Q (local output) = (FD-M) (local demand supplied by local producers) +X (foreign demand supplied by local producers)
This means that production will adapt to demand (which itself can depend on the availability of products).
o But we could also suppose that:
The producers chose to limit their output at a level actually lower than demand, because additional production would
bring negative marginal profits. In this case Q will be fixed, and we could have:
Q = fixed, X = f(WD), FD = f(economy), M = FD - Q + X
8
This would also be absurd in terms of household behavior.
23
o Or the country can only import in foreign currency, which it obtains through exports.
X = f(WD), M = f(X), Q = fixed, FD = Q+(M-X)
1.2.3 PARAMETERS
Parameters can be defined as scalars with a varying value. The only formal difference with exogenous variables is that
they lack a time dimension9.
Two types of parameters can be considered, according to the way their value is established:
• Those estimated by reference to the past: starting from a theoretical but fully defined formula including
unknown parameters, the model builder will seek the values which provide the formulation closest to
observed reality, according to a certain distance. This means using "econometrics".
• Those decided by the model builder: economic theory or technical considerations can supply a priori
assumptions concerning a particular behavior. For instance, if a Central Bank uses a standard Taylor rule to
decide its interest rate, its sensitivity to the inflation level should be 0.5. A special case will be represented by
a control variable, giving (without changing the formulation) a choice between several types of independent
behaviors.
The distinction is not as clear as it may seem: in particular, if estimation fails to provide an economically coherent
result, the model builder can be driven to accept the value of a parameter, if it is consistent with economic theory.
And even if not, to choose the nearest value within the theoretical range. For instance, an indexation of wages on
inflation by 1.1 can lead the modeler to apply 1, if the difference is not significant.
With a as a vector of parameters (â estimated) the system becomes:
𝒇(𝒙, 𝒚, 𝒂̑ , 𝒂) = 𝟎
And in our example, one could estimate the influence of world demand on exports, for example by supposing that
relative variations are proportional (or equivalently that the elasticity of exports to world demand is constant).
𝜟𝑿/𝑿 = 𝒂 ⋅ 𝜟𝑾𝑫/𝑾𝑫
9
In EViews, modifying a parameter value applies to the current model, and taking it into account calls for a new
compilation, making the new version operational. This is both tedious and error-prone. One might consider replacing
parameters by series with a constant value, which gives access to the much more manageable “scenario” feature.
24
where a should be close to unity, if the share of the country on the world market is stable10.
But if the estimated coefficient not significant, we can get back to:
𝛥𝑋/𝑋 = 𝛥𝑊𝐷/𝑊𝐷
This choice could also have been made from the start for theoretical reasons, or to ensure the long-term stability of
the model.
Clearly, to estimate a parameter it is necessary to define entirely the associated formula.
1.2.4 THE RANDOM TERM
In practice, the behavior of agents does not answer exactly to formalized functions, and the formulation obtained by
estimation will not reproduce the reality. It will only approximate this behavior, using elements which conform to
some economic theory, each of them providing a significant contribution to the changes in the explained variable. The
number of estimated parameters will then generally be much lower than the size of the sample, or the number of
observed values. In practice, adding elements to the explanation can:
• In the favorable cases, improve the quality of the explanation given by the elements already present, which
can now concentrate on their natural role, instead of trying to participate in the explanation of other
mechanisms in which their efficiency is limited11.
• But the new element can compete with the others in explaining a mechanism in which they all have some
competence, limiting the improvement and leaving the sharing of the explanation rather undetermined (and
therefore limiting the significance of the coefficients).1213
In practice, these correlation problems will always appear, sometimes very early, and generally before the fifth or
sixth element. Beyond that figure, the precision of individual coefficients will decrease, and global quality will improve
less and less.
This means that a typical econometric equation will contain a maximum of four parameters, while variables will be
known on fifty to one hundred quarters.
10
In our model WD stands for world trade (including its expansion), not the aggregate demand of countries.
11
Just like a worker which has to use his time on two tasks, and is really qualified for one. For example, if an excellent
musician but average lyricist is teamed with a good lyricist, the quality of songs (both music and lyrics) will improve.
12
This can be a problem for the model if the two competing elements have a different sensitivity to a particular
variable. For instance, if one is sensitive to a tax rate, the other not: then the role of the tax rate will be undetermined.
13
If two workers with the same profile complete a task together, it is difficult to evaluate their individual contribution.
One might have rested the whole period.
25
It will be therefore necessary, to formulate an exact model, to accept the presence of non-zero additional terms
(residuals). If one believes in the model, this residual should be interpreted as a random perturbation without
economic meaning14. But if the equation is badly specified, it will also come from other sources: omitting a relevant
variable, replacing it by another less relevant, or choosing the wrong form for the equation 15.
The fault will not always lie with the model builder, who might not have been able to apply his original ideas. The
variables he needs may not be precisely measured, or only with a slightly different definition, or they may not be
available at all, as in, for example, the goals or expectations of a given agent.
Practically speaking, one will often suppose that this residual follows a random distribution, with a null average, a
constant standard error, and residuals independent across periods.
Our formulation becomes therefore, in the general case, noting u the vector of residuals:
𝒇(𝒙, 𝒚, 𝒂̑ , 𝒂, 𝒖) = 𝟎
In the example, if we want to represent changes in household consumption as a constant share of total production
variations, we will write:
𝑪𝑶 = 𝒂 ⋅ 𝑸 + 𝒃 + 𝒖
or rather, if we want u to have a constant relative influence:
𝑪𝑶/𝑸 = 𝒂 + 𝒖
As we shall see later, the second equation avoids heteroscedasticity problems.
1.2.5 RESIDUALS VERSUS ERRORS
14
One can also take into account that the relationship is not exact. For instance, that an value of an elasticity is only
very close to constant.
15
Of course, as we have said before, one is never able to estimate the « true » equation. This remark should apply to a
large conceptual error, leading to behaviors distinctly different from an acceptable approximation of reality.
26
It is probably the time to bring an important issue about the nature of econometrics.
When he considers a behavioral equation, the economist can have two extreme positions.
• He believes the behavior can be exactly specified according to a formula, which is affected by an error term
with a given distribution (maybe a white noise, or a normal law). With an infinite number of observations, we
would get an exact measurement of the parameters, and therefore of the error (which remains of course)
and its distribution.
• He thinks that the concept he wants to describe is linked to some other economic elements, but the relation
is only an application, of which any formula represents only an approximation. To this application a random
term can also be added, if one believes that the replication of the same explanatory elements will bring a
different result. Additional observations will only get a better mapping.
The debate is made more complex by several facts:
• The data on which he wants to base his estimation is not measured correctly. One cannot expect the
statisticians to produce error free information, for many reasons: measurement errors, inappropriate sample,
mistaken concepts...
• Even if measured correctly, the concepts he is going to use are not necessarily the right ones. For instance, a
given behavior should be applied to the only firms which do make profits, a separation which is not available
at the macroeconomic level.
• The discrete lags which he will apply to these concepts are not the right ones either. For instance, it might be
known that an agent considers the price index of the last month, but only quarterly data is available.
• The estimation period is not homogenous, and this cannot be explained by the data. For instance, the mood
of consumers (and their consumption behavior) can evolve without any link to measurable economic
elements.
From the above issues, the logical conclusion should be:
• The first position is illusory, and to a point which is impossible to measure (of course).
• But we have to take it if we want to apply econometric methods.
This means that in the following text we shall put ourselves in the first position, but we will always keep in mind the
true situation and give to the difference between the concept and its estimation the less ambitious name of
“residual”.
1.2.6 FORMULATIONS
We shall now consider the form of the equations. Let us first approach the time dimension.
1.2.7 THE TIME DIMENSION
Variables in economic models have generally a time dimension, which means they are known through discrete values,
almost always with a constant periodicity: generally annual, quarterly or monthly series. This means we will consider
models in discrete time.
There are exceptions, however. The most frequent applies to micro-economic models, describing the behavior of a
panel of individual firms or households, and the dimension will correspond to items in a set. Sometimes they will be
ordered, using the level of one variable, such as the income level for a set of households. Time can be introduced as
an additional dimension, but possibly with a varying interval, either predetermined (phases of the moon) or
unpredictable (periods of intense cold).
27
1.2.7.1 Consequences of the discretization
The time discretization of variables will be introduced in several ways, leading to:
• really instantaneous variables, measured at a given point in time: the capital on the 31st of December at
midnight, in an annual model (defined as a stock variable).
• averages: the average level of employment observed during a period.
• flows: the goods produced during a period.
The same economic concept might appear under several forms: inflation and price level, stock of debt and balance for
the period, average and end-of-period employment levels, plus net job creations. For one household, we can consider
the revenue, its yearly change, and the global revenue accumulated during its existence.
1.2.7.2 The seasonality
When models have a less than yearly periodicity, some series can present a specific distortion depending on the sub-
period inside the year. This can come from changes in the climate: in winter the consumption of electricity will
increase due to heating and lighting, but construction will be mostly stopped. It can be due to social issues: the
concentration of holidays in the summer months can reduce production, and the coming of Christmas will increase
consumption (in Christian countries). We are going here to provide a basic sketch of the problem, leaving a more
serious description to specialized books like Ladiray and Quenneville (2001).
Using unprocessed data can lead to problems: for instance, the level of production in the second quarter will be lower
than what we could expect from labor and capital levels. This will disturb estimations and make model solutions more
difficult to interpret.
Two solutions can be considered:
• Introducing in the estimated equations “dummy variables associated to each sub-period.
• Extracting from the series their seasonal component and producing a completely new set of values.
Of course, one should not mix the two types of techniques in the same equation (or model).
The second method will be favored, as it also solves the interpretation problem.
Several techniques are available, the most well-known being Census-X13 ARIMA, developed by the US Census Bureau
and Statistics Canada16. But TRAMO-SEATS17 is also a common choice. Both are available under EViews.
One must be aware that this process often reduces the statistical quality of estimations. For instance, if demand is
particularly high in the last quarter of each year, and imports follow, seasonally adjusting both series will make the link
less clear, bringing less precise results. Even more obviously, the relation between demand for heating and
temperature will lose power from seasonal adjustment.
These examples show the main issue: in a one-equation model, three situations are possible:
The dependent variable contains a seasonal component, in addition to truly economic features. For instance,
agricultural production will be lower in winter, even if the same level of labor, land, fertilizer, machinery is available....
16
https://www.census.gov/srd/www/x13as/
17
http://www.bde.es/webbde/es/secciones/servicio/software/econom.html
28
Truly, at the same time, the use of fertilizer will decrease, and probably of labor too, but in a lower way. One either
adjust this variable or introduce dummy elements. The internal quality of the relationship should be the same, but the
statistical one will improve in appearance, through the correlation of the unadjusted dependent variable with the
explanatory dummies.
On the contrary, if all the seasonal explanation comes from the seasonality of explanatory elements, seasonally
adjusting is not necessary, and even reduces the quality of estimations (with the variability of elements). One could
use raw series to estimate an imports equation, using demand, rate of use of capacities and price competitiveness as
explanatory elements.
But what is true for one equation does not apply to the whole model. One cannot mix the two types of series, and this
means seasonally adjusting will prevail in practice.
1.2.7.3 Static and dynamic models
To determine the equilibrium for a given period, some models will use only variables from this period: we shall call
them static models. They correspond to the formulation
𝑓𝑡 (𝑥𝑡 , 𝑦𝑡 , 𝑎, 𝑢𝑡 ) = 0
The most frequent case is that of input-output models, which use a matrix of "technical coefficients" to compute the
detailed production associated to a given decomposition of demand into categories of goods, which itself depends
only on instantaneous elements.
𝑄 = 𝐴 ⋅ 𝐹𝐷
(A representing an n-by-n square matrix)
𝐹𝐷 = 𝑓(𝑄)
On the contrary, dynamic models use variables from other periods.
• The reasons are quite numerous. They can be:
• theoretical: some agents will be supposed to base their behavior on the observation of the past. Firms will
increase their prices if the profits of the previous quarter have been too low. Or they will build their
expectations of demand growth on the previous evolutions of the same variable. These two examples
illustrate the main issues: using the past to create an image of the future (backward looking expectations), or
to measure a previous gap between actual and target values, which the agent will try to close in the present
period.
29
• institutional: the income tax paid by households can be based on their income of the previous period (this
was the case in France, until 2019).
• technical: if a model considers a variable and its growth rate, computing one from the other considers the
previous level.
One observes that each of these justifications supposes that influences come only from previous periods: one will
speak of (negatively) lagged influences.
The formulation becomes therefore:
𝑓𝑡 (𝑦𝑡 , 𝑦𝑡−1 , . . . . , 𝑦𝑡−𝑘 , 𝑥𝑡 , 𝑥𝑡−1 , . . . . , 𝑥𝑡−𝑙 , 𝑎, 𝑢𝑡 ) = 0
Let us go back to our model. We can observe already an undisputable lagged influence: most of present capital will
come from the remaining part of its previous level. Any other case is still undecided. However, without going too deep
into economic theory, one can think of several lagged influences:
• For household consumption, we have already considered that adapting to a new level of revenue takes some
time. This means it will depend on previous levels. If we detailed it into products, the previous level can have
a positive influence (some consumptions are habit - forming) or a negative one (generally, people do not buy
a new car every quarter):
𝐶𝑂𝑡 = 𝑓(𝐶𝑂𝑡−1 , 𝐶𝑂𝑡−2 , . . . . , 𝐻𝑅𝐼𝑡 )
• Firms invest to adapt their productive capacities to the level of production needed in the future. We can
suppose that they build their expectations on past values.
𝑰𝒕 = 𝒇(𝑸𝒕 , 𝑸𝒕−𝒍 , 𝑸𝒕−𝟐 , . . . . )
It is interesting to note that the previous formulation could be simplified, eliminating any lag larger than one by the
addition of intermediate variables:
𝐟𝐭 (𝐲𝐢,𝐭 , 𝐲𝐣,𝐭−𝐤 ) = 𝟎
(where yi and yj represent variables, indexed by time t and t-k)
is equivalent to
30
𝑓𝑡 (𝑦𝑖,𝑡 , 𝑧𝑗,𝑡 ) = 0
𝑧1,𝑡 = 𝑦1,𝑡−1
𝑧2,𝑡 = 𝑧1,𝑡−1 (= 𝑦𝑗,𝑡−2 )
........
𝑧𝑘−1,𝑡 = 𝑧𝑘−2,𝑡−1 (= 𝑦𝑗,𝑡−𝑘+1 )
𝑧𝑘,𝑡 = 𝑧𝑘−1,𝑡−1 (= 𝑦𝑗,𝑡−𝑘 )
in which a lag of k periods on a single variable has been replaced by k one period lags on as many variables (including
new ones).
The same method clearly allows eliminating lagged exogenous variables.
On the investment equation of the example, this would give:
𝐼𝑡 = 𝑓(𝑄𝑡 , 𝑄1𝑡 , 𝑄2𝑡 , 𝑄3𝑡 )
𝑄1𝑡 = 𝑄𝑡−1
𝑄2𝑡 = 𝑄1𝑡−1
𝑄3𝑡 = 𝑄2𝑡−1
But if this method simplifies the theoretical formulation, it has the obvious disadvantage of artificially increasing the
size of the model and reducing its readability, without producing additional information. Its interest is reserved to
specific studies. For instance, assessing model dynamics can call for the linearization of the model according to
present and lagged variables. The above transformation will limit the matrix system to two elements (with lags 0 and
1), which will make further formal computations easier, and independent from the number of lags.
It also allows us to use a simplified formulation in subsequent presentations:
𝑓𝑡 (𝑦𝑡 , 𝑦𝑡−1 , 𝑥𝑡 , 𝑎, 𝑢𝑡 ) = 0
31
1.2.7.4 A particular case: rational expectations
It has appeared natural, in previous examples, to consider only negative lags. This will happen if we suppose that the
anticipation of agents relies only on the observation of the past (and the present) 18.
To justify positive lag formulations, it is necessary to suppose:
• That agents have the possibility, by their present decisions, to determine the future values of some variables
(and the associated behavior can be formalized).
• That agents perfectly anticipate the future (perfect expectations).
• That the expectation by agents of specific evolutions has for consequence the realization of these evolutions
(self-fulfilling expectations).
• That agents build their expectations on the behaviors of the other agents19, for which they know the
properties (rational expectations). Basically, this means that they are able to apply the model controlling the
economy (but not necessarily know its formulas), and the decision process defining its assumptions. For
instance, they can forecast the investment program of the Government (depending on economic conditions),
they know how firms and households will react, and they know the links between these elements (they are
able to consider the supply-demand equilibrium). Actually, this is rather called “model consistent
expectations”.
• However, they do not necessarily know the unexplained part of the behaviors (which can be associated with
the random term). If know only their distribution, we shall speak of stochastic rational expectations. EViews
does not provide this feature at present (only one or the other), although this should appear in a future
version. They also do not have to know the actual formulas, just be able to compute them.
You do not have to believe in rational expectations to apply them. Producing alternate simulations with different
assumptions on expectations will improve greatly the insight in one particular model or on economic mechanisms in
general. We shall present this later on a specific case.
1.2.7.5 Other case: continuous time models
This also is a very specific area: some theoretical models will be formulated as a system of equations where variables
appear as a function of continuous time, and variations (or growth rates) become exact derivatives. One ends up then
with a system of differential equations, which one can be led to integrate.
These models seldom evolve beyond a theoretical stage, if only for lack of statistical information.
But some operational models, describing for instance the stock exchange, can reduce their periodicity to a daily or
even shorter value.
1.2.8 LINEARITY
We will consider here the linearity relative to variables. The linearity relative to coefficients will appear in the chapter
on estimation.
18
This use of proxies is made necessary by the absence of direct measurement of anticipations. Exceptionally, they
can be obtained by surveys, leading to a specific estimation.
19
Including the State.
32
The potential linearity of a model represents a very important property for its analysis as well as its solution. But first
we must define the notion of linearity, which can be more or less strict.
The most restrictive will be:
𝑨 ⋅ 𝒚𝒕 + 𝑩 ⋅ 𝒚𝒕−𝟏 + 𝑪 ⋅ 𝒙𝒕 + 𝒃 + 𝒖𝒕 = 𝟎
but one can let matrix elements change with time:
𝐴𝑡 ⋅ 𝑦𝑡 + 𝐵𝑡 ⋅ 𝑦𝑡−1 + 𝐶𝑡 ⋅ 𝑥𝑡 + 𝑏𝑡 + 𝑢𝑡 = 0
a definition again less restrictive will suppose linearity relative to the sole endogenous variables:
𝐺(𝑥𝑡 , 𝑎)𝑦𝑡 + 𝐻(𝑥𝑡 , 𝑎)𝑦𝑡−1 + 𝐽(𝑥𝑡 , 𝑎) + 𝑢𝑡 = 0
or even relative to the endogenous of the period:
𝐺(𝑦𝑡−1 , 𝑥𝑡 , 𝑎) ⋅ 𝑦𝑡 + 𝐻(𝑦𝑡−1 , 𝑥𝑡 , 𝑎) + 𝑢𝑡 = 0
Using the multiplier as an example, we can already show that these properties affect the computation of derivatives
of model solutions. We will detail later the consequences on convergence properties.
The first property tells that it does not depend on the initial equilibrium, or the period considered. Multiplying the
shock by a given factor will have a proportional effect. It is enough to compute it once to know it once and for all.
In the second case, the multiplier will depend only on the period. Starting from different base assumptions will not
change the consequences of a given change.
In the third case, the multiplier will depend also on the exogenous values (and the coefficients). It has to be re-
computed each time these elements change (or have changed in the past except for one period ahead solutions), but
can be stored until the next time they do.
The last case is similar to the third one. But convergence will be affected (see later).
1.2.8.1 Practical cases of non-linearity
33
It is obvious enough that a single non-linear equation makes the model non-linear, according to one of the previous
definitions. Reasons for non - linearity are multiple; one will find in particular:
• Expressions measured in growth rates (therefore possibly linear relative to the endogenous of the period).
For example, the growth rate of wages can depend on inflation.
• Expressions formulated as elasticities (generally integrated into logarithms). One will suppose for example
that imports and domestic demand show proportional relative variations.
• Ratios entering in behavioral equations.
• Equations using elements at current prices, computed as the product of a quantity by a deflator (which shows
the evolution of the price compared to a base year). For example, the trade balance will be obtained as the
difference between the products of exports and imports at constant prices by their respective deflators.
Sometimes this distinction is purely formal, and an adequate variable change will allow the return to a linear
formulation. However, if we consider the whole model, replacing by its logarithm a variable computed in elasticities
will only transfer the problem if the level appears also in the model.
Thus, in our general example, if one uses for the exports equation the formulation:
𝐿𝑜𝑔(𝑋) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑊𝐷) + 𝑏
one can very well introduce variables 𝐿𝑋 = 𝐿𝑜𝑔(𝑋) and𝐿𝑊𝐷 = 𝐿𝑜𝑔(𝑊𝐷), which will make the equation
linear:
𝐿𝑋 = 𝑎 ⋅ 𝐿𝑊𝐷 + 𝑏
But it will be necessary, to introduce exports in the supply - demand equilibrium:
𝑄 + 𝑀 = 𝐹𝐷 + 𝑋
to add the non - linear equation
𝑋 = 𝐸𝑥𝑝(𝐿𝑋)
34
Therefore, most economic models presenting a minimum of realism will not be linear. But numerical computations
will generally show that even for models including many formal non - linearities, the approximation by a linearized
form around a model solution (denoted by an asterisk):
(𝜕𝑓𝑡 ⁄𝜕𝑦𝑡 )(𝑦𝑡 − 𝑦𝑡∗ ) + (𝜕𝑓𝑡 ⁄𝜕𝑦𝑡−1 )(𝑦𝑡−1 − 𝑦𝑡−1
∗ )
+ (𝜕𝑓𝑡 ⁄𝜕𝑥𝑡 )(𝑥𝑡 − 𝑥𝑡∗ ) = 0
is acceptable for general purposes.
On the other hand, the stability of the derivatives with time is much more questionable.
Let us suppose the formulation for imports is:
𝐿𝑜𝑔(𝑀𝑡 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝐹𝐷𝑡 ) + 𝑏
Linearizing it around a particular solution (noted *), we get
(𝑀𝑡 − 𝑀𝑡∗ )/𝑀𝑡∗ = 𝑎 ⋅ (𝐹𝐷𝑡 − 𝐹𝐷𝑡∗ )/𝐹𝐷𝑡∗
Or
(𝑀𝑡 − 𝑀𝑡∗ ) = 𝑎 ⋅ 𝑀𝑡∗ /𝐹𝐷𝑡∗ ⋅ (𝐹𝐷𝑡 − 𝐹𝐷𝑡∗ )
which will represent an adequate linear approximation of the connection between M and FD, provided that M and FD
do not move too far away from their base value20. This base value might represent a reference path, from which
actual values differ due to a change in assumptions.
But, if we restrict further the expression to a constant influence (linearity to constant coefficients),
(𝑀𝑡 − 𝑀𝑡∗ ) = 𝑎 ⋅ (𝐹𝐷𝑡 − 𝐹𝐷𝑡∗ )
20
In other words, if the terms of the derivative are negligible beyond the first order.
35
the approximation can be accepted only if the ratio M / FD does not change too much with time. This is not generally
true: the expansion of international trade has led, and still leads, to a sustained growth of the share of imports in
domestic demand, for most countries. The ratio M * / FD * will grow strongly with time, and the last formulation will
be quite inadequate for forecasts.
1.2.9 OTHER PROPERTIES
1.2.9.1 Continuity
We consider here the continuity of the whole set of endogenous variables relative to assumptions (exogenous
variables, parameters). It is almost never verified formally, but should only be considered within the set of acceptable
solutions (and assumptions).
For instance, most models use ratios, which is acceptable if the denominator can never become null (like the
productivity of labor measured as the ratio of production to employment). Or using logarithms to link imports to
demand requires (logically) that those elements are strictly positive. In other words, a fully linear model can produce a
negative GDP, but this does not make it less operational if this value is associated with absurd assumptions or
coefficients.
So even if all models show non-continuity potential, it should never occur in practice. We can think of only three
cases:
• The model framework is correct, but something is wrong with its elements: the numerical assumptions, the
estimated coefficients.
• The algorithm used for solving the model leads to absurd values (more on this later).
• The behavioral equations are wrongly specified. As we also shall see later, it can be dangerous to put
together elements without a previous assessment of the associated mechanisms (for instance using
logarithms as a natural solution).
It is necessary, however, to distinguish these absurd cases from those where the discontinuity applies to the derivative
of a variable differentiable by pieces, as we are going to see in the following paragraph.
1.2.9.2 Differentiability
It is less necessary, but its absence can lead to problems in the system solving phase, as well as in the interpretation of
results.
Separating from the previous criteria is not always straightforward, as the non-derivability of one variable can
correspond to the discontinuity of another: a discontinuous marginal productivity can make the associated production
non-differentiable at points of discontinuity.
Returning to the example, we could formalize household consumption in the following manner:
• They receive a constant share - a - of production Q.
• Under an income threshold - R - they consume a share c0.
• On the supplement they consume a share c1.
The consumption equation will become:
36
𝐶𝑂𝑡 = 𝑐0 ⋅ 𝑎 ⋅ 𝑄𝑡 + (𝑐1 − 𝑐0) ⋅ 𝑚𝑎𝑥( 0, (𝑎 ⋅ 𝑄𝑡 − 𝑅𝑡 ))
At the point Q = R / a, CO is not differentiable (the derivative to the left is c0.a, to the right c1.a). And the sensitivity of
consumption to income is not continuous.
This derivative is not purely formal: it defines the marginal propensity to consume (consumption associated to a
unitary income increase), which can appear itself in the model, at least as a descriptive element.
At the household level, the evolution of income tax as a function of revenue (with rates associated to brackets) would
represent another example, determining disposable household income.
1.2.9.3 Existence of a solution
It is obviously necessary for the model to have a solution, at least when it is provided with acceptable assumptions21.
But the potential absence of a solution is present in many formal systems, including linear models. This absence of
solution is generally logically equivalent to the existence of an absurd solution, as one can illustrate on the following
case.
Let use consider a model with n+1 endogenous variables: X (dimension n) and x (a single variable). We shall describe it
as f, a vector of formulas (dimension n+1), in which x appears as an argument of a logarithm,
𝑓(𝑥, 𝑋, 𝑙𝑜𝑔( 𝑥)) = 0
If none of the positive values of x ensures the solution of the complete model, it has no solution.
In other words, taking the argument of the logarithm as a parameter
𝑓(𝑥, 𝑋, 𝑙𝑜𝑔( 𝛼)) = 0
and making it vary in R+, solving the associated model on x and X will never provide a value of x equal to this
parameter.
The model has obviously no solution.
21
Refusing to provide a solution for absurd assumptions should rather be considered as a quality.
37
But if the model builder has used a formulation in logarithms, he has probably not considered letting the argument
take negative values. By replacing the logarithm by some other expression giving similar values, we would probably
have obtained a solution. But if the variable remains negative, this solution would have been unacceptable.
To illustrate this case, we are going to reduce the usual model to a three equations version.
Production adapts to demand corrected by imports and exports, the last being exogenous:
[1] 𝑄 + 𝑀 = 𝐹𝐷 + 𝑋
as for demand, one supposes that its relative variations are proportional to those of production:
[2] 𝐿𝑜𝑔(𝐹𝐷) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑄) + 𝑏
And imports are a share of demand
[3] 𝑀 = 𝑐 ⋅ 𝐹𝐷
Let us suppose that one has obtained by estimation in the past: a = 1.05 and b > 0, justified by a level and growth of
demand generally superior to production, obviously associated to imports greater (and growing faster) than exports.
Now, let us produce a forecast.
The model can be reduced into:
[1’] 𝐹𝐷 = 𝑄/(1 − 𝑐) − 𝑋
(from (1) and (3))
[2’] 𝐹𝐷 = 𝑄 𝑎 ⋅ 𝑒𝑥𝑝( 𝑏)
38
from (2)
and
[3’] 𝑋/𝑄 = (1/(1 − 𝑐) − 𝑄𝑎−1 ⋅ 𝑒𝑥𝑝( 𝑏))
Obviously, if Q grows (as a-1 = 0.05), the negative element will become eventually higher that the positive one, which
means that Q can only be negative, which is impossible as it enters in a logarithm in equation (2). The model has no
solution.
Of course, these mathematical observations have an economic counterpart. In the long run, final demand cannot
grow continuously faster than production, if imports are a share of demand and exports are fixed. Assumptions,
therefore, are not consistent with the estimated formula.
One will notice that the absence of solution is due here to the implicit adoption of a condition verified numerically on
the past, but not guaranteed in general. This will be in practice the most frequent case.
1.2.9.4 Uniqueness of the solution
The uniqueness of the solution, for given (and reasonable) values of parameters and assumptions, is also very
important. Indeed, we do not see how one could use a model which leaves the choice between several solutions,
except maybe if this freedom has a precise economic meaning.
In practice, most models are highly nonlinear if you look at the equations, but the linear approximation is rather
accurate within the domain of economically acceptable solutions. This limits the possibility of multiple equilibria: if the
system was fully linear, and the associated matrixes regular, there would be indeed a single solution. However, as we
move away from this domain, the quasi linearity disappears, and we cannot eliminate the possibility of alternate
solutions, probably far enough from the reasonable solution to appear quite absurd. Fortunately, if we start
computations inside the domain, an efficient algorithm will converge to the acceptable equilibrium, and we will never
even know about any other.
The most significant exception will be that of optimization models, which look for values of variables giving the best
result for a given objective (for example the set of tax decreases which will produce the highest decrease in
unemployment, given a certain cost): if several combinations of values give a result equal in quality 22, this lack of
determination will not undermine the significance of the solution. The existence of several (or an infinity of) solutions
will represent an economic diagnosis, which will have to be interpreted in economic terms23.
Another case appears when the formula represents the inversion of another formula giving a constant value, at least
on a c
ertain interval. For example, if over a certain threshold of income households save all of it:
22
For instance, if the model is too simple to differentiate the role of two taxes.
23
provided the algorithm used for solving the model is able to manage this indetermination.
39
𝐶𝑂 = 𝑚𝑖𝑛( 𝑓(𝑄), 𝐶𝑂∗ )
Then the income level associated with CO * will represent the total set of values higher than the threshold.
In the general case, the main danger appears in sensitivity studies: if one wants to measure and interpret the
economic effects of a modification of assumptions, the existence of a unique reference simulation is an absolute
necessity.
Finally, finding several solutions very close to each other might come from purely numerical problems, due to the
imprecision of the algorithm: any element of the set can then be accepted, if the difference is sufficiently low.
1.2.9.5 Convexity (or concavity)
The convexity of the system, which is the convexity of the evolution of each endogenous variable with each
exogenous variable and parameter taken individually (or of a linear combination of them), can be requested by some
algorithms, especially in optimization. In practice it is very difficult to establish, and even rarely verified. At any rate,
this characteristic is linked to the definition of variables, and a single change of variables might make it disappear.
1.2.10 CONSTRAINTS THE MODEL MUST MEET
In addition to its theoretical validity, the model will have to meet a set of more technical constraints.
1.2.10.1 Global compatibility
Constraints of compatibility will bear in practice:
a - on the endogenous between themselves: one cannot let the model compute variables independently if they are
linked by a logical relationship, accounting or theoretical. For example, if the consumer price enters in the
determination of the wage rate, it also will have to be influenced directly by the (estimated) price of local production.
Or the employment level has to affect household revenue and consumption through a sequence of links.
Accounting balances must be verified: once household revenue has been computed as a sum of elements, an increase
in consumption must produce the associated decrease in savings.
Maybe the most important issue lies with the « supply = demand » identity, which will have to enforced both in at
constant and current prices. This can lead either to use one of its elements to balance the equation, or to distribute
the residual over the global set of elements on one side. By formulating total supply and demand as:
n
O = Oi
i=1
And
40
m
D = Dj
j=1
One will use for instance, either
m-1
Dm = O - D j
j=1
Or one will correct the set of demand variables by multiplying each of them by the uncorrected ratio O / D.
In most cases the equilibrium at constant prices will be enforced automatically. It can be written as:
Local production + Imports = Local demand + Exports
Or identifying intermediate consumption:
Local GDP + Intermediate consumption + Imports = Local final demand + Intermediate consumption + Exports
• With only one product, intermediate consumption can be discarded, and one will generally use the equation
to compute GDP, checking that it does not get higher that productive capacity24 .
• With several products, we must consider as many equilibrium equations, in which the supply of intermediate
consumption goods sums inputs needed for production of the good, and the demand for intermediate
consumption goods sums the intermediate uses of the good itself.
𝑄𝑖 + ∑𝑗 𝐼𝐶𝑗,𝑖 + 𝑀𝑖 = 𝐹𝐷𝑖 + ∑𝑗 𝐼𝐶𝑖,𝑗 + 𝑋𝑖
24
This can be obtained by a share of imports growing with constraints on local productive capacity.
41
If we suppose that returns to scale are constant, the vector of value added by good will come from a matrix
transformation. The constraint on capacity will be achieved in the same way as above (provided a capacity equation
can be obtained).
Defining ci,j as the quantity of good i needed to produce one unit of good j, we get:
𝑄𝑖 + ∑𝑗 𝑐𝑗,𝑖 𝑄𝑖 + 𝑀𝑖 = 𝐹𝐷𝑖 + ∑𝑗 𝑐𝑖,𝑗 𝑄𝑗 + 𝑋𝑖
Or in matrix terms
𝑄 + 𝐶 ⋅ 𝑄 + 𝑀 = 𝐹𝐷 + 𝑡𝐶 ⋅ 𝑄 + 𝑋
or
Q = ( I − C + t C) −1 ( FD + X − M )
Using this framework will automatically enforce the supply-demand equilibrium for all goods.
In practice, most of the problem comes from the equilibrium at current prices. If demand prices are computed
individually using behavioral equations, there is no chance the equilibrium will be met. The process described earlier
will in practice correct the prices. With S and D as supply and demand elements at constant prices, ps and pd as the
associated deflators, we can compute the global values as:
SV= ∑𝑛i=1 ps𝑖 𝑆𝑖
DV= ∑𝑚
j=1 pd𝑗 𝐷𝑗
The first option will compute a specific price
42
m-1
pd𝑚 = (SV- ∑ DV𝑗 )/𝐷𝑚
j=1
and the second
pd𝑗′ = (SV/ ∑ pd𝑗 𝐷𝑗 )pd𝑗
j=1
where the “pd” elements are the independently computed demand prices, and the “pd'” elements the corrected
values.
The correcting factor:
r = SV/ ∑𝑚
j=1 pd𝑗 𝐷𝑗
can also be written as
r = r (SV/ ∑𝑚
j=1 pd'𝑗 𝐷𝑗 )
which with
pd 'j = r pd j
gives a set of equations ensuring the equilibrium. As “r” measures the potential discrepancy between supply and
demand, one must check that it is not too different from one.
The following issues appear:
43
• With the first method, which element should be used to balance the system? The choice is between
o A small and unimportant variable, to reduce the consequences for model properties; perhaps even a variable
which has absolutely no influence on the rest of the model.
o A variable with large value, to reduce the correcting factor
• The second method represents an extreme application of the first one, where all variables on one side are
affected in the same proportional way.
Actually, none of the solutions dominates clearly, the worst being in our sense the very first, which is the same as
accepting de facto an imbalance, hidden but with potentially damaging consequences. Also, the second could be
associated with a converging economic process, while the first can have no economic interpretation whatsoever.
In fact, one should concentrate on limiting the size of the correction itself. One could represent the problem as
eliminating dust: instead of storing it in a specific location (under a carpet), or spreading it all over the place, the best
solution is clearly to reduce its production as much as possible. This means that the initially computed prices should be
designed to give naturally close values to global supply and demand.
b - on exogenous-> endogenous connections: Connections must be formulated with care. For example, if the social
contributions rate is defined as an exogenous variable in the model, it has to enter in all computations of contribution
levels. In particular, it cannot coexist with an exogenous representation of contributions, or one using an estimated
coefficient.
To avoid this type of error, a systematic study of model properties must be undertaken before any operational
application: in our example, this would mean checking that an increase of the social contribution rate has all expected
effects on State revenues as well as on the accounts and behaviors of concerned agents.
Also, the true exogenous concept should be decided. Concerning contributions, the decision variable is clearly its rate,
while the associated revenue is influenced by endogenous prices and employment.
c - on the exogenous between themselves: one should avoid defining two variables as exogenous if they are linked (in
any direction) by a logical relationship. If possible, one should endogenize one of them by formalizing this connection.
Let us suppose for example that a model for France uses two exogenous measures of the prices established by its
foreign competitors: in foreign currency and in Euros (with a fixed exchange rate). To take into account an increase of
foreign inflation, these two variables will have to be modified simultaneously. This is as best more complex, and can
lead to errors if one is not careful enough, while it can be avoided simply by endogenizing the price in Euros as the
product of the price in foreign currency by the (exogenous) exchange rate.
However, establishing such links is not always possible. For instance, in a national model, foreign prices and foreign
production are exogenous, but also clearly influenced by each other. But the nature and importance of the link are
highly variable. For instance, a decrease in foreign production can produce world deflation,25 while inflation can
reduce exports and production. To describe them completely one should have to resort to a foreign or world model.
An intermediary solution could be to establish a set of linear multipliers linking these elements, but generally the
model builder himself will take care of the problem by producing a set of consistent assumptions (with perhaps some
help from specialists of the world economy, or from a separate model).
25
This is the case for the MacSim world model we shall present later.
44
d - on endogenous-> exogenous connections: they are obviously proscribed, because contrary to the preceding links
the model builder cannot master them. They will be found in some models, however, through the presence of the
following exogenous:
• Elements measured in constant terms, while they should change with economic activity.
• Deflators, which should depend on other deflators.
• Elements measured in current terms, for both reasons.
If the associated model can possibly produce correct estimates and even forecasts, it runs the risk of showing
abnormal sensitivity properties. Let us take an example:
Let us suppose household income HI is composed
• Of the wage revenue, computed as the product of employment by the wage rate: LT . W.
• of other exogenous revenues
Salaries will be indexed perfectly on prices:
𝑊 = 𝑊𝑅 ⋅ 𝐶𝑃𝐼
One will have therefore:
𝑅𝐻𝐼 = 𝑊𝑅 ⋅ 𝐶𝑃𝐼 ⋅ 𝐿𝑇 + 𝐻𝐼𝑄
This equation might perform well in forecasts. But if a change in the assumptions makes prices increase, the
purchasing power of total wages will remain unchanged, but for the complement HIQ it will be reduced in the same
proportion as the price rise:
𝛥(𝐻𝐼𝑄/𝐶𝑃𝐼) = −(𝛥𝐶𝑃𝐼/𝐶𝑃𝐼) ⋅ (𝐻𝐼𝑄/𝐶𝑃𝐼)
One can question this assumption. Some elements in non-wage revenue (social benefits, rents, firm owner’s profits,
independent workers revenue) are more or less indexed, and can even be over indexed in the case of interest
payments (the interest rate should increase with inflation). Others, associated to differed payments (dividends,
income tax) will not change immediately. The global sensitivity to prices is not clear, but a null value is obviously not
correct.
We will face the same problem with a change in GDP:
45
𝛥(𝐻𝐼𝑄/𝑄) = −(𝛥𝑄/𝑄) ⋅ (𝐻𝐼𝑄/𝑄)
where we cannot suppose that revenue does not change (grow) with economic activity. Some elements do not or
show a limited sensitivity (pensions) but dividends and the revenue of owners of small firms certainly do.
In conclusion, even when a variable measured at current prices has no theoretical content, it should not be kept
exogenous, especially if it can be supposed to grow at constant prices. It is general better to consider as exogenous its
ratio to another variable, supposed to follow the same trend (in the absence of idea, one can use plain GDP). The
model equation will compute the variable by applying the exogenous ratio. This is also can be valid for variables at
constant prices (which generally increase with production), to the exception of decision variables identified as such.
In the case above, one could write:
𝐻𝐼 = 𝑊𝑅 ⋅ 𝐶𝑃𝐼 ⋅ 𝐿𝑇 + 𝐶𝑃𝐼 ⋅ 𝑄 ⋅ 𝑟_ℎ𝑖𝑞
in which the introduction of Q links additional revenue to the global growth of the economy.
1.2.10.2 Homogeneity
If some equations in a model do not meet homogeneity constraints, this endangers its properties, particularly its
sensitivity to shocks. Let us quote some cases:
• Linear relationships between values and quantities. The equation:
CO (consumption at constant prices) = a HRI (current income) + b is not only absurd from a theoretical viewpoint,
but will lead in the long-term to a level of savings
𝑆 = 𝐻𝑅𝐼 − 𝐶𝑃𝐼 ⋅ (𝑎 ⋅ 𝐻𝑅𝐼 + 𝑏)
that will become clearly negative over a certain price level.
• Mixing logarithms and levels. Similarly, the equation:
𝐶𝑂 = 𝑎 ⋅ 𝐿𝑜𝑔(𝐻𝑅𝐼) + 𝑏
46
(this time the two elements will be measured in quantities) makes the ratio CO / HRI decrease to 0, and therefore the
savings rate to 1, when HRI grows indefinitely.
This last example shows however a limit to the argument: on short periods the equation can present a satisfactory
adjustment, as the consumption to income ratio (propensity to consume, complement to 1 of the savings rate)
decreases effectively with income. It is the speed of the decrease, and its long-term evolution, which is questioned
here.
1.2.10.3 Constants with dimension
The problem is identical to that of the exogenous with dimension. It invites a careful study of the theoretical content
of the constant. Furthermore, as most variables grow with time, the influence of the constant will generally decrease
or even disappear in practice. We shall address this issue later, on a practical case.
1.2.11 NORMALIZATION AND IDENTIFICATION
Once equations have been estimated, the problem of normalization remains. We have seen that very often the
estimated formula will not explain a variable, but an expression (logarithm, growth rate, ratio, or a more complex
expression). But some simulation algorithms will request a model a specific form, called “identified”, in which a single
untransformed variable appears on the left-hand side:
𝑦𝑡 = 𝑓𝑡 (𝑦𝑡 , 𝑦𝑡−1 , 𝑥𝑡 , 𝑎, 𝑢𝑡 )
This means the model builder might have, after estimation, to transform the formulation: this operation is called the
normalization of the model.
The advantage is double:
• The application of some solution algorithms is made easier. In some cases (Gauss-Seidel), this form is actually
requested.
• This type of formulation allows a better interpretation of the process determining the equilibrium, provided
each equation can be interpreted as a causal relation. If the equation describes a behavior, the economist
should have placed to the left the element it is supposed to determine, conditional on the elements on the
right. This is what we can (and will) do naturally in our example. For instance, the equation describing the
choice by households of their consumption level will place naturally the variable "consumption" to the left.
The vast majority of equations will take naturally an identified form. Sometimes, a simple transformation will be
necessary, however. Perhaps the most frequent nonlinear operator is the logarithm, associated with the integration of
a formula in elasticities.
𝑑𝑥/𝑥 = 𝑓(. . . )
represents
47
𝐿𝑜𝑔(𝑥) = ∫ 𝑓(. . . . ) ⋅ 𝑑𝑥
In this case, one just needs to replace:
𝐿𝑜𝑔(𝑥) = 𝑓(. . . . )
by
𝑥 = 𝑒𝑥𝑝( 𝑓(. . . . ))
If you use EViews26, the software will do it for you. You can write the equation using the first form, and the software
will normalize the equation itself, computing x. This is also true if the left-hand element contains several variables, but
allows straightforward normalization. The most frequent cases are:
A change in logarithm: 𝐿𝑜𝑔(𝑥𝑡 /𝑥𝑡−1 ) = 𝑓(. . . . )
A growth rate: (𝑥𝑡 − 𝑥𝑡−1 )/𝑥𝑡−1 = 𝑓(. . . . )
A ratio: 𝑥𝑡 /𝑦𝑡 = 𝑓(. . . . )
To choose which variable to compute, EViews will take the first variable in the specification of the equation. This
simple method will be applied even if the variable has been identified as computed by a previous equation. For
instance, in our model, if we introduce the estimation of imports M, then state:
𝑀 + 𝑄 = 𝐹𝐷 + 𝑋
EViews will give an error message, as M appears to be computed twice.
26
Or most packages of the same type.
48
Moreover, when an equation is forecasted individually, one can chose between the computation of the left-hand term
and the element which determines it, for instance M or log (M) for our imports equation.
However, EViews does not solve analytically any equation for the variable. For instance:
𝑀/(𝑄 + 𝑀) = 𝑓(. . . . )
will be translated into:
𝑀 = (𝑄 + 𝑀) ⋅ 𝑓(. . . . )
introducing a non-recursive process over M.
In any event, normalizing the general equation
𝑓(𝑦, . . . . ) = 0
is possible by adding on both sides the same variable, which gives:
𝑦 = 𝑦 + 𝑓(𝑦, . . . . )
However the convergence of a model defined in this manner is often difficult to obtain (for instance if “f” is positively
linked to y). In that case, one can use (the value for “a” can be negative):
𝑦 = 𝑦 + 𝑎 ⋅ 𝑓(𝑦, … . )
Stronger simplifications are sometimes possible and will be approached with the numerical solution process.
49
Identification is not always economically straightforward: in our example, when balancing demand and supply, we can
observe that three last variables (Final demand, Exports and Imports) are going to be determined by their own
equation (the sum of its elements for the first, estimated equations for the others). This means that balancing must be
done through GDP, and we must write the equation as:
𝑄 + 𝑀 = 𝐹𝐷 + 𝑋
or
𝑄 = (𝐹𝐷 − 𝑀) + 𝑋
which makes its theoretical content clearer as: production must (and can) satisfy both exports and the non-imported
part of domestic demand.
50
1.2.12 CONCLUSION
It must be clear by now that the formal definition of the whole set of equations represents with the estimation of
behavioral equations an iterative and simultaneous process:
• Behavioral equations start from an initial theoretical formulation to evolve gradually to their final form by
reconciling this theory with data and estimation results.
• Accounting equations have been defined as precisely as possible in the preliminary phase, to establish a
coherent framework, but they often will have to adapt to the evolution of behavioral equations. Let us
suppose for example that the first estimation results suggest excluding from the econometric explanation of
exports their agricultural component, setting it as exogenous: a new equation and variable will appear, and
the equation for total exports will become an identity.
51
2 CHAPTER 2: MODEL APPLICATIONS
1. We shall now give a panorama of applications using models. Comments will be centered on the example of economic
models, and more particularly on the macro-economic ones. But most of the observations can be transposed to the
general case.
2. For each of these applications, technical details shall be left to the "implementation" part (chapter 7). To understand
these practical aspects of the use of models, one must first know about the way they are built, described later in
chapters 4 to 8.
2.1 OPERATIONAL DIAGNOSES
3. The most natural use of a model seems to be the evaluation of the economic future, whether as its most probable
evolution or as the consequences of some decisions. Assumptions concerning the future will be injected into the
model, and its solution will produce the requested diagnosis. Thus, one will seek to anticipate the evolution of the
main aggregates of the French economy until the year 2020, taking into account assumptions on the evolution of
international economy.
2.1.1 DIFFERENT TYPES OF DIAGNOSES: SCENARIOS AND SHOCKS
Two types of forecasts can be considered: scenarios and shocks.
• In a scenario, one is interested in absolute results, and associating to a full set of assumptions a future
evolution of the economic equilibrium. One might seek to obtain
o forecasts based on the most probable assumptions
o forecasts associated to a given set (like a party’s program)
o an evaluation of the scope of potential evolutions
o assumptions allowing to reach specific economic targets.
• On the contrary, with a shock, one starts from a base simulation (often called "reference forecast" or
“baseline”), or a simulation on the historical period, and measures the sensitivity of the economic equilibrium
to a change of assumptions. Two economic paths will then be compared (on the past, one of them can be the
historical one).
These shocks can be more or less complex, from the modification of a single assumption to the testing of a new
economic policy27.
These two techniques, scenarios and shocks, before the production of any operational policy diagnosis, will play an
important role in the model validation process.
2.1.2 ADVANTAGES OF MODELS
Now that we have described the characteristics of models and their basic use, we shall discuss the advantages they
bring (and their failings too).
Relative to the diagnosis provided by a human expert, advantages common to all models will:
27
However, this new policy should stay within the economic framework of the original model.
52
• Guarantee the accounting coherence of the resulting equilibrium.
• Consider a practically unlimited number of interdependent influences.
• Provide an explicit formalization of behaviors, allowing an external user to interpret them.
• Produce an exact and instantaneous computation of associated formulas.
• Adapt immediately the full system to a local change of theoretical formulation.
but also
• Allow the stability of reasoning, for human users of an unchanged model.
• Provide the possibility of formal comparisons with other models.
This forecasting ambition was already the basis for the construction of the first models. But this type of use has
benefited (since the 1970s) from some evolutions:
• The progress of economic theory, allowing the formalization of more sophisticated mechanisms, better
adapted to the observed reality.
• The progress of econometrics, giving access to the statistical method that will produce the most reliable
formulation associated with a given problem, and to test more complex assumptions.
• The improvement of numerical algorithms, both for computation speed, and solving more complex systems.
• The simultaneous improvement of computation hardware allowing to process problems of growing size, by
increasingly complex methods.
• The progress of modelling science, in producing models better adapted to the original problem, facilitating
the production of assumptions, and reducing the cost of reaching acceptable solutions.
• The production of computer software specialized in model building, increasingly efficient, user-friendly, and
connected with other packages.
• The improvement of the reliability of data, and the growth of the available sample, regarding both the scope
of series and the number of observations (years and periodicity)28.
• The easier communication between modelers, through direct contact and forums, allowing to communicate
ideas, programs and methods, and to get the solution to small and large problems already addressed by
others.
2.1.3 A CERTAIN REASSESSING
However, the use of models has engendered criticism from the start, using often the term « black box », describing
the difficulty in controlling and understanding a set of mechanisms often individually simple but globally very complex.
In recent decades criticism has mounted to the point of calling for a global rejection of traditional (“structural”)
models. Surprisingly, critics often find their arguments in the above improvements. One can find:
A utilitarian critique: models have proven unable to correctly anticipate the future. If this observation has appeared
(in the beginning of the eighties), it is obviously not because the quality of models has declined. But information on
model performance is more accessible (some systematic studies have been produced), and the fluctuations following
28
However, the size of samples does not necessarily grow with time. In a system of national accounts, the base year
has to be changed from time to time, and the old data is not necessarily converted.
53
the first oil shock have made forecasting more difficult. In periods of sustained and regular growth, extrapolating a
tendency is very easy for experts as well as for models. Paradoxically, the emergence of this criticism has followed,
rather than preceded, the increasingly direct intervention of model builders and their partners in forecasting results.
An econometric critique: modern techniques require a quantity and a quality of observations that available samples
have not followed. A gap has opened between estimation methods judged by econometrics theoreticians as the only
ones acceptable, and methods really applicable to a model29.
A theoretical critique: the development of economic theory often leads to sophisticated formulations that available
information has difficulty to validate. And in any event many areas present several alternate theories, between which
any choice runs the risk of being criticized by a majority of economists. Thus, in the monetary area, going beyond a
basic framework leads to rely on information unavailable in practice, or on formulations too complex to be estimated.
A mixed critique: users of models are no longer passive clients. They criticize formulations, as to their estimated
specification, or their numerical properties. This evolution is paradoxically favored by the improvement of the logical
interpretation of economic mechanisms, itself fathered essentially by economic knowledge (even the economic
magazine articles use implicit macroeconomic relations) and modelling practice (the population of clients includes
more and more previous model builders or at least followers of courses on modelling). One could say that model users
ask the tool to go beyond their own spontaneous diagnosis, and they want this additional information to be justified.
It is clear that these criticisms grow in relevance as the goal grows in ambition. Forecasts are more vulnerable than
simple indicative projections, which seek to cover the field of the possible evolutions. As for policy shock studies, they
are not prone to errors on the baseline assumptions, if we discount non-linearities30.
This relevance will also depend on credit granted to results. One can use figures as such, or be content with orders of
magnitude, or even simply seek to better understand global economic mechanisms by locating the most influential
interactions (possibly involving complex causal chains). In our sense, it is in this last aspect that the use of models is
the most fruitful and the least disputable31.
2.2 THEORETICAL MODELS
Contrary to previous models, theoretical models may be built for the single purpose of formalizing an economic
theory. It may be sufficient to write their equations, associating to a theoretical behavior a coherent and complete
system. Reproducing the observed reality is not the main goal of these models, and it is not mandatory to estimate
parameters: one can choose an arbitrary value, often dictated by the theory itself. In fact, this estimation will often be
technically impossible, when some of the variables used are not observed statistically (the goals or expectations of
agents for example).
29
Actually, the sample size required by present techniques (50 or better 100 observations) limits the possibility of
estimating equations using deflators or variables at constant prices. Even using quarterly data, separating values into
prices and volumes is quite questionable 15 years from the base period.
30
With a linear model, the consequence of a shock depends only on its size, not on the simulation it starts from.
31
One example is the impact of a decrease in local tariffs. Ex-ante it increases imports (a negative demand shock). Ex-
post it decreases local factor costs (with cheaper investment and cheaper labor, indexed on a lower consumption
price). This leads to more local capacity and competitiveness, both on the local scene (limiting the imports increase),
and the foreign one. In most models, GDP decreases then grows.
The full interpretation of such a shock provides a lot of information, even if one remains at the non-quantitative level.
54
However, even based on an artificial series and arbitrary parameters, the numerical simulation of these models can be
interesting. Actually, the formulas are often so complex that solving the model numerically will be necessary to
observe its solutions as well as properties (such as the sensitivity of solutions to assumptions and to coefficients).
2.3 QUANTIFIED SMALL MODELS
2.3.1 WITH SCIENTIFIC PURPOSE
These models represent an intermediate case. One seeks a realistic representation of the economy, adapted to
observed reality, but sufficiently simple to accept the application of complex analysis methods (and the interpretation
of their results). In addition to scientific research, this study can be done to measure and to analyze properties of an
operational model on a simplified representation (in the eighties Minims, then MicroDMS have been used to
characterize the Dynamic Multi Sectorial model of INSEE).
There are two categories of methods:
• “External” methods will use model simulations to observe its quantitative properties, and infer a descriptive
comment, both statistical and economic.
• “Internal” methods seek to explain properties of the model by its structural characteristics, using
mathematical tools. This does not necessarily call for actual simulations.
2.3.2 WITH AN EDUCATIONAL PURPOSE
Although often of the same type as the ones above, these models try to present economic mechanisms as complete as
possible, based on real data, under an interpretable and concise form. If necessary, one will favor the message
contained in the presentation over the respect of statistical criteria.
This is the case of the MacSim package, allowing students to interpret international mechanisms and interactions.
55
3 CHAPTER 3: MODEL TYPES
We shall now try to establish a classification of models, focusing on the link between the model’s characteristics and
the goal for which it has been built.
3.1 THE FIELD
The field described by a model is characterized by the variables it computes, but also by assumptions it takes into
account.
In the economic model subset, we can consider:
• A geographical field: national models, multinational models, world models. These last can be built in two
ways: by putting together preexisting national models, with potentially quite different structures, or by
building simultaneously country models of identical structure, possibly with a single team of modelers. We
shall deal with this later.
• A theoretical field: the theory used for the formalization of the model may or may not approach specific
economic aspects. A Keynesian model might limit the treatment of monetary aspects. A short-term model
will not formalize demographic evolutions.
• A field of units: a model might present only variables at constant prices, or physical quantities like barrels of
oil or number of pigs.
• A field of agents: a model will describe the behavior of a single agent: households, the State, firms.
• A field of goods: a model might consider only the production and the consumption of one good, for example
energy. An energy model can use physical units.
There are other types of fields. However, the distinction is not always easy: some models will describe summarily a
global field, except for a certain aspect on which it will concentrate. An energy model, to consider interactions with
the rest of the economy, will have to model it also, but not in the same detail. And it can mix physical units (barrels of
oil or gigawatts) with national accounts elements, with obvious conversion problems.
On the other hand, it will always be possible, and made easier by some modelling packages, to change (actually to
restrict) at the time of simulation the scope of the model. The distinction is then no longer permanent: a multi-
national model can be used to simulate a complete evolution of the world economy, but its user can also restrict
calculations to the evolution of a group of countries or even a single one, the other elements being fixed. One can
simulate a model of the real economy with or without additional monetary features. Or a model using normally
rational expectation elements can drop them to become purely backward looking.
3.2 THE SIZE
The history of modelling shows that for a long period new models generally have seen their size grow, for the reasons
cited earlier: the progress of model-building techniques, the increased availability of data, the faster computer
computations. Additionally, for any given model, size increases regularly in the course of its existence, as new team
members want to add their contribution.
However, the last decades have seen a trend favoring a return to models of limited size. Productivity improvements,
requested from teams of model builders, are less and less compatible with the management of a large model. Despite
the progress of model-building techniques, the desire to reduce costs and delays conflicts with the size, especially (but
not only) regarding human operations: elaboration of assumptions and interpretation of results.
56
Also, the use of a very detailed model can make individual estimations and specifications look too expensive. The
attractiveness of a calibrated CGE model will increase.
Finally, the desire to reply to critics comparing models to "black boxes" leads model builders to look for more explicit
and manageable instruments.
3.2.1 QUASI-ACCOUNTING MODELS
However, paradoxically, the need for detailed explanations, the availability of more detailed data, and the increased
power of computers (both in speed and size of manageable problems) has led to the development of more detailed
(often extremely detailed) tools: the Quasi-Accounting versions, considering generally a large number of products
(possibly hundreds), with a limit depending only on the availability of data. The framework is generally a full input-
output table.
Of course, econometrics are no longer applicable, and formulations are most often rather crude, with behaviors
established as exogenous ratios. But this also makes specifications clearer and more manageable, and the properties
easier to control. Also, the request for large samples is less strong.
This issue will be treated in a specific part. Among the cases we will present, we will consider a model of more than
15 000 equations, which can be summarized by collapsing the dimensions into a 50 equations presentation.
3.2.2 DETERMINANTS OF THE SIZE
Determinants of the size of the model will be:
• The size of the field covered (see above).
• The degree of aggregation, which can be
o vertical: number of operations considered (for example one can distinguish several types of subsidies, or
social benefits),
o or horizontal: number of agents listed; one can distinguish more or less sectors of firms, or categories of
households.
The degree of aggregation will not be inevitably uniform: an energy model will use a particularly fine detail for energy
products.
In fact the same model can appear under several versions of different size, depending especially on the degree of
aggregation. Each version has then its proper area of utilization: detailed forecasts, quick simulations, mathematical
analysis, and educational uses.
Thus at the end of the 1980s, the 3000 equation D.M.S model (Dynamics Multi Sectorial) used by INSEE for its
medium-term forecasts had two companion versions of reduced size: Mini - DMS (200 equations), used for some
operational projections and analysis which did not require detailed products, and Micro - DMS (45 equations), with an
essentially educational purpose.
This distinction has lost most of its validity, however, following the reduction of the size of operational models.
57
3.3 THE HORIZON
3.3.1 FOR FORECASTING 32
If a model is designed for forecasting, its horizon will be defined at the construction of the model. It will be strongly
linked to its general philosophy and to the set of mechanisms it implements. A long-term model will be little
interested in circumstantial phenomena (such as the lags in the adjustment of wages to prices), while a short-term
one will not consider the longest trends (such as the influence of the economic situation on demography).
These differences seem to discard elaborating a model that can be used for both short - and long-term projections.
But we shall see that strong reasons, in particular econometric, have made this option appear as the most natural in
the present context. We will develop them when we address periodicity, in paragraph 3.4.
In any case, one can find a certain asymmetry in the relevance of this observation. If long-term models can neglect
intermediate periods if they do not show significant fluctuations, simulation of the periods beyond the operational
horizon can evidence future problems, already present but not visible in the short-term.
It is clear:
• That treating medium or even short-term problems calls for a model with stabilizing properties, which can
only be controlled through long-term simulations. This includes in particular controlling the existence and
speed of numerical convergence and evidencing cyclical properties.
• That observing long-term properties has to be complemented by the intermediary profile. Again, a long-term
stabilization can be obtained through a monotonous or cyclical process.
3.3.2 FOR MODEL ANALYSIS
Here, the horizon depends on the type of analysis one wants to produce. Often, to analyze a model built with a given
forecasting horizon, simulation over a longer period must be obtained. Even more than for forecasts, analytic shocks
will show and explain anomalies that were not apparent in the normal projection period, but had already a
significantly harmful influence. We shall stress these issues later.
3.3.3 A TENTATIVE CLASSIFICATION
One could use the following classification:
• Short-term models: 1 quarter to 2 years.
• Medium-term models: 4 to 7 years.
• Long-term models: 10 years and more.
Obviously, for a dynamic simulation, the full path, including intermediate values, is of interest.
3.4 THE PERIODICITY
32
One shall notice that we can use several words to characterize these exercises: forecasts, projections, scenarios,
simulations. It all depends on the purpose for which the test was made, and perhaps the trust allowed to the results.
We favor the last term, which unfortunately has to be completed into: « simulation over future periods ».
58
The periodicity of a model is linked to the mechanisms it seeks to study and therefore to its horizon.
Short-term models require a short periodicity to consider circumstantial phenomena: delays in the wage indexation
on prices, progressive adjustment of the consumption level to an increase of income.
Long-term models can use a sparser periodicity, less for theoretical reasons (long-term behavior can be described by a
short-periodicity model), than for technical ones: this choice will reduce constraints on the availability of series,
facilitate the production of assumptions, and limit simulation costs.
However, we shall see that the use of “modern” econometrics methods calls for a short periodicity, for all kinds of
models, as soon as estimations are considered.
This means that the main determinant of model periodicity comes from the data. Countries which produce quarterly
national accounts use quarterly models, which allow them to apply modern techniques with some comfort, and
produce both short and long-term studies. Of course, results can be summarized later in yearly tables.
When only yearly accounts are available, the techniques become more simplistic, and true short-term applications are
not possible. Unfortunately, this applies most often to countries with a short history of statistics, making the problem
the harder.
3.5 OTHER MODELS
We have essentially concentrated on the macro-economic model case. One can also find:
• Micro-economic models: describing the behavior of firms, of households.
These models will sometimes be more theoretical, calling for optimization computations (such as cost minimization)
or to elements of strategy (game theory). They will often be estimated on survey data, with very large samples
• Non-economic models: they can apply to biology, physics, chemistry, astronomy, meteorology, ecology,
process control, and so on.... and can be used to evaluate the consequences of building a dam, controlling a
manufacturing process, looking for the best organization of a project, describing a biological process. These
models will often be conceived not as a formalized equation system, but as the maximization of a criterion
under some constraints, or as a system of propositions connected by logical operators.
59
4 CHAPTER 4: GENERAL ELEMENTS
This part of the book describes the process of development, use and management of a model, taking special interest
in technical aspects and particularly computer-oriented features. Application to EViews will be presented in detail, but
most of the teachings can be applied to other packages, including those which are not dedicated to econometric
structural modelling.
But let us give first a quick description of the organization of the model building process.
4.1 THE STAGES IN THE PROCESS
4.1.1 PREPARING THE MODEL
The first step in the building of any model is producing a draft which ensures some compatibility between available
data (wherever it might come from) and the type of model its builder has in mind (goal, field, nature of the variables,
underlying theory).
Knowing the scope of available data, the economist will define a model framework for which values can be attributed
to all variables, either by using available elements, by computation or as a last resort to expert advice (including the
modeler itself). This means that a first decision has to be made as to the field described by the model, the variables
used as assumptions, and the variables it shall compute. Moreover, he must divide the equations into identities, which
set indisputable links between variables, and equations describing the behavior of agents, for which the final
formulation will be based on past evolutions of the associated elements. In the course of model building, this status
can change.
The first task will be to gather, by reading from files and transforming the data, the full set of variables needed by the
model, to define the form of the identities, and give a first assessment of the behaviors he intends to describe. He
shall check for which periods the necessary data is known, and that on these periods identities hold true. If some
elements are not available, he will use the best proxies he can get. And if this also fails, he will use his imagination.
He can also make a first economic analysis of the framework implied by model specifications. This is greatly helped by
EViews which can give essential information on the model’s logic, even in the absence of any data.
4.1.2 ESTIMATION
The second phase will look for a satisfying description of the behavior of agents, by checking economic theory against
available data. The modeler shall define a set of formulations with unknown parameters, compute for each
formulation the values which give the best explanation of past evolutions, and make his selection, using as criteria
both statistical tests and compliance to economic theory. This process can call for the introduction of new variables, or
changes in some definitions, which will mean reformulating some identities.
Of course, both individual and global consistencies must be applied. For instance, using a Cobb-Douglas production
function implies considering the global cost in the equation for the output deflator.
4.1.3 SOLVING AND TESTING OVER THE PAST.
Once the full model is defined, one can try to solve it.
60
• One shall first check for consistency the set of equations, data and parameters, by applying each formula
separately on the sample period. If the estimation residuals have been introduced as additional elements, the
process should give the historical values in all cases.
• One shall then simulate dynamically the full model on the same period, setting (temporarily) the residuals to
zero. This will show if considering current and lagged interactions does not amplify too much the estimation
errors, both on the current period and with time. Using an error correction framework should limit the risk of
divergence.
• Finally, the reactions of the equilibrium to a change in assumptions, for instance the exogenous component
of demand, will be measured. The results will be compared with the teachings of economic theory, and what
is known of values given by other models, moderated by the characteristics of the country. However, one
should not spend too time here, as simulations over the future will provide a much better context.
Discovering discrepancies can lead to changes in some elements of the model, including the set of its variables. This
means going back to step 1 or 2.
4.1.4 SOLVING AND TESTING OVER THE FUTURE
Once the model has passed all tests on the past, further tests will be conducted, under conditions more
representative of its actual use: on the future. For this, values will have to be established for future assumptions.
Again, the sensitivity of the model to shocks will be studied, this time with a longer and smoother base, better
associated with future use. As to the reliability of baseline results, one can rely this time on stochastic simulations.
The results of this step can of course show the necessity to revert to a previous stage, including the introduction of
new data, changing causalities, or re-estimation. To limit the number of backward steps, one should introduce in the
original data set all potential variables, and decide on behavioral equations considering the global properties.
4.1.5 USING THE MODEL FOR FORECASTS AND POLICY STUDIES
Finally, the model will be considered as fit for economic studies: forecasts and economic policy analysis.
From now on, we shall suppose we are using a dedicated package like EViews. But for people who till model through a
spreadsheet), most of our observations will stim apply.
4.2 HOW TO ORGANIZE THE DEVELOPMENT OF THE MODEL
Let us now consider the organization of the model production task.
To create a model, two extreme types of organization can be considered:
• Methodical option:
The model builder
o Specifies completely a coherent model (including accounting equations), precisely separating assumptions
from results.
o Looks for the necessary series.
o Estimates behavioral equations.
o Uses the consequent model.
Applying such a framework is obviously illusory, as many backtrackings will be necessary in practice:
61
o Some series will show up as unavailable, and it will be necessary to replace them or to eliminate them from
formulations. Thus, in the absence of series for interests paid by firms one will have to be content with profits
before interests.
o Some estimations will give unsatisfactory results: it will be necessary to change formulations, to use
additional or alternate series. Thus, a formulation in levels might have to be replaced by a formulation in
logarithms (constant elasticities), or in growth rates. Or one will be led to explain the average monthly wage
instead of the hourly wage, and to introduce in this last explanation the evolution of the minimal wage. For
an oil producing country, it will appear necessary to identify oil (and non-oil products) in both production and
exports.33
o New ideas will appear during estimation. For example, a recent article on the role of Foreign Direct
Investment might lead to test an original formulation.
o Formal errors are going to be identified. Thus, an element (a type of pension) might have been forgotten
from households’ income.
o Some variables defined as assumptions are going to appear sufficiently influenced by results to see their
status modified.
o Some causalities will be questioned when observing numerical properties.
o Simultaneities will have to be replaced by lagged influences.
o The size (or even the sign) of the response to changes in the assumptions will be inconsistent with theory.
o The model will not converge, and the specifications will be considered as the cause.
• Improvisation
To the contrary, a model builder can
o establish general options for the model structure and theoretical framework,
o produce some formulations independent from each other,
o estimate them by accessing to separate series,
o And gradually connect selected elements by completing the model with linking identities and the data set
with the necessary exogenous variables.
This framework will be even less effective, if only because the number of single operations on equations and series
will present a prohibitive cost. Furthermore, enforcing the accounting and theoretical coherence of the model could
prove difficult, and the modelling process might never converge at all to a satisfying version.
• The optimal solution is of course intermediate:
o Define as precisely as possible the field and the classification of the model.
o Define its general theoretical options and its goal.
o Obtain, create and store the total set of presumably useful series, with no limitations.
o Establish domains to estimate, specify associated variables and set formal connections, especially accounting
relations.
o Undertake estimations
o And go through changes (hopefully limited) until an acceptable form is obtained.
It is clear that this type of organization is all the easier to implement if:
• The size of the model is small: it is possible to memorize the total set of variable names for a thirty equations
model, but for a large model a formal documentation will be necessary, produced from the start and updated
33
Actually, this should have been evident from the start.
62
regularly. This framework should be discussed in detail by the modelling team and as many outsiders as
possible.
• The number of concerned persons is small (the distinction comes essentially between one and several): for a
team project, the role of each participant and his area of responsibility have to be clearly defined. Especially,
physical changes (on both data and model specifications) should be the responsibility of one individual, who
will centralize requests and apply them. And modifications must be clearly stated and documented, through a
formal process.
Individual modifications of the model can be allowed, however, provided a base version is preserved. Thus several
members of a team of model builders can test, one a new production function, another an extended description of
the financial sector. But even in this case updates will often interfere, at the time modifications generated in separate
test versions are applied to the base one. For instance, a new definition of the costs of wages and investment, which
define the optimal shares of labor and capital in the production function, will influence the target in the price
equation.
63
5 CHAPTER 5: PREPARING THE MODEL
We shall now start with the first (and probably most important task): preparing the production of the model.
One might be tempted to start actual model production as soon as possible. But it is extremely important to spend
enough time at the start evaluating the options and choosing a strategy. Realizing much later that he has chosen the
wrong options, the modeler is faced by two bad solutions: continuing a process leading to a subpar model, or
backtracking to the point where the choice was made.
This can concern
• The organization of tasks, like producing at first single country models, for a world modelling project.
• Economic issues, like choosing the complexity of the production function.
• Accounting issues, like deciding the decomposition of products, or the distinction into agents.
• Technical ones, like the number of letters identifying the country in a world model series name.
5.1 PREPARING THE MODEL: THE FRAMEWORK
At the start of the model building process, the modeler (or the team) has at least:
• General ideas about the logic of the model he wants to build.
• Information about the set of available data.
Actually, things can be more advanced:
• The data can be directly available, as a computer file, but not necessarily in the format needed by the
modelling package. Many databases (like the World Bank’s World Development Indicators) are stored on the
producer’s website in Excel or CSV format. In more and more cases, access can be provided from inside
EViews, but this is not necessarily the best option.
• Equations may have already been established, either as formulas or even estimated items, if the modeling is
the continuation of an econometric study, produced by the modeler or another economist.
In any case, the first stage in the process should lead to:
• A fully defined set of equations, except for the actual estimated formulas.
• The corresponding set of potentially relevant data.
Obviously, these two tasks are linked, as equations are established on the basis of available data, and the data is
produced to fit the model equations. This means that they are normally processed in parallel. However, it is quite
possible:
• To produce most of the data before the equations are defined. Some concepts (the supply - demand
equilibrium at constant and current prices, employment, the interest rates) will certainly appear in the
model. But some model-specific variables might have to wait.
• To produce the model specification before any data is available. Of course, writing an identity, or stating the
equation to be estimated, does not require data. It is only the application (checking the identity is consistent
or estimating the equation) which does. But one must be reasonably sure that the data will be available, or
that there will be a reasonable technique totargettarget estimate it.
One can even produce a first version of the program transforming into model concepts the original data, once these
concepts are completely defined, but before any data is technically available (one just needs their definition).
64
One can compare the situation with the building of a house: one can draw the plans before the equipment is
purchased, but its eventual availability (at the right time) must be certain. And the goods can be obtained before the
plans are completely drawn (but the chance of having to use them must be reasonably high)34.
One can even imagine the data using a random process, and apply an estimation program, without considering the
results but checking for the presence of technical mistakes.
These options are not optimal in the general case, but they can help to gain time. Most modelling projects have a
deadline, and once the work force is available, the tasks should be processed as soon as possible, if one wants to have
the best chance of meeting it.
One can question the feasibility of producing a full set of equations before any estimation. What we propose is to
replace the future formulations by a “declaration of intent” which states only the dependent variable on the left, and
the explanatory elements on the right. For each equation, the format should be as close as possible to:
Variable=f*(sum of the explanatory variables)
For instance, for exports X depending on world demand WD, the rate of use of capacities UR and price
competitiveness COMPX, one will use:
scalar f
X = f*(WD+UR+COMPX)
The first statement avoids having f considered as an exogenous element.
The advantages of defining a full model are numerous:
• The modeler will be able to check by sight the logic of his model
• The text can be provided to other economists for advice
• The full list of requested variables can be established, allowing to produce a complete transfer program
• Processing the equations through EViews will give interesting advice on several elements. Double clicking on
the “model” item one will get:
o The equations (from ‘Equations” or “Print View”).
34
As there is a cost to the goods. For free or quasi-free data, the chance can be lowered.
65
The grammatical acceptability of equations will be checked: for instance, if the number of left and right
parenthesizes is indeed the same. Erroneous equations will appear in red in “Equations”.
Also, the fact that each endogenous variable is computed only once. The second occurrence will also appear
in red.
o The variables.
66
Note: if a variable is currently overridden, its name will appear in red.
The most important information will come from the list of exogenous: one might find elements which should
have been determined by the model, according to its logic. In general, this will mean one has forgotten to
state the associated equation. Also, some elements might appear, which should not belong to the model.
Normally this corresponds to typing errors.
o The block structure:
Number of equations: 14
Number of independent blocks: 3
Number of simultaneous blocks: 1
Number of recursive blocks: 2
Block 1 : 2 Recursive Equations
prêt(3) x(13)
Block 2: 11 Simultaneous Equations (1 feedback var)
ic(10) led(4) le(5)
lt(6) rhi(7) co(9)
ih(8) i(2) fd(11)
m(12) q(1)
Block 3: 1 Recursive Equations
k(14)
67
It decomposes the set of equations into a sequence of blocks, either recursive (each variable depends only on
preceding elements) or simultaneous (variables are used before they are computed). If one is going to succeed in
estimating equations which follow the same logic as intended in the preliminary version, the block structure described
at this stage will be already fully representative of the future one. One can detect:
• Abnormal simultaneities: a causal loop might appear, which is not supported by economic theory behind the
model.
• Abnormal recursive links: a block of equations containing a theoretical loop (the wage price loop, the
Keynesian cross) can appear as recursive. This can come from a forgotten equation, a typing error…
Practical operational examples will be given later.
In any case, observing the causal structure of the model will give some preliminary information about its general logic,
and its potential properties.
5.1.1 PRINT VIEW
To check on the model specifications, you can use the “View > Print View” will display the following window:
With “OK” we get a presentation more versatile than “Text”.
One will note that it is possible to decide on the number of significant digits, which produces clearer displays in a
document (the default is 8).
68
5.2 PREPARING THE MODEL: SPECIFIC DATA ISSUES
Let us detail the process.
5.2.1 TYPES OF DATA
In the case of a national macroeconomic model, the needed data can be:
• National Accounts elements: operations on goods and services, transfers between agents, measured in value,
at constant prices, or at the prices of the previous year. The producer will generally be the national statistical
office. For France it would be INSEE (the National Institute for Statistics and Economic Studies).
• The corresponding deflators.
• Their foreign equivalents, using the accounting system and the corresponding base year of the particular
country, or rather a synthesis produced by an international organism (OECD, International Monetary Fund,
EuroStat....).
• Variables in a greater detail, possibly measured in physical quantities (oil barrels, tons of rice). They can come
from a public or private agency, or from the producers themselves. In France energy elements would come
from the Observatory of Energy.
• Monetary and financial data, coming mostly from the local National Bank (in France the Bank of France or the
European Central Bank. ...), or from an international bank (OECD, World Bank, International Monetary Fund,
EBRD, ADB…).
69
• Data on employment or unemployment. One can get detailed labor statistics (by age, qualification, sex...)
from the US Bureau of Labor or the French “Ministère du Travail”.
• Demographic data: population, population in age of working, age classes (INSEE in France).
• Survey data: growth and investment prospects according to firm managers, productive capacity, living
conditions of households (coming from public or private institutes).
• Qualitative elements: the fact of belonging to a specific set, meeting a specific constraint.
• Micro economic models will generally use survey data (households, firms) with sometimes a time dimension
(panels, cohorts) and possibly include some of the above elements as global indicators.
As the area of application of models is unlimited, the field of potentially relevant data is also. A model on the economy
of transportation would include technical data on the railway system and on distances between cities, an agricultural
model meteorological data and information on varieties of species.
5.2.2 THE ACCESS TO DATA
The medium through which data can be obtained will play an important role. Accessing the necessary data takes into
account several features:
• the mode of transmission
• the format used
• the institutional aspects
We shall treat them in turn, then present the most usual cases.
5.2.3 THE MODE OF TRANSMISSION
Several options are available for transferring the data to the model.
5.2.3.1 Physical transmission
Data can be obtained from a physical support, either commercially produced or created for the purpose. This can be
either a CD or DVD-ROM, or another rewritable media such as an USB key, or a memory card. For instance, INSEE
provides CD-ROMs containing the full National Accounts.
5.2.3.2 E-mail transmission
Files can be transferred from a user to another by e-mail, as an attachment to a message.
5.2.3.3 Sharing files
One can share files using Google Drive, Dropbox, or other means.
The advantage is for participants to a project to share elements in real time. Of course, one must be careful with their
status, between read only and read+write. This requires some organization between team members.
In any case, one can transfer the shared file to his computer, allowing any changes.
5.2.3.4 Downloads from a site
70
Files can be downloaded from a website, commercial or not
An extensive survey of the data available online for free, compiled by John Sloman at the Economics, will be found at
the address:
https://www.economicsnetwork.ac.uk/data_sets
with direct access to the relevant sites.
In our experience of building single or multi-country macro econometric models, we are mainly using:
• The OECD Main Economic Indicators.
• The World Bank Development Indicators.
Both institutions provide a large set of series, covering most of the fields associated with our models. There are
differences, however:
• The OECD set covers a limited number of countries:
OECD Economies
Austria, Belgium, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary,
Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Mexico, Netherlands, New, Zealand,
Norway, Poland, Portugal, Slovak, Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, United, Kingdom, United,
States, Euro area (17 countries), OECD Total.
World
Non-OECD Economies
Argentina, Brazil, Bulgaria, China (People's Republic of), Colombia, Costa Rica, India, Indonesia, Romania, Russia, South
Africa
Dynamic Asian Economies
Rest of the World
Other oil producers
Each set contains 122 quarterly series, and 221 yearly. They begin in 1962 and end at present (April 2020) in 2017.
The data is available in CVS format, easily converted to Excel and then transferred to EViews.
• The World Bank covers a much larger set of countries (267 including groups), and series (1429), but:
o The periodicity is annual, starting in 1960 at the earliest.
o Some important elements are lacking, in particular employment and capital (available in principle in the
OECD set).
The list of countries and groups can be found at:
The reason for the large number is that the fields covered are much wider, with a focus on sociological issues,
including for instance (potentially):
71
o People using safely managed drinking water services, rural (% of rural population)
o Women who believe a husband is justified in beating his wife when she burns the food (%)
The set is accessed at:
https://datacatalog.worldbank.org/dataset/world-development-indicators
which displays a page containing:
Clicking on “Download” creates a Zip file for a complete Excel sheet (110 Mb).
Creating an EVIews file containing pages associated with countries is rather straightforward, but involves a little
programming (50 lines). The text is available from the author.
Of course, the number of actual values can vary and can be zero.
Other interesting sites are:
• The IMF site
http://data.imf.org/
Very interesting for financial data, of course, such as the countries’ Balance of Payments. But most of this information
is duplicated on the World Bank site (in annual terms, however).
• The International Labor Organization
https://ilostat.ilo.org/data/
for detailed labor series, and employment series missing for the World Bank.
• and for French data, one can access the INSEE site www.insee.fr.
5.2.3.5 Attribute import and export
72
If documentation is attached to series, it can be imported along with the values and the series names.
5.2.3.6 Using World Bank data
Now you can access the World Bank data from inside a session, through the menu options.
First use:
File > Open> Database
EViews displays this menu in which you select: World Bank Database.
73
5.2.3.7 Accessing INSEE series
For the French users (or those interested in the French economy), the official INSEE series can be downloaded.
From EViews, one should access:
File>Open database>INSEE SDMX database, which opens:
74
The information can be obtained at:
https://www.insee.fr/en/information/2868055
5.2.3.8 Other media
In less and less frequent cases, some data will not be available in magnetic form: series will be found in printed or
faxed documents, or obtained directly from other experts, or fixed by the user (who then plays the role of expert).
This data will have generally to be entered by hand, although a direct interpretation by the computer through optical
character recognition (OCR) is quite operational (but this technique calls for documents of good quality).
In this case it is essential not to enter figures directly into the model file, but to create a preliminary file (such as an
Excel sheet or even and ASCII file) from which the information will be read. This separates the modelling process from
the production of “official” information.
5.2.3.9 Sharing files
It is now quite easy to share files through Internet. This is useful in two cases:
You participate in a project with other modelers.
You follow a project and receive information from the manager.
5.2.3.10 Using the cloud
5.2.3.10.1 Cloud drive support
Although this feature can look unrelated to modelling, it can be quite useful for researchers communicating together
on some project. In particular the files can be shared on a Google Drive.
For instance, for a Google Drive, you can work through a Cloud directory, using:
Open>Foreign Data as Workfile>New Location>
The location can be:
75
• Dropbox.
• Google Drive.
• One Box
• Box.
Authorization can be (has to be) managed.
Then authorize and get access to the whole directory.
76
5.2.3.11 Change of format
As indicated above, the original data format is generally different from the one used by the model-building software.
In the worst cases, transfer from one software program to another will call for the creation of an intermediate file in a
given format, which the model-building software can interpret. The Excel format is the most natural intermediary, as it
is read and produced by all packages. In that case, it is not necessary to own a copy of the package to use its format.
In the very worst cases, it is always possible to ask the first program to produce a character file (in the ASCII standard)
which can, with minimal editing, be interpreted by the second program as the sequence of statements allowing the
creation of the transferred series, including data and definitions35.
However, the situation has improved in the last years, as more and more packages provide a direct access to the
formats
Access
Armes-TSD
Binard
dBase
Excel (through 2003)
Excel 2007 (xml)
EViews Workfile
Gauss Dataset
GiveWin/PcGive
HTML
Lotus 1-2-3
ODBC Dsn File
ODBC Query File
35
For instance, the sequence:
use 1970 to 2007
read x
----- values ----
end
can be translated easily by a word processor into
smpl 1970 2007
series x
----- values ----- ;
77
ODBC Data Source
MicroTSP Workfile
MicroTSP Mac Workfile
RATS 4.x
RATS Portable / TROLL
SAS Program
SAS Transport
SPSS
SPSS Portable
Stata
Text / ASCII
TSP Portable
5.2.3.12 Institutional issues
Of course, one must also consider the relationship between the data producing and modelling institutions. The most
technically complex transfers do not necessarily occur between separate institutions. A commercial contract might
give the modelling institution direct access (through one of the above means) to information managed by a data
producing firm, under the same software format, while a large institution might still use CD-ROMs as a medium
between separate units.
However, one must also consider the cost of establishing contracts, including perhaps some bartering between data
producing and study producing institutions.
5.2.3.13 How to manage the access to several sources
As a general principle, one should favor using a single source. But this is not always possible. In that case, one should
define a primary source, and take from the alternate ones the only additional series. The main problems might come
from:
• Deflators and values at constant prices using a different base year.
• Financial and labor data bases sharing elements with national accounts.
• Variables measured in physical units (tons, square meters) having their counterparts in values.
In all these cases, the priority is the consistency of model equations, based on the data from the primary source.
Additional elements must be adjusted to provide this consistency. This applies in particular to the balance of
equilibrium equations (supply = demand), or sums (total demand= sum of its components).
In addition, for operational models designed to produce official studies and in particular forecasts, it is essential that
the results concur with the official local statistics. As forecasts are presented mostly as growth rates (GDP and inflation
for example), but provide also the last statistical (official) level, the first value in the forecast must be consistent with
both. If the model is built on an outside source, the forecast must be corrected accordingly. This issue will be
developed when we present the forecasting task.
For instance, let us suppose that the limits on the availability of local statistics availability forces the modeler to use on
external source (like the WDI from the World Bank) to produce a full model, but the local statistical office provides
basic elements like GDP. If the model forecast starts in 2020
5.2.4 PREPARING THE DATA FOR THE TRANSFER
78
Let us now define the best organization for transferring data from the original source to the software (we shall use
EViews as an example).
We must guarantee several things:
• The original data must remain available
• It must be updated easily.
• Transfer must be as easy as possible.
To achieve these goals, the best organization should be:
• Storing the original file under another name.
• In this file, creating a new page.
• Copying the original series into this page, using “copy with link” for Excel.
We shall suppose the original data is organized as a matrix (or a set of matrixes) with series either in lines or columns.
If not, an additional intermediary phase can be needed 36.
• Insert if necessary a line of series names above the first period data (or a column left of the first column).
It does not matter if the matrix does not start in cell B2. Just insert as asked.
• Read it in EViews using import or copy.
This guarantees that:
• The original data is not modified.
• Updates are easy: just copy the new page into the original one (and drag cells in the second page if new
observations have appeared).
The only change in the EViews transfer programs concerns the sample period.
In recent versions, the import statement memorizes the reference to the original (Excel) file. EViews will detect if a
change is made and propose updating the page in the workfile accordingly.
5.2.5 THE PRELIMINARY PROCESSING OF SERIES
Very often the nature of available series is not really adapted to the needs of the model. A preliminary processing is
then necessary. This can apply to several features.
5.2.5.1 Time transformations
Most of the time the series the model builder will access have the right periodicity. Individual exceptions can occur.
New series will have to be computed (inside the modelling package).
The change can be undertaken in two directions: aggregating and disaggregating.
36
For instance, quarterly data can appear in yearly lines of four columns.
79
5.2.5.1.1 Aggregation
The easiest case happens if the available periodicity is too short. The nature of the variable will lead naturally to a
method of aggregation, giving the exact value of the series:
If we call Xt the aggregated variable in t, and xt, i the variable of the sub - period i in t, we can consider the following
techniques:
• Sum, for a flow (such as the production of a branch).
n
X t = x t ,i
i =1
• Average, for a level (such as unemployment for a given period).
n
X t = 1 / n x t ,i
i =1
• First or last value, for a level at a given date (for example the capital on the first day of a year will come from
the first day of the first quarter). This will apply to stock variables.
𝑋𝑡 = 𝑥𝑡,1
or
𝑋𝑡 = 𝑥𝑡,𝑛
5.2.5.1.2 Disaggregation
When moving to a shorter periodicity, EViews provides a large list of options, depending on the nature (flow, level or
stock) of the variable.
The following table is copied from the EViews Help and applies to the “c=” modifier in the copy statement.
For instance:
copy(c=q) quart\x
80
can copy the yearly series X into the quarterly page quart using a quadratic smoothing, in such a way that the average
of the quarterly values matches the yearly one.
arg Low to high conversion methods: “r” or “repeata” (constant match average), “d” or “repeats”
(constant match sum), “q” or “quada” (quadratic match average), “t” or “quads” (quadratic match
sum), “linearf” (linear match first), “i” or “linearl” (linear match last), “cubicf” (cubic match first),
“c” or “cubicl” (cubic match last), “pointf” (point first), “pointl” (point last), “dentonf” (Denton
first), “dentonl” (Denton last), “dentona” (Denton average), “dentons” (Denton sum), “chowlinf”
(Chow-Lin first), “chowlinl” (Chow-Lin last), “chowlina” (Chow-Lin average),“chowlins” (Chow-Lin
sum), “litmanf” (Litterman first), “litmanl” (Litterman last), “litmana” (Litterman average),
“litmans” (Litterman sum).
rho=arg Autocorrelation coefficient (for Chow-Lin and Litterman conversions). Must be between 0 and 1,
inclusive.
5.2.5.1.3 Smoothing
Smoothing represents a particular case: preserving the same periodicity as the original series, but with the constraint
of a regular evolution, for example a constant growth rate. Instead of n free values, the choice is reduced to the value
of one (or maybe two) parameters.
EViews provides a large set of methods, some very sophisticated, the most popular being the Hodrick-Prescott and
Holt-Winters methods. The methodology and syntax are explained in detail in the User’s Manual.
5.2.5.1.4 Seasonal adjustment
As we explained before, one method for dealing with variables presenting a seasonality is to eliminate it, and work
with seasonally adjusted series.
Several algorithms can be considered, the best known being probably Census X-13-ARIMA and TRAMO-SEATS, both
available in EViews.
Obviously, one should not mix original and adjusted series in the same set of model variables.
5.2.5.2 Change of classification
We have already considered this problem when we addressed the fields of models.
Changing categories will usually correspond to aggregation. In the case of economic models, this will apply essentially
to:
• Economic agents: one might separate more or less precisely households’ categories (following their income,
their occupation, their size...), types of firms (according to their size, the nature of their production...),
Government institutions (central and local, social security, agencies...).
• Products (production can be described in more or less detail).
• Operations (one can separate social benefits by risk, by controlling agency, or consider only the global value).
• Geographical units (a world model can aggregate countries into zones).
5.2.5.3 Formal transformations
81
Some variables needed by the model will not be available as such, but will have to be computed from existing series
by a mathematical formula. For example, the rate of use of production capacity will be defined as the ratio between
effective production and capacity, coming possibly from different sources. Or the relative cost of wages and capital
(used for defining the optimal production process) will take into account the price of the two factors, but also the
interest rate, the depreciation rate, the expected evolution of wages, and some tax rates.
5.2.6 UPDATES
Once adapted to needs of the model builder, series often will have to be modified.
Changing the values of existing series can have several purposes:
• Correcting a formal error, made by the model builder or the producer of series: typing errors, or errors of
concept.
• Lengthening the available sample: new observations will keep showing up for the most recent periods.
• Improving information: for the last known years, series in the French National Accounts appear in succession
as provisional, semi-final and final.
• Changing the definition of some variables. For instance, the work of private doctors in State hospitals can
move from the market to the non-market sector or vice-versa.
One can also add to the data set a completely new series
• Which has appeared recently as useful to the model.
• Which has been made available by access to a new source of information, or the creation by data builders of
a new, more interesting, concept.
This multiplicity of possible changes prohibits the global set of series used by the model to remain the same even for a
short period. Adapting constantly model specifications (in particular the estimated equations) to this evolution would
ask a lot from the model builder to the detriment of more productive tasks. This means one should limit the frequency
of reconstitutions, for the operational data set (for example once or twice per year for an annual model, or every
quarter for a quarterly one), with few exceptions: correcting serious mistakes or introducing really important
information.
Without doubt, the best solution is actually to manage two sets of data, one updated frequently enough with the last
values, the other built at longer intervals (the periodicity of the model for example). This solution allows to study in
advance, by estimations based on the first set, the consequences of the integration of new values on the
specifications and properties of the next model version.
5.2.7 SUPPRESSIONS
It is beneficial to delete in the bank those series which have become useless:
• This allows to gain space.
• Searches will be faster.
• The bank will be more coherent with the model.
• The model builder will have less information to memorize, and the architecture of the bank will be easier to
master (one will have guessed that this is the most important feature, in our sense).
Useless series that are preserved too long lead to forgetting what they represent, and their later destruction will
require a tedious identification process.
82
For EViews, this presents an additional interest: the elements in the workfile will be display in a single window, and it
is essential for this window to concentrate as many interesting elements as possible.
5.2.8 THE DOCUMENTATION
Similarly, investment in the documentation of series produces quick returns. It can concern:
• The definition, possibly on two levels: a short one to display titles in tables or graphs, and a long one to fully
describe the concept used.
• The source: original file (and sheet), producing institution and maybe how to contact the producer.
• The units in which the series is measured
• Additional remarks, such as the quality and status (final, provisory, estimated) of each observation.
• The date of production and last update (hours and even minutes also can be useful to determine exactly
which set of values an application has used). This information is often recorded automatically by the
software.
• If pertinent, the formula used to compute it.
Example: Wage rate = Wages paid / (employment x Number of weeks of work x weekly work duration).
EViews allows to specify the first four types, using the label command, and produces automatically the last two.
For example, a series called GDP can be defined through the sequence:
GDP.label(c)
GDP.label(d) Gross Domestic Product at constant prices
GDP.label(u) In 2014 Euros
GDP.label(s) from the Excel file accounts.xls produced by the Statistical Office
GDP.label(r) 2019Q4 is provisory
Which clears the contents, gives the definition, describes units, the source, and adds remarks.
• In addition, from version 8, EViews allows to introduce one’s own labels, for instance the country for a
multinational model, the agent for an accounting one, or the fact that a series belongs to a particular model.
For instance, you can use:
HI.label(agent) Households
MARG.label(agent) Firms
83
• If the workfile window screen is in “Display+” mode, you can sort the elements according to their
characteristics. In addition to the name, the type and the time of last modification (or creation) you have
access to the description.
Moreover, if you right click on one of the column headings, and choose “Edit Columns” you can display additional
columns for any of the label types, including the ones you have created.
This can prove quite useful, as it allows you to filter and sort on any criterion, provided you have introduced it as a
label.
This criterion can be for instance:
o The agent concerned
o The country
o The association with a given model
o The formula in the model
o The formula used to create the series (if any)37
o The type within this model (exogenous, endogenous, identity, behavior…)
o The sub-type: for exogenous it can be policy, foreign, structural. For endogenous it can be behavior or
identity.
Once the display is produced, it can be transferred to a table, which can be edited (lines, fonts…) and used for
presentations.
For instance, one can produce a table for a model, with columns for type, agent, units, source, identity / behavior….
This table can be sorted using any of the criteria.
These new functions allow table production to be integrated in the modelling process, a very powerful information
tool for both model development and documentation.
For instance, you could use:
F_HDI.label(d) Disposable income
U_MARG.label(d) Margins
F_HDI.label(model) France small
U_MARG.label(model) USA small
F_HDI.label(agent) Households
U_MARG.label(agentl) Firms
37
You can also use the “source”
84
One of the main interests of this feature is to create a table(using “freeze”). This table can then be sorted according to
any of the criteria.
These definitions follow the series as they are moved through the workfile, or even to an external file.
5.2.9 CONSEQUENCES ON WORK ORGANIZATION
Let us now give some specific considerations on data management.
In the general case, the model builder will be confronted with a large set of series of more or less various origins.
Optimal management strategy might appear to vary with each case, but in fact it is unique in its main feature: one
must produce a file, in the standard of the model building software, and containing the series having a chance to be
useful for the model.
This is true even if the global set of necessary series is produced and managed on the same computer or computer
network, using the same software (the task of transfer will be simply made easier): it is essential that the model
builder has control over the series he uses, and especially that he manages changes (in particular updates of series in
current use). Interpreting a change in model properties (simulations, estimations), one must be able to dismiss a
change in the data as a source, except if this change has been introduced knowingly by the model builder himself38.
Such an organization also makes the management of series easier. In particular, limiting the amount of series in the
bank, apart from the fact that it will save computer time and space, will make the set easier to handle intellectually.
Concerning the scope of the series, two extreme options can however be considered:
• Transferring in the model bank the whole set of series that have a chance (even if a small one) to become
useful at one time to the development of the model39.
• Transferring the minimum, then adding to the set according to needs.
If a median solution can be considered, the choice leans strongly in favor of the first solution. It might be more
expensive initially, in human time, and in size of files, but it will prove generally a good investment, as it avoids often a
costly number of limited transfers, and gives some stability to the bank as well as to its management procedures.
5.2.10 THE PRACTICAL OPTIONS
For models managed by institutions or research groups, the most frequently found organization is a team working
through computers connected through Internet, where storage and synchronization services like Google Drive allow
to share files inside a project. A rigorous work organization is needed to manage the elements of the project,
between work in progress for which current information must be provided, and documented final versions, which can
38
This remark is a particular application of the general principle « let us avoid potential problems which can prove
expensive in thinking time ».
39
Even if they are not considered for actual model variables. For instance, one can be interested in comparing the
capital output radio of the modelled country with those of other countries.
85
only be unfrozen by a controlling manager. Not meeting these principles can lead very quickly to a confusing and
unmanageable situation.
The final version can be made available online to followers, along with the associated documentation and examples. If
the follower has access to the relevant modeling software, direct access to the files can be provided.
In the case of an operational project (like allowing Government economists to produce forecasts and studies) access
can be provided through a front end, which does not require any knowledge of the model management software. This
is the case for EViews.
As to the original data, it can come from distant sources like the website of the World Bank, or of the statistical office
of the model’s country. One might in some cases access directly the data sets of the provider from inside a model-
building session (this is the case in EViews for the World Bank’s WDI). The producers of modelling packages are giving
a high priority to this type of option.
One must however pay attention to format incompatibilities, especially if the operating system is different (Windows
and its versions, Linux, UNIX, Macintosh...)40.
5.3 ACCESSING INTERNATIONAL DATA BASES
In this chapter, we describe the use of data provided by international organizations and institutes.
We will not try to be comprehensive: this is a formidable task, certainly more time consuming than the production of
the present book.
This is done much better by various institutes, which the reader can find easily with a simple Google search. One very
good instance is:
https://www.economicsnetwork.ac.uk/data_sets
Also, a portal giving access to all UN-related sites is:
https://data.un.org/
where you can get access to all the main sites, and also some specific domains, like agriculture and rural development.
Our purpose will rather be the following: to help the producer of a new model to complete the data base he needs for
this task.
So we will rather focus on the technical process, limited to the most promising options in our opinion.
Two cases must be considered:
• The model applies to a single country, which represents the interest of the builder.
40
Most modelling packages work actually under Windows, except freeware like R.
86
This is particularly relevant if he belongs to an official organization, and is responsible for providing studies (maybe
forecasts) applied to the country’s economy. In this case, he will have access to the official data for this country, and
his model must conform to it, if only to provide results in the official format for data and concepts.
But this is also true if the builder is a local independent (maybe a PHD student). He will be more familiar with the
context and have more direct access to local resources.
However, he will need some foreign information, if only to produce some assumptions on foreign demand and prices.
In this case, any source of information is adequate. Detail will only be useful for building detailed scenarios on the
evolution of the world economy, and its consequences on local growth. For instance, a Vietnamese modeler might
require a description of external trade identifying Chinese growth, to establish the model assumptions.
But the classifications, base years and accounting systems can remain different.
• The model applies to a group (like the European Union) or the world (a set of groupings covering the whole
world).
In this case compatibility between models requires access to a single source. The choice must be made at the start,
based on a detailed study of the advantages of all solutions available.
We will rather select the most and focus on the technical access to the main sources, with practical examples. A very
comprehensive list (giving access to the various sites) will be found in:
5.3.1 THE WORLD BANK
In our opinion, the main sources for international data are:
• The World Bank World Development Indicators
• The OECD Economic Perspectives.
And as a potential complement:
• The IMF and ILO databases.
We will focus on the World Bank “World Developments Indicators” under EViews.
5.3.1.1 The main criteria for the choice
The two main options provide enough information to build a single product model. We have used both to build two
multi country models OECD data for the MacSim project detailed above, and the World Bank WDI for a fifteen country
ECOWAS model, presently in the final stage of development.
One can wonder why we did not use a single source, which would have simplified the process, in particular the
programming.
The reasons are practical, and the choice has been obvious:
The OECD data uses a quarterly periodicity, essential to describe the dynamics of the MacSim developed economies
and applying sophisticated econometric techniques such as cointegration.
The countries in the ECOWAS model are not considered developed enough to be described in the OECD data set.
87
As to the additional sources:
The ILO data set is extremely detailed, but limited to the field of labor: employment, unemployment, revenue and
costs. However, if fills gaps on wages, a problem for the WDI and less for OECD.
The IMF data set is logically more detailed on financial series, although many of them can be found in the WDI, with
an annual periodicity, however.
5.3.2 THE ORIGINAL SERIES
At this time (22nd June, 2020) The World Bank makes available, on the site
https://datatopics.worldbank.org/world-development-indicators/
The compressed file:
The series are contained in the page “Data” unfortunately as a single page, with all the series (1442 of them) for all the
countries (266 of them, including a number of groupings) and the periods 1960 to 2020 (in principle).
As shown below, Columns A to D contain:
• The country name
• The 3-letter acronym
• The definition
• The name using “.” as separator
If used as such, EViews will replace (conveniently) the dots by underscores (“_”).
The following columns contain the data, starting in 1960.
88
The list of the countries and series are given in annex, as separate Excel tables.
Obviously, many of the observations are missing, sometimes entirely.
5.3.3 CREATING A SINGLE COUNTRY WORKFILE
If your purpose is to model a single country, no true programming is necessary.
You just have to:
• Locate the country series in the page “Data” of the WDI data set.
• Copy the set in a separate Excel file.
• Create a one-page EViews workfile.
• Import the data into the page by menu or program.
In addition, you can attach the definitions to the series, using the elements of column C of the same page. One can use
the first 1431 cells, which apply to all countries. This calls for a little editing.
5.3.4 WORKING ON SEVERAL (OR ALL) COUNTRIES)
We will now present our method to produce a set of EViews workfiles, in which each page contains the whole set of
1341 WDI series associated with a single country or group.
Each page will:
• Use the name of the three-letter acronym for the country or group, used by the World Bank.
• Contain all the WDI series for that country or region.
• Give access to the definition of the variables (both short and long ones).
This is what we have done, and the provided programs will do. As we have considered that the size of a complete file
was too big, we divided the countries into 8 regions, following the World Bank’s own partition (column H of sheet
Countries).
1 : EAP : EAST ASIA AND PACIFIC
2 : ECA : EUROPE AND CENTRAL ASIA
3 : LAC : LATIN AMERICA COUNTRIES
4 : MEA : MIDDLE EAST AND AFRICA
5 : NAM : NORTH AMERICA AND MEXICO
89
6 : SAS : SOUTH ASIA
7 : SSA : SUB SAHARIAN AFRICA
8 : OTH : OTHERS (such as OECD members…)
In addition, the region groupings (like the page for South Asia as a whole) will be found in the corresponding regional
data set.
The list of countries and sub-groups will be found in the related annex.
5.3.4.1 The method
Although this is not really needed (after all, the files are there), we shall briefly describe the method.
• We create a file called indic.xls with 263 lines (for each country or group) and two columns: the acronym and
the number. We read it as a matrix.
• We modify the original World Bank WDI Excel file by separating the “Data” sheet into 6 parts, to meet a
constraint on the number of lines. At this moment, only 65536 (=2 16) lines can be read, which means 45
countries. Our pages will contain only 40.
• We chose a region.
• We create a workfile for the region.
• We run an EViews program which checks for the presence of the region number, in each of the 6 pages in
sequence.
• If the acronym meets the number in indic:
o We create a page with the acronym name.
o We read the following 1343 series into the page.
• We repeat the process until the end of the sheet is met.
This sequence is available to any user, after editing the program for any changes in the number of countries and
series.
5.3.4.2 the definitions
At the same time as a series is created, a full description is attached. It contains:
• A short description, useful for tables
• The topic of the series, useful for selections
• A longer description, to fully understand the concept
• The source, including the organization and the publication.
The four elements appear when a single series is displayed (using the “sheet” option).
The “Display+” option presents one series per line, including the short description. The other elements can be
displayed too, by right-clicking on the top bar, selecting “Edit columns” and clicking in the appropriate boxes (the
“Type” and “Last Update” columns can be dropped at the same time).
90
5.3.4.3 The problems
Although the World Bank provides a lot of information, sometimes in great detail, some very important elements are
not present. In particular, these elements are clearly requested if you want to build a general econometric model.
They are, in order of decreasing importance:
• Employment and wages.
These series are clearly required, as they enter the wage-price loop, the production function and the households and
firms accounts.
• Capital.
This is required too, but not readily available in most data banks. There are ways to compute it, depending on the
related information available.
• Intermediate consumption.
This is needed to compute total demand (which defines imports) and the production price (which defines the trade
prices and competitiveness).
• Housing investment.
This is a part of demand.
• Non-wage revenue of households.
This enters household revenue and influences consumption.
• Social contributions.
This affects the revenue of all agents, and the cost of labor (thus the value-added deflator and the capital labor ratio in
case of substitution).
91
5.3.4.4 The program
' A program for creating a workfile with one page
' for each of the "East Asia and Pacific" countries
cd "d:\eviews\__world_bank_2020"
' We close the workfiles
' This workfile contains the single page "countries" with artificial series for all the acronyms (each with the "NA" value);
' This can be done easily using the following list
close wb_all
close wb_eap_2020
' We open the base workfile
' and save it as wb_eap (for "East Asia and Pacific"
open wb_all
save wb_eap_2020
' We include a file with a subprogram for creating the series characteristic (short definition, long definition, topic,
source)
include def_2020
' We create the acronym series again
pageselect countries
delete(noerr) *
group g_countries
for %1 ABW AFG AGO ALB ANO ARB ARE ARG ARM ASM ATG AUS AUT AZE BDI BEL BEN BFA BGD BGR BHR
BHS BIH BLR BLZ BMU BOL BRA BRB BRN BTN BWA CAF CAN CEB CHE CHI CHL CHN CIV CMR COD COG COL
COM CPV CRI CSS CUB CUW CYM CYP CZE DEU DJI DMA DNK DOM DZA EAP EAR EAS ECA ECS ECU EGY
EMU ERI ESP EST ETH EUU FCS FIN FJI FRA FRO FSM GAB GBR GEO GHA GIB GIN GMB GNB GNQ GRC GRD
GRL GTM GUM GUY HIC HKG HND HPC HRV HTI HUN IBD IBT IDA IDB IDN IDX IMN IND IRL IRN IRQ ISL ISR ITA
JAM JOR JPN KAZ KEN KGZ KHM KIR KNA KOR KWT LAC LAO LBN LBR LBY LCA LCN LDC LIC LIE LKA LMC LMY
LSO LTE LTU LUX LVA MAC MAF MAR MCO MDA MDG MDV MEA MEX MHL MIC MKD MLI MLT MMR MNA MNE
MNG MNP MOZ MRT MUS MWI MYS NAC NAM NCL NER NGA NIC NLD NOR NPL NRU NZL OED OMN OSS PAK
PAN PER PHL PLW PNG POL PRE PRI PRK PRT PRY PSE PSS PST PYF QAT ROU RUS RWA SAS SAU SDN SEN
SGP SLB SLE SLV SMR SOM SRB SSA SSD SSF SST STP SUR SVK SVN SWE SWZ SXM SYC SYR TCA TCD TEA
TEC TGO THA TJK TKM TLA TLS TMN TON TSA TSS TTO TUN TUR TUV TZA UGA UKR UMC URY USA UZB VCT
VEN VGB VIR VNM VUT WLD WSM XKX YEM ZAF ZMB ZWE
if @isobject(%1)=0 then
genr {%1}=na
g_countries.add {%1}
endif
pagedelete(noerr) {%1}
next
' We read the file indic
' creating a vector with the category for each acronym
' from to 1 to 8
' Here 1 = East Asia and Pacific
' This file can be adapted to any case
' For instance using 1 for relevant countries and 0 for others
' one can restrict the workfile to a subset
vector(263) indic
indic.read(type=excel,b1,s=indic) indic.xls
92
' We
!j=1
' The number of countries and groups
!n=263
' The number of countries in each Excel page (1 to 7)
!p=40
!l=1
while !n>0
' The starting ilne for a country's series
' We initialize by “2” if there is an items definition line
‘ 1 otherwise
!k= 2
' We process the "p" next countries
' in page number "j"
' "l" is the number of the country
' and %1 the associated acronym
'
for !i=1 to !p
pageselect countries
%1=g_countries.@seriesname(!l)
' We delete the page preventively
pagedelete(noerr) {%1}
' If the country belongs to the set
if indic(!l)=1 then
' We create the page
pagecreate(page={%1}) a 1960 2020
' We read the 1431 series from the EXcel sheet number "j"
' starting in cell e{!k}
smpl 1960 2020
read(type=excel,e{!k},s=Data{!j},t) wdiexcel20.xls 1431
' Once data is read we associate definitions
call def
' We close the page
close {%1}
endif
' If the country does not belong we do nothing
pageselect countries
' We increment the starting line in the sheet
93
!k=!k+1442
' and the number of countries treated (relevant or not)
!l=!l+1
next
' We increment the sheet number
!j=!j+1
' "n" is he number of remaining countries
!n=!n-!p
' If it is lower than the number of countries in a sheet
' we have reached the end of the countries set
if (!n<!p) then
' The number of remaining countries is now "n"
!p=!n
endif
wend
' We save the workfile
wfsave wb_eap_2020
5.3.5 THE ALTERNATE SOURCES
The simplest and most accurate way to fill the gaps in the WDI data is the access to an alternate source. We shall
concentrate on the first case, clearly the most important.
Wages and employment can be obtained from:
5.3.5.1 The International Labor Organization.
This institution (www.ilo.org) looks like the most logical source. It does provide information for 234 countries (almost
all countries in the world) and 433 indicators.
There are two main options:
• Downloading a whole set of data, using “Bulk download”.
94
The explanations are given in the last file:
https://www.ilo.org/ilostat-files/Documents/ILOSTAT_BulkDownload_Guidelines.pdf
• Downloading the data through Excel.
If you are currently accessing the ILO data by Internet, the Excel menu will be modified automatically to include a
“ILOSTAT” item, as follows (sorry for the French):
As you can see, you can access data for a country or a subject.
If you chose “Subject” you get the following menu:
95
which leads to the menu:
You can choose the time period and periodicity:
This will start a download for all available series, for all countries for which at least one series is present.
96
The results can appear as:
5.3.5.2 The OECD (Organization for Economic Cooperation and Development.
First, access to this information requests registration and logging in, at least in some cases. But this is free, and advises
you periodically (not too often) on various updates. And by stating your preferences, you can direct this information.
This is a very interesting source (stats.oecd.org). It provides data from 2000 to 2018, sometimes 2019 (at present).
The only drawback is the limitation to the 40 following countries:
(excluding for instance all African countries except South Africa).
After logging in, you will have access to the following menu, with the Economic Outlook the most interesting elements
formacromodelers.
You can also access it directly busing:
https://stats.oecd.org/viewhtml.aspx?datasetcode=EO107_INTERNET_2&lang=en
97
Using “Customize” you can decide on the countries, the topics and the time span.
Then you can ask from the selected elements to be downloaded as and Excel or CSV file.
Remember that to read a CSV file in Excel you should not try to open it directly, but rather open a blank sheet and use:
Data >Obtain data>From file>From Text/CSV.
Select the OECD EO107 (or later) downloaded file.
and select “Comma” as a delimiter (“Virgule’ in French).to organize the information into colums.
98
5.3.5.3 The United Nations, in particular the Economic Commission for Europe.
The site:
https://w3.unece.org/PXWeb/en
gives you access to the following menu:
If you chose “National Accounts” you get:
99
The countries are (63 including groupings):
European Union-28, Euro area-19, EECCA, CIS-11, North America-2, UNECE-52, Western Balkans-6, Albania, Andorra,
Armenia, Austria, Azerbaijan, Belarus, Belgium, Bosnia and Herzegovina, Bulgaria, Canada, Croatia, Cyprus, Czechia,
Denmark, Estonia, Finland, France, Georgia, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Kazakhstan,
Kyrgyzstan, Latvia, Liechtenstein, Lithuania, Luxembourg, Malta, Monaco, Montenegro, Netherlands, North
Macedonia, Norway, Poland, Portugal, Republic of Moldova, Romania, Russian Federation, San Marino, Serbia,
Slovakia, Slovenia, Spain, Sweden, Switzerland, Tajikistan, Turkey, Turkmenistan, Ukraine, United Kingdom, United
States, Uzbekistan.
Once you have made your selection, you can save it in different format (through “Save as”), the most practical being
probably Excel .xlsx.
100
5.3.5.4 The International Monetary Fund.
After signing in (free) on the site:
https://www.imf.org/en/Data
you are given access to the following files, for 193 countries and 11 groups:
101
The reference will be created as something like:
https://data.imf.org/?sk=388DFA60-1D26-4ADE-B505-A05A558D9A42&sId=1479329328660
and propose the files (which we will not present individually):
You can ask for the download of a given file. It will be sent using you registered e-mail address, for instance as:
If you want data for a specific country (like Gabon) and table (like the Balance of Payments), you can use:
102
………
which you can export to Excel (not CSV!) using the corresponding menu item.
103
5.3.6 BACK TO OUR EXAMPLE
Now that we know the principles, let us see how to apply them to the case we have defined earlier. To avoid switching
between distant pages, we shall repeat its presentation.
1. In the example, our economist has decided to build a very simple model of a country’s economy, which includes
the following elements: Based on their production expectations and productivity of factors, firms invest and hire
workers to adapt productive capacity. However, they exert some caution in this process, as they do not want to be
stuck with unused elements.
2. Productive capital grows with investment but is subject to depreciation.
3. The levels actually reached for labor and capital define potential GDP.
4. They also need intermediate products (proportional to actual GDP), and adapt inventories, from the previous
level.
5. Households obtain wages, based on total employment (including civil servants) and a share of Gross Domestic
Product. They consume part of this revenue and invest another (in housing).
6. Final demand is the sum of its components: consumption, productive investment, housing investment,
inventories, and government demand. Total demand includes also intermediate consumption. Final and total demand
are the sum of their components
7. Imports are a share of local total demand, final or intermediate. But the fewer capacities remain available, the
more imports will be called for.
8. Exports follow world demand, but the priority of local firms is satisfying local demand. They are also affected by
capacity constraints.
9. Supply is equal to demand.
104
We have voluntarily kept the framework simple, as our purpose is only explanatory at this time. However, the model
we are building has some economic consistency, and can actually represent the nucleus for further extensions which
we shall present later.
We shall also suppose that the following data is available in an Excel file called FRA.XLS, selected from OECD’s
Economic Perspectives data set. Series are available from the first quarter of 1962 to the last of 2010. However, the
set contains a forecast, and the historical data ends in 2004.
A note: the reason for using older data is not laziness in updating the statistics. The period we are going to consider is
the most interesting as it includes years of steady growth (1962 to 1973), followed by uncertainty following the first oil
shock. This justifies too using French data: the main point here is using actual data for a non-exceptional country.
When we move to a more operational case our data set will include later periods.
The reason for the “FRA” prefix is to identify series for France in a large set of countries, representing all the OECD
members as well as some groupings.
They use the following units:
Values: Euros
Deflators: base 100 in 1995.
Volumes (or quantities): Millions of 1995 Euros
Populations: persons
FRA_CGV Government Consumption, Volume
FRA_CPV Private Consumption, Volume
FRA_EG Employment, Government
FRA_ET Total Employment
FRA_CGV Government Consumption, Volume
FRA_CPV Private Consumption, Volume
FRA_EG Employment, Government
FRA_ET Total Employment
FRA_FDDV Final Domestic Demand, Volume
FRA_CGV Government Consumption, Volume
FRA_CPV Private Consumption, Volume
FRA_EG Employment, Government
FRA_ET Total Employment
FRA_FDDV Final Domestic Demand, Volume
FRA_GAP Output Gap
FRA_GDPTR Potential Output, Total Economy at Current Prices
FRA_GDPV Gross Domestic Product (Market prices), Volume
FRA_IBV Gross Fixed Cap Form, Business Sector, Volume(Narrow Definition)
FRA_ICV Intermediate consumption, Volume
FRA_IGV Government Investment, Volume
FRA_IHV Investment in Housing, Volume
FRA_ISKV Increase in stocks. volume
FRA_KBV Capital Stock, Business
FRA_MGSV Imports Goods and Services, N.A. Basis, Volume
105
FRA_PCP Deflator, Private Consumption
FRA_TDDV Total Domestic Demand, Volume
FRA_WSSS Compensation of Employees
FRA_XGSV Exports Goods and Services, N.A. Basis, Volume
FRA_XGVMKT Exports Goods and Services, Market Potential, Volume
FRA_YDRH Real Household Disposable Income
Applying the principles, we have defined above calls for:
• Creating the model specifications
• Identifying the variables in the model.
• Separating them into endogenous and exogenous.
• Writing down the full identities.
• Establishing each behavioral equation as an identity, presenting in the simplest way the variable it defines,
and the explanatory elements.
• Creating the associated series, from the available data.
• Transferring the elements already available into model series using the names allocated to them.
• Specifying formulas computing the remaining elements
Now that we have obtained the data, we can move to the two tasks: transform it to fit the model needs, start
specifying the model equations.
It should be clear that this will have to be done through a set of stored statements in a readable language (a program).
This option will allow:
• Establishing an apparently consistent set of statements, which can be controlled visually.
• Locating errors and introducing corrections as simply and clearly as possible.
• Storing subsequent versions, including the last and most correct one, until a satisfying version is established.
• Replicating this process with the smallest amount of work.
• Displaying the steps in the process as clearly as possible, introducing comments.
• Once a satisfying stage has been reached, memorizing the actions for later use (especially if the modeling
project faces breaks, short or long).
• Allowing external users to master the current state of operations, to evaluate the present stage of
development of the project.
The program can be inserted with comments, making the sequence of tasks and the role of individual commands
clearer, and allowing to warn of the presence of local problems and the way they have been processed. This is
especially useful for a team project, for which the name of the author should also be included.
Under EViews, two other methods are available:
• Using a sequence of menu and sub-menu functions,
• Typing commands without saving them, directly from the command window.
These two methods fail on all criteria. The record of the tasks is not available, which means errors are difficult to
detect. Reproducing the task, whether to correct errors or to update specifications or data, calls for a new sequence of
menu selections or typed statements41.
41
However, one can copy the sequence of statements entered in the command window into a program file.
106
The obvious choice is even comforted by three features provided by recent versions:
• You can run part of a program, by selecting it with the mouse (in the usual Windows way), clicking on the
right button, and choosing “Run Selected”.
This is generally more efficient than the previous method of copying the selected part into a blank program, and
running it. However, the new method does not allow editing, useful when one wants to run a selected AND modified
set.
• Symmetrically one can exclude temporarily from execution part of a program, by “commenting it out”. To do
this, one should select the relevant part, click on the right button, and choose “Comment Selection”. To
reactivate the statements, one should select them again and use “Uncomment Selection”. One can also type
a single quote (‘) before the statement.
This can be a little dangerous, especially if you (like myself) have the reflex of saving the program before each
execution. To avoid destroying the original, one can save first the modified program under another name42.
• Finally, one can ask a column of numbers to be displayed left of the program lines. This is particularly efficient
if you use the “Go To Line” statement43.
This is done by first unwrapping the command lines. Each command will use one line regardless of its length, which
can be a little annoying for very long ones. Then (and only then) one can ask for the line numbers.
• So actually, the only option is the one we proposed above: defining a program producing all the necessary
information, and the framework of equations which is going to use it. But the ordering of the tasks can be
questioned, as we have started explaining earlier. Until both are completed, the job is not done, but they are
technically independent: one does not need the physical model to create the data, or series filled with values
to specify the equations. This means that one can consider two extreme methods:
• Creating all the data needed by the model, then specifying the model.
• Specifying all the model equations, and then producing the associated data.
The criterion is the intellectual feasibility of the ordered sequence of tasks.
Clearly the first option is not realistic, as writing down the equations will surely evidence the need for additional
elements. The second is more feasible, as one does not need actual series to write an equation. But as the definition
of the equation processes, one has to check that all the addressed elements are or will be available in the required
form, either as actual concepts (GDP) or transformations of actual concepts (the budget deficit in GDP points calls for
the deficit and GDP series). If a concept appears to be lacking, one will have to: use an alternate available element (a
“proxy”), establish an assumption, look in alternate bases not yet accessed, or simply eliminate the element from the
model.
This shows that if producing both sets can be done in any order, there is a preference for specifying the equations
first, with a general knowledge on data availability. If the data set is not ready, but its contents are known, it is
possible to write down the equations and ask the software to proceed the text. The user will be told about possible
42
Only once of course.
43
However, you have to be careful to update the numbers when the program changes.
107
syntax errors, about the nature of the variables (endogenous / exogenous), and the architecture of his model. This will
lead to early model corrections, allowing to gain time and avoiding taking wrong directions later. And if the model
specifications are still discussed, it is possible to build a first version of the associated data set, which will be updated
when the model is complete.
In practice, especially in the simplest cases, one can also start defining the program with two blank paragraphs, and fill
them with data and equation creating statements until both are complete. The eight original paragraphs in our model
specifications can be treated one by one (not necessarily in the numerical order) filling separately the data and
equation generating blocks with the associated elements.
Actually, among the above proposals we favor two alternate techniques:
• Model then data: Specifying first the full model, checking that all elements used can be produced either
directly or through a formula. Then producing the full set of data, preferably through a direct transfer or a
transformation.
• Model and data: Producing the equations in sequence, or related block by related block, and establishing
simultaneously the statements which create all the series they need.
5.3.6.1 Application to our example
Let us now show on our example how the process can be conducted using the second method, probably more
adapted to such a small model (one cannot expect to switch between the two processes too many times).
We shall first present the process in general (non-EViews) terms, treating each case in sequence, and presenting both
the equations and the statements generating the associated variables. To make thinks clearer, the equations will be
numbered, and the creation statements will start with “>>”.
Also, the endogenous variable will use uppercase characters, the exogenous lowercase. This has no impact on
treatments by EViews, but will make interpretation clearer for the model builder and especially for his
(1) Based on their production expectations and productivity of factors, firms invest and hire workers.
This defines two behavioral equations for factor demand, in which employment (let us call it LE) and Investment
(called I) depend on GDP, called Q.
(1) LE=f(Q)
(2) I=f(Q)
We need:
>> IP=FRA_IBV
>> Q =FRA_GDPV
108
But for LE, we face our first problem. Private employment is not directly available. However, we have supposed that
total employment contained only public (government) and private. This means we can use:
>> LE=FRA_ET-FRA_EG
In another case, private and public employment could have been available, but not the total, which would have been
computed as a sum. This highlights the fact that computation and economic causality need not be related.
(2) Productive capital grows with investment but is subject to depreciation.
Capital K, measured at the end of the period, is defined by an identity. Starting from the initial level, we apply a
depreciation rate (called dr) and add investment. The equation is written as:
(3) K(t)= K(t-1).(1-dr(t)) + I(t)
Defining K at the end of the period would only change notations.
We need the data for K
>> K=FRA_KBV
And we get dr by inverting the formula:
>> dr=((K(-1) + IP) – K) / K(-1)
In other words, dr will be the ratio, to the initial capital level, of the difference between two levels of capital: the
value we would have obtained without depreciation, and the actual one.
(3) The levels actually reached define potential production.
Capacity (called CAP) depends on factors LE and K
109
(4) CAP(t)=f(LE(t), K(t))
It can be computed directly as:
>> CAP=FRA_GDPVTR
which rather represents a “normal” GDP value considering the present level of factors.
The direct availability of this concept as a series represents the best case, not often met in practice. Later in the text
we shall address the alternate techniques available in less favorable situations.
(4) They need inputs, and also build inventories.
Intermediate consumption can be defined as proportional to GDP, using the actual value. This means that at any level
of production, each unit produced will need the same amount of intermediary products.
(5) IC = r_icq . Q
For inventories, we will estimate its change:
(6) CI=f(Q)
For this, we need to compute:
>> IC=FRA_ISKV
>> r_icq=IC/Q (or FRA_ISKV/FRA_GDP)
>> CI=FRA_CIV
(5) Households obtain wages, based on total employment (including civil servants) and a share of Gross
Domestic Product. They consume part of this revenue.
110
Now we need to define total employment, by adding government employment (called lg) to LE.
(7) LT=LE+lg
The new series are obtained by:
>> LT=FRA_ET
>> lg=FRA_EG
Now we have to compute household revenue, which we shall call R_HI. We shall suppose that the same wage applies
to all workers, and that the non-wage part of Household revenue is a given share of GDP, a series called r_rhiq. This
gives:
(8) RHI = wr . LT + r_rhiq . Q
Actually the above assumption, while simplistic, is probably not too far from the truth. The sensitivity to GDP of the
elements included in this heterogeneous concept can be low (such as pensions, or interests from long-term bonds),
high (the revenue of small firm owners, with fixed costs and variable output), or medium (self-employed working in
the same capacity as wage earners).
Household consumption is given by applying to RHI the complement to 1 of a savings rate which we shall call sr. For
the time being, the savings rate is exogenous:
(9) CO = RHI . (1 – sr)
Housing investment is also a share of RHI, which we shall call r_ih.
(10) IH= r_ih . RHI
The new variables are RHI, wr, r_rhiq, sr, IH and r_ih.
111
RHI is given simply by:
>> RHI = FRA_YDRH
Let now compute the real wage rate wr. This is done through the following computation.
Dividing FRA_WSSS by FRA_ET gives the individual nominal value, which we divide again by FRA_CPI/10044 to get the
real value45.
>> wr = (FRA_WSS/FRA_ET)/(FRA_CPI/100)
(parenthesizes are added for clarity).
r_rhi will be obtained as the ratio to GDP of household revenue minus wages
>> r_rhi = (RHI – wr . LT) /Q
44
The OECD deflators are measured as 100 in the year 1995. This means that 1995 the average of values and volume
is the same.
2.98E+11
2.97E+11
2.96E+11
2.95E+11
2.94E+11 GDP at current prices
GDP at constant prices
2.93E+11
2.92E+11
1995q1 1995q2 1995q3 1995q4
45
Considering the above list of available series, one can observe other optiosn aren possible.
112
Consumption and housing investment will be obtained directly:
>> CO=FRA_CPV
>> IH=FRA_IHV
Computing the savings rate and r_ih will use the inversion of the associated equation:
sr =(RHI-CO)/RHI
or
sr=(FRA_YDRH-FRA_CPV)/FRA_YDRH
(savings divided by revenue)
>> r_ih=IH/RHI
Or
>> FRA_IHV/FRA_YDRH
(6) Final demand is the sum of its components: consumption, productive investment, housing investment,
inventories, and government demand. Total demand includes also intermediate consumption.
(11) FD=IP+CO+IH+gd+CI
(12) TD = FD + r_ic . Q
113
We need to compute gd as the sum of FRA_IGV and FRA_CGV.
>> gd = FRA_IGV+FRA_CGV
>> FD = FRA_TDDV
>> r_ic = FRA_ICV/FRA_GDPV
(7) Imports are a share of local demand («domestic demand»). But the less capacities are still available, the
more an increase in demand will have to be imported.
This calls for:
(13) UR=Q/CAP
(14) M=f(FD+IC,UR)
We need to compute:
>> UR=Q/CAP (its definition)
>> M=FRA_MGSV
(8) Exports will essentially depend on World demand. But we shall also suppose that if tensions appear
(through UR) local firms will switch some of their output to local demand, and be less dynamic in their
search for foreign contracts.
(15) X=f(WD, UR)
We need:
>> X=FRA_XGSV
>> WD=FRA_XGVTR
114
(9) Supply is equal to demand.
The supply-demand equation will for the moment use the following implicit form:
(16) Q + M = FD + X
(all variable values are obtained earlier)
We can now reorder the framework of our model into the following elements:
[1] LE =f(Q)
[2] IP=f(Q)
[3] K= K-1 (1-depr) + IP
[4] CAP=f(LE, K-1)
[5] IC=r_icq . Q
[6] CI=f(Q)
[7] LT=LE+lg
[8] RHI = wr . LT + r_rhiq . Q
[9] CO = (1-sr) . RHI
[10] IH = r_ih . RHI
[11] FD = CO + IH + IP + CI + gd
[12] TD = FD + r_ic . Q
[13] UR = Q/CAP
[14] M=f(TD, UR)
[15] X=f(wd, UR)
[16] Q + M = FD + X
115
Endogenous variables
I Firms investment.
LE Firms employment.
K Firms (productive) capital
CAP Potential output
LT Total employment.
CI Change in inventories
IC Intermediate consumption
IH Housing investment.
CO Household consumption.
FD French final demand
TD French total demand
M French Imports.
RHI Household real income.
UR Rate of use of capacities
X French Exports.
Q Gross Domestic Product
exogenous variables
depr Depreciation rate of capital
gd State consumption and investment.
lg Public employment
r_ih Share of Housing investment in Household revenue.
r_rhiq Share of GDP transferred to Households, in addition to wages
wd World demand normally addressed to France.
116
r_icq Ratio of intermediate consumption to GDP
wr Real average wage rate
One observes:
o That we have indeed as many equations as variables to compute.
o That we have separated behavioral equations and identities.
o That accounting identities are completely defined.
o That on the other hand the form of behavioral equations is still vague, although the explanatory elements are
known (at least as a first guess).
This distinction is normal. As we have already indicated, identities generally represent a mandatory formal connection,
while conforming behavior equations to economic theory is not so restrictive.
• Computing formulas
By considering the formulas we have obtained, we can see that most of the data needed is available directly, so a
simple transfer should be enough. We might even have considered using the original names. But as our model will
apply only to France, there is no reason to keep the prefix, which helped to identify the French data inside a much
larger multi-country file. And one might decide (rightly in our sense) that our names are clearer.
The correspondences are:
Q = FRA_GDPV
CAP = FRA_GDPVTR
CI = FRA_ISKV
LT = FRA_ET
LG = FRA_EG
FD = FRA_TDDV
CO = FRA_CPV
RHI = FRA_YDRH
I = FRA_IBV
IH = FRA_IHV
WD = FRA_XGVMKT
X = FRA_XGSV
M = FRA_MGSV
117
Only eight elements are lacking, seven of them exogenous variables:
gd =FRA_IGV+FRA_CGV Government demand
UR =Q/CAP Rate of use of capacities (endogenous)
depr = ((K(t-1) + IP) – K(t))/K(t-1) Depreciation rate of capital
r_ic =IC/Q Ratio of intermediate consumption to GDP
r_ih =IH/RHI Share of Housing investment.
r_rhiq = (RHI – wr . LT) /Q Non-wage households revenue: share of GDP
sr =(RHI-CO)/RHI Savings rate
wr = (FRA_WSS/FRA_ET)/(FRA_CPI/100) Real average wage rate
In real cases, this kind of computation will be used often. One must be aware of one important issue:
The use of these formulas is logically distinct from the definition of model equations. The only reason we need them is
to produce the historical values of series not yet available. If the statisticians had made a comprehensive job (and if
they knew the requirements of the model) they would have provided the full set, and no computation would have
been necessary (just a changes in names).
So these two types of formulas have completely different purposes
• Applying the computation statements ensures that all the requested data is available. By associating
formulas to missing elements, they allow to produce the full set required for simulation and estimation. If the
data was already available in the right format, and the names given to the variables were acceptable, no
statement would be necessary. And one can check that in our case, most of the computations are actually
direct transfers, which allow to create a model element while retaining the original series.
Actually, one could question the necessity of having a full set of historical values for endogenous variables. These will
be computed by the model, which will be simulated on the future anyway. The reasons for producing a full set are the
following:
o Estimation will need all the elements in the associated equations.
o Controlling the consistency of identity equations with the data is a prerequisite before any simulation;
otherwise, we may start with a flawed model set.
o Checking that the model gives accurate simulations on the past will need all the historical elements.
o Many equations use lagged values. This requires actual values preceding the starting simulation date.
These formulas can include original data, transformed data computed earlier in the program, or simply assumptions.
For instance:
o GDP has been drawn directly from the original set.
o The depreciation rate is computed using the sequence of capital values, and investment.
o In the absence of other information, the target for inflation can be set to 2%.
118
• The model equations establish a logical link between elements, which will be used by the model to produce a
consistent equilibrium. This means that if the formula for computing variable A contains variable B, variable A
is supposed to depend on B, in economic terms.
This is obviously true for estimated equations. For instance, the wage rate can depend on inflation, or exports on
world demand. But this is also true for identities:
Household revenue is the sum of its elements. If one grows, revenue changes in the same way (ex-ante, of course).
Basically, we suppose that some behaviors apply in the same way to every element of revenue, whatever its source.
If household consumption is estimated, savings are the difference between revenue and consumption.
It is extremely important to understand this issue, at the start of any modeling project.
It is quite possible however that the same formula is present in both sets. For instance, we might not have values for
FD, and we believe that CO, I, IH and gd represent the whole set of its components. In this case the formula:
FD = CO + IP + IH + gd
will be used both to compute historical values of FD and to define FD in the model.
This introduces an obvious problem: if we make a mistake in the formula, or we use the wrong data, there is no way to
detect it.
119
5.3.6.2 The EViews program
Let us now consider how the above task can be produced. We want to create:
• A workfile for all model elements
• An image of the model, with fully defined identities, and indications as the intended estimated equations,
• The associated data.
5.3.6.2.1 The workfile
• First, we need a work file. In EViews, all tasks are conducted in memory, but they apply to the image of a file
which will contain all the elements managed at a given moment.
We can create the file right now (as a memory image) or start from a pre-existing one, in which case the file will be
transferred from its device into memory.
Some precautions have to be taken.
o First, only one version of the file must be open in memory. As we state elsewhere, EViews allows the user to
open a second version (or even third, and fourth…) of a file already opened. Then changes can be applied
only to one of the memory versions, such as series generation and estimations.
This is obviously46 very dangerous. At the least, one will lose one of the set of changes, as there is no way to transfer
elements from an image to the other. Of course, each file can be saved under a different name, but this does not
allow merging the changes47. At the worst, one will forget the allocation of changes to the files, and one or both will
become inconsistent, the best option being to avoid saving any of them, and to start afresh.
This means one should:
• In command mode, check that no file of the same name is opened, and close it if necessary.
• In program mode (the case here) make sure that no file is open at first. This calls for an initial “CLOSE”
statement, which will not succeed most of the time 48 but will guarantee that we are in the required situation.
o Second, a new project must start from a clean (or empty) workfile. For an initial file to contain elements is at
best confusing, at worst dangerous. For instance, series with the same name as elements in our project can
already be present with a different meaning (GDP for a different country?), and available for a larger period.
Allowing EViews to estimate equations over the largest period available will introduce in the sample
irrelevant values.
A simple way to solve the problem is to delete any existing element, through the statement:
46
This is only a personal opinion.
47
Providing this option does not look impossible.
48
With fortunately no error message.
120
DELETE *
which will destroy any pre-existing item, except for the generic C (generic vector of coefficients) and RESID (generic
series of residuals) which are created automatically with the work file and cannot be deleted.
There is only one acceptable case for pre-existing elements: if the work file contains some original information,
provided to the user by an external source. But even in this case the file has to be saved first, to allow tracing back the
steps to the very beginning in which only this original information was present, in its original form.
In any case, in EViews, the possibility to define separate pages (sheets) inside the work file solves the problem. As we
have seen earlier, one can just store the original data in one page and start building the transformed data in a blank
one, logically linked to the original.
First principle of modeling: always organize your work in such a way that if step n fails, you can always get back to
the result of step n-1.
First principle of modeling (alternate version): Always organize your programs in such a way that you can produce
again all the elements associated with the present situation.
This (long) discourse leads to the following statements:
CLOSE small
WFCREATE(page=model) small Q 1970Q1 2005Q4
DELETE *
Applying them guarantees:
• That the file small.wf1 is open in memory with the needed characteristics, for a page called “model”.
• That only one version is open (provided no more than one was open previously, of course, but we shall
suppose you are going to follow our suggestions).
• That the page is empty (actually it contains only C and RESID).
5.3.6.2.2 The data
Now that we have a work file, we must fill it with the necessary information.
The original information is represented by the 72 series in the FRA.XLS49 Excel file. We shall import them using the
IMPORT statement. This statement is quite simple (see the User’s manual for detailed options):
49
EViews allows also to read Excel 2010 .xlsx files (but not to produce them).
121
READ fra.xls 72
But beware: even if the Excel file contains dates (in the first column or line) this information is not taken into account.
What is used is rather the current sample, defined by the last SMPL statement. Fortunately, in our case, the current
sample, defined at workfile creation, is the same as the one in the Excel file. But this will not always be the case:
better to state the SMPL before the READ.
SMPL 1962Q1 2010Q4
READ fra.xls 72
Second principle of modelling: if introducing a (cheap) statement can be useful, even extremely seldom, do it now.
One also has to be careful about the orientation of series: normally they appear as columns, and data starts from cell
B2 (second line, second column). Any other case has to be specified, as well as the name of the sheet for a multi-sheet
file.
An alternate (and probably better) option
If the follow the above method, all the data will be transferred to the “model” page. This makes things easier in a way,
as all information will be immediately available. But
• The separation between original and model data will not be clear.
• The stability of the original data is not guaranteed.
• As the set original series is probably larger, most of the screen will be occupied by elements no longer useful.
Of course, one can separate original and model data by using a prefix for the first type. But it is even better to
separate the two sets physically. This can be done through the “link” EViews function.
Instead of loading the original series in the model page, a specific page is created (named for instance “oecd”) in
which the data is imported.
Then in the model page the model variables are declared as “linked”, and a link is defined with the original series in
the “OECD” page.
The associated syntax will be presented later.
5.3.6.2.3 The model
• Now, we need to define the model on which we shall work. Producing a model starts with the statement:
MODEL modelname
122
Let us call our model _fra_1.
A trick: starting the name of important elements by an underscore allows them to be displayed at the upper corner of
the workfile screen, avoiding a tedious scrolling if the number of elements is large. For very important elements (like
the model itself) you can even use a double underscore.
The statement
MODEL __fra_1
defines __fra_1 as the “active” model.
Two cases can be considered:
o The model does not exist. It is created (with no equations yet).
o The model exists. It is opened, with its present equations.
The second option is dangerous in our case, as we want to start from scratch. To make sure of that, the most efficient
(and brutal) technique is to delete the model first, which puts us in the first case.
DELETE _fra_1
MODEL _fra_1
This introduces a slight problem, however. In most cases (including right now) the model does not exist, and the
DELETE statement will fail. No problem, as what we wanted is to make sure no model preexisted, and this is indeed
the situation we obtain. But EViews will complain, as it could not perform the required task. And if the maximum
number of accepted errors is 1 (the default option) the program will stop.
o It is better to specify the “noerr” option, which accepts failure of the statement without error message.
DELETE(noerr) _fra_1
MODEL _fra_1
o We can change the default number of accepted errors.
123
Another way to avoid this situation is obviously to set the maximum number of errors to more than 1. This is done by
changing the number in the “Maximum errors before halting” box in the “Run program” menu. If you want this option
to apply to all subsequent runs, you have to tick in the “Save options as default” box.
Actually, if you have followed the principle above, there is no risk in proceeding in a program which produced error
messages, even valid ones. You have saved the elements associated to the initial situation, and even if you forgot to
do that, you can always repeat the steps which led to it.
The advantage of this option:
o The program will continue after irrelevant error messages.
o You can produce artificial errors, which can be quite useful as flags50.
o The messages can be associated to several logically independent errors, which can be corrected
simultaneously, leading faster to a correct version.
Now, which number should we specify? In my opinion, depending on the model size, from 1000 to 10000. The number
has to be higher than the number of potential errors, as you want to get as close as possible to the end of the
program. Of course, you will never make 10000 logical errors. But the count is made on the number of error
messages. And in a 2000 equations model, if you have put all the endogenous to zero and you compute their growth
rates, this single mistake will generate 2000 messages.
The only drawback is that if your program uses a loop on the number of elements of a group, and this group could not
be created, the loop will run indefinitely with the message:
Syntax error in "FOR !I=1 TO G.@COUNT"
You will have to wait for the maximum number to be reached.
50
The message associated with a real error will locate it between the preceding and following artificial errors.
124
• Introducing the equations.
Now that we have a blank model, we can introduce the equations one by one. The text of these equations has already
been defined; we just need to state the EViews commands.
This is done through the APPEND statement.
The first one will define investment:
_fra_1.append IP=f*(Q)
Clearly the syntax
o Contains the statement “append”
o Uses the model name on the left, with a dot.
o Adds the text of the equation on the right, with a separating blank.
We must now explain the syntax of our equation.
o At this moment, we expect the model to explain the decision on investment by the evolution of GDP. This
seems quite logical, but we have not decided between the possible forms of the theoretical equation, and we
have not checked that at least one of these equations is validated by all required econometric tests.
o But at the same time we want EViews to give us as much information as possible on the structure of our
model: simultaneities, exogenous parts...
o The best compromise is clearly to produce a model which, although devoid of any estimated equation,
nevertheless presents the same causal relationships as the (future) model we consider.
The simplest choice should be, as if we were writing model specifications in a document or on a blackboard, to state:
IP=f(Q)
Unfortunately, EViews does not accept an equation written in this way. It will consider we are using a function called f,
with the argument Q. As this function does not exist, the equation will be rejected.
The trick we propose is to put an asterisk between “f “and the first parenthesis, which gives
IP=f*(Q).
And state f as a scalar (to avoid confusion with an additional exogenous).
125
If more than one explanatory variable is used, such as in the productive capacity equation, we would like to write:
CAP=f*(LE,K)
Again, this is not accepted by EViews, and we can write instead:
CAP=f*(LE+K)
One just has to state his conventions, and you are welcome to use your own.
However, dropping the f is dangerous, such as in:
M=FD+TD
This will work too, but the equation can be confused with an actual identity, quite misleading in this case.
The complete set of equation statements is:
_fra_1.append LE =f*(Q)
_fra_1.append I=f*(Q)
_fra_1.append K = K(-1)*(1-depr) + I
_fra_1.append CAP=f*(LE+ K)
_fra_1.append IC=r_icq * Q
_fra_1.append CI=f*(Q)
_fra_1.append LT=LE+lg
_fra_1.append RHI = wr * LT + r_rhiq .*Q
_fra_1.append CO = (1-sr) * RHI
_fra_1.append] IH = r_ih * RHI
126
_fra_1.append FD = CO + IH + IP + CI + gd
_fra_1.append TD = FD + r_ic * Q
_fra_1.append UR = Q/CAP
_fra_1.append M=f*(TD+UR)
_fra_1.append X=f*(wd+UR)
_fra_1.append Q + M = FD + X
They produce a 16 equations model called _fra_1. After running the statements, an item will be created in the
workfile, with the name “_fra_1” and the symbol “M” (in blue).
Double-clicking on this item will open a special window, with the list of equations:
• Text (with the icon ”TXT” on the left),
• Number (in the order of introduction in the model).
• Economic dependencies: the dependent variable on the left, the explanatory on the right, using actually the
syntax we could not apply earlier. Lags are not specified, as we shall see later. So K is presented as depending
on K.
Actually, three other display modes are available, using the “View” button:
• Variables: shows the variables (endogenous in blue with “En”, exogenous in yellow with “X”). For the
endogenous, the number of the equation is given. This allows locating the equation in the model text, which
is useful for large models.
The “dependencies” button gives access to a sub-menu, which allows to identify the variables depending on the
current one (Up) and which influences it (Down).
For instance, for FD, “Up” will give TD and Q, “Down” will give CO, I, G, and IH
Of course, exogenous will only be allowed the “Up” button.
The “Filter” option will allow selecting variables using a specific “mask”. For instance, in a multi-country model the
French variables can be identified with FRA_*, provided one has used such a convention.
• Source text: this is basically the text of the model code. We shall see that this changes with estimated
equations.
• Block structure: this gives information on the logical structure of the model (detailed later in Chapter 7).
We get:
o The number of equations.
o The number of blocks, separated into simultaneous and recursive.
o The contents of each block.
127
For the time being, let us only say that a simultaneous block contains interdependent elements. For any couple of
elements in the block, a path can lead from the first to the second, and vice-versa. Of course, this property does not
depend on the ordering of equations inside the block.
EViews gives also number of feedback variables (this will be explained later too).
On the contrary, a recursive block can be ordered (and EViews can do it) in such a way that each variable depends only
(for the present period) on previously defined variables.
This information is useful to improve the understanding of the model, to locate inconsistencies and to correct
technical problems.
EViews can detect errors if:
o A variable is defined twice
o The syntax of an equation is wrong (a parenthesis is lacking for instance)
and allow the user to observe errors himself if:
o Normally endogenous elements appear as exogenous: the equation for the variable has been forgotten, or
written incorrectly.
o Elements foreign to the model appear: variables have been misspelled.
o A loop appears where there should be none.
o Or (more likely) an expected loop does not appear: for instance a Keynesian model is described as recursive,
or a model for two countries trading with each other can be solved as two independent blocks.
All these errors can be detected (and corrected) without calling for the data. This can speed up the building process,
especially if the data is not yet produced.
For the production of series, there are two options.
If the original and model series share the same page, one will simply use the “genr” statement, in the sequence.
genr Q=FRA_GDPV
genr CAP= FRA_GDPVTR
genr CI =FRA_ISKV
genr IC=FRA_ICV
genr LT =FRA_ET
genr LG =FRA_EG
genr FD = FRA_TDDV
genr CO =FRA_CPV
genr RHI =FRA_YDRH
128
genr IP = FRA_IBV
genr IH = FRA_IHV
genr WD = FRA_XGVMKT
genr GD = FRA_IGV+FRA_CGV
genr X =FRA_XGSV
genr M =FRA_MGSV
genr r_icq= FRA_IC/FRA_Q
genr r_ih=IH/RHI (or FRA_IHV/FRA_YDRH)
genr r_rhiq=(RHI-WR*LT)/Q
genr sr=(RHI-CO)/RHI
genr UR=Q/CAP
genr wr=FRA_WSSS//FRA_LT (FRA_PCP/100)
genr rdep=((K(-1)+IP)-K)/K(-1)
If the original series are managed in their own page (a better option in our opinion), one will use:
for %1 Q CAP CI IC LT LG FD CO RHI I IH WD X M
link {%1}
next
Q.linkto oecd\FRA_GDPV
CAP.linkto oecd\FRA_GDPVTR
CI.linkto oecd\FRA_ISKV
IC.linkto oecd\FRA_ICV
LT.linkto oecd\FRA_ET
LG.linkto oecd\FRA_EG
FD.linkto oecd\FRA_TDDV
CO.linkto oecd\FRA_CPV
129
RHI. oecd\FRA_YDRH
I.linkto oecd\FRA_IBV
IH.linkto oecd\FRA_IHV
WD.linkto oecd\FRA_XGVMKT
X.linkto oecd\FRA_XGSV
M.linkto oecd\FRA_MGSV
GD .linkto oecd\FRA_IGV+FRA_CGV
However, a problem remains for GD, the sum of the two original variables FRA_IGV and FRA_CGV. The LINK
function allows to refer to single variables and not functions (as Excel does). Until EViews 8 you had two options.
Creating links to the original elements in the model page.
LINK FRA_IGV
LINK FRA_CGV
FRA_IGV.linkto oecd\FRA_IGV
FRA_CGV.linkto oecd\FRA_CGV
genr gd=FRA_IGV+FRA_CGV
Or computing a FRA_GDV variable in the original page.
genr FRA_GDV=FRA_IGV+FRA_CGV
And linking it
LINK GD
GD.linkto oecd\FRA_GDV
130
But it is also possible to refer to variables in a different page, as
page_name\variable name
This means you can use the much simpler method :
genr GD= oecd\FRA_IGV+oecd\FRA_CGV
Of course, the same method could have been used for single variables.
genr Q=oecd\FRA_GDPV
genr CAP=oecd\FRA_GDPVTR
genr CI=oecd\FRA_ISKV
genr IC=oecd\FRA_ICV
genr LT=oecd\FRA_ET
genr LG=oecd\FRA_EG
genr FD=oecd\FRA_TDDV
genr CO=oecd\FRA_CPV
genr RHI. oecd\FRA_YDRH
genr I=oecd\FRA_IBV
genr IH=oecd\FRA_IHV
genr WD=oecd\FRA_XGVMKT
genr X=oecd\FRA_XGSV
genr M=oecd\FRA_MGSV
genr GD =oecd\FRA_IGV+oecd\FRA_CGV
131
it all depends on if you want changes in the original series to be applied automatically, or to control the process
through GENR51. But if the series is not present (like GD) in the original data, a GENR statement is called for anyway.
Now we have produced a first version of the model, and the associated data. As the behaviors have not been
established, we obviously cannot solve it. But we can check two important things:
o The data required for estimation is present.
o The data is consistent with the identities.
These conditions are needed to start estimation, the next stage in the process. The first one is obvious, the second less
so. But inconsistencies in identities can come from using a wrong concept for a variable, of computing it wrongly. If
this variable is going to be used in estimation, whether as dependent or explanatory, the whole process will be based
on wrong elements.
o The time spent in estimation will be lost.
o This time will probably be longer than usual, as it is generally more difficult (sometimes impossible) to find a
good fit based on wrong data (fortunately?).
o If a good fit is found, the associated equation can remain in the model for a long time (if not indefinitely), and
all the subsequent results will be invalidated. If one is honest, discovering the error later means that a lot of
work will have to be done again, including possibly published results.
This test can be conducted through a very simple technique: the residual check
5.3.6.3 A first test: checking the residuals in the identities
At this point, asking for a solution of the model cannot be considered. However, some controls can be conducted,
which do call for a very specific “simulation”. This technique is called “residual check”.
This method will compute each formula in the model using the historical values of the variables. This can be done by
creating for each equation a formula giving the value of the right-hand side expression (using the GENR statement in
EViews). However, there is a much simpler method, provided by EViews.
If we consider a model written as:
𝑦𝑡 = 𝑓(𝑦𝑡 , 𝑦𝑡−1 , 𝑥𝑡 , 𝛼̂)
with y and x the vectors of endogenous and exogenous variables.
We can perform a very specific “simulation”, in which each equation is computed separately using historical values.
Technically this means:
• Breaking down the model into single equation models, as many as there are equations.
51
Of course, this will also increase the size of the workfile.
132
• Solving each of these models at the same time but separately, using as explanatory values the historical ones.
If we call these historical values 𝑦𝑡0
It means we shall compute:
0
𝑦𝑡 = 𝑓(𝑦𝑡0 , 𝑦𝑡−1 , 𝑥𝑡 , 𝛼̂) + 𝑒𝑡
This method will control:
• For identities, the consistency between data and formulation.
• For the behavioral equations, the availability of the variables requested by the contemplated estimations. But
one gets no numerical information (actually the method we are proposing will give a zero value).
Actually, EViews allows the use of an expression on the left-hand side. This applies also here; the comparison being
made between the left and right expressions.
The interest of this method is obvious: if the residual in the equation is not zero, it means that there is at least one
error in that particular equation. Of course, the problem is not solved, but its location is identified. We shall see later
that this method is even more efficient for a fully estimated model, and we shall extend our discussion at that time.
It would be illusory, however, to hope to obtain a correct model immediately: some error diagnoses might have been
badly interpreted, and corrections badly performed. But even if the error has been corrected:
• There could be several errors in the same equation
• The correcting process can introduce an error in another equation that looked previously exact, but
contained actually two balancing errors. Let us elaborate on this case.
Let us consider our example. If we had used for housing investment the value at current prices:
genr IH=FRA_IH
Then the equation for FD
_fra_1.append FD = CO + IH + IP + CI + gd
will not hold true, but the one for IH
_fra_1.append IH = r_ih . RHI
133
will, as the computation of r_ih as the ratio of IH to RHI will compensate the error by another error.
If we correct the error on IH without correcting r_ih, the IH equation will now appear as wrong, while its actual
number of errors has decreased from 2 to 1.
This means achieving a set of all zero residuals might take a little time, and a few iterations, but should converge
regularly until all errors have disappeared52.
5.3.6.3.1 The types of error met
The residual check allows diagnosing the following errors
• Failure to solve
o syntax error (call to a non-existent function, unbalanced parentheses).
o series with the right name, but unavailable, either completely (they have not been obtained), or partially
(some periods are lacking).
o bad spelling (call to a non-existent series)
• Non-zero residuals
o bad spelling (call to the wrong series).
o errors of logic. This can be more or less serious, as it can come from a purely technical error: forgetting a
term for example, or from a conceptual error: stating an unverified theoretical identity.
o data error: badly entered information, badly computed series, information coming from non-coherent
sources, or from different versions of the same bank.
• Non-verified behavioral equations (or with erroneous residual). This issue will be applicable (and addressed)
later.
Observing error values can give clues as to their origin:
• If some periods give a correct result:
o At the base year (where elements at constant and current prices are identical): the price indexes could be
mistaken for one another, or values could be mistaken for volumes.
o If a variable in the formula is null for these periods, it could be responsible.
o Otherwise, it could come from a typing error (made by the user or the data producer).
o Or if it appears in the last periods, the provisory elements could be inconsistent.
• Observing the magnitude of the error also can be useful: a residual exceeding the normal economic
magnitude (1000% for example) should come from a specification error: bad operator, inversion of
52
Unless the modeler allows some identities to hold true only approximately.
134
coefficients, mistaking values for values per capita. A low residual will often come from confusion between
two close concepts (the consumption price deflator excluding or including VAT).
• For additive equations, a missing or extra element may be identified by comparing the residual to the actual
values of variables. For instance, if the error on final demand for 2010Q1 is 56734 and this is the actual value
of housing investment.
• If the sign of the error is constant (and especially if the order of magnitude is similar across periods), the error
could come from the absence of an element, a multiplication by a wrong factor, or a missing positive
influence.
• If several errors have identical values, they should have the same origin. This is the case when values are
mistaken for volume, if the share the same deflator.
• If two variables show roughly identical errors with the opposite sign, this can come the fact that one of them
has erroneous values and explains the other.
For instance, if historical values for Q are overestimated, the relative error on UR and Q will be similar with different
signs.
UR = Q/CAP
Q + M = FD + X
5.3.6.3.2 Processing errors
Diagnosing errors in the residual check phase can lead back to different phases of the modelling process:
• Data management: the data obtained from external producers is not consistent, for a full series or for specific
observations (this happens!).
• Production of model data: using the wrong original series, using a wrong computation.
Example: using a variable at current prices instead of constant or forgetting an element in a sum.
• Specification of the model (badly written equations).
Example: forgetting the term for Housing investment in the definition of demand. But if the same error was made
when computing the series, the two errors will compensate each other.
• Estimation (modified series since estimation, bad coefficients).
Example: an error in the imports equation shows that the explanatory series for domestic demand has been changed
since estimation.
Applying this process a number of times will be necessary to produce a coherent model.
135
5.3.6.3.3 Back to the example
Producing a residual check is quite easy in EViews: one just has to specify the option “d=f” in the SOLVE statement:
_fra_1.solve(d=f)
Of course, as all equations will be computed separately, all information must be available on the current sample
period, including the values of the endogenous variables (which should be explanatory somewhere else). Contrarily to
computations and estimations, EViews does not adapt the simulation process to the feasible period (this seems rather
logical).
As the model is recursive (super-recursive?) computation gives the result directly, and no element describing the
solving method is needed (we shall see them later).
However:
• One should specify the name given to the computed variables.
Every time EViews has to solve a model, the name given to the results will be built from the original name of the
variable, with the addition of a suffix (a prefix is also possible but less manageable in our opinion). This avoids
destroying the original information, and allows comparing alternate solutions.
The prefix is specified using the statement:
modelname.append @all suffix
(Remember: append adds text to the model, an identity equation is only a special case of text).
In our case, applying the suffix “_C” calls for:
_fra_1.append @all _C
The equation for FD will give FD_C, which we can compare with the actual values of FD.
Computing the differences between actual and computed values can be done in a loop, using the syntax described
later. The elements in the loop can be defined “by hand” but it is more efficient to use the “makegroup” statement.
_fra_1.makegroup(a,n) groupname @endog
136
_fra_1.makegroup(a,n) groupname @exog
In our case:
_fra_1.makegroup(a,n) g_vendo @endog
_fra_1.makegroup(a,n) g_vexo @exog
Two remarks:
• You surely wonder about the reason for the (a,n). This modifies the default options of the “makegroup”
statement, which would produce a group with the baseline names (in our case with _C added) and leave out
the actual names. Stating (a,n):
o Introduces the actual names (a for actual)
o Eliminates the baseline ones (n for no baseline)
It would be best to restrict the computations to the identities. The residuals on the “estimated” have no meaning: as
the “f” scalar is null, the right-hand side will be computed as zero, and the percentage error as 100% as 100*(value -
0)/value. But being able to compute the whole model proves that estimations can be conducted on that period.
One can create two sub-groups by
group g_vbeha CI I LE M X
group g_viden CAP CO FD IC IH K LT Q RHI TD UR
Or
group g_vbeha CI I LE M X
_fra_1.makegroup(a,n) g_viden @endo
g_viden.drop CI I LE M X
This creates first a full group g_viden, then eliminates the estimated from it.
137
This last technique is clearly inefficient here, but will be much more with a 500 equations model with 50 estimated
ones (a more usual situation).
However, both techniques call for a user-defined list, which will have to be updated each time the variable set is
modified, something we want to avoid: we propose using a more complex, but automatic one.
A tip: A visual check is made difficult by the relative imprecision of EViews, which often produces small residuals for
exact equations. In scientific format, these residuals appear as numbers with high negative exponents, which are hard
to identify. One solution is to move to a fixed decimal presentation, by selecting a zone (in the “spreadsheet” view)
then using the right mouse button to access “display format” then “fixed decimal”.
A simpler solution to observe if there is no error is to display all the residuals as a single graph, and look, not at the
series (they should move around in some Brownian motion) but at the scale: both maximum and minimum must be
very small.
Another idea is to transfer the residuals to Excel and sort the sheet (unfortunately EViews does not sort a sheet across
series on the values at a given period). The non-negligible elements should appear at the top and the bottom
according to their sign and the sorting order. Then one can delete the small errors in the middle (less than 0.001%?).
As error correction progresses, the number of remaining lines should decrease.
This technique takes more time but allows to identify immediately and fully the faulty elements.
5.3.6.4 Group members view
Managing groups (very useful in modelling) is more flexible and organized. The Preview function (see above) applies to
the group as a set.
However, for very simple tasks (like adding and element to a group) using the command window (in which the
previous group specification is available) can actually prove faster.
5.3.6.4.1 A trick: generating the groups of identities and behavioral
You certainly have realized by now (and you knew it probably before anyway) that one should avoid as much as
possible having to edit the text of modeling programs, each time changes have been made earlier in the process. This
represents at best extra work, at worst a source of error. We have just violated this principle, by separating by
ourselves the endogenous into behavioral and identity.
This will introduce problems, in particular in large models: the initial process will be tedious and error prone, and one
will have to remember to update the list every time the model structure changes.
We propose a simple technique to avoid this, and make the initial separation and its updating automatic. It is based on
the presence of the “f” scalar in the behavioral equations.
We just have to:
o Simulate the model with the option “d=f” and f=1, saving the results under a given suffix
o Set f to 2 and update the model (this is necessary for EViews to take into account the change).
o Simulate the model again with f=2 and another suffix.
o Create empty groups of estimated and identity variables.
o Produce a loop over the whole group of endogenous, and test each time if the results of the two simulations
are different.
If they are, add the variable to the list of estimated elements.
138
If not, to the list of identity elements.
Note: when we move to actual estimated formulas, we will introduce a residual appending the suffix “_ec” to the
name of the variable. We will use the same technique applying a change to this element.
We can use the following program (for the period 2000 – 2002). We suppose that any percentage error higher than
0.00001 denotes an error.
_fra_1.makegroup(am) g_vendo @endog
_fra_1.makegroup(an) g_vexo @exog
group g_varia g_vendo g_vexo
group g_vbeha ‘creates an empty group
group g_viden ‘creates an empty group
smpl 1986Q1 2004Q4
_fra_1.append assign @all _c
scalar f=0
solve(d=f) _fra_1
scalar f=1
_fra_1.update
_fra_1.append assign @all _d
solve(d=f) _fra_1
for !i=1 to g_vendo.@count
%1=g_vendo.@seriesname(!i)
series pf_{%1}=100*({%1}_d-{%1}_c)/({%1}_c+({%1}_c=0))
if @max(@abs(pf_{%1}))>1e-5 then
g_vbeha.add {%1}
else
g_viden.add {%1}
endif
next
139
This sequence calls for some explanation.
• The loop (“for” to “next”) is reproduced for each variable in the list g_vendo. The number of these variables is
g_vendo.@count (For EViews, x.@count is an integer scalar containing the number of elements in group x).
o !i is the rank of the variable in the group g_vendo (from 1 to g_vendo.@count).
o %1 receives as a character string the contents of g_vendo.@seriesname(!i) , the name of the variable in group
g_vendo, with rank !i.
o The subsequent formulas replace %1 by its string value, and brackets are dropped leaving the characters in the
statement.
For regular users of EViews, or people familiar with programming, the above was probably clear. For others, this is the
time to give very basic information about EViews programming (even if this is not the purpose of this book).
5.3.7 USING LOOPS AND GROUPS IN EVIEWS
In the programs we are going to present, intensive use is made of two elements: groups and loops.
5.3.7.1 Groups
Groups are named elements which refer to a set of objects (which can be series, series expressions but also other
objects), allowing to treat them either as a whole or in sequence.
The statement creating a group is
group name-of-the group list-of-elements
For instance
group g x y z x/y
will create a group named g containing the three series x, y and z and the ratio of x to y.
The element must be series of expression, but one can cheat by creating artificial series with the name of the
requested element.
One can:
• Group groups
• Add and drop elements from groups:
140
g.add a
will add the series a to the group g.
g.drop x
will drop the series x from the group g.
Two useful elements can be associated with the group:
g.@count is a scalar which contains the number of elements of group g.
g.@seriesname is a character vector which contains the names of the series in group g.
Finally, groups can be created through a mask:
group g_fra fra_* will create a group from all the elements starting with fra_ an underscore.
group g_GDP ???_GDP will create a group from all the GDPs of OECD countries (using three characters as a
label).
group g_3 ???_* will create a group from all the elements starting with three characters, then
Groups can be used to display a list of series, as spreadsheet or graph, by double-clicking on its name in the workfile
window (where they appear as a blue “G” symbol) or calling for it.
The default display is a spreadsheet format, but one can move to graphs using the “View” button then “graph”, or
even editing the list of elements by “View” + “group members”.
Managing groups has been made more flexible and organized in the last versions. The Preview function (see later)
applies to the group as a set.
However, for very simple tasks (like adding and element to a group) using the command window (in which the
previous group specification is available) can actually prove faster.
5.3.7.2 Loops
EViews allow two kinds of loops:
• By element
• s (a list or a group)
The syntax is:
141
for %parameter list-of-variables or group-name
block of statements including {%parameter} or %parameter
next
The block of statements will be repeated in sequence for each element in the list, which will then replace the
parameter.
The presence of brackets around the parameter changes its status. With brackets the associated characters are
included in the statements, then the brackets are dropped. Without brackets the parameter is considered as a
character string variable.
For instance, with
%1="111"
The statement
genr xxx={%1}
will give to the series xxx the value 111,
while
xxx=%1
will create a character string with the value “111”
The statement
genr xxx=%1
142
will be illegal as it tries to transfer a character string to a series.
We get the message:
can not assign string expression to numeric variable in "GENR XXX="111""
On the other hand, the statement:
%2=%1+"333"
Will create a “111333” string, while
%2={%1}+"333"
will be illegal as it mixes strings and values:
Scalar assigned to string in "%2=111+"333""
• By integer number.
The syntax is:
for !name=first-integer to second-integer by third integer
block of statements including {!parameter}
next
The block of statements will be repeated in sequence from first-integer to second-integer, incrementing if necessary
by third-integer, the value replacing the parameter.
Integers can be negative. If third-integer is omitted, the increment will be 1.
143
This type of loop can also be applied to a group
for !integer=1 to group-name.@count
%1=group-name.@seriesname(!integer)
block of statements including !integer, %1, {%1}
next
• group-name.@count is the number of elements in the group group-name.
• %1 receives the contents of group-name.@seriesname(!i) , the name of the variable in group group-name,
with rank !integer.
5.3.8 COMPARING WORKFILES : THE WFCOMP ARE COMMAND
During the modelling process, you often have to compare two sets of information.
In particular, you might want to:
o Make sure that two sets of data are identical. This applies to the results of a program you are running again,
maybe after a long delay.
o Control the evolution of historical values for a model data set, showing for instance which equations will have
to be estimated again.
o Summarize the results of a residual check, showing for which equations the right-hand side (using historical
values of the explained variable) is different from the right hand side (the result of the computation). By
setting a tolerance level slightly higher than zero (for instance 0.0001) one can restrict the display to the
errors deemed significant.
o Or you just might want to know which elements of a set are present in another set, for instance which
available series are actually used by one model.
This can be done easily, using the wfcompare command.
You can compare elements between workfiles and pages inside the same workfile. EViews will display one line per
element, in which will be stated its relation, between: unchanged, modified (numerically), added, deleted,
replaced (logically, the last case applies for instance to a linked variable have been modified). A filter can be
applied.
For series, a tolerance level can be set, under which the series are not considered modified. The display will tell how
many periods show a higher difference.
By default, all elements will be displayed, but one can restrict the case (for instance, to all variables present in both
pages with a difference higher than the criterion).
Equations and models are not compared but appear in the list.
The syntax of the wfcompare command is:
144
wfcompare(tol=criterion,list=comparison_type) list_of_compared_series list_of_reference series
For more details you should refer to the EViews Help.
For instance, if you want to compare all French series (starting with “FRA_”) between the pages “base” and
“updated”, for a tolerance level of 0.00001 one will state:
wfcompare(tol=1E-5,list=m) updated\fra_* base\fra_*
145
6 CHAPTER 6 THE ESTIMATION OF EQUATIONS
We now have
• A full description of the framework of the model, in which all the identities are completely specified, and the
intents in terms of behaviors are described as clearly as possible.
• A full database containing all the series in the present model, endogenous and exogenous, with their
description.
We have also checked that:
• The specification of identities is consistent with the available data.
• The information obtained on the structure of the model (causalities, interdependencies) is consistent with
our economic ideas.
Both the list of variables and equations are available as printable documents.
The next stage is obviously to replace each of the tentative behaviors by actual ones, validated both by economic
theory and statistical criteria.
6.1 THE PROCESS OF ESTIMATION
What we are proposing is not a book on econometrics, and anyway we will never be as knowledgeable, by far, as the
EViews team and collaborators, both in terms of theory and ability to teach it (remember that one of them is Robert
Engle…).
This means we will not approach the theoretical aspects of the subject, leaving the reader to the use of the books we
propose in our bibliography, or even to the EViews Help manuals, which can be actually used as teaching tools, as they
are both comprehensive and very progressive in their approach.
But once the modeler is familiar with the concepts, their application to an actual case53 is not straightforward at all.
This means we think this book can bring a very important contribution: showing how these methods can be used in
the process of building our models. The reader will learn how, in very specific cases, first very basic then more
operational econometrics can be used (or not used), considering the information he has and the goal he is pursuing.
We shall also show the role econometrics take in the process, not as a single task between data gathering and
simulations, but as a recurrent partner in the iterative process of building a working model.
We shall not only give examples working smoothly from the start, but show also how econometrics can be set aside,
and how, in some cases, an initial failure can be transformed into success, with some imagination 54.
53
One in which he is not playing with data, but actually oblided to succeed.
54
Remember David Hendry’s four golden rules of econometrics: 1.Think brilliantly, 2.Be infinitely creative, 3.Be
outstandingly lucky, 4.Otherwise, stick to being a theorist.
146
6.2 SPECIFIC ISSUES
Nevertheless, we feel it will be useful to start with two cases, which are not generally treated by manuals, and can
lead to wrong decisions, or wrongly evaluating the results of tests.
We shall use a very practical approach.
6.2.1 THE R2 OR R-SQUARED
The statistic called "R2" or "R-squared" is the most commonly used to judge the global quality of an estimation. It is
defined by the following formula.
R2 = ∑𝑇t=1(𝑥̂𝑡 - x̄ )2 / ∑𝑇t=1(𝑥𝑡 - x̄ )2
This statistic can therefore be interpreted as the share of the variance of the observed variable x explained by the
estimated formula.
A geometrical explanation also can be used: if we consider the space of variables (dimension T = number of
observations), the estimation method will consist in minimizing the distance between the explained variable and the
space (the plane or hyper plane) generated by the vectors of explanatory series, using combinations of parameter
values.
Especially, if the formula is linear relative to estimated parameters and contains a constant term, we can consider the
estimation is based on the difference of variables (explained and explanatory) to their means. In this case, minimizing
the Euclidian distance will lead (as can be seen on the graph) the vector (𝑦̂𝑡 - y𝑡 ) to be orthogonal to the space and
therefore to the vector(𝑦̂𝑡 - ȳ ). These two elements represent the non-explained and explained part of(𝑦𝑡 - ȳ ), the
variance of which is the sum of their squares. The R 2 can be interpreted as the square of the cosine of the angle
between the observed and adjusted series: the closer the R 2 is to 1, the smaller the angle will be and the higher the
share of the estimated variable in the explanation of the total variance. The explanation will be perfect if y-ȳ belongs
to the space, and null if the perpendicular meets the space at the origin.
147
obs 3
y
x1-x1
y
y x2-x2
obs 2
obs 1
If the equation presents no constant term, the same reasoning can be applied, but this time the mean is not
subtracted. However, the R2 no longer has the same meaning: instead of variances, direct sum of squares will be used.
We will not go further in the explanation of this test, concentrating instead on its practical properties.
6.2.1.1 Questioning the R-squared
One must be very careful when using the R2 statistic.
• it is favored by trends in variables, independently from economic significance.
The R2 statistic will be all the higher as the explained variable and at least one of the explanatory variables present a
time trend according to the rank of the observation. Thus components of each of these variables on axes of
observations will grow in the same or opposite direction (from highly negative to highly positive or the reverse), and
give associated vectors very close orientations. In the above graph, the components of variables on the axes will be
more or less ordered according to the numbering of the axes themselves. The first observations will be the most
negative, then values will grow through zero and reach the most positive ones in the end. The same goes if the
ordering follows opposite directions: the estimation will evidence a negative link.
In this case, even in the absence of a true causal relationship between variables, the orientation of the vectors will be
similar to a given multiplicative factor, and the R2 test will seem to validate the formulation. And most time series (like
values, quantities or prices), generally present a growing trend, giving this phenomenon a good chance to happen. For
example, if we apply the previous equation for French imports:
148
(1) 𝐿𝑜𝑔(𝑀𝑡 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑇𝐷𝑡 ) + 𝑏 + 𝑢𝑡
Replacing TD by any steadily growing (or decreasing) variable55 will give a “good” R2, better maybe than actual French
demand.
Actually, it can be shown that testing for each OECD country the estimation of its imports as a function of the demand
of any country, the “true” equation does not come always as the best, although it is never far from it.
• It gives misleading diagnoses when comparing estimations explaining different elements.
This happens in particular when we explain the same concept using a different transformation.
Let us consider our equation 14, as
(14) 𝛥𝐿𝑜𝑔(𝑀𝑡 ) = 𝑎 ⋅ 𝛥𝐿𝑜𝑔(𝑇𝐷𝑡 ) + 𝑏 + 𝑣𝑡
We can see that the time trend has disappeared from both series, and any correlation will come from common
deviations around this trend (or rather common changes in the value from one period to another). This is of course a
much better proof of a link between the two elements (independently from autocorrelation).
To put the two formulations on equal grounds, they must explain the same element. For this, one can just modify the
new equation into:
𝐿𝑜𝑔(𝑀𝑡 ) = 𝐿𝑜𝑔(𝑀𝑡−1 ) + 𝑎 ⋅ 𝛥𝐿𝑜𝑔(𝑇𝐷𝑡 ) + 𝑏 + 𝑣𝑡
Compared to the initial formula, this transformation will not change the explanation56, as obviously the minimization
of the sum of squared residuals represents the same process. The only modified statistic will be the R2, which will
increase a lot, as an identical element with a high variance (compared to that of Log(Mt)) has been added on both
sides.
55
Like Australian demand, or the price of a pack of cigarettes in Uzbekistan.
56
Before estimation EViews will move the lagged term to the left.
149
The choice between the two formulations should not rely on the R 2 but on the autocorrelation of the residual: if ut is
not correlated one should use (1), if it is one should try (2). But in any case the issues will be solved by error correction
models and cointegration, which we shall address later.
The following two formulations are equivalent an indeed give exactly the same results, except for the R-Squared
statistic.
Dependent Variable: DLOG(M)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/21/20 Time: 19:25
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
DLOG(M)=C_M(1)*DLOG(FD)+C_M(2)
Coefficient Std. Error t-Statistic Prob.
C_M(1) 1.941998 0.092941 20.89494 0.0000
C_M(2) 0.001169 0.001359 0.860131 0.3910
R-squared 0.725731 Mean dependent var 0.015300
Adjusted R-squared 0.724068 S.D. dependent var 0.028994
S.E. of regression 0.015231 Akaike info criterion -5.519130
Sum squared resid 0.038275 Schwarz criterion -5.481788
Log likelihood 462.8473 Hannan-Quinn criter. -5.503974
F-statistic 436.5986 Durbin-Watson stat 1.720498
Prob(F-statistic) 0.000000
Dependent Variable: LOG(M)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/21/20 Time: 19:23
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
LOG(M)=LOG(M(-1))+C_M(1)*DLOG(FD)+C_M(2)
Coefficient Std. Error t-Statistic Prob.
C_M(1) 1.941998 0.092941 20.89494 0.0000
C_M(2) 0.001169 0.001359 0.860131 0.3910
R-squared 0.999457 Mean dependent var 24.31343
Adjusted R-squared 0.999454 S.D. dependent var 0.651772
S.E. of regression 0.015231 Akaike info criterion -5.519130
Sum squared resid 0.038275 Schwarz criterion -5.481788
Log likelihood 462.8473 Hannan-Quinn criter. -5.503974
F-statistic 303832.9 Durbin-Watson stat 1.720498
Prob(F-statistic) 0.000000
6.2.2 THE CONSTANT TERM
When observing the validity of individual influences, one element plays a very specific role: the constant term.
This element can have two purposes:
150
• To manage the fact that the equation does not consider elements as such, but the deviations from their
means. In ordinary least squares, even if the final result is a linear formulation of the variables and a
constant term, the process actually
o computes the mean
o substracts it from the variable
o uses the deviations to estimate a formula with no constant57
o recombines estimated coefficients and means into a constant
This constant is an integral part of the process. It should be included every time at least one of the
explanatory elements does not have a zero mean.
• To describe an economic mechanism.
Let us give an example for the first case: if imports have a constant elasticity to demand, we will estimate:
𝛥𝑀𝑡 /𝑀𝑡 = 𝑎 ⋅ 𝛥𝑇𝐷𝑡 /𝑇𝐷𝑡
Or
𝐿𝑜𝑔(𝑀𝑡 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑇𝐷𝑡 ) + 𝑏
but the estimation process will first use the difference to the average to get “a”
𝐿𝑜𝑔(𝑀𝑡 ) − 𝐿𝑜𝑔(𝑀̄) = 𝑎 ⋅ (𝐿𝑜𝑔(𝑇𝐷𝑡 ) − 𝐿𝑜𝑔(𝑇𝐷))
or
𝐿𝑜𝑔(𝑀𝑡 /𝑀̄) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑇𝐷𝑡 /𝑇𝐷)
Then the constant
57
As all elements in the formula have zero mean, the sum of the residuals will also
151
𝑏 = 𝐿𝑜𝑔(𝑀) − 𝑎 ⋅ 𝐿𝑜𝑔(𝑇𝐷)
We can see in particular the consequences of a change in the units (thousands, millions, billions...). The constant term
will absorb it, leaving “a” unchanged. In the absence of “b”, “a” will get a different value, for no economic reason.
Of course, the more “b” is significant, the more its absence will be damaging to the quality of the estimation (and the
more “a” will be affected). But this is no reason to judge “b”. We can see this as weighting an object with a balance:
the two platters never have the same weight, and even if the damage decreases with the difference, it is always useful
to correct it. And in our case there is no cost (actually it makes things cheaper, as the cost of the decision process
disappears).
It is not frequent for the constant term to have a theoretical meaning. The majority of such cases come from a formula
in growth rates or variations, where the constant term will be associated with a trend.
The only justification for the absence of a constant term is when this theoretical constant is null. In this case,
observing a significant value becomes a problem, as it contradicts the theory. We shall give an example soon.
6.3 APPLICATIONS: OUR MODEL
Let us now apply the above principles to our sample model.
In our model, we have to estimate five equations, for which we have already ideas about their logic:
• The change in inventories, employment and investment should depend on GDP
• Exports and imports should depend on the associated demand (world and domestic) and availability of
potential supply.
We shall use each of these equations to illustrate a specific aspect of estimation.
• The change in inventories: general elements, homoscedasticity, presence of a constant term.
• Employment: stationarity, error correction models.
• Investment: the necessity to establish a consistent theoretical equation prior to estimation.
• Exports: autoregressive processes, cointegration, long-term stability.
• Imports: going further on cointegration and long-term stability.
Each of our formulations will be based on very simple economic ideas, and we shall select a specification which
complies with both econometric tests and economic consistency. They are also chosen in a way which should allow
them to merge harmoniously into the model we are building. However, it should be clear that
• Other simple formulations could probably be built on the same sample, with equivalent or maybe better
quality. having
• Using another sample (another country for instance) the same economic ideas could lead to different
formulations (not only different coefficient values).
• Other economic ideas could be applied, with the same complexity.
• To produce a truly operational model, the present framework would have to be developed in a large way. We
will present such developments later.
152
However the model we are building represents in our sense a simplified but consistent summary of the general class
of models of this type. Reading descriptive documents for any operational structural model, one will meet many of the
ideas we are going to develop.
6.3.1 THE CHANGE IN INVENTORIES
We shall use this simplest estimation to present the basic features of EViews estimation, and also stress the necessity
for homoscedasticity.
Our formulation will suppose simply that firms want to achieve a level of stocks proportional to their production (or
GDP). For a particular producer, this should be true both for the goods he produces and for the ones he is going to use
for production. For instance, a car manufacturer will allow for a given delay between production and sale (maybe
three months, which will lead to an inventory level of 1/4 th of annual production). And to be sure of the availability of
intermediary goods (like steel, tires, electronic components and fuel for machines in this case) he will buy the
necessary quantity (proportional to production) sometime in advance.
We shall suppose that firms have achieved, at the previous period, an inventory level IL representing a number of
semesters of production:
𝑳𝑰𝒕−𝟏 = 𝒂 ⋅ 𝑸𝒕−𝟏
And they want to keep this level at the present period:
𝑳𝑰𝒕∗ = 𝒂 ⋅ 𝑸𝒕
𝑳𝑰𝒕∗ = 𝑰𝑳𝒕
Then the change in inventory will represent:
𝑪𝑰𝒕 = (𝑳𝑰𝒕 − 𝑳𝑰𝒕−𝟏 ) = 𝒂 ⋅ 𝜟𝑸𝒕
This means that contrary to the general case this equation should not include a constant term. Its presence would call
for a trend (and a constant) in the equation in levels, with no economic justification. It would also introduce a
problem: adding a constant to an explanation in constant Euros would make the equation non-homogenous.
Even then, the equation faces a problem, concerning the residual: between 1963 and 2004, French GDP has been
multiplied by 4. We can suppose the level of inventories too (maybe a little less with economies of scale and improved
management techniques).
153
It is difficult to suppose that the unexplained part of the change in inventories is not affected by this evolution. As the
variable grows, the error should grow. But to apply the method (OLS), we need the residual to have a constant
standard error. Something must be done.
The simplest idea is to suppose that the error grows at the same rate as GDP, which means that if we measure the
change in inventories in proportion to GDP, we should get a concept for which the error remains stable. Of course, we
shall have to apply the same change to the right-hand side, which becomes the relative change in GDP.
To avoid causality problems (for a given semester, demand for IC is partly satisfied by Q) we shall use the previous
value of Q.
The equation becomes:
𝑪𝑰𝒕 /𝑸𝒕−𝟏 = 𝒂 ⋅ 𝜟𝑸𝒕 /𝑸𝒕−𝟏
6.3.1.1 The basic EViews estimation features
As this is our first example, when shall use it to present the basic estimation features.
Actually, the technique will be different according to the stage in the estimation process: whether we are exploring
several individual formulations, looking for the best option both in statistical and economic terms, or we have already
elected the best one, and want to merge it into our model.
We shall start with the first situation.
The simplest way to estimate an equation under EViews is through the menus, using in succession:
Quick > Estimate equation
A window appears, in which one has to type the formula.
154
In the case of ordinary least squares, this can be a list of elements separated by blanks, in our case:
We can also use
IC/Q(-1)=c(1)*D(Q)/Q(-1)
Of course, the two methods give exactly the same results (in the first case, the “c” vector will also be filled with the
estimated coefficient).
The default method will be Least Squares, appropriate in our case. If the equation was not linear in the coefficients,
the second presentation would be automatically called for.
One will note that
• A constant term has to be introduced explicitly (as an additional element called “C”).
• EViews allows to specify a sample, which will be applied only to the particular equation (the current sample is
not modified). This is quite useful if some periods have to be excluded from the estimation. This will happen
for instance if they are deemed not to follow the estimated behavior (like pre – transition data for Central
European countries, China or Vietnam), or observations are also provided over the future (OECD’ Economic
Perspectives completes the historical data with the results of its forecasts over the next three years).
• On the contrary, one does not have to care about leaving in the sample periods for which estimation is not
possible, due to missing elements or the impossibility to compute a term (for instance the logarithm of a
negative value). EViews will eliminate by itself the corresponding periods (and tell you about the reduced
sample).
In our case we can use the sample:
155
smpl 1962q1 2004q4
which means that we consider data from the first quarter of 1962 to the last of 2004.
If the equation is linear in coefficients, EViews recognizes this property, and does not try to iterate on the coefficients,
as it knows the values found are the right ones.
Using the “Ok” button gives the following results
Dependent Variable: CI/Q(-1)
Method: Least Squares
Date: 03/21/20 Time: 20:46
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
D(Q)/Q(-1) 0.183685 0.034776 5.281999 0.0000
R-squared 0.057924 Mean dependent var 0.002150
Adjusted R-squared 0.057924 S.D. dependent var 0.006806
S.E. of regression 0.006606 Akaike info criterion -7.195844
Sum squared resid 0.007243 Schwarz criterion -7.177173
Log likelihood 601.8529 Hannan-Quinn criter. -7.188266
Durbin-Watson stat 0.695764
We can see that EViews gives the sample used (the relevant periods of our sample). Estimation starts in 1963q2, (with
data starting in 1963Q1) and ends in 2004q4.
• We get also the number of periods, and the time and date.
• The other elements are the usual statistics, described earlier. The most important are:
o The R-squared, the Durbin-Watson test and the Standard Error of regression for global elements.
o The coefficient, the t-Statistic and the probability for reaching the coefficient value if the true coefficient is
null with the estimated standard error.
In our case:
• The R-Squared is very low, even if the extreme variability and the absence of trend of the left-hand element
plead in favor of the explanation58.
However, as with almost all homogenous estimations, a simple interpretation is available, through the standard error:
as the explained variable is measured in points of GDP, this average error represents 0.68 points.
58
If we knew the values for IL, its estimation would get a better R 2 (due to the colinearity of LI and Q). But we would
be led to estimate an error correction model on IL, anyway. We have seen the advantage of this formulation, but for
the quality to extend to the whole model, all equations must be of this type.
156
• The coefficient is very significant. The probability of reaching 0.107 for a normal law with mean 0 and
standard error .00126 is measured as zero. Of course, it is not null, but it is lower than 0.00005, and probably
much so.
• But the Durbin-Watson test takes an unacceptable value, even if the absence of a constant term (and the
subsequent non-zero average of residuals) makes its use questionable.
• The graph of residuals is the second important element for diagnosis. It shows the evolution of actual and
estimated series (top of the graph, using the right-hand scale) and of the residual (bottom, using the left hand
scale, with lines at + and – 1 standard error). This means that inside the band residuals are lower than
average, and higher outside it. Of course, it gives only a relative diagnosis.
.04
.02
.00
.03
.02 -.02
.01 -.04
.00
-.01
-.02
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
The graph shows (in our opinion) that the equation provides some explanation, but some periods (1975-1980 in
particular) present a large and persistent error
In addition to the display of estimation results and graph of residuals, EViews creates several objects:
• A vector of coefficients, contained in the “C” vector. The zero values or the results from the previous
regression are replaced59 .
• A series for the residuals, contained in the “RESID” variable. The “NA” values or the results from the previous
regression are replaced60.
• A tentative equation, called “Untitled” for the moment, and containing the developed formula, with “C” as
the vector of coefficients, with numbers starting from 1. In our case, the formula is obviously
IC/Q(-1)=c(1)*D(Q)/Q(-1)
59
But if the present regression contains fewer coefficients than the previous ones, the additional elements are not put
to zero.
60
But this time, residuals from previous equations are given either computed values or « NA ».
157
Any subsequent estimation will replace this equation by the new “Untitled” version.
• EViews provides also several options, accessed from the menu, and which can be useful:
o View gives three representations of the equation: the original statement, and two formulas including
coefficients as parameters (the above “c” type) or as values.
o “Print” allows printing the current window: to a printer, to a text file (using characters, which saves space but
reduces readability, especially for graphs), or to a graphics RTF file. This last option might call for a
monochrome presentation, which is obtained through the « Monochrome » template (the last of the general
Graph options).
o “Name” allows creating the equation as a named item in the workfile, with an attached comment. It is
important to use it immediately after the estimation, as the temporary equation (named “untitled”) will be
replaced by the next estimation.
However, inserting an underscore ("_") before the name proposed will place the equations in the first positions of the
working window.
EViews proposes as a standard name “EQ” followed by a two-digit number, following the lowest one unused at the
moment.
There are two options:
Give a name representative of the equation (like “EQ_X3U” for the third equation estimating X as influenced
by the rate of use).
Accept the EViews suggestion and rely on the attached comment for the explanation.
Personally we favor the second option:
• It is simpler and more natural to use.
• It allows placing all the equation in the same workfile (and window) location.
• It avoids defining a complex and maybe unclear naming method.
The comment zone is much wider and can follow any format, including blanks and special characters.
Actually the item saved is more complex than the actual formula. Double-clicking on it shows that it contains the full
representation, including the residual (and actually the standard errors of the coefficients, even if they are not
displayed).
158
o Forecast produces a series for the estimated variable (or the estimated left-hand expression, generally less
interesting), and an associated graph with an error band (and a box with the statistics).
6.3.1.2 An alternate technique: using the command window
Instead of using Quick>Estimate, one can work directly through the command window. One just has to add “ls” before
the formula.
ls IC/Q(-1)=c(1)*D(Q)/Q(-1)
This has several advantages:
• By copying and editing the current equation on the next line of the command box, entering changes is made
much easier.
• After a session of estimations, the set can be copied into a program file and reused at will. Management of a
set of alternate versions is much easier.
• One can control the size of characters. This is quite interesting when working with a team, or making a
presentation, as the normal font is generally quite small.
• The only drawback is sample definition: it has to be entered as a command, not as an item in the “estimate”
panel.
6.3.1.3 Other possible specifications
Let us go back to our estimated formula. If we are not satisfied with the previous results, we can try alternate options,
without changing the economic background.
First, one can observe a very clear outlier for the second quarter of 1968. Economists familiar with French history are
aware that this corresponds to the May 1968 “revolution” when student demonstrations turned into general strikes,
paralyzing the economy as not only factories closed but also transportation came to a standstill, and available goods
could not be delivered to the firms which needed them.
As production was much more reduced than demand, in an unexpected way, satisfying it had to call for inventories,
and even a lower production needed intermediate goods (in particular oil and coal) which were not available due to
these transportation problems.
The situation came back to normal in the third quarter.
This will lead us to introduce a “dummy” variable, taking the value 1 in the second quarter only. We can observe
already the gain from our “time” variable: we can introduce it explicitly, without having to create a specific element
which will have to be managed in every simulation, including forecasts.
159
Dependent Variable: CI/Q(-1)
Method: Least Squares
Date: 03/21/20 Time: 23:00
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
D(Q)/Q(-1) 0.141900 0.037159 3.818766 0.0002
T=1968.25 -0.019887 0.007058 -2.817550 0.0054
R-squared 0.101169 Mean dependent var 0.002150
Adjusted R-squared 0.095722 S.D. dependent var 0.006806
S.E. of regression 0.006472 Akaike info criterion -7.230859
Sum squared resid 0.006911 Schwarz criterion -7.193517
Log likelihood 605.7767 Hannan-Quinn criter. -7.215703
Durbin-Watson stat 0.614546
s
The results do improve.
6.3.1.4 Introducing distributed lags
One could assume that the change in inventories does not depend on the present change in GDP, but rather on the
sequence of past changes, due both to inertia on firms behavior, and technical difficulties in implementing decisions.
Let us first consider the last five periods leaving the coefficients free. We get;
Dependent Variable: CI/Q(-1)
Method: Least Squares
Date: 03/21/20 Time: 23:22
Sample (adjusted): 1964Q2 2004Q4
Included observations: 163 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
@PCH(Q) -0.015392 0.043290 -0.355545 0.7227
@PCH(Q(-1)) 0.058955 0.036838 1.600411 0.1115
@PCH(Q(-2)) 0.166449 0.034885 4.771425 0.0000
@PCH(Q(-3)) 0.177331 0.037380 4.743971 0.0000
@PCH(Q(-4)) 0.073800 0.035119 2.101456 0.0372
T=1968.25 -0.037652 0.006491 -5.801017 0.0000
R-squared 0.386670 Mean dependent var 0.002280
Adjusted R-squared 0.367137 S.D. dependent var 0.006800
S.E. of regression 0.005409 Akaike info criterion -7.565217
Sum squared resid 0.004594 Schwarz criterion -7.451337
Log likelihood 622.5652 Hannan-Quinn criter. -7.518983
Durbin-Watson stat 0.753603
Not only most of the explanations are not significant, but the value of the first one (maybe the most important) takes
the wrong sign.
To make the set of lagged coefficients smoother, we can constraint them to follow a polynomial on the lags.
The syntax for this element is :
160
PDL(variable, number of lags, degree of the polynomial, conditions).
Here we shall use :
PDL(@pch(Q),4,3,2)
which implies
• A maximum lag of 4.
• A polynomial of degree 3.
• A zero value for the last coefficient (just beyond the last lag).
We have also introduced:
• The lagged value of the dependent variable (representing additional inertia).
• A constant term associated with a trend in the level of inventories.
Dependent Variable: CI/Q(-1)
Method: Least Squares
Date: 03/21/20 Time: 23:46
Sample (adjusted): 1964Q2 2004Q4
Included observations: 163 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
CI(-1)/Q(-2) 0.612895 0.063632 9.631865 0.0000
T=1968.25 -0.026807 0.005099 -5.256900 0.0000
C -0.002149 0.000591 -3.636950 0.0004
PDL01 0.054922 0.027450 2.000823 0.0471
PDL02 -0.028519 0.022629 -1.260305 0.2094
PDL03 0.017477 0.009394 1.860457 0.0647
R-squared 0.636943 Mean dependent var 0.002280
Adjusted R-squared 0.625380 S.D. dependent var 0.006800
S.E. of regression 0.004162 Akaike info criterion -8.089560
Sum squared resid 0.002720 Schwarz criterion -7.975679
Log likelihood 665.2991 Hannan-Quinn criter. -8.043325
F-statistic 55.08772 Durbin-Watson stat 2.022998
Prob(F-statistic) 0.000000
Lag Distribution of... i Coefficient Std. Error t-Statistic
0 0.21940 0.03934 5.57655
1 0.10561 0.02824 3.73939
2 0.05492 0.02745 2.00082
3 0.03919 0.02780 1.40989
4 0.03026 0.02738 1.10538
Sum of Lags 0.44938 0.07327 6.13293
The results are rather satisfactory, with a nice profile for the reconstructed coefficients, and generally significant
explanations. The dummy element provides also a much better explanation.
161
.04
.02
.00
.015 -.02
.010 -.04
.005
.000
-.005
-.010
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
6.3.1.5 Preparing the equation for the model
Once an equation has been selected for introduction in the model, a different strategy should be used.
If we use the estimated formula, we will face several problems:
• It is not simple to link the equation name with its purpose, which makes the process unclear and forbids to
use any automated and systematic process.
• The C vector is used by all equations is only consistent with the last estimated one.
• The residuals cannot be managed simply
• Instead, we propose the following organization, deriving all elements from the name of the dependent
variable, though a systematic transformation:
• Naming the equation after the estimated variable.
For instance we can call our equation EQ_CI.
• Using the developed specification, with explicit coefficients.
• Naming the coefficient vector after the estimated variable.
For instance we can call it C_CI. Of course this calls for its creation, with a high enough dimension:
coef(10) c_ci
(we chose 10 as a round number which we know we shall never reach).
• Introducing an additive explicit residual, named after the estimated variable. The reason is the following.
162
o It is essential for a model to estimate and simulate the same equation. Of course two versions can be
maintained, one being copied into the other after each new estimation. This is:
Tedious.
Difficult to manage.
Error-prone.
It is much better to use a single item. However this faces a problem: one wants access to the residual, in particular
for forecasts as we shall see later. And the estimation calls for no residual.
The solution is quite simple: introduce a formal residual, but set it to zero before any estimation.
• Work through a program
This allows:
o Visual control over the specification.
o Easy replication of the estimation (for instance if the data has changed).
o Easy introduction of marginal changes.
o Documentation of the economic context (by introducing comments in the program).
In our case we shall use:
coef(10) ec_ci
genr ec_ci=0
equation eq_ci ci/q(-1)=c_ci(1)*@pch(q)+c_ci(2)*@pch(q(-1))+ec_ci
genr ec_ci=resid
6.3.2 INVESTMENT: THE NECESSITY TO ESTABLISH A CONSISTENT THEORETICAL EQUATION
PRIOR TO ESTIMATION
In this estimation, we shall stress the importance of establishing a sound economic framework before any estimation.
The basic economic idea is quite simple: the purpose of investment is
• To replace capital discarded from wear or obsolescence.
• To allow a higher level production, facing an increase of demand.
Without proceeding further on theory, many formulations can be considered. For instance, investment could have a
constant elasticity to GDP, maybe with an error correction term including capital...
In our sense, trying for the best estimation without considering the economics behind the formula, and especially its
consequences for model properties, is rather irresponsible. For instance, using the logarithm of investment is quite
dangerous. Its value can change in very high proportions, and if we go back to the microeconomic foundation of this
163
behavior, its value could very well be negative, as some firms are led to disinvest from time to time, by selling more
capital than they buy.
For instance, the following equation seems to work quite well:
𝜟𝑳𝒐𝒈(𝑰𝒕 ) = 𝒂 ⋅ 𝜟𝑳𝒐𝒈(𝑸𝒕 ) + 𝒃 ⋅ 𝑳𝒐𝒈(𝑰𝒕−𝟏 /𝑸𝒕−𝟏 ) + 𝒄 ⋅ 𝒕 + 𝒅
The results are:
Dependent Variable: DLOG(I)
Method: Least Squares
Date: 03/22/20 Time: 11:49
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
DLOG(Q) 1.835512 0.106843 17.17945 0.0000
LOG(I(-1)/Q(-1)) -0.046829 0.020085 -2.331615 0.0209
C -0.931254 0.301928 -3.084362 0.0024
T 0.000415 0.000138 3.012286 0.0030
R-squared 0.651895 Mean dependent var 0.008590
Adjusted R-squared 0.645488 S.D. dependent var 0.028010
S.E. of regression 0.016677 Akaike info criterion -5.325856
Sum squared resid 0.045336 Schwarz criterion -5.251174
Log likelihood 448.7090 Hannan-Quinn criter. -5.295544
F-statistic 101.7498 Durbin-Watson stat 2.152872
Prob(F-statistic) 0.000000
.2
.1
.0
.06
.04 -.1
.02
-.2
.00
-.02
-.04
-.06
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
Everything seems to go well: the statistics are quite good (even the Durbin-Watson test), the signs are right, the graph
shows a really strong fit. However, when we merge the equation into a model, its simulation properties will be
affected by the base solution: even a very high increase in GDP will have a low impact of the absolute level of
164
investment if it this level was very low in the previous period. And investment can show huge variations as it
represents a change in a stable element, capital.
One can guess that although linking investment (a change in capital) to the change in production seems a natural idea,
jumping to the above formulation was moving a little too fast. In general, one should be naturally reticent in taking
the logarithm of a growth rate, itself a derivative.
The right starting approach is to clarify the economic process through a full logical formalization.
Let us suppose that production follows a “complementary factors” function, which means that to reach a given level
of productive capacity, fixed levels of capital and employment are required, and a reduction in one factor cannot be
compensated by an increase in the other. This means obviously that the less costly process (optimal) is the one which
respects exactly these conditions.
Combinations of K and L
such that:CAP=CAP1
K1 CAP=CAP1
L1m L1 L
With “pk” productivity of capital, and “pl” productivity of labor, we get :
𝑪𝑨𝑷𝒕 = 𝒎𝒊𝒏( 𝒑𝒌𝒕 ⋅ 𝑲𝒕−𝟏 , 𝒑𝒍𝒊 ⋅ 𝑳𝒕 )
(the “t-1” means that we shall use the level of capital reached at the end of the previous period).
Actually, for a given level of employment, there is always some short-term leverage on production, at least at the
macroeconomic level. Temporarily increasing labor productivity by 1% can be easily achieved through extra hours, less
vacations, cancelled training courses...
This means capital will be the only limiting factor in the short-term.
The capacity equation can be simplified into:
165
𝑪𝑨𝑷𝒕 = 𝒑𝒌𝒊 ⋅ 𝑲𝒕−𝟏
Now let us define the rate of use of capacities:
𝑼𝑹𝒕 = 𝑸𝒕 /𝑪𝑨𝑷𝒕
𝑪𝑨𝑷𝒕 = 𝒑𝒌𝒕 ⋅ 𝑲𝒕−𝟏 = 𝑸𝒕 /𝑼𝑹𝒕
*
Now let us suppose firms actually want to reach a constant target utilization rate UR , and expect a production
𝑎
level𝑄𝑡+1 . Then by definition:
𝑲∗𝒕 = 𝑪𝑨𝑷∗𝒕 /𝒑𝒌∗𝒕 = 𝑸𝒂𝒕+𝟏 /𝑼𝑹∗ /𝒑𝒌𝒕+𝟏
𝑲𝒕 = 𝑪𝑨𝑷𝒕 /𝒑𝒌𝒕 = 𝑸𝒕 /𝑼𝑹𝒕 /𝒑𝒌𝒕
And defining tx(z) as the growth rate of z:
𝑡𝑥 ∗ (𝐾𝑡 ) ≈ 𝑡𝑥 ∗ (𝐶𝐴𝑃𝑡 ) − 𝑡𝑥(𝑝𝑘𝑡 ) ≈ 𝑡𝑥 𝑎 (𝑄𝑡 ) − 𝑡𝑥 ∗ (𝑈𝑅𝑡 ) − 𝑡𝑥(𝑝𝑘𝑡 )
This means that the target growth rate of capital can be decomposed as the sum of three terms, one with a positive
influence:
• The expected growth rate of production
and two negative ones:
• The target growth rate of the rate of use: if the firms feel their capacities are 1% too high for the present level
of production, they can reach the target by decreasing capital by 1% even if production is not expected to
change.
• The growth rate of capital productivity: if it increases by 1%, 1% less capital will be needed.
But the element we need is investment. To get it we shall use the definition.
166
𝐾𝑡 = 𝐾𝑡−1 ⋅ (1 − 𝑑𝑟𝑡 ) + 𝐼𝑡
which can be written as
𝑡𝑥(𝐾𝑡 ) = −𝑑𝑟𝑡 + 𝐼𝑡 /𝐾𝑡−1
This gives finally:
𝐼𝑡∗ /𝐾𝑡−1 = 𝑡𝑥(𝐾𝑡 ) = 𝑑𝑟𝑡 + 𝑡𝑥 ∗ (𝐾𝑡 ) = 𝑑𝑟𝑡 + 𝑡𝑥 𝑎 (𝑄𝑡 ) − 𝑡𝑥 ∗ (𝑈𝑅𝑡 ) − 𝑡𝑥(𝑝𝑘𝑡 )
In other words:
• If firms expect production to grow by 2.5%, capacities should adapt to that growth
• But if they feel their capacities are under-used by 1%, their desired capacity will only increase by 1.5%.
• If capital productivity is going to increase by 0.5%, they will need 0.5% less capital.
• But once capital growth has been defined, they also have to compensate for depreciation (5% is a reasonable
value).
In summary, the accumulation rate (the ratio of investment to the previous level of capital) would be:
2.5 - 1 - 0.5 + 5 = 6%.
If we suppose
• That the depreciation rate is constant, as well as the rate of growth of capital productivity,
• That production growth expectations are based on an average of the previous rates,
And we consider as the rate of use the ratio of actual GDP to a value obtained under normal utilization of factors,
which leads to a unitary target, we get the simplified formula:
𝐼𝑡∗ /𝐾𝑡−1 = 𝑎 + ∑𝑛𝑖=0 𝛼𝑖 ⋅ 𝑡𝑥 𝑎 (𝑄𝑡−𝑖 ) − 𝑡𝑥 ∗ (𝑈𝑅𝑡+1 )
With
167
∑𝑛𝑖=0 𝛼𝑖 = 1
finally, we can suppose, as we shall do also for employment, that the desired growth of capital is only partially reached
in practice, either because firms react cautiously to fluctuations of demand, or because they are constrained by
investment programs covering more than one period, from the decision to the actual installation of investment
goods.61
And we shall leave free the coefficients:
𝐼𝑡∗ /𝐾𝑡−1 = 𝑏 ⋅ 𝐼𝑡−1 /𝐾𝑡−2 + (1 − 𝑏) ⋅ (𝑎 + 𝑐 ⋅ ∑𝑛𝑖=0 𝛼𝑖 ⋅ 𝑡𝑥 𝑎 (𝑄𝑡−𝑖 ) − 𝑑 ⋅ 𝑡𝑥 ∗ (𝑈𝑅𝑡 ))
The results are rather satisfactory, with the right sign and acceptable statistics for all explanatory elements. This was
not obvious, as their strong correlation (both use Q in the numerator) could have made it difficult for the estimation
process to separate their role.
Dependent Variable: I/K(-1)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 13:01
Sample (adjusted): 1977Q4 2004Q4
Included observations: 109 after adjustments
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*(UR_STAR-UR)/UR+C_I(3)*.125*Q/Q(-8)
+C_I(4)
Coefficient Std. Error t-Statistic Prob.
C_I(1) 0.941668 0.011760 80.07071 0.0000
C_I(2) -0.002311 0.000863 -2.679352 0.0086
C_I(3) 0.045795 0.013397 3.418178 0.0009
C_I(4) -0.004818 0.001653 -2.915711 0.0043
R-squared 0.988698 Mean dependent var 0.020830
Adjusted R-squared 0.988375 S.D. dependent var 0.002417
S.E. of regression 0.000261 Akaike info criterion -13.63100
Sum squared resid 7.13E-06 Schwarz criterion -13.53224
Log likelihood 746.8897 Hannan-Quinn criter. -13.59095
F-statistic 3061.832 Durbin-Watson stat 1.649042
Prob(F-statistic) 0.000000
61
In this model, we suppose there is no delay between the acquisition of investment (with impact on demand and the
supply-demand equilibrium) and the participation of this investment to the production process.
168
.026
.024
.022
.020
.0008
.018
.0004
.016
.0000
-.0004
-.0008
78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
The graph of residuals shows that the quality of the explanation grows with time, and is especially good for the last
periods. This is rather important for simulations over the future, and one can wonder what we would have done if the
sample had been reversed, and the initial residuals had applied to the last periods.
We will deal with this problem of growing errors on recent periods when we address forecasts.
The equation we have built is not only satisfactory by itself, but we can expect it to provide the model with adequate
properties. In particular, the long-term elasticity of capital to production is now unitary by construction. Starting from
a base simulation, a 1% permanent shock on Q will leave the long run value of UR unchanged62. This gives the same
relative variations to production, capacity and (with a constant capital productivity) capital.
The coefficients “a” and “b” determine only the dynamics of the convergence to this target.
Actually we have estimated a kind of error-correction equation, in which the error is the gap between actual and
target capacity (the rate of use).
We hope to have made clear that to produce a consistent formulation, in particular in a modelling context, one must
start by establishing a sound economic background.
6.3.3 EMPLOYMENT: STATIONARITY, ERROR CORRECTION MODELS, BREAKPOINT TEST.
6.3.3.1 The economic framework
Of course, the employment equation should follow also a complementary factors framework.
In the previous paragraph, we have shown that in this framework the element determining capacity is the sole capital,
while firms could ask from workers a temporary increase in productivity, high enough to ensure the needed level of
production63. Adapting employment to the level required to obtain a “normal’ productivity target will be done by
steps.
62
As the left-hand side represents the (fixed) long-term growth rate of capital.
63
This is true in our macroeconomic framework, in which the changes in production are limited, and part of growth is
compensated by increases in structural productivity (due for instance to more capital-intensive processes). At the firm
169
This means estimating employment will allow us to apply elements on error correction models in a very simple
framework.
We shall suppose that firms:
• Know the level of production they have to achieve.
• Know also the level of production which should be achieved by each worker under normal circumstances (in
other term his normal productivity).
From these two elements they can determine the normal number of workers they need.
But they do not adapt the actual employment level to this target, and this for:
• Technical reasons: between the conclusion that more employees are needed and the actual hiring 64, firms
have to decide on the type of jobs called for, set up their demands, conduct interviews, negotiate wages,
establish contracts, get authorizations if they are foreign citizens, maybe ask prospective workers to train...
Of course this delay depends heavily on the type of job. And this goes also for laying out workers.
• Behavioral reasons: if facing a hike in production, firms adapt immediately their employment level to a higher
target, they might be faced later with over employment if the hike is only temporary. The workers they have
trained, maybe at a high cost, have no usefulness at the time they become potentially efficient. And laying
them out will call generally for compensations....
6.3.3.2 The formulas: stationarity and error correction
We should realize that we are facing an error correction framework, which we can materialize as:
“Normal” labor productivity does not depend on economic conditions. It might follow a constant trend over the
period, such as:
𝐿𝑜𝑔(𝑝𝑙𝑡 ) = 𝑎 + 𝑏 ⋅ 𝑡
Firms use this target to define “normal” employment:
𝐿𝐸𝑡∗ = 𝑄𝑡 /𝑝𝑙𝑡∗
They adapt actual employment to this target with some inertia:
level, employment can produce bottlenecks. This will be the case if a sudden fashion appears for particular goods
requiring specialized craftspeople, even if the tools and machines are available for buying.
64
But not the start of actual work: what we measure is the number of workers employed, even if they are still training
for instance.
170
𝛥𝐿𝑜𝑔(𝐿𝑡 ) = 𝛼 ⋅ 𝛥𝐿𝑜𝑔(𝐿∗𝑡 ) + 𝛽 ⋅ 𝐿𝑜𝑔(𝐿∗𝑡−1 /𝐿𝑡−1 ) + 𝛾 + 𝜀𝑡
We recognize here the error correction framework presented earlier, which requires:
𝐿𝑜𝑔(𝐿∗𝑡 /𝐿𝑡 ) to be stationary.
But does not have to be unitary. However, if we follow the above reasoning, its value should be between 0 and 1,
and probably significantly far from each of these bounds.
To estimate this system we face an obvious problem: pl* is not an actual series (LE* either, but if we know one we
know the other).
But if we call “pl” the actual level of productivity (Q/LE) we can observe that:
𝐿𝑜𝑔(𝐿∗𝑡 /𝐿𝑡 ) = 𝐿𝑜𝑔((𝑄𝑡 /𝑝𝑙𝑡∗ )/(𝑄𝑡 /𝑝𝑙𝑡 )) = −𝐿𝑜𝑔(𝑝𝑙𝑡∗ /𝑝𝑙𝑡 )
The stationarity of 𝐿𝑜𝑔(𝐿∗𝑡 /𝐿𝑡 ) is equivalent to that of 𝐿𝑜𝑔(𝑝𝑙𝑡∗ /𝑝𝑙𝑡 )
Now it should be obvious that if pl* and pl have a trend, it must be the same, actually the trend defining completely
pl*. If not, they will diverge over the long run, and we will face infinite under or over employment. So target
productivity can be identified using the trend in the actual value, if it exists.
This means we can test the stationarity of the ratio as the stationarity of actual productivity around a trend, a test
provided directly by EViews.
We can expect a framework in which actual productivity fluctuates around a regularly growing target, with cycles
which we do not expect to be too long, but can last for several periods65.
6.3.3.3 The first estimations
First, we compute actual labor productivity
genr PROD = Q / LE
65
Which will create (acceptable) autocorrelation in the difference to the trend.
171
and regress it on time:
ls log(PROD) c t
to get the structural productivity trend.
Results are quite bad. Of course productivity shows a significant growth, but the standard error is quite high (more
than 5 %). More important, the graph of residuals and the auto-correlation test show that we are not meeting the
condition we have set: that observed productivity fluctuates around a trend, with potential but not unreasonably long
cycles.
Dependent Variable: LOG(Q)
Method: Least Squares
Date: 03/22/20 Time: 13:16
Sample (adjusted): 1963Q1 2004Q4
Included observations: 168 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -25.54456 0.829672 -30.78877 0.0000
T 0.026054 0.000418 62.30088 0.0000
R-squared 0.958986 Mean dependent var 26.14374
Adjusted R-squared 0.958739 S.D. dependent var 0.323533
S.E. of regression 0.065719 Akaike info criterion -2.595031
Sum squared resid 0.716947 Schwarz criterion -2.557841
Log likelihood 219.9826 Hannan-Quinn criter. -2.579938
F-statistic 3881.400 Durbin-Watson stat 0.036326
Prob(F-statistic) 0.000000
26.8
26.4
26.0
.1
25.6
.0
25.2
-.1
-.2
-.3
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
The problem apparently lies in the fact that the average growth rate is consistently higher in the first part of the
period, and lower later. Seen individually, each sub-period might seem to meet the above condition.
172
From the graph, we clearly need one, and probably two breaks. One will observe that the first period follows the first
oil shock, and the beginning of a lasting world economic slowdown. The reason for the second break is less clear
(some countries like the US and Scandinavia show a break in the opposite direction).
For choosing the most appropriate dates, we can use two methods:
• A visual one: 1973 and 1990 could be chosen, possibly plus or minus 1 year.
• A statistical one: the most appropriate test is the Chow breakpoint test, which diagnoses if the introduction
of one or more breaks improve the explanation. To make our choice automatic, we shall consider three
intervals, and apply the test to all reasonably possible combinations of dates from those intervals. As we
could expect, all the tests conclude to a break. But we shall select the combination associated to the lowest
probability (of no break), which means the highest likelihood ratio66. Of course, this criterion works only
because the sample and the number of breaks remain the same.
The best result corresponds actually to 1972q3 and 1992q4, as shown in this table of log-likelihood ratios.
The equation for structural productivity is
log(prle)=c_prle(1)+c_prle(2)*(t-2004)+c_prle(3)*(t-1972 .50)*(t<1972 .50)+c_prle(4)*(t-1992.75)*(t<1992.75)
Dependent Variable: LOG(PRLE)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 15:32
Sample (adjusted): 1963Q1 2004Q4
Included observations: 168 after adjustments
LOG(PRLE)=C_PRLE(1)+C_PRLE(2)*(T-2004)+C_PRLE(3)*(T-T1)
*(T<T1)+C_PRLE(4)*(T-T2)*(T<T2)
Coefficient Std. Error t-Statistic Prob.
C_PRLE(1) 9.839846 0.002212 4447.971 0.0000
C_PRLE(2) 0.009382 0.000262 35.76099 0.0000
C_PRLE(3) 0.022782 0.000457 49.89498 0.0000
C_PRLE(4) 0.015405 0.000361 42.64593 0.0000
R-squared 0.999093 Mean dependent var 9.466339
Adjusted R-squared 0.999077 S.D. dependent var 0.307411
S.E. of regression 0.009341 Akaike info criterion -6.485251
Sum squared resid 0.014310 Schwarz criterion -6.410871
Log likelihood 548.7611 Hannan-Quinn criter. -6.455064
F-statistic 60233.59 Durbin-Watson stat 1.553017
Prob(F-statistic) 0.000000
66
the highest F gives the same conclusion
173
9.8
9.4
.04 9.0
.02
.00 8.6
-.02
-.04
-.06
-.08
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
One will note:
• That we have introduced no residual, contrary to our usual practice.
• That we have introduced reversed trends, which stop after a while instead of starting inside the period.
o Target productivity is not a behavior.
The first element is quite logical: what we are estimating for the model is not actual productivity (this is given in the
model by an identity, dividing actual GDP by employment). We are looking for the exact formula for target
productivity, prone to error only because we have not enough information to produce the true value. If the sample
grew, or the periodicity shortened, the precision would improve constantly. The residual might not decrease, but it
does not represent an error, rather the gap between the actual and “normal” values of labor productivity. Whereas, in
a normal behavioral equation, the residual corresponds to an error on the variable, and cannot be decreased
indefinitely, as the identification of the role of explanatory elements becomes less reliable with their number.
o Partial trends should apply to past periods.
The reason here is purely technical. Our model is designed to be used on the future. So it is essential to make the
forecasting process as easy as possible.
If the partial trends are still active in the future, we shall have to manage them simultaneously. We can expect that we
want to control the global trend of labor productivity, if only to make it consistent with our long-term evolutions of
GDP (which should follow world growth) and employment (which should follow population trends). Obviously,
controlling a single global trend is easier than a combination of three trends.
Also, the last trend is the most important for interpretation of model properties, and it is better to make it the easiest
to observe.
On the other hand, our technique has no bad points, once it has been understood.
Finally, the reason for breaking the trend in 2004 is also associated with handling of its future values. If the global
coefficient is changed, this will be the period for a new break, and this is the best period to introduce it.
We can summarize the above in the following graph.
174
10
-10
-20
-30
-40
-50
65 70 75 80 85 90 95 00 05 10
T-2004
(T-1972.50)*(T<1972.50)
(T-1992.75)*(T<1992.75)
We can see that in the beginning three trends apply, then two, then after 2004 only the global (blue) trend is
maintained.
The results look quite good, both in the validation of coefficients and the graphs. We are presenting the program
version, which will be introduced in the model (as a identity )67.
However, we observe a very high residual in the second quarter of 1968. As we are estimating a trend, this is not the
place for considering a one-period outlier. We will come back to this problem when we estimate employment itself.
Now we must test the stationarity of the residual. We shall use the Dickey Fuller test (or Phillips – Perron).
First we need to generate from the current RESID a variable containing the residual (the test is going to compute its
own RESID, so it is not possible to test on a variable with this name).
• In program form, the test is conducted by:
genr res_prle=resid
uroot(1,p) res_prle
uroot(h,p) res_prle
• Using menus, one has to
o Display the variable
o Select View>Unit root test
o Choose the method.
67
This is not absolutely needed, as a variable depending only on time can be considered exogenous and computed
outside the model. But we want to be able to change the assumption in forecasts, and this is the easiest way.
175
Null Hypothesis: RES has a unit root
Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=13)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -7.029197 0.0000
Test critical values: 1% level -3.469933
5% level -2.878829
10% level -2.576067
*MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation
Dependent Variable: D(RES)
Method: Least Squares
Date: 03/22/20 Time: 15:55
Sample (adjusted): 1963Q3 2004Q4
Included observations: 166 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
RES(-1) -0.675886 0.096154 -7.029197 0.0000
D(RES(-1)) -0.163434 0.074306 -2.199455 0.0293
C 0.000183 0.000677 0.270373 0.7872
R-squared 0.420548 Mean dependent var 2.58E-05
Adjusted R-squared 0.413438 S.D. dependent var 0.011383
S.E. of regression 0.008718 Akaike info criterion -6.628887
Sum squared resid 0.012389 Schwarz criterion -6.572646
Log likelihood 553.1976 Hannan-Quinn criter. -6.606059
F-statistic 59.15010 Durbin-Watson stat 2.024797
Prob(F-statistic) 0.000000
The test concludes very strongly the stationarity of the residual.
Null
The Hypothesis:
values RES_PRLE
of target productivity has aemployment
and desired unit root are given by:
Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=11)
t-Statistic Prob.*
176
Augmented Dickey-Fuller test statistic -5.191811 0.0000
Test critical values: 1% level -3.522887
5% level -2.901779
10% level -2.588280
genr log(prle_t)= c_prle(1)+c_prle(2)*t+c_prle(3)*(t-1972.50)*(t<1972.50)+c_prle(4)*(t-
1992.75)*(t<1992.75)
genr led=q/prle_t
Now, as to the estimation of employment itself, LE will be estimated (using here the developed form) by:
equation eq_le.ls dlog(LE)=c_le(1)*dlog(LED)+c_le(2)*log(LED(-1)/LE(-1))+c_le(3)
where LED is equal to Q/prle_t, the trend obtained in the previous equation.
It is now time to consider the 1968 residual. With such a high value, there should be some economic explanation.
Indeed, the behavior of firms did not follow normal lines. They believed (rightly) that the decrease in production they
were facing was quite temporary. For them, laying out the corresponding number of workers was not reasonable, as it
would cost severance payments, and when things went back to normal there was no reason they would find as
efficient and firm-knowledgeable workers as before.
This means labor productivity decreased, then increased to get back to normal.
Dependent Variable: DLOG(LE)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 16:35
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
DLOG(LE)=C_LE(1)*DLOG(LED)+C_LE(2)*LOG(LED(-1)/LE(-1))
+C_LE(3)+C_LE(4)*((T=1968.25)-(T=1968.50))+EC_LE
Coefficient Std. Error t-Statistic Prob.
C_LE(1) 0.295420 0.030602 9.653533 0.0000
C_LE(2) 0.199058 0.024568 8.102254 0.0000
C_LE(3) 0.000475 0.000190 2.495656 0.0136
C_LE(4) 0.017596 0.003294 5.342659 0.0000
R-squared 0.443499 Mean dependent var 0.000756
Adjusted R-squared 0.433257 S.D. dependent var 0.003228
S.E. of regression 0.002430 Akaike info criterion -9.178280
Sum squared resid 0.000962 Schwarz criterion -9.103598
Log likelihood 770.3864 Hannan-Quinn criter. -9.147968
F-statistic 43.30049 Durbin-Watson stat 1.087096
Prob(F-statistic) 0.000000
The results are rather significant, except for the last coefficient.
177
LE
19,200,000
18,800,000
18,400,000
18,000,000
17,600,000
17,200,000
16,800,000
16,400,000
1965 1970 1975 1980 1985 1990 1995 2000
Following the reasoning made earlier, c_le (3) (or rather c_le(3)/c_le(2)) should represent the logarithm of the long-
term gap between the target employment and the level reached. This gap will be significant if both:
• Employment shows a trend (the target is moving), which means that GDP and target productivity show
different trends.
• A difference between the growths of GDP and target productivity is not compensated immediately (the value
of c_le(1) is different from one)
The second condition is clearly met, but for the first the answer is dubious. Instead of trend, one rather observe a
break in the level. Nevertheless, the coefficient diagnosed as significant.
As to the first coefficients, they are quite significant, maybe lower than expected.
Dependent Variable: DLOG(LE)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 16:59
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
DLOG(LE)=C_LE(1)*DLOG(LED)+C_LE(2)*LOG(LED(-1)/LE(-1))
+C_LE(3)+EC_LE
Coefficient Std. Error t-Statistic Prob.
C_LE(1) 0.166334 0.020297 8.195005 0.0000
C_LE(2) 0.216267 0.026322 8.216159 0.0000
C_LE(3) 0.000598 0.000204 2.931151 0.0039
R-squared 0.346046 Mean dependent var 0.000756
Adjusted R-squared 0.338071 S.D. dependent var 0.003228
S.E. of regression 0.002626 Akaike info criterion -9.028889
Sum squared resid 0.001131 Schwarz criterion -8.972877
Log likelihood 756.9122 Hannan-Quinn criter. -9.006155
F-statistic 43.39114 Durbin-Watson stat 0.894407
Prob(F-statistic) 0.000000
178
.010
.005
.000
-.005
.010 -.010
-.015
.005
.000
-.005
-.010
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
This will not be true in the US case. We will use another data base, this time bi-yearly.
First, the “1992” break exists too, but it is now positive, as shown here:
179
-2.6
-2.8
-3.0
.08 -3.2
-3.4
.04
-3.6
.00
-.04
-.08
1960 1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
Second, employment has grown substantially over the sample period, which means that a constant term is called for:
180
.04
.02
.02
.00
.01
-.02
.00
-.04
-.01
-.02
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
6.3.3.4 A virtuous dummy element
Reverting to the French case, the outlier observed earlier (and in the equation for the change in inventories) remains.
The year 1968 presents a strong negative residual in the first semester, and a negative one for the last. As we are now
considering a dynamic behavior, this is the time for treating this problem.
As stated earlier, French people (and people familiar with French post-war history) will certainly recall the May 1968
“student revolution” which lasted roughly from March to June. During that period, the French economic process was
heavily disturbed, in particular the transportation system, and GDP decreased (by 7.6% for the quarter). If the
equation had worked , employment would have decreased too, especially as productivity growth was quite high. On
the contrary, it remained almost stable.
The explanation is obvious: firms expected the slump to be purely temporary, and activity to start back after a while
(actually they were right, and GDP grew by 7.5% in the next semester, due in part to the higher consumption allowed
by “Grenelle68” wage negotiations, very favorable to workers). They did not want to lay out (at a high cost) workers
whom they would need back later with no guarantee to find the same individuals, familiar with the firms’ techniques.
So the employment level was very little affected.
This means that the global behavior does not apply here, and the period has to be either eliminated from the sample,
or rather treated through a specific dummy variable, taking the value 1 in the first semester and –1 in the second
(when employment increased less than the growth in GDP would call for).
This case is rather interesting: some economists could be tempted to introduce dummies just because the equation
does not work, and indeed the results will be improved (including in general the statistics for the explanatory
variables). This can probably be called cheating. On the contrary, not introducing the present dummy can be
considered incorrect: we know that the local behavior did not follow the formulation we have selected, so it has to be
modified accordingly.
68
From the location of the Ministry of Employment where negotiations where conducted.
181
The global results are slightly improved, and the first coefficient increases significantly, meaning that the adaptation of
employment to the target is more effective at first. The introduction of the element was not a negligible issue.
Dependent Variable: DLOG(LE)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 17:15
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
DLOG(LE)=C_LE(1)*DLOG(LED)+C_LE(2)*LOG(LED(-1)/LE(-1))
+C_LE(3)+C_LE(4)*((T=1968.25)-(T=1968.50))+EC_LE
Coefficient Std. Error t-Statistic Prob.
C_LE(1) 0.295420 0.030602 9.653533 0.0000
C_LE(2) 0.199058 0.024568 8.102254 0.0000
C_LE(3) 0.000475 0.000190 2.495656 0.0136
C_LE(4) 0.017596 0.003294 5.342659 0.0000
R-squared 0.443499 Mean dependent var 0.000756
Adjusted R-squared 0.433257 S.D. dependent var 0.003228
S.E. of regression 0.002430 Akaike info criterion -9.178280
Sum squared resid 0.000962 Schwarz criterion -9.103598
Log likelihood 770.3864 Hannan-Quinn criter. -9.147968
F-statistic 43.30049 Durbin-Watson stat 1.087096
Prob(F-statistic) 0.000000
.010
.005
.000
.008
.004 -.005
.000 -.010
-.004
-.008
-.012
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
We shall use:
equation eq_prle.ls(p) log(prle)=c_prle(1)+c_prle(2)*t+c_prle(3)*(t-1973)*(t<1973)
+c_prle(4)*(t-1992)*(t<1992)
equation eq_le.ls(p) dlog(le)=c_le(1)*dlog(led)+c_le(2)*log(led(-1)/le(-1))+c_le(3)
+c_le(4)*((t=1968.5)-(t=1968))+ec_le
182
genr ec_le=resid
and as we have to introduce two new variables
g_vendo.drop led prle_t
g_vendo.add led prle_t
Note: the reason for the initial “drop” statement is to avoid the duplication of the elements inside the group, in case
the procedure is repeated. If the elements are not present, nothing happens.
6.3.4 EXPORTS: AUTOREGRESSIVE PROCESS, COINTEGRATION, LONG-TERM STABILITY.
Estimating exports will be simpler from the theoretical side. We shall use it as an example for introducing first
autoregressive processes, then cointegration.
Let us first start with the simplest idea: exports show a constant elasticity to world demand. In other words:
𝛥𝑋/𝑋/(𝛥𝑊𝐷/𝑊𝐷) = 𝑎
or by integration:
𝐿𝑜𝑔(𝑋) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑊𝐷) + 𝑏
Estimation should give to a value close to unity, as world demand is measured as the normal demand addressed to
France by its clients, and takes into account:
• The expansion of international trade.
• The types of goods France exports,
• The structure of the countries to which France naturally exports.
For instance, both luxury goods and Germany have a higher share in this indicator, compared to the global world
market. As a weighting of normal imports, it takes also into account the growing importance of international trade.
Indeed the coefficient we obtain is close to unity (but significantly different).
183
Dependent Variable: LOG(X)
Method: Least Squares
Date: 03/22/20 Time: 17:39
Sample (adjusted): 1974Q1 2004Q4
Included observations: 124 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(WD) 0.887012 0.004704 188.5550 0.0000
C 2.940895 0.115151 25.53944 0.0000
R-squared 0.996580 Mean dependent var 24.64867
Adjusted R-squared 0.996552 S.D. dependent var 0.446469
S.E. of regression 0.026216 Akaike info criterion -4.428916
Sum squared resid 0.083846 Schwarz criterion -4.383427
Log likelihood 276.5928 Hannan-Quinn criter. -4.410437
F-statistic 35552.98 Durbin-Watson stat 0.480527
Prob(F-statistic) 0.000000
25.6
25.2
24.8
.12
24.4
.08
24.0
.04
23.6
.00
-.04
-.08
74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
However the low value of the Durbin-Watson test indicates a strongly positive autocorrelation of residuals, and
invalidates the formulation. The graph shows indeed long periods with a residual of the same sign, even though the
two variables share quite often common evolutions.
Let us try to eliminate auto-correlation, supposing that the residual is actually:
𝑒𝑡 = 𝜌 ⋅ 𝑒𝑡−1 + 𝑢𝑡
where should be significant (positive here), and u(t) independent across time.
The simplest idea is to transform the equation
184
𝐿𝑜𝑔(𝑋𝑡 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑊𝐷𝑡 ) + 𝑏 + 𝑒𝑡
which is also true in the previous period.
𝐿𝑜𝑔(𝑋𝑡−1 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑊𝐷𝑡−1 ) + 𝑏 + 𝑒𝑡−1
We can multiply the second equation by , and subtract it from the first:
𝐿𝑜𝑔(𝑋𝑡 ) − 𝜌 ⋅ 𝐿𝑜𝑔(𝑋𝑡−1 ) = 𝑎 ⋅ (𝐿𝑜𝑔(𝑊𝐷𝑡 ) − 𝜌 ⋅ 𝐿𝑜𝑔(𝑊𝐷𝑡−1 )) + 𝑏 ⋅ (1 − 𝜌) + 𝑒𝑡 − 𝜌 ⋅ 𝑒𝑡−1
The residual for the new equation is the uncorrelated u:
𝐿𝑜𝑔(𝑋𝑡 ) − 𝜌 ⋅ 𝐿𝑜𝑔(𝑋𝑡−1 ) = 𝑎 ⋅ (𝐿𝑜𝑔(𝑊𝐷𝑡 ) − 𝜌 ⋅ 𝐿𝑜𝑔(𝑊𝐷𝑡−1 )) + 𝑏 ⋅ (1 − 𝜌) + 𝑢𝑡
6.3.4.1 Introducing an autoregressive process
To estimate the above formula, it is not necessary to establish the full equation (which calls for a full non–OLS
specification, as it is not linear in the coefficients).
One can very well use the same presentation as for ordinary least squares, introducing in the estimation window the
additional term AR(n), n representing the autocorrelation lag, in our case 1:
ls Log(X) log(WD) c ar(1)
But the application to the developed formula is also quite simple:
equation eq_lx.ls Log(X)=c_x(1)*log(WD)+c_x(2)+[ar(1)=c_x(3)]
185
Dependent Variable: LOG(X)
Method: ARMA Conditional Least Squares (Marquardt - EViews legacy)
Date: 03/22/20 Time: 17:50
Sample (adjusted): 1974Q2 2004Q4
Included observations: 123 after adjustments
Convergence achieved after 1 iteration
LOG(X)=C_X(1)*LOG(WD)+C_X(2)+[AR(1)=C_X(3)]
Coefficient Std. Error t-Statistic Prob.
C_X(1) 0.884418 0.012700 69.64166 0.0000
C_X(2) 3.003843 0.311465 9.644243 0.0000
C_X(3) 0.757250 0.059233 12.78433 0.0000
R-squared 0.998533 Mean dependent var 24.65426
Adjusted R-squared 0.998509 S.D. dependent var 0.443924
S.E. of regression 0.017142 Akaike info criterion -5.270440
Sum squared resid 0.035263 Schwarz criterion -5.201850
Log likelihood 327.1320 Hannan-Quinn criter. -5.242579
F-statistic 40847.79 Durbin-Watson stat 2.304652
Prob(F-statistic) 0.000000
Inverted AR Roots .76
The results are rather satisfactory: the first coefficient retains the theoretical value, the new coefficient is significant,
the global precision is much improved (see also the graph) and the DW test is closer to satisfactory.
25.6
25.2
24.8
.06 24.4
.04 24.0
.02
23.6
.00
-.02
-.04
-.06
74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
However our formulation is a little too simplistic. We want exports to decrease with the rate of use of capacities,
representing the fact that if firms are already using selling most of their potential production, they will be to be less
dynamic in their search for foreign markets (more on this later).
Let us introduce the rate of use UR in the formula.
We get:
186
Dependent Variable: LOG(X)
Method: ARMA Conditional Least Squares (Marquardt - EViews legacy)
Date: 03/22/20 Time: 18:00
Sample: 1978Q1 2004Q4
Included observations: 108
Convergence achieved after 13 iterations
LOG(X)=C_X(1)*LOG(WD)+C_X(2)*LOG(UR)+C_X(3)+[AR(1)=C_X(4)]
Coefficient Std. Error t-Statistic Prob.
C_X(1) 0.381166 0.088426 4.310555 0.0000
C_X(2) 0.219096 0.083380 2.627682 0.0099
C_X(3) 21.58771 36.98678 0.583660 0.5607
C_X(4) 0.998913 0.006440 155.1013 0.0000
R-squared 0.998621 Mean dependent var 24.74243
Adjusted R-squared 0.998581 S.D. dependent var 0.399745
S.E. of regression 0.015058 Akaike info criterion -5.517431
Sum squared resid 0.023582 Schwarz criterion -5.418092
Log likelihood 301.9413 Hannan-Quinn criter. -5.477153
F-statistic 25100.00 Durbin-Watson stat 2.144309
Prob(F-statistic) 0.000000
Inverted AR Roots 1.00
The relevant coefficients are significant, the average error lower 69 (1,5%), the Durbin-Watson test acceptable, but
there is a problem: the sign for the new element is wrong (and unsurprisingly, the coefficient for World demand is
now much too low.
Let us not despair. If this old fashioned tool did not work, let us try a more up to date: cointegration.
6.3.4.2 Applying cointegration under EViews
Just as for stationarity, we will not develop the theory behind the method, leaving it to actual econometricians, an
excellent source of information being actually the EVIews manual. We sill also rely on basic cointegration theory, and
not on any of the recent developments.
Let us just say that cointegration is actually a simple extension of stationarity to a set of two or more variables. To
establish cointegration between two elements, one has to prove that in the long run these elements move together,
maintaining a bounded “distance” (or rather that a linear combination is bounded), while the value of each of the two
elements is unbounded (a necessary condition).
For a group of more than two elements to be cointegrated, no subset of this group must have this property
(stationarity of no single element and cointegration of no subset).
If we want to go beyond intuition, the reason for the last condition is that if a cointegrating relation is evidenced
between elements, some of which are already cointegrated, one can always recompose the encompassing equation
into the true cointegrating equation (considered as a new stationary variable) and other variables.
69
As the logarithm measures relative evolutions, an absolute error on a logarithm is equivalent to a relative error on
the variable itself.
187
For instance,
if
𝑎⋅𝑥+𝑏⋅𝑦+𝑐⋅𝑧
is tested as a cointegrating equation, but:
𝑎 ⋅ 𝑥 + 𝑏′ ⋅ 𝑦
is too (we can use the same a as a cointegrating equation is known to a given factor), then
𝑎⋅𝑥+𝑏⋅𝑦+𝑐⋅𝑧
is equivalent to:
(𝑎 ⋅ 𝑥 + 𝑏′ ⋅ 𝑦) + (𝑏′ − 𝑏) ⋅ 𝑦 + 𝑐 ⋅ 𝑧
three new elements, one of which is stationary, which forbids us to test cointegration on the three.
So the two properties must be checked: moving together means both “moving” and “together”.
Using images rather related to stationarity (as they apply to the actual difference of two elements, without weighting
coefficients) we can illustrate the concept as
• Astral bodies moving in outer space and linked together by gravity. Their distance is bounded but their
position relative to each other is unknown within those bounds, and we do not know if one is leading the
other.
• Human beings: if they are always close to each other, they can be decided to be related (love, hate,
professional relationship). But only if they move: if they are in jail, a small distance means nothing.
In our example, the first idea could be to test cointegration between X, WD and UR. But to ensure the stability of our
long-term simulations, we need exports to have a unitary elasticity to WD. If this is not the case, when X reaches a
constant growth rate, it will be different from that of WD: either France will become the only exporter in the world (in
relative terms) or the role of France in the world market will become infinitely negligible. Both prospects are
unacceptable (the first more than the second, admittedly).
188
This constraint can be enforced very easily by considering only in the long run (cointegrating) equation the ratio of X
to WD, which we shall link to the rate of use. We will test cointegration between these two elements.
Let us first test their stationarity. We know how to do it (from the estimation of employment).
In a program, this can be done through:
UROOT(1,p) Log(X/WD)
UROOT(1,p) Log(UR)
Note: if we use menus, we should first display the group (either by selecting elements in the workfile window or
creating a group). The default display mode is “spreadsheet” but the “View” item proposes other modes, among them
“Unit root test” and “cointegration test” (if more than one series is displayed at the same time).
189
Null Hypothesis: LOG(X/WD) has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 1 (Automatic - based on SIC, maxlag=12)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -2.515780 0.3201
Test critical values: 1% level -4.046072
5% level -3.452358
10% level -3.151673
*MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation
Dependent Variable: D(LOG(X/WD))
Method: Least Squares
Date: 03/22/20 Time: 19:06
Sample: 1978Q2 2004Q4
Included observations: 107
Variable Coefficient Std. Error t-Statistic Prob.
LOG(X(-1)/WD(-1)) -0.157329 0.062537 -2.515780 0.0134
D(LOG(X(-1)/WD(-1))) -0.265307 0.092520 -2.867562 0.0050
C 0.038142 0.016339 2.334424 0.0215
@TREND("1978Q2") -0.000260 0.000117 -2.217517 0.0288
R-squared 0.172932 Mean dependent var -0.001515
Adjusted R-squared 0.148842 S.D. dependent var 0.017713
S.E. of regression 0.016341 Akaike info criterion -5.353556
Sum squared resid 0.027505 Schwarz criterion -5.253637
Log likelihood 290.4152 Hannan-Quinn criter. -5.313050
F-statistic 7.178750 Durbin-Watson stat 1.937887
Prob(F-statistic) 0.000201
190
Null Hypothesis: LOG(UR) has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 10 (Automatic - based on SIC, maxlag=12)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -2.399050 0.3778
Test critical values: 1% level -4.054393
5% level -3.456319
10% level -3.153989
*MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation
Dependent Variable: D(LOG(UR))
Method: Least Squares
Date: 03/22/20 Time: 19:08
Sample (adjusted): 1980Q3 2004Q4
Included observations: 98 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(UR(-1)) -0.091478 0.038131 -2.399050 0.0186
D(LOG(UR(-1))) 1.464814 0.094658 15.47473 0.0000
D(LOG(UR(-2))) -1.665424 0.170251 -9.782189 0.0000
D(LOG(UR(-3))) 1.785757 0.212274 8.412515 0.0000
D(LOG(UR(-4))) -1.737132 0.246525 -7.046472 0.0000
D(LOG(UR(-5))) 1.618277 0.251123 6.444148 0.0000
D(LOG(UR(-6))) -1.586227 0.255319 -6.212736 0.0000
D(LOG(UR(-7))) 1.412498 0.236345 5.976417 0.0000
D(LOG(UR(-8))) -1.047248 0.214050 -4.892542 0.0000
D(LOG(UR(-9))) 0.719400 0.157728 4.561021 0.0000
D(LOG(UR(-10))) -0.292147 0.096418 -3.030009 0.0032
C -0.019947 0.008365 -2.384508 0.0193
@TREND("1978Q2") 3.72E-05 3.38E-05 1.101800 0.2737
R-squared 0.793872 Mean dependent var 0.000238
Adjusted R-squared 0.764772 S.D. dependent var 0.018172
S.E. of regression 0.008813 Akaike info criterion -6.502075
Sum squared resid 0.006603 Schwarz criterion -6.159171
Log likelihood 331.6017 Hannan-Quinn criter. -6.363378
F-statistic 27.28050 Durbin-Watson stat 2.006954
Prob(F-statistic) 0.000000
These first tests show that both UR and the ratio of exports to world demand cannot be considered stationary, even
around a trend: the T statistic is too low, and the estimated probability for the coefficient to zero is too high 70.71
70
Not extremely high, however.
191
Let us now see if the two elements are co integrated, using the Johansen test.
For EViews this calls for:
coint(option,p) list-of-variables or group-name
Option represents the type of cointegration tested:
a No deterministic trend in the data, and no intercept or trend in the cointegrating equation.
b No deterministic trend in the data, and an intercept but no trend in the cointegrating equation.
c Linear trend in the data, and an intercept but no trend in the cointegrating equation.
d Linear trend in the data, and both an intercept and a trend in the cointegrating equation.
e Quadratic trend in the data, and both an intercept and a trend in the cointegrating equation.
s Summarize the results of all 5 options (a-e).
In our case, we shall use option d (trend in the cointegrating equation, no trend in the VAR)
coint(d,p) log(x/wd) log(ur)
71
We can observe that in a traditional least squares estimation, the same T value would give the opposite diagnosis.
192
Date: 03/22/20 Time: 19:09
Sample (adjusted): 1979Q1 2004Q4
Included observations: 104 after adjustments
Trend assumption: Linear deterministic trend (restricted)
Series: LOG(X/WD) LOG(UR)
Lags interval (in first differences): 1 to 4
Unrestricted Cointegration Rank Test (Trace)
Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.211670 31.14976 25.87211 0.0100
At most 1 0.059815 6.414546 12.51798 0.4095
Trace test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.211670 24.73521 19.38704 0.0076
At most 1 0.059815 6.414546 12.51798 0.4095
Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I):
LOG(X/WD) LOG(UR) @TREND(62Q2)
51.27274 34.10694 0.080146
-18.80755 21.73616 -0.041619
Unrestricted Adjustment Coefficients (alpha):
D(LOG(X/WD)) -0.005934 0.002273
D(LOG(UR)) -0.003137 -0.001939
1 Cointegrating Equation(s): Log likelihood 622.6250
Normalized cointegrating coefficients (standard error in parentheses)
LOG(X/WD) LOG(UR) @TREND(62Q2)
1.000000 0.665206 0.001563
(0.12689) (0.00013)
Adjustment coefficients (standard error in parentheses)
D(LOG(X/WD)) -0.304258
(0.07800)
D(LOG(UR)) -0.160833
(0.05275)
Obviously, cointegration is accepted (a message says so). But we can also:
• Understand the logical process:
EViews tests first if there is no cointegration. If this is accepted (if it shows a high enough probability), the process
stops. But here this is refused, as the probability (that there is no cointegration) is too low.
193
In this case, there is at least one cointegrating equation, and EViews proceeds to testing if there is more than one. This
cannot be refused here, as this time the probability is too high.
We have at least one relation, and at most one: we have one.
If the second assumption (at most 1 relation) had not been rejected, there would be at least two and we would have
to continue (there cannot be more relations than variables, however).
Evidencing more than one relation is problematic, maybe worse for the model builder than finding no relation (in
which case we can always proceed with adding new elements). Even if a cointegrating equation has no implications on
causality between elements, we generally intend to include it in a single dynamic formula (a VAR), which does explain
a given variable. With two equations, we are stuck with a parasite one, which will be difficult if not impossible to
manage in the context of the model (if we stop at econometrics, the problem is smaller).
• Look at the probabilities:
We can observe if the existence (or the rejection) of at least one relation is barely or strongly accepted (and also of
only one for that matter).
• Observe the coefficients in the cointegrating equation.
The equation introduces a tradeoff between several concepts (here the share of French exports in world demand and
their rate of use). We have always an idea on the sign of the relationship, and also on an interval of economic validity.
There is no guarantee that the value will follow these constraints. It can even happen that the right sign obtained by
Ordinary Least Squares will become wrong when cointegration is tested on the same relation.
Here it is not too difficult to judge on the soundness of the explanation.
First, the sign is right: exports go down when the rate of use goes up (the sign is positive but both elements are on the
same side of the formula).
The size of the coefficient is more difficult to judge. The derivative of the equation relative to Q gives:
(X)/X= -0.665 (UR)/UR= -0.665 ( (Q)/Q - (CAP)/CAP.
In the short run, CAP does not change:
(X)/X= -0.665 (UR)/UR= -0.665 (Q)/Q
(X) = -0.665 X/Q (Q)
Let us suppose an increase in Q of 1 billion Euros coming only from local demand FD. In 2004, the share of exports in
French GDP was 32%. The exports target will decrease by
194
X = 0.665 * 0.32 = 213 millions of Euros.
The substitution effect looks quite reasonable.
Of course:
• In the long run, capacities will build up, UR will get back to its base value (we know that from the investment
equation) and the loss will disappear.
• The changes in Q can come also from X, introducing a loop. This means UR might not be the best
representative element. Perhaps we should restrict UR to the satisfaction of local demand (but this looks
difficult to formulate).
6.3.4.3 Once cointegration has been evidenced
Two things have to be done:
• Storing the cointegrating equation and its parameters.
• Estimating the VAR (the dynamic equation) and creating the associated element.
The first task should be easy, as EViews does display the requested equation, with its values. However:
• This equation is not available as an item.
• The coefficients are not available as a vector (or as scalars).
There is a trick, however, which solves this problem. One can estimate a VAR, in other terms a system which includes
both a dynamic equation and a cointegrating equation. This is not directly useful for us, as
• This requires to use the same variables in both forms, in other words to extend the unitary elasticity
assumption to the dynamic equation, which is neither needed statistically nor realistic from an economic
point of view (as we have seen when estimating employment).
• The output does not provide information on the quality of the cointegration, an essential element in our
process.
But the good point is that by estimating a VAR using the same elements as the tested cointegration, we get the same
cointegrating coefficients, and this time they are stored in a vector! We can then specify the cointegrating equation
using these elements.
This has the extremely high advantage of allowing to establish a program which will adapt automatically to any change
in the data, an essential element in all operational modelling projects.
For this particular example, after:
coint(d,p) log(x/wd) log(ur)
195
we shall use
vary _var_x.ec(d,p) 1 1 log(x/wd) log(ur)72
to create and estimate the VAR, and store the coefficients in a vector by accessing the first line of matrix _var_x.b (not
displayed in the workfile window):
vector(10) p_x
p_x(1)=_var_x.b(1,1)
p_x(2)=_var_x.b(1,2)
p_x(3)=_var_x.b(1,3)
The cointegrating equation will be:
0= p_x(1)*log(x/wd)+p_x(2)*log(ur)+p_x(3)*@trend(60S1)
Actually the first parameter is not really needed, as it is equal by construction. We think using it makes the equation
clearer.
Estimating the dynamic equation calls for the computation of the residual in the cointegrating equation:
genr res_x= p_x(1)*log(x/wd)+p_x(2)*log(ur)+p_x(3)*t
Then we estimate the VAR, releasing the constraint on the unitary elasticity of X to WD. In principle, the coefficient
should be positive for WD, negative for UR, and we can introduce a lag structure for both elements.
Unfortunately, the coefficient for UR is not significant here, maybe because of a delay in its influence.
72
The figures 1 and 1 indicate the scope of lagged variations of the left-hand side variable in the VAR which will be
added to the right-hand side. Here it will be 1 to 1 (or 1 with lag 1). The options used are consistent with the ones we
have used in coint (which have been determined automatically by EViews).
196
Dependent Variable: DLOG(X)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 19:18
Sample (adjusted): 1978Q3 2004Q4
Included observations: 106 after adjustments
DLOG(X)=C_X(4)*DLOG(WD)+C_X(6)*RES_X(-1)+C_X(7)
Coefficient Std. Error t-Statistic Prob.
C_X(4) 0.494017 0.084894 5.819245 0.0000
C_X(6) -0.091424 0.038151 -2.396342 0.0184
C_X(7) 0.019601 0.006141 3.191585 0.0019
R-squared 0.266535 Mean dependent var 0.011661
Adjusted R-squared 0.252293 S.D. dependent var 0.017116
S.E. of regression 0.014800 Akaike info criterion -5.560452
Sum squared resid 0.022562 Schwarz criterion -5.485071
Log likelihood 297.7039 Hannan-Quinn criter. -5.529900
F-statistic 18.71468 Durbin-Watson stat 2.032482
Prob(F-statistic) 0.000000
.06
.04
.02
.06
.00
.04
-.02
.02
-.04
.00
-.02
-.04
78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
6.3.5 IMPORTS: GOING FURTHER ON COINTEGRATION AND LONG-TERM STABILITY.
In a single country model, the rest of the world is exogenous, and imports and exports have to be estimated
separately, following of course the same guidelines:
• Imports will depend on demand and capacity utilization, through constant elasticities.
However, the definition of demand is not straightforward. For exports, we just had to consider global imports from
each partner country in each product, and compute an average using a double weighting: the role of these partners in
French exports, and the structure of products exported by France.
197
This was possible because we considered the rest of the world as exogenous, and did not try to track the origin of its
imports.
Now imports can come from three endogenous elements:
• Local final demand, such as foreign cars.
• The intermediate goods necessary to local firms to satisfy this local demand. For cars it will be electronics,
steel, energy to run the machines…
• Identically, intermediate goods necessary to produce exported goods.
• But not finished goods used to satisfy foreign demand: France does not re-export significantly goods without
transformation (contrary to Hong Kong for instance).
Basically, two methods can be considered:
• We can consider final demand and exports. Obviously, they do not have the same impact on imports (due to
the absence of re-exports). We can generate a global demand by applying to exports a correcting factor.
Under rather standard assumptions (same import share in all uses, unitary ratio of intermediate consumption
to value added), this factor can be set at 0.5.
• We can also define intermediate consumption, and add it to final demand to get the total demand of the
country, a share of which will be imported. This method is obviously more acceptable from the economic
point of view. Unfortunately it relies on the computation of intermediate consumption, a variable less
accurately measured, and sensitive to categories and the integration of the productive process73.
We have chosen this last method nevertheless, favoring economic properties over statistical reliability.
Dependent Variable: LOG(M)
Method: Least Squares
Date: 04/25/20 Time: 12:16
Sample (adjusted): 1962Q1 2004Q4
Included observations: 172 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 1.986960 0.013282 149.6030 0.0000
C -30.78521 0.378155 -81.40902 0.0000
R-squared 0.992462 Mean dependent var 25.78331
Adjusted R-squared 0.992417 S.D. dependent var 0.723716
S.E. of regression 0.063021 Akaike info criterion -2.679148
Sum squared resid 0.675173 Schwarz criterion -2.642549
Log likelihood 232.4067 Hannan-Quinn criter. -2.664299
F-statistic 22381.05 Durbin-Watson stat 0.069582
Prob(F-statistic) 0.000000
73
For instance, if good A (say cotton) is used to produce good B (unprinted fabric) which gives good C (printed fabric),
both A and B will be counted as intermediary consumption. If the fabric is printed as the same time it is produced,
only A will be counted. If we consider value added, the total amount will not change, just the number of elements.
198
Dependent Variable: LOG(M)
Method: ARMA Maximum Likelihood (BFGS)
Date: 04/25/20 Time: 12:18
Sample: 1962Q1 2004Q4
Included observations: 172
Convergence achieved after 8 iterations
Coefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 2.027497 0.063354 32.00281 0.0000
C -31.90108 1.786973 -17.85202 0.0000
AR(1) 0.976587 0.020587 47.43628 0.0000
SIGMASQ 0.000271 2.29E-05 11.81244 0.0000
R-squared 0.999480 Mean dependent var 25.78331
Adjusted R-squared 0.999471 S.D. dependent var 0.723716
S.E. of regression 0.016642 Akaike info criterion -5.312913
Sum squared resid 0.046529 Schwarz criterion -5.239715
Log likelihood 460.9105 Hannan-Quinn criter. -5.283215
F-statistic 107738.0 Durbin-Watson stat 1.836241
Prob(F-statistic) 0.000000
Inverted AR Roots .98
27.0
26.5
26.0
25.5
.08 25.0
24.5
.04
24.0
.00
-.04
-.08
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
The fit is quite good including the sensitivity to fluctuations (which was the least we could expect), and the
autocorrelation eliminated. But we face a problem: the high value of the coefficient.
Now the question is: should the growth of demand be the only explanation for the growth of imports? In other words,
is there not an autonomous force increasing the weight of foreign trade, independently from growth itself? Or: if
demand did not grow for a given period, would imports stay stable, or keep part of their momentum? The associated
formula would present:
• A more or less unitary elasticity of imports to demand.
• A positive trend representing the regular growth of world trade.
Applying this idea, we get:
199
Dependent Variable: LOG(M)
Method: Least Squares
Date: 04/25/20 Time: 12:32
Sample (adjusted): 1962Q1 2004Q4
Included observations: 172 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 1.608779 0.056271 28.58972 0.0000
C -42.37624 1.719498 -24.64454 0.0000
T 0.011273 0.001640 6.872870 0.0000
R-squared 0.994108 Mean dependent var 25.78331
Adjusted R-squared 0.994039 S.D. dependent var 0.723716
S.E. of regression 0.055878 Akaike info criterion -2.913993
Sum squared resid 0.527683 Schwarz criterion -2.859095
Log likelihood 253.6034 Hannan-Quinn criter. -2.891719
F-statistic 14257.73 Durbin-Watson stat 0.098040
Prob(F-statistic) 0.000000
Dependent Variable: LOG(M)
Method: ARMA Maximum Likelihood (BFGS)
Date: 04/25/20 Time: 12:33
Sample: 1962Q1 2004Q4
Included observations: 172
Convergence achieved after 7 iterations
Coefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 2.116669 0.100512 21.05895 0.0000
C -26.91516 7.657330 -3.514954 0.0006
T -0.003789 0.004770 -0.794421 0.4281
AR(1) 0.980164 0.021535 45.51418 0.0000
SIGMASQ 0.000269 2.30E-05 11.72939 0.0000
R-squared 0.999483 Mean dependent var 25.78331
Adjusted R-squared 0.999471 S.D. dependent var 0.723716
S.E. of regression 0.016653 Akaike info criterion -5.305020
Sum squared resid 0.046312 Schwarz criterion -5.213523
Log likelihood 461.2317 Hannan-Quinn criter. -5.267897
F-statistic 80700.22 Durbin-Watson stat 1.843649
Prob(F-statistic) 0.000000
This does not work. Either we get autocorrelation, or a negative (but not significant) trend.
The problem with our formulation is actually very clear: in terms of model properties, it is reasonable to suppose that
in the short run, an increase in final demand will increase imports beyond their normal share, by generating local
bottlenecks on domestic supply. But the explanatory element should be the rate of use of capacities. At it is fixed in
the long run, the share should go back to normal with time.
A rate of use of 85% (a normal value over the whole economy) does not mean that all firms work at 85% capacity, in
which case they can easily move to 86% if needed. It means that the rates of use follows a given distribution, some at
lower than 85%, some higher, some at 99%, and a finite number at 100% (see graph).
200
Demand and the rate of use
.6
Baseline
An increase in demand
.5
.4
probability
.3
.2
.1
.0
2500 5000 7500 10000 12500 15000
The rate of use (in 10000ths)
An increase in demand will move the curve to the right, and more firms will face the limit: an increase of 1% will meet
it halfway for firms starting from a 99.5% rate of use. The additional demand, if clients do not accept local substitutes,
will have to be supplied by imports.
However, local firms will react to this situation, and try to gain back lost market shares by increasing their capacities
through investment: this is the mechanism we have described earlier. In our small model, the long-term rate of use is
fixed: the sharing of the additional demand will come back to the base values. These values can increase with time
due to the expansion of world trade.
Our formula will make imports depend on total demand and the rate of use:
𝛥𝑀/𝑀 = 𝑎 ⋅ 𝛥𝑇𝐷/𝑇𝐷 + 𝑏. 𝛥𝑈𝑅/𝑈𝑅
Where
𝐼𝐶 = 𝑡𝑐 . 𝑄
𝑇𝐷 = 𝐹𝐷 + 𝐼𝐶
201
And tc is the quantity of intermediary consumption units required to produce one unit of GDP.
By integration, we get
𝐿𝑜𝑔(𝑀) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑇𝐷) + 𝑏 ⋅ 𝐿𝑜𝑔(𝑈𝑅) + 𝑐
Introducing the rate of use, we get:
Dependent Variable: LOG(M)
Method: Least Squares
Date: 04/25/20 Time: 12:41
Sample (adjusted): 1963Q2 2004Q4
Included observations: 167 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 1.982239 0.013327 148.7350 0.0000
LOG(UR) 1.118627 0.187822 5.955783 0.0000
C -30.65432 0.379702 -80.73265 0.0000
R-squared 0.993071 Mean dependent var 25.82899
Adjusted R-squared 0.992987 S.D. dependent var 0.683591
S.E. of regression 0.057248 Akaike info criterion -2.865037
Sum squared resid 0.537488 Schwarz criterion -2.809025
Log likelihood 242.2306 Hannan-Quinn criter. -2.842303
F-statistic 11752.39 Durbin-Watson stat 0.099373
Prob(F-statistic) 0.000000
Dependent Variable: LOG(M)
Method: ARMA Maximum Likelihood (BFGS)
Date: 04/25/20 Time: 12:43
Sample: 1963Q2 2004Q4
Included observations: 167
Convergence achieved after 7 iterations
Coefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 2.012993 0.082841 24.29940 0.0000
LOG(UR) 0.160941 0.127756 1.259755 0.2096
C -31.50111 2.352651 -13.38962 0.0000
AR(1) 0.972438 0.023201 41.91354 0.0000
SIGMASQ 0.000245 2.23E-05 10.96716 0.0000
R-squared 0.999473 Mean dependent var 25.82899
Adjusted R-squared 0.999460 S.D. dependent var 0.683591
S.E. of regression 0.015881 Akaike info criterion -5.400439
Sum squared resid 0.040859 Schwarz criterion -5.307085
Log likelihood 455.9366 Hannan-Quinn criter. -5.362549
F-statistic 76850.10 Durbin-Watson stat 1.820626
Prob(F-statistic) 0.000000
202
Dependent Variable: LOG(M)
Method: ARMA Maximum Likelihood (BFGS)
Date: 04/25/20 Time: 12:45
Sample: 1963Q2 2004Q4
Included observations: 167
Convergence achieved after 9 iterations
Coefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
LOG(TD) 1.931196 0.201264 9.595350 0.0000
LOG(UR) 0.240297 0.208748 1.151137 0.2514
C -33.98000 7.679693 -4.424657 0.0000
T 0.002421 0.006392 0.378777 0.7054
AR(1) 0.967930 0.024792 39.04209 0.0000
SIGMASQ 0.000245 2.26E-05 10.85188 0.0000
R-squared 0.999473 Mean dependent var 25.82899
Adjusted R-squared 0.999457 S.D. dependent var 0.683591
S.E. of regression 0.015935 Akaike info criterion -5.388806
Sum squared resid 0.040881 Schwarz criterion -5.276782
Log likelihood 455.9653 Hannan-Quinn criter. -5.343338
F-statistic 61066.95 Durbin-Watson stat 1.818857
Prob(F-statistic) 0.000000
Inverted AR Roots .97
No formulation is acceptable on all counts, with no autocorrelation, a significant explanation of the rate of use, and a
positive significant trend (to say nothing of the demand coefficient).
Actually, we are facing a usual problem : the two explanations (deviations of demand from its trend, and output gap)
are strongly correlated.
One idea can be to force the first coefficient to unity, and estimate the ratio of imports to demand.
203
Dependent Variable: LOG(M/TD)
Method: ARMA Maximum Likelihood (BFGS)
Date: 04/25/20 Time: 13:07
Sample: 1963Q2 2004Q4
Included observations: 167
Convergence achieved after 7 iterations
Coefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
LOG(UR) 1.129681 0.096782 11.67241 0.0000
C -60.03575 5.793368 -10.36284 0.0000
T 0.028911 0.002922 9.893121 0.0000
AR(1) 0.969395 0.018275 53.04455 0.0000
SIGMASQ 0.000268 2.33E-05 11.47345 0.0000
R-squared 0.997749 Mean dependent var -2.664699
Adjusted R-squared 0.997693 S.D. dependent var 0.345836
S.E. of regression 0.016610 Akaike info criterion -5.311368
Sum squared resid 0.044692 Schwarz criterion -5.218015
Log likelihood 448.4993 Hannan-Quinn criter. -5.273478
F-statistic 17951.06 Durbin-Watson stat 1.721525
Prob(F-statistic) 0.000000
Inverted AR Roots .97
-2.0
-2.4
-2.8
.08
-3.2
.06
.04 -3.6
.02
.00
-.02
-.04
-.06
1965 1970 1975 1980 1985 1990 1995 2000
Residual Actual Fitted
This works quite well on all counts, and ensures that the growth of imports and demand will converge in the long run,
once the trend has been suppressed, and the rate of use has stabilized through the investment equation.
We could stop here. But as in the exports case, we can try to separate the behavior into short-term and long-term, in
other words to apply a n error correction framework. To represent this process, we need a formula which:
• Enforces a long-term unitary elasticity of imports to the total demand variable, with a positive additional
effect of the rate of use.
• Allows free elasticities in the short run.
We shall start by testing the cointegration between the share of imports in demand: M/TD and the rate of use.
Before, we test the stationarity of M/TD, or rather its logarithm (UR has already been tested):
204
It is strongly contradicted by the Dickey Fuller test:
Null Hypothesis: LOG(M/TD) has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 1 (Automatic - based on SIC, maxlag=14)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -2.921357 0.1582
Test critical values: 1% level -4.007084
5% level -3.433651
10% level -3.140697
*MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation
Dependent Variable: D(LOG(M/TD))
Method: Least Squares
Date: 03/22/20 Time: 20:09
Sample (adjusted): 1963Q3 2010Q4
Included observations: 190 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
LOG(M(-1)/TD(-1)) -0.061830 0.021165 -2.921357 0.0039
D(LOG(M(-1)/TD(-1))) 0.271652 0.067991 3.995397 0.0001
C -0.188041 0.066624 -2.822416 0.0053
@TREND("1962Q1") 0.000421 0.000149 2.828711 0.0052
R-squared 0.108959 Mean dependent var 0.007611
Adjusted R-squared 0.094587 S.D. dependent var 0.019241
S.E. of regression 0.018308 Akaike info criterion -5.142088
Sum squared resid 0.062347 Schwarz criterion -5.073730
Log likelihood 492.4984 Hannan-Quinn criter. -5.114397
F-statistic 7.581507 Durbin-Watson stat 2.018032
Prob(F-statistic) 0.000082
Now we test cointegration of LOG(UR) and LOG(M /TD)
It fails!
205
Date: 03/22/20 Time: 20:12
Sample (adjusted): 1979Q1 2010Q4
Included observations: 128 after adjustments
Trend assumption: Linear deterministic trend (restricted)
Series: LOG(M/TD) LOG(UR)
Lags interval (in first differences): 1 to 4
Unrestricted Cointegration Rank Test (Trace)
Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None 0.056385 11.83231 25.87211 0.8234
At most 1 0.033818 4.403551 12.51798 0.6832
Trace test indicates no cointegration at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None 0.056385 7.428760 19.38704 0.8694
At most 1 0.033818 4.403551 12.51798 0.6832
Max-eigenvalue test indicates no cointegration at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I):
LOG(M/TD) LOG(UR) @TREND(62Q2)
-15.37135 -8.656178 0.136527
27.20696 -15.09145 -0.171007
Unrestricted Adjustment Coefficients (alpha):
D(LOG(M/TD)) 0.001320 -0.001626
D(LOG(UR)) 0.004203 0.001371
1 Cointegrating Equation(s): Log likelihood 731.9615
Normalized cointegrating coefficients (standard error in parentheses)
LOG(M/TD) LOG(UR) @TREND(62Q2)
1.000000 0.563137 -0.008882
(0.43358) (0.00089)
Adjustment coefficients (standard error in parentheses)
D(LOG(M/TD)) -0.020290
(0.01466)
D(LOG(UR)) -0.064610
(0.02652)
Let us not despair. Diagnosis an absence of cointegration is not such bad news74, as it allows us to proceed further. If a
set of two variables does not work, why not a set of three?
Now which additional element could we consider? The natural candidate comes both from theory and from the data:
74
Identifying two equations would be much worse, especially in a modelling framework.
206
If demand is present but local producers have no capacity problems, how can foreign exporters penetrate a market?
Of course, through price competitiveness, in other words by decreasing the import price compared to the local one.
This observation is confirmed by the data. Let us regress the import-demand ratio over the rate of use, consider the
residual (the unexplained part) and compare it to the ratio of import to local prices: we observe a clearly negative
relation.
.24
.20
Residual of log(m/td) over log(ur)
Log of import price competitiveness
.16
.12
.08
.04
.00
-.04
-.08
78 80 82 84 86 88 90 92 94 96 98 00 02
We observe a clear correlation, especially if we consider variations around a negative trend.
After having tested of course:
• Non-stationarity of Log(COMPM)
• Non- cointegration of Log(COMPM) with both Log(UR) and log(M/TD) individually,
We can test the cointegration of the three elements:
It works!!
207
Date: 03/22/20 Time: 20:17
Sample (adjusted): 1979Q3 2004Q4
Included observations: 102 after adjustments
Trend assumption: Linear deterministic trend (restricted)
Series: LOG(M/(FD+CT*Q)) LOG(UR) LOG(COMPM)
Lags interval (in first differences): 1 to 3
Unrestricted Cointegration Rank Test (Trace)
Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.240307 46.08716 42.91525 0.0233
At most 1 0.122323 18.05342 25.87211 0.3403
At most 2 0.045452 4.744804 12.51798 0.6333
Trace test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.240307 28.03374 25.82321 0.0252
At most 1 0.122323 13.30861 19.38704 0.3038
At most 2 0.045452 4.744804 12.51798 0.6333
Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I):
LOG(M/(FD+... LOG(UR) LOG(COMPM) @TREND(62Q2)
-38.34645 58.18449 -21.17524 0.221992
40.54223 -13.27895 14.45687 -0.274847
2.500665 4.998751 15.34349 0.024427
Unrestricted Adjustment Coefficients (alpha):
D(LOG(M/(F... 0.002033 -0.003559 -0.000202
D(LOG(UR)) -0.004216 -0.002450 -0.000251
D(LOG(COM... -0.000361 0.000809 -0.001906
1 Cointegrating Equation(s): Log likelihood 982.2613
Normalized cointegrating coefficients (standard error in parentheses)
LOG(M/(FD+... LOG(UR) LOG(COMPM) @TREND(62Q2)
1.000000 -1.517337 0.552209 -0.005789
(0.16825) (0.08279) (0.00025)
Adjustment coefficients (standard error in parentheses)
D(LOG(M/(F... -0.077941
(0.04357)
D(LOG(UR)) 0.161669
(0.04152)
D(LOG(COM... 0.013852
(0.03721)
2 Cointegrating Equation(s): Log likelihood 988.9156
Normalized cointegrating coefficients (standard error in parentheses)
LOG(M/(FD+... LOG(UR) LOG(COMPM) @TREND(62Q2)
1.000000 0.000000 0.302737 -0.007052
(0.14837) (0.00043)
2080.000000 1.000000 -0.164414 -0.000832
(0.11811) (0.00035)
Adjustment coefficients (standard error in parentheses)
D(LOG(M/(F... -0.222221 0.165519
(0.05989) (0.06405)
D(LOG(UR)) 0.062329 -0.212769
(0.05869) (0.06277)
D(LOG(COM... 0.046648 -0.031760
Of course we have to consider the coefficients in the equation. They describe:
• An apparently high sensitivity of imports to the rate of use (but remember the investment equation will
stabilize it in the end).
However the true effect is not so high. If the equation was applied to the short-term, with a share of imports in total
demand of 0.15 (the value for 2004), we would get:
(M) = 1 x 0.15 (TD) +1.52 x 0.15 (Q)
(M) = 0.15 (TD) +0.23 (Q)
And if we do not consider the change in exports:
(FD - Q) = 0.15 (FD + ct . Q) +0.23 (Q)
(Q) = 0.85 / 1.52 (FD) = 0.56 (FD)
A quite acceptable multiplier for France.
• A much higher sensitivity than in the autoregressive formula.
We can now test the VAR:
209
Dependent Variable: DLOG(M)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 03/22/20 Time: 20:26
Sample (adjusted): 1978Q4 2004Q4
Included observations: 105 after adjustments
DLOG(M)=C_M(5)*DLOG(FD+CT*Q)+C_M(6)*DLOG(UR)+C_M(7)
*DLOG(COMPM)+C_M(8)*RES_M(-1)+C_M(9)+EC_M
Coefficient Std. Error t-Statistic Prob.
C_M(5) 1.676415 0.236019 7.102896 0.0000
C_M(6) 0.213352 0.065300 3.267272 0.0015
C_M(7) -0.105752 0.078397 -1.348931 0.1804
C_M(8) -0.102809 0.025468 -4.036803 0.0001
C_M(9) -0.275761 0.068639 -4.017567 0.0001
R-squared 0.614643 Mean dependent var 0.011899
Adjusted R-squared 0.599229 S.D. dependent var 0.016169
S.E. of regression 0.010236 Akaike info criterion -6.279400
Sum squared resid 0.010477 Schwarz criterion -6.153021
Log likelihood 334.6685 Hannan-Quinn criter. -6.228189
F-statistic 39.87492 Durbin-Watson stat 2.059633
Prob(F-statistic) 0.000000
.06
.04
.02
.03 .00
.02 -.02
.01 -.04
.00
-.01
-.02
-.03
80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
The results are rather acceptable, including the graph. By the way, it shows that rejecting a low R-Squared is not
justified when the dependent element shows a high variability.
6.3.6 BACK TO THE RESIDUAL CHECK
The method we are using for storing equations has an additional advantage. Now that the residuals have been
introduced with their estimated values, all the equations should hold true. The checking process can now be extended
to all the endogenous variables.
Theoretically the estimated equations should be consistent with the data, as merging with the model the actual
estimated equations ensures the consistency. However:
• If the package does not provide this direct storing, or if the equation had to be normalized by the modeler,
editing the formulation could introduce errors.
• The storing of coefficients may have been done badly.
• The text, series or coefficients may have been modified by the user after estimation
210
• One could have accessed other series or coefficients than the ones used by the estimation (for example one
can seek them in another bank including series of similar names).
The reasons for a non-zero residual are less numerous than for identities. They can come only from the fact that
equation elements have changed since the last estimation.
Obviously, the main suspect is the data. New data series are due to be inconsistent with the previous estimation,
whether it has been updated (moving to a more precise version) or corrected (suppressing an error).
Actually, in EViews, applying a new version of an equation to a model requires, in addition to its estimation, to actually
merge it again into the model. This will create a new compiled version, without need to explicitly update the model.
Anyway, in our opinion, applying a new estimation should call for a full model re-creation. This is the only way to
guarantee a clear and secure update.
For our model, the statements for the residual check will be the following:
' We check the residuals
smpl 1980Q1 2002S1
__fra_1.append assign @all _c
solve(d=f) _fra_1
for !i=1 to g_vendo.@count
%2=g_vendo.@seriesname(!i)
genr dc_{%2}={%2}-{%2}_c
genr pc_{%2}=100*dc_{%2}/({%2}+({%2}=0))
next
The solution series will have the suffix “_c”, and the residuals the prefix “dc_” for the errors in levels, and “pc_” for the
relative errors.
We can now present the framework of the model.
211
Architecture of the small model
Scrapping
rate
Firms
Total Civil
GDP employmt
employmt servants
Invest
Capital
ment Household Real
revenue Wage rate
share of GDP
Change
inventor
Capacity
Housing Share of
Investmt revenue
Rate of
use
Consump Savings
Final rate
demand tion
exports imports
Governmt
demand
World Price
demand compet
6.3.7 THE PRESENT MODEL
In EViews notations, its specifications are:
[1] cap = pk * k(-1)
[2] ur = q / cap
[3] q + m = fd + x
[4] i/k( - 1) = 0.944*I( - 1)/K( - 2) - 0.00256*(ur-ur_star)/ur + 0.0428*.125*q/q( - 8) - 0.00448+ ec_i
[5] ic = ct * q
[6] log(prle_t) = 9.8398+ 0.009382* (t - 2002) + 0.02278* (t - t1) * (t<t1) + 0.01540* (t - t2) * (t<t2)
[7] led = q / prle_t
[8] DLOG(le) = 0.2954*DLOG(led) + 0.1990*LOG(led( - 1)/le( - 1)) + 0.0004748* + 0.01759**((T = 1968.25) - (T =
1968.50)) + ec_le
[9] lt = le + lg
[10] rhi = wr * lt + r_rhiq * q
[11] co = rhi * (1 - sr)
[12] ih = r_ih * rhi
212
[13] ci/q( - 1) = -0.02680*(T = 1968.25) + 0.6128*ci( - 1)/q( - 2) - 0.0021490 + 0.2193*@PCH(q) + 0.1056*@PCH(q( - 1))
+ 0.05492*@PCH(q( - 2)) + 0.03918*@PCH(q( - 3)) + 0.03026*@PCH(q( - 4)) + ec_ci
[14] fd = co + i + gd + ci + ih
[15] td = fd + ic
[16] res_m = log(m / (fd + ct * q)) ‘ 1.517 * log(ur) +-0.552 * log(compm) + 0.00579 * (@trend(60:1) * (t<=2004) +
@elem(@trend(60:1) , "2004q4") * (t>2004))
[17] DLOG(m) = 1.676*DLOG(td) + 0.2133*DLOG(ur) - 0.1057*DLOG(compm) - 0.1028*res_m( - 1) - 0.2757+ EC_M
[18] res_x = p_x(1) * log(x / wd) + 0.9815 * log(ur) + 0.001441 * (@trend(60:1) * (t<=2004) + @elem(@trend(60:1) ,
"2004q4") * (t>2004))
[19] DLOG(x) = 0.4940*DLOG(wd) - 0.09142*res_x( - 1) + 0.01960+ ec_x
[20] k = k(-1) * (1 - dr) +i
213
7 CHAPTER 7: TESTING THE MODEL THROUGH SIMULATIONS OVER THE PAST
We have now achieved the production of a model, for which:
• All the identities are consistent with the data.
• All estimations are in our opinion acceptable from a statistical point of view.
• All the coefficient values and equation specifications seem consistent with economic theory75.
• All the necessary connections between model elements are present.
• The estimated equations have been built in such a way that they should provide a steady state growth, and a
long-term solution for the full model.
This does not mean our model is acceptable.
• The consistency of the data might hide errors compensating each other.
• Some complex equations might have hidden wrong individual properties.
• Connections between concepts might have been forgotten or wrongly specified.
• The growth rates provided naturally by equations could be the wrong ones.
• Some error correction processes could be diverging instead of converging.
• Assembling individually acceptable formulations might create a system with unacceptable properties.
• Exogenous elements might be linked with each other.
• Theoretical balances might not have been specified.
And finally:
Another modeler would have obtained a different model, with potentially different properties.
Let us just give an example for a problem: if in the short run increasing government demand by 1000 creates 800
consumption and 600 investment, while exports do not change and imports increase by 300, the individual equations
might look acceptable, but the model will diverge immediately through an explosive multiplier effect (800 +600-300 =
1100).
Our next goal will be to control:
• That the model gives a numerical solution
• That it can be used for forecasts.
• That it can be used for policy analysis.
Actually, our first tests will rather answer the reverse question:
• Is there some indication that the model is unsuitable for forecasts, and for policy analysis?
In our opinion, it is only later, by simulations over the future (its normal field of operation, actually) that we can really
validate the use of a model. But as usual, problems should be diagnosed as soon as possible. And the availability of
actual data increases strongly the testing possibilities.
75
In some cases we might have been obliged to calibrate the values.
214
Finally, the errors evidenced at this stage might help to build a better forecasting structure, before any attempt on the
future.
Let us first address the process of solving the model.
7.1 THE SOLUTION
To solve the model we need to apply a method (an “algorithm”). Let us present the different options.
7.1.1 GAUSS-SEIDEL
This is the most natural algorithm: one often uses Gauss-Seidel without knowing it, like M. Jourdain (the Bourgeois
Gentilhomme) makes prose.
The method starts from initial values. They can be the historical values on the past, on the future the values computed
for the previous period or for an alternate base simulation. The whole set of equations will be applied in a given order,
using as input the most recent information (computations replace the initial values). This gives a new set of starting
values. The process is repeated, using always the last information available, until the distance between the two last
solutions is small enough to be considered negligible. One will then consider that the solution has been reached.
Let us formalize this process.
Considering the model
𝑦𝑡 = 𝑓𝑡 (𝑦𝑡 , 𝑦𝑡−1 , 𝑥𝑡 , 𝛼̂)
Where 𝑦𝑡 is the vector of endogenous variables at period t.
As only present values will change during computation, we will not consider the other elements, and will drop the
time index.
𝑦 = 𝑓(𝑦)
We will use it to define the particular endogenous, and an exponent to define the iteration count.
a - We start from y0, value before any computation, setting the number of iterations to zero.
b - We add 1 to the number of iterations (which we shall note k); this gives to the first iteration the number 1.
k
c -We compute y i from i = 1 to n, taking into account the i-1 values we have just produced. This means we compute:
215
𝑦𝑖𝑘 = 𝑓(𝑦1𝑘 , . . . , 𝑦𝑖−1
𝑘
, 𝑦𝑖𝑘−1 , . . . . , 𝑦𝑛𝑘−1 )
(at the first iteration, explanatory elements will take the starting value y 0 if their index is higher than the computed
variable)76.
d – At the end of the process, we compare yk and yk-1: if the distance is small enough for every element (using a
criterion we shall present) we stop the process, and use the last value as a solution. If not, we check if we have
reached the maximum number of iterations, in which case we accept the failure of the algorithm, and stop. Otherwise
we resume the process at step b.
Clearly, this algorithm requests an identified model, with a single variable on the left hand side (or an expression
containing a single simultaneous variable).
7.1.2 RITZ-JORDAN
The Ritz-Jordan method is similar to the one above: it simply abstains from using values computed at the current
iteration:
𝑦 𝑘 = 𝑓(𝑦 𝑘−1 )
Refusing to consider the last information, it looks less efficient than Gauss-Seidel. In our opinion, its only interest
appears when the model faces convergence problems: it makes their interpretation easier by reducing the
interferences between variables.
This method is not provided by EViews.
7.1.3 NEWTON AND ITS VARIANTS
Contrary to the two above, the Newton method applies naturally to non-identified formulations. It represents
actually a generalization to an n-dimensional case of the well-known method using a sequence of linearizations to
solve a single equation.
Let us consider the model:
𝑓𝑡 (𝑦𝑡 , 𝑦𝑡−1 , 𝑥𝑡 , 𝛼̂) = 0
76
This means only variables which are used before they are computed must be given values for initialization. We shall
come back to this later.
216
that we will simplify as above into:
𝑓 (𝑦) = 0
The linearization of f around a starting solution gives, by calling “fl” the value of f linearized:
(𝜕𝑓/𝜕𝑦)𝑦=𝑦0 ⋅ (𝑦 − 𝑦 0 ) = 𝑓𝑙(𝑦) − 𝑓(𝑦 0 )
Solving the system for fl (y) = 0 leads to:
𝑦 = 𝑦 0 − (𝜕𝑓/𝜕𝑦)−1 0
𝑦=𝑦 0 ⋅ 𝑓(𝑦 )
or with an identified system:
𝑦 − 𝑓(𝑦) = 0
we would get naturally:
𝑦 = 𝑦 0 − (𝐼 − 𝜕𝑓/𝜕𝑦)−1 0 0
𝑦=𝑦 0 ⋅ (𝑦 − 𝑓(𝑦 ))
217
The Newton method (one equation)
f(y0)
f(y1)
f(y2)
y2 y1 y0
Linearizing the model again, around the new solution y 1, and solving the new linearized version of the model, we
define an iterative process which, as the preceding, will stop when the distance between the two last values gets small
enough. Implementing this method is more complex: in addition to inverting a matrix, each iteration involves the
computation of a Jacobian. This can be done in practice in two ways:
• Analytically, by determining from the start the formal expressions of derivatives. At each iteration, we shall
compute them again from the present values of variables. This method supposes either undertaking the
derivation "by hand" with a high probability of errors, or having access to an automatic formal processor, a
program analyzing the text of equations to produce the formal expression of their derivatives. To a high initial
cost, this method opposes a simplification of computations during the iterative process 77.
EViews allows both methods78.
• By finite differences, determining separately each column of the Jacobian by the numerical computation of a
limited first order development, applied in turn to each variable. One computes the y vector using the base
values, then for starting values differing only by a single variable, and compares the two results to get a
column of the Jacobian. One will speak then of a method of secants, or pseudo-Newton.
The derivate formulation becomes in this case:
77
However, changing some model specifications calls for a new global derivation (or a dangerous manual updating).
78
Unfortunately, the associated code is not apparently available to the user, which would allow interesting
computations.
218
f ( y
n
i
k
+ e j y j ) − f ( y k ) / y j ( y j − y kj ) = fli ( y ) − f i ( y k )
j =1
where ej is a vector of dimension n (number of endogenous), with 1 in position i and 0 otherwise.
In other words, the element of the Jacobian:
fi / y j will be approximated by fi ( y k + e jy j ) − f ( y k ) / y j
One will have only to compute the y vector n+1 times: one time with no modification and one time for each of the
endogenous variables.
The expensive part of this algorithm being clearly the computation of the Jacobian and its inversion, a variant will
consist in computing it only each m iterations. The convergence will be slower in terms of number of iterations, but
the global cost might decrease.
EViews provides another alternative: Broyden’s method, which uses a secant method and does not require to
compute the Jacobian at each step. As we shall see later, this method proves often very efficient.
7.1.3.1 The identified case
If the model is in “identified” form:
𝑦 = 𝑓 (𝑦)
the Newton algorithm will be applied to
𝑦 − 𝑓(𝑦) = 0
and the Newton formula becomes:
219
𝑦 = 𝑦 0 − (𝐼 − 𝜕𝑓/𝜕𝑦)−1 0 0
𝑦=𝑦 0 ⋅ (𝑦 − 𝑓(𝑦 ))
or
𝑦 = (𝐼 − 𝜕𝑓/𝜕𝑦)−1 0 0
𝑦=𝑦 0 ⋅ (𝑓(𝑦 ) − (𝜕𝑓/𝜕𝑦)𝑦=𝑦 0 ⋅ 𝑦 )
This does not change the technical process.
7.1.4 BROYDEN’S METHOD
Broyden’s method (also called secant method) computes the Jacobian only once, in the same way as Newton’s,
and computes a new value of the variable accordingly.
After that, it updates the Jacobian, not by derivation, but by considering the difference with the previous solution,
and the direction leading from the previous solution to the new one.
The formula for updating the Jacobian is:
𝐽𝑡+1 =𝐽𝑡 + (𝐹(𝑥𝑡+1 ) − 𝐹(𝑥𝑡 ) − 𝐽𝑡 ∆𝑥𝑡 ). ∆𝑥 ′ 𝑡 /(∆𝑥 ′ 𝑡 . ∆𝑥𝑡 )
where J is the Jacobian, F the function which should reach zero, and x the vector of unknown variables.
Let us clarify all this with a graph based on the single equation case.
220
We can see that the direction improves with each iteration, less than Newton but more than Gauss-Seidel (for
which it does not improve at all).
Otherwise the method shares all the characteristics of Newton’s, in particular its independence to equation
ordering. It takes generally more iterations, but each of them is cheaper (except for the first).
We shall see on a set of practical examples that on average it looks like the most efficient option on the whole,
both in terms of speed and probability of convergence 79. But the diagnosis is not so clear cut.
7.1.5 ITERATIONS AND TEST OF CONVERGENCE
Methods described above have a common feature: starting from initial values, they apply formulations to get a new
set. The process is repeated until the two last sets are sufficiently close to be considered as the solution of the system.
One cannot identify the difference between two iterations with the precision actually reached (or the difference to
the solution). This is valid only for alternate processes. For monotonous ones, it actually can be the reverse: the slower
the convergence, the smaller the change in the criterion from one iteration to the other, and the higher the chance
that the criterion will be reached quite far from the solution. As to cyclical processes, they can reach convergence
mistakenly at the top or bottom of a cycle.
79
The most important feature in our opinion.
221
values
iterations
convergence
So one could question the use of this type of method, by stressing that the relative stability of values does not mean
that the solution has been found. However, one can observe that if the values do not change, it means that the
computation which gave a variable would give the same result with the new values of its explanatory variables: it
means also that the equation holds almost true (but that very different values have the same property).
In this case, it is clear that we do not get the exact solution. This criticism should not be stretched too much: the
precision of models is in any event limited, and even supposedly exact algorithms are limited by the precision of
computers.
7.1.5.1 The general options
For the algorithm to know at which moment to stop computations, we shall have to establish a test.
In fact, the only criterion used in practice will consider the variation of the whole set of variables in the solution, from
an iteration to the other.
It can be measured, for each variable:
in relative values:
𝑑𝑖 = |(𝑦𝑖𝑘 − 𝑦𝑖𝑘−1 )/𝑦𝑖𝑘−1 |
or in levels:
222
𝑑𝑖 = |𝑦𝑖𝑘 − 𝑦𝑖𝑘−1 |
As to the condition for accepting convergence, it can be defined:
by variable: 𝑑𝑖 < 𝑐𝑖 , ∀𝑖
on the whole set: 𝑑𝑖 < 𝑐𝑖 , ∀𝑖
or sometimes through a global measure: 𝑓(𝑑) < 𝑐
Generally one will choose a criterion in relative value, each error being compared with a global criterion. This value
will have to be small compared to the expected model precision (so that the simulation error will not really contribute
to the global error), and to the number of digits used for results interpretation.
The most frequent exception should correspond to variables which, like the trade balance, fluctuate strongly and can
even change sign: here the choice of a criterion in level seems a natural solution, which will avoid a non-convergence
diagnosis due to negligible fluctuations of a variable around a solution (by pure chance) very small.
For example, if the convergence threshold is 0.0001 in relative value, convergence will be refused if solutions for the
US trade balance alternate by chance between - 1 billion current US Dollars and - 1.0002 billion80, while a difference of
200 000 Dollars, at the scale of US foreign trade, is obviously very small. And this difference, which represents less
than one millionth of US exports and imports, might never be reduced if the computer precision guarantees only 8
significant figures81.
In practice we shall see that the test could be restricted to a subset of variables in the model, the convergence of
which extends mathematically to global convergence.
The value given to the criterion can depend:
• On the algorithm used:
o In case of Gauss-Seidel, each additional digit bears roughly the same cost. The convergence is qualified as
linear.
o In case of Newton, the number of digits gained increases with the iterations: beyond the minimum level (say
0.01%) a given gain is cheaper and cheaper (this will be developed later). The convergence is called quadratic
80
There is no risk for this in present times.
81
Exports and imports will be precise to the 8th digit, but the difference, a million times smaller, to the 2nd only.
223
• On the type of simulation:
o For a forecast, one will not be too strict, as we all know the precision is quite low anyway. Forecasting growth
of 2.05% and 2.07% three years from now delivers the same message, especially as the actual growth might
materialize as 1% (more on forecast precision later).
o For a shock analysis, especially if the shock is small, the evaluation of the difference between the two
simulations is obviously more affected by the error: decisions increasing in GDP by 0.07% and 0.09% will not
be given the same efficiency.
• And perhaps on the stochastic character:
In a stochastic simulation, it is essential that the consequence for the solution of introducing small random residuals is
precisely associated with the shock, and not on the simulation process.
As to the number of iterations, it will be used as a limit, after which we suppose that the model has no chance to
converge. In practice one never considers stopping an apparently converging process, just because it has taken too
much time. So the only relevant case is when the process is not progressing, because it is oscillating between two or
more solutions, and the deadlock has to be broken. Reverting to the use of damping factors (described later) should
solve the problem in the Gauss-Seidel case.
7.1.5.2 The EViews options
Testing convergence under EViews is not very flexible: the only option allowed is the level of the (relative)
convergence criterion, and it will apply to all variables.
One can also decide on the maximum number of iterations. For most models, after 1000 iterations, convergence
becomes rather improbable. But just to make sure, one can set an initial high number. Observing the actual number
required can allow to improve the figure.
7.1.6 STUDY OF THE CONVERGENCE
We are now going to show how the choice of the algorithm affects the convergence process.
Let us begin by stating the problem, and introducing some definitions.
7.1.6.1 The incidence matrix
The incidence matrix of an n-equation model
f (y, ...) = 0 (n endogenous variables, n equations)
will be defined as the Boolean matrix A (dimension n by n) such that
• Ai,j = 1 if the variable yj appears formally, through its unlagged value, in the equation of rank i.
• Ai,j = 0 otherwise.
224
We will suppose the model to be normalized, therefore put under the form:
𝑦 − 𝑓(𝑦) = 0
where the variable yi will appear naturally to the left of the equation of rank i: the main diagonal of the matrix will be
composed of 1s.
The definition of the incidence matrix, as one can see, depends not only on the model, but also on the ordering of
equations, actually the one in which they are going to be computed.
The formal presence of a variable in an equation does not necessarily mean a numerical influence: it could be affected
by a potentially null coefficient, or intervene only in a branch of an alternative. Often we will not be able to associate
to a model a unique incidence matrix, nor a matrix constant with time, except if one considers the total set of
potential influences (the matrix will be then the Boolean sum of individual Boolean matrices).
One will also notice that defining the incidence matrix does not require the full formulations, or the knowledge of
variable values. We simply need to know the list of variables which appear in each explanation, as well as their
instantaneous or lagged character 82.
Application to our model
To apply this technique to our model, we can rely on the block structure provided by EViews, through access to:
(double-click)>View>Block structure,
which gives in our case:
Number of equations: 20
Number of independent blocks: 3
Number of simultaneous blocks: 1
Number of recursive blocks: 2
Block 1: 3 Recursive Equations
cap(1) prle_t(6) x(19)
Block 2: 14 Simultaneous Equations (1 feedback var)
ic(5) ci(13) led(7) le(8)
td(15) m(17) q(3)
Block 3: 3 Recursive Equations
res_m(16) res_x(18) k(20)
All these elements are consistent with the model graph.
We can use these elements to improve the ordering.
82
Following our methodology, the incidence matrix can be produced before any estimation.
225
First, we can use the above separation to move the three predetermined variables at the beginning, and the three
post determined at the end, which give the following matrix:
cap prle_t x q ur ic ci i led le lt rhi ih co fd td m res_x res_m k
cap 1
prle_t 1
x 1
q 1 1 1 1
ur 1 1 1
ic 1 1
ci 1 1
i 1 1 1
led 1 1 1
le 1 1
lt 1 1
rhi 1 1 1
ih 1 1
co 1 1
fd 1 1 1 1 1
td 1
m 1 1 1 1
res_x 1 1
res_
m 1 1 1
k 1 1
We can see that the model has been divided into three parts:
• A three equation block, with elements which do not depend on the complement, or on subsequent variables
in the same block. The variables in this can then be computed once and for all, in a single iteration, at the
beginning of the solving process. Actually they do not depend on any variable in the same block, but this is
not required.
This property is called recursiveness, and the block is usually named the prologue.
We can see that variables can belong to this block for various reasons:
o prle_t depends only on time. The only reason for introducing its equation is to allow easy modification in
forecasts.
o cap depends on an exogenous and a predetermined variable.
o x should depend on the rest of the equilibrium (through UR) but this link has not been evidenced
statistically, leaving only the instantaneous influence of the exogenous WD.
In practice, however, respecting the convergence threshold will need two iterations, the starting value being different
from the solution found, unless the recursivity is known from the start, and the first solution accepted without
control83.
83
Which is of course the case for EViews.
226
• A three equations block, in which elements do not affect the rest of the model, and do not depend on
subsequent variables in the same block. These variables can be computed after all the others, once and for
all in one pass. Again, they do not depend on any variable in the same block, but this is not necessary. The
only condition is that they do not depend on elements computed later (or that the matrix is lower-
triangular).
In this block we find:
o The residuals for the cointegration equations, which will only be corrected at the next period.
o The end-of-period capital, which obviously cannot affect the equilibrium for the period.
We shall see later another important category: variables with a purely descriptive role, like the government deficit in
GDP points
• The rest of the model is simultaneous, and sometimes called the heart. We can check on the model graph
that for any given couple of variables in the set, there is at least one sequence of direct causal relationships
leading from the first to the second, and vice versa. This means also that exogeneizing any element (at a
value different from the model solution of course) all the other elements will be affected.
We can now try to better interpret the simultaneity in the heart. The first stage is observing the presence of loop
variables.
The incidence matrix allows defining loop variables, as variables that enter in an equation of rank lower than the one
that defines them, or will be used before they are computed. In matrix notations, we shall have:
for variable j, ∃ Ai,j = 1 such as 𝑖 < j
The variables appearing as an explanatory factor in their own equation of definition also will have to be added to this
set. But in practice this case is rather rare.
Let us look at our incidence matrix. Two loop variables are present: FD and M. The reason is that they are used to
compute Q, in an equation which appears at the beginning (of the heart).
Actually X should also be present, but as UR appears only through its lagged value, and WD is exogenous, its exact
value can be computed immediately, which means it is located in the prologue. In a way it is now technically
exogenous (considering only same period relationships).
Of course, a model can contain a sequence of non-recursive blocks. This will happen for instance for two subsequent
non-recursive blocks if elements of the second depend on elements in the first, but not vice-versa. Between the two
blocks, a recursive one can be introduced.
We shall see examples of this situation when we deal with more complex models.
The definition of the set of loop variables presents the following interest: if this set is empty, the model becomes
recursive, which means that the sequential calculation of each of the equations (in the chosen order) gives the
solution of the system. Values obtained at the end of the first iteration will satisfy the global set of equations, none of
these values being questioned by a later modification of an explanatory element. And a second iteration will not
modify the result.
227
This favorable case is rare enough 84. However, one can often identify simultaneous subsets (or « blocks » ) with a
recursive structure relative to each other, such that the p first equations of the model are not influenced by the last n
– p (as we have shown on our example). The process of simulation can be then improved, as it will suffice to solve in
sequence two systems of reduced size, allowing to gain time as the cost of solution grows more than proportionally
with the number of equations. This property is evident for Newton, where the cost of both Jacobian computation and
inversion decrease, less for Broyden and even less for Gauss-Seidel, where the only proof comes from practice.
It is obvious that discovering the above properties and performing the associated reordering are interesting for the
model builder, as they allow to improve the organization of the solution process, and therefore reduce computation
time. This process will also allow to detect logical errors, for example by evidencing the recursive determination of an
element known as belonging to a loop (such as investment in the Keynesian loop). Most software packages, including
EViews, take care of this search and the associated reorganization, but further improvement may be sought in the
solving process by modifying by hand the order of equation computations.
In the light of previous observations, one can look:
• For the best block-recursive structure.
• Inside each block, for the order which permits the fastest convergence, or ensures its highest probability.
The first goal is indisputable, and in fact the easiest to realize from the algorithmic viewpoint. The separation found is
unique, but some orderings of blocks can be equivalent (for example equations using only exogenous or lagged
elements can be placed in any order).
The second is much less obvious and in any case more complex. One will seek generally to minimize the number of
loop variables. The cost of this technique will depend on the ambition: the search for one set of loop variables from
which we cannot eliminate an element (Nepomiaschy and Ravelli) is cheaper than the search for all orderings with the
smallest possible number of elements (Gilli and Rossier). The first type of set will be called minimal, the second
minimum. In fact, minimizing the number of loop variables might not be a good preparation for the use of the Gauss -
Seidel algorithm, as we will see later.
EViews determines automatically the block structure of the model (which is de facto optimal, even if other
organizations exist). As to reordering inside the simultaneous blocks, if it does not apply an optimization algorithm, it
determines the loop variables associated with a given ordering (actually associated to the initial one) and places the
associated equations at the end of the block. The efficiency of this last action is questionable, as it means that in a
given iteration all computations use the previous value of loop variables, delaying somewhat the impact of “new”
information.
For instance, in our model, we can reduce the number of loop variables by transferring the equation for Q to the end
of the heart:
84
And the associated model is probably quite poor.
228
prle_ res_
cap t x ur ic ci i led le lt rhi ih co fd td m q res_x m k
cap 1
prle_
t 1
x 1
ur 1 1 1
ic 1 1
ci 1 1
i 1 1 1
led 1 1 1
le 1 1
lt 1 1
rhi 1 1 1
ih 1 1
co 1 1
fd 1 1 1 1 1
td 1
m 1 1 1 1
q 1 1 1 1
res_x 1 1
res_
m 1 1 1
k 1 1
Now Q is the only loop variable (but a strong one as it appears 6 times in that capacity).
The new ordering is:
Prologue
[1] cap = pk * k(-1)
[2] log(prle_t) = 9.8398+ 0.009382* (t - 2002) + 0.02278* (t - t1) * (t<t1) + 0.01540* (t - t2) * (t<t2)
[3] DLOG(x) = 0.4940*DLOG(wd) - 0.09142*res_x( - 1) + 0.01960+ ec_x
Heart
[4] ur = q / cap
[5] i/k( - 1) = 0.944*I( - 1)/K( - 2) - 0.00256*(ur-ur_star)/ur + 0.0428*.125*q/q( - 8) - 0.00448+ ec_i
[6] led = q / prle_t
[7] DLOG(le) = 0.2954*DLOG(led) + 0.1990*LOG(led( - 1)/le( - 1)) + 0.0004748* + 0.01759**((T = 1968.25) - (T =
1968.50)) + ec_le
229
[8] ic = ct * q
[9] lt = le + lg
[10] rhi = wr * lt + r_rhiq * q
[11] co = rhi * (1 - sr)
[12] ih = r_ih * rhi
[13] ci/q( - 1) = -0.02680*(T = 1968.25) + 0.6128*ci( - 1)/q( - 2) - 0.0021490 + 0.2193*@PCH(q) + 0.1056*@PCH(q( - 1))
+ 0.05492*@PCH(q( - 2)) + 0.03918*@PCH(q( - 3)) + 0.03026*@PCH(q( - 4)) + ec_ci
[14] fd = co + i + gd + ci + ih
[15] td = fd + ic
[16] DLOG(m) = 1.676*DLOG(td) + 0.2133*DLOG(ur) - 0.1057*DLOG(compm) - 0.1028*res_m( - 1) - 0.2757+ EC_M
[17] q + m = fd + x
Epilogue
[18] res_x = p_x(1) * log(x / wd) + 0.9815 * log(ur) + 0.001441 * (@trend(60:1) * (t<=2004) + @elem(@trend(60:1) ,
"2004q4") * (t>2004))
[19] res_m = log(m / (fd + ct * q)) ‘ 1.517 * log(ur) +-0.552 * log(compm) + 0.00579 * (@trend(60:1) * (t<=2004) +
@elem(@trend(60:1) , "2004q4") * (t>2004))
[20] k = k(-1) * (1 - dr) +i
and solving the model will be done in the following way (we shall see later the corresponding statements):
Model: _FRA_1
Date: 03/23/20 Time: 11:57
Sample: 2002Q1 2004Q4
Solve Options:
Dynamic-Deterministic Simulation
Solver: Gauss-Seidel
Max iterations = 5000, Convergence = 1e-08
Scenario: Baseline
Solve begin 11:57:47
2004Q1 Block 1 - 3 eqns Solved (recursive block)
2004Q1 Block 2 - 14 eqns Convergence after 12 iterations
2004Q1 Block 3 - 3 eqns Solved (recursive block)
2004Q2 Block 1 - 3 eqns Solved (recursive block)
2004Q2 Block 2 - 14 eqns Convergence after 12 iterations
2004Q2 Block 3 - 3 eqns Solved (recursive block)
2004Q3 Block 1 - 3 eqns Solved (recursive block)
2004Q3 Block 2 - 14 eqns Convergence after 12 iterations
2004Q3 Block 3 - 3 eqns Solved (recursive block)
2004Q4 Block 1 - 3 eqns Solved (recursive block)
2004Q4 Block 2 - 14 eqns Convergence after 12 iterations
2004Q4 Block 3 - 3 eqns Solved (recursive block)
230
Solve complete 11:57:47
It is now possible to compare convergence properties of the different methods, beginning with the Gauss-Seidel
algorithm.
One can see that, for the Gauss-Seidel algorithm, only loop variables influence (by their starting values) the result of
the current iteration. Let us note yb the vector of loop variables (complement yc), associated to the current ordering.
One could consider replacing in turn each occurrence of a variable already computed by its current expression
(possibly already transformed). This will give a model in which only loop variables appear on the right side:
𝑦 = 𝑔(𝑦𝑏 )
And for the convergence process, we have only to consider in the y b elements of the result:
𝑦𝑏 = 𝑔(𝑦𝑏 )
One has just to solve the new (smaller) model, and y c will be obtained in one pass at the end of the process.
Our model gives:
[4] ur = f4(q)
[5] ic = f5(q)
[6] ci = f6(q)
[7] i = f7(q,f4(q))
[8] led = f8(q)
[9] le = f9(f8(q))
[10] lt=f(f10(f9(f8(q)))
[11] rhi = f11(f(f10(f9(f8(q))),q)
[12] ih = f12(f11(f(f10(f9(f8(q))),q))
[13] co = f13(f12(f11(f(f10(f9(f8(q))),q)))
231
[14] fd = f14(f13(f12(f11(f(f10(f9(f8(q))),q))),f7(q,f4(q)), f6(q), f12(f11(f(f10(f9(f8(q))),q)))
[15] td = f(f14(f13(f12(f11(f(f10(f9(f8(q))),q))),f7(q,f4(q)), f6(q), f12(f11(f(f10(f9(f8(q))),q))),f5(q))
[16] m = f15(f14(f13(f12(f11(f(f10(f9(f8(q))),q))),f7(q,f4(q)), f6(q), f12(f11(f(f10(f9(f8(q))),q))), f5(q), f4(q))
Even for this simple example, producing the total set of explicit relations would have been difficult if not impossible.
But if g is formally unknown, its values can be computed for any y𝑏 (in a typical Gauss- Seidel iteration).
Thus, in our example, the successive application (in the given order) of the formulas in the f system associates to an
initial value of Q a numerical value of the associated g (but not its solution of course). No other initial value has to be
considered.
Again, this shows that for the convergence, one has only to consider the reduced model:
y𝑏 = g(𝑦𝑏 ) or y𝑏k+1 = g(𝑦𝑏𝑘 )
and that the control of convergence can be limited to these variables.
Let us use this observation to interpret the convergence of the algorithm. We shall linearize the process around the
solution:
y 𝑘 - y* = g(𝑦 k-1 ) - g(y*) ≈ (𝜕g/ ∂𝑦)y=y k-1 (𝑦 k-1 - y*)
This approximation, if it can be used in practice (which supposes either that one is near the solution, or that the model
is linear enough), shows that the vector of errors is multiplied at each iteration by the Jacobian. If the Jacobian is
stable enough in the process, this makes convergence almost linear, and the cost of obtaining an additional solution
digit will be independent from its position in the representation of the variable. And in order for the algorithm to
converge effectively, the generally necessary and always sufficient condition is that when one raises the Jacobian
matrix to a power growing to the infinite, its value goes to zero. An equivalent condition is that the every eigenvalue
of the matrix is strictly lower than 1 in module (one will say that the spectral radius is lower than unity). Thus, in our
example, convergence will be assured if:
∆g(q)/q <1
232
One sees that minimizing the number of loop variables, if it reduces the size of the Jacobian, has no specific reason to
reduce the spectral radius85.
Actually it frequently happens that the reordering does not modify in any way the convergence process, as it affects
only the point at which it enters in an otherwise unchanged loop. This is the case in our example: by transferring the Q
equation to the end, we start with the value for Q in the workfile and not the one which balances the supply-demand
equilibrium using the initial values for X, M and FD. But from then on the process is exactly the same. Even more: if the
data values meet the equilibrium (they do most of the time), computing Q will not change its value, and the process
will be identical even in its figures.
One can even (with no more proof) have the reverse intuition: by concentrating the process on a limited number of
variables, one might increase their individual role.
In addition, using an automatic algorithm transfers reorganization control from a logical economist (who has probably
chosen a logical choice in terms of causality) to a mathematical blind tool.
Obviously, if the number of loop variables is high, the probability of convergence would be very low if coefficients
were chosen randomly. In practice, fortunately, one can associate to the mathematical convergence process a more
economic one, the iterative progress to a balanced equilibrium by combining the behaviors of agents (examples can
be found in the Keynesian loop, or in the WS-PS wages-prices loop). The probability of success is much improved, and
most models converge if they use reasonable assumptions and a reasonable starting point 86.
Let us illustrate this point by simplifying to the extreme our usual model, taking only into account GDP (Q), private
demand (C) and government demand (g). Our model does not consider external trade.
(1) C = a Q
(2) Q = C + g
and
0<a<1
With the Gauss-Seidel algorithm, this model will always converge to the solution Q*, as for the unique loop variable
(Q):
𝑄 𝑘 − 𝑄 ∗ = 𝑎 ⋅ (𝑄 𝑘−1 − 𝑄 ∗ )
85
In simpler words: the highest modulus found in the set of eigenvalues.
86
Not to converge for improbable assumption values should be considered as a good point.
233
and |a| < 187
Let us now invert causalities:
(1) C = Q - g
(2) Q = C / a
This time the model diverges:
𝑄 𝑘 − 𝑄 ∗ = 1/𝑎 ⋅ (𝑄 𝑘−1 − 𝑄 ∗ )
and
1
𝑎𝑏𝑠 ( ) < 1
𝑎
but this is normal, as to a convergent economic framework (the Keynesian multiplier) we have substituted a
framework where, to simplify:
• Households consume what remains of supply once exogenous demand is satisfied.
• Firms seek to maintain supply in a constant ratio (higher than 1) with consumption.
If such a framework does not converge mathematically, it the same logically: although there exists obviously a solution
(the same as higher), an initial error on Q will apply identically to C, then will amplify on Q.... And the same instability
will be found if we shock public demand (computing the public expenditure multiplier).
One can observe also that with a >1, the first model would diverge (as secondary effects on production would be
larger than their source). And the second would converge (as a change in consumption is now reduced on production):
the two errors compensate each other.
Formalizing the problem, with the first framework, an initial change in government demand will lead at iteration k, to
the cumulated variation:
87
With a>=1 the model would diverge, which could be expected from an economic point of view.
234
𝛥𝑄 𝑘 = (1 + 𝑎 + 𝑎2 =. . . . +⥂ 𝑎𝑘 ) ⋅ 𝛥𝑔 = (1 − 𝑎𝑘−1 )/(1 − 𝑎) ⋅ 𝛥𝑔
which converges to
ΔQk = 1/(1 − a) ⋅ Δg
if a < 1
and with the second
Δ𝑄 𝑘 = (1 + 1/𝑎 + (1/𝑎)2 +. . . . +(1/𝑎𝑘 )) ⋅ 𝛥𝑔 = (1 − (1/𝑎)𝑘−1 )/(1 − (1/𝑎))𝛥𝑔
which converges only if a > 1
The convergence of one model is equivalent to the divergence of the other.
7.1.6.2 Accelerating Gauss-Seidel: damping factors
An efficient technique to accelerate convergence, or to force naturally diverging models to reach a solution, is to
introduce one or more «damping factors». This technique is based on the following elementary observation, inspired
by the above.
If y is solution of
𝑦 = 𝑓(𝑦)
then it will be also the solution of
𝑦 = 𝐷 ⋅ 𝑓(𝑦) + (I-D) ⋅ 𝑦
235
where y represents the vector of endogenous and D an identity matrix.
It goes similarly for g:
if yb is a solution of
𝑦𝑏 = 𝑔𝑏 (𝑦𝑏 )
then it will be also of:
𝑦𝑏 = 𝐷 ⋅ 𝑔𝑏 (𝑦𝑏 ) + (I-D) ⋅ 𝑦𝑏
But the new formulation modifies the Gauss - Seidel solving process, by introducing inertia on starting values y b.
In practice that will mean initializing the iteration k, not by:
y𝑏k-1
but by
𝐷𝑏 ⋅ 𝑦𝑏𝑘−1 + (I-D𝑏 ) ⋅ 𝑦𝑏𝑘−2
The change of starting values plays a role only if variables are used before they are computed during the iteration.
Therefore damping factors will be applied to loop variables only.
One shows easily that by applying damping factors, we are moving towards a Newton type method, at least in simple
cases. Indeed, in a model with a single loop variable, the introduction of damping factors means using:
ykb = D gb (yk-1 k-1
b ) + (I - D) yb
236
or
ykb - yk-1
b = D (gb (yk-1 k-1
b ) - yb )
This is equivalent to the Newton method if one chooses
D = (I - g b / y b ) -1 y b = y kb -1
or the inverse of the Jacobian of:
𝑦𝑏 − 𝑔𝑏 (𝑦𝑏 ) = 0
This method, more or less efficient following the stability of the Jacobian, will give the solution in a single iteration in
the case of a linear model.
Let us complete our presentation by a graphic example.
We will use the simple formal model
(1) y = c x + d
(2) x = - a y + b
237
“damped” equation
x = l (-a y + b) + (1- l) x
y damping
process y=cx+d
and now it
converges
diverges initial equation
x = -a y + b
x
x0
The basic process clearly diverges, but if we apply a convenient damping factor (we have chosen a value close to 0.5)
we can make it converge. What we have done is simply replacing one of the equations (y = - ax + b) by another which
associates to a given value of y a combination, with constant weights, of the values of x associated to the same y by
the two equations. The solution is not changed, but the convergence process is modified.
One could have reached the exact value, by making the new “curve” vertical
x = l (-a y + b) + (1- l) x
l=1/(1−ac)
y
y=cx+d
initial equation
x = -a y + b
x
x0
Applying damping factors is not however so simple.
238
• It can be done only through trial and error, excepts if one solves the problem described above, which is
actually more complex than solving the model itself by the Newton method88.
• It is only truly efficient for values between 0 and 1 (alternate convergence). If the two curves have slopes with
the same orientation, making convergence monotonous, the method becomes more difficult to handle, as an
efficient damping factor must now be chosen outside the interval (although it can work, as the following
diagram shows).
y diverges
y= - a x + b x= c y + d
y= df (-a x + b) + (1-df) y
converges
x* x
•
And when the size of the model grows, the searching process becomes rapidly complex.
88
One can also concentrate on one loop variable, considered as the most important, and measure its changes for
three consecutive iterations. A very rough approximation of the convergence factor can be obtained by:
𝑎 = (𝑥 𝑘 − 𝑥 𝑘−1 )/(𝑥 𝑘−1 − 𝑥 𝑘−2 )
and one can use as a damping factor the value:
𝜆 = 1/(1 − 𝑎)
239
7.1.6.2.1 More complex models
If the equivalence to Newton cannot be extended to the case of several loop variables, it is only because the method
restrains D to a diagonal matrix. One could however consider:
A Looking for the diagonal matrix "closest " to the Jacobian, according to a given measure.
B Decomposing the vector y on the eigenvector basis of the Jacobian, which would allow to diagonalize the process.
y 𝑘 - y k-1 ≈ (𝜕g /𝜕𝑦)y=yk-1 (𝑦 k-1 - y k-2 )=V D V -1 (𝑦 k-1 - yk-2 )
V -1 (y k - y k-1 ) DV -1 (y k-1 - y k-2 )
with D diagonal matrix of the eigenvalues of the Jacobian.
A relative stability of the Jacobian (therefore a good linearity, or an initialization value close to the solution) is then
necessary89.
• Let us see now how the above considerations influence the convergence of the Newton algorithm.
The usefulness of loop variables, and the minimization of their number, is now indisputable, if one decides to compute
the Jacobian by finite differences. Indeed one just has to know starting values for yb to compute g(yb) and the
Jacobian 𝜕𝑔𝑏 /𝜕𝑦𝑏 (this last by Gauss-Seidel iterations),
• Only the part 𝜕g/ ∂𝑦𝑏 is not identically null.
• And only the part 𝜕𝑔𝑏 /𝜕𝑦𝑏 is going to affect the calculation of the new value of yb.
We just have to apply the Newton formula to the loop variables alone. Once again, only the values of loop variables of
the previous iteration (and their influence on the Jacobian of g) will play a role in the iterative process.
This has obviously the advantage of limiting computations, for the Jacobian itself (calculation of the only non-zero
columns) as well as for its inversion.
The usefulness of taking into account loop variables looks clear. The reality is less so:
89
Experience shows that the Jacobian of the linear differentiation is much less stable than the relative one:
D =(I - ∂𝑔𝑏 /𝜕𝑦𝑏 )-1 𝑦 k-1
𝑏 =y𝑏
240
• The Jacobian matrix associated to the initial formulation f is very sparse (the average number of same period
explanatory variables per equation does not grow too fast with the size of the model, and stays generally in
the order of 3 or 4). This means the Newton method applied to f can use an inversion algorithm specific to
sparse matrices, appreciably more effective. The formulation in g concentrates the process of determination
on some variables, but the implicit formulation complexity produces a very dense Jacobian matrix, which can
be inverted only by a general algorithm.
• As to the calculation of the matrix itself, it will be obviously more expensive for f than for g, if one uses finite
differences. But the Jacobian matrix associated to the f vector can also be determined analytically: then one
will just have to compute independently each non zero element of the Jacobian, according to a formula
generally simpler than the initial equations90. If the average number of instantaneous variables per equation
is lower than the number of loop variables, the global cost will decrease. This reasoning is however only valid
if one makes abstraction of the (important) cost of the initial formal derivation91, which supposes also a
certain stability of the formulation of the model.
As for the comparison with Gauss-Seidel, one can expect a lower number of iterations, but each of them will be clearly
much more expensive.
Let us see how the Newton process converges.
For this let us develop to the second order the derivation formula:
y k+1 -g(𝑦 k+1 ) - (𝑦 𝑘 -g(𝑦 𝑘 )) = (I-𝜕g / 𝜕𝑦)y=y𝑘 (𝑦 k+1 - y 𝑘 ) - 𝑇 (𝑦 k+1 - y 𝑘 )(𝜕 2 g / 𝜕𝑦 2 )y=y 𝑘 (𝑦 k+1 - y 𝑘 )
plus a third order term.
Using the linearization to determine yk+1 eliminates the central terms, leaving:
yk+1 - g(y k+1 ) = - T
( y k+1 - yk ) (2g / y2 ) y=yk ( y k+1 - yk ) + e3 ( y k+1 - yk )
and as y k+1 -y 𝑘 is a linear function of y 𝑘 -g(𝑦 𝑘 ) (again by construction), the error y 𝑘+1 -g(𝑦 𝑘+1 ) is a quadratic
function of y 𝑘 -g(𝑦 𝑘 ).
Similarly, the distance of y to the solution y* is a quadratic function of the distance at the previous iteration, as the
derivation to the first order has given:
90
For instance, when derivation is applied to a sum.
91
EViews is able to compute the analytical derivatives. However, it does not provide the formulas.
241
(𝑦 k+1 -g(𝑦 k+1 )) - (y*-g(y*)) [=0] =(I-𝜕g / 𝜕𝑦)y=y𝑘 (𝑦 k+1 - y ∗ )
The convergence of the algorithm is not therefore uniform: in fact it is going to accelerate with the number of
iterations, as the linearized form gets nearer to the derivative at the solution. One will say that it is quadratic in the
case of a formal derivation, but only supralinear if one uses finite differences, because one does not compute the
exact derivative. The Newton algorithm should be the more advisable as the precision requested from the solution
gets higher.
7.1.6.3 Comparison of the relative efficiencies
Let us consider first the ability to find the solution. The convergence of the Gauss-Seidel algorithm appears quite
doubtful, if it based on several iterative interacting processes (the number of loop variables) all of which must be
converging. In other terms, a matrix must have all its eigenvalues lower than unity in modulus. It is however more
frequent than one can expect. As we have seen earlier, the numerical process is generally linked to a set of economic
causalities, determining a normally convergent process. Actually, if this process does not converge, the model itself
should probably be tested for economic consistency.
The Newton algorithm on the contrary puts no condition on the numerical intensity of relationships, but on its
variability (in practice the stability of the Jacobian and in a lesser measure the Hessian). This condition is generally
verified in the case of econometric models. In particular, the convergence will be immediate for a linear model
showing a unique solution. The Newton algorithm will then be advisable for models presenting convergence
difficulties, or for model builders wanting to free themselves from this concern, which remains always present and
sometimes crucial (for instance if they frequently face deadline problems).
On the other hand, considering the speed of convergence, the choice is much less obvious. If the Newton method
converges in less iterations, generally between three and five (remember that convergence accelerates with
iterations), each iteration is far more expensive. In fact this supplement can be measured, in the case of a derivation
using finite differences: each iteration takes as many Gauss-Seidel iterations as the model contains loop variables (plus
one), as well as a matrix inversion (growing with the dimension as a third degree polynomial). One can therefore
minor the total cost, whatever the method, by a number of Gauss-Seidel iterations.
In practice one observes that the number of iterations necessary for convergence, whatever the method employed,
grows slowly enough with the size of the model. On the other hand the number of loop variables is more or less
proportional to it.
We shall give now the time taken by each of the methods (in case of success):
• Gauss – Seidel
• Newton with analytical derivations
• Newton with numerical approximation
• Broyden with analytical derivations
• Broyden with numerical approximation.
The test was conducted on a Toshiba Qosmio F60-10J, using an Intel I5 chip and 4 GB memory.
For 10 models of growing size, all built or contributed by the author during cooperation projects and seminars:
• France : a very small French model, resembling the one we are using here
• France cf : a small French model with complementary factors
242
• France cd : a small French model with Cobb-Douglas. Both these models will be presented later.
• Vietnam : a small Vietnamese model with one product
• Algeria 1 : a small Algerian model with one product (+oil in some cases)
• Algeria 5 : a five product Algerian model
• Algeria QAM : a 19 product Quasi-Accounting Algerian model
• Vietnam : a 3 product Vietnamese model
• World : a 12 country world model
• China : a 3 product, 4 region model
The first three models are available in the files accompanying the book.
With the following characteristics:
• Prod : number of products
• Coun : number of products or regions
• Per : number of periods
• Equ : number of equations
• Block size : size of the largest non – recursive block
• Feedback : number of feedback (or loop) variables
Countries prod coun per equ Block size
Feedback
Gauss-Seidel
Newton A Newton N Broyden A Broyden N
France 1 1 q 18 13 2 0,002 failed failed 0,003 0,004
France cf 1 1 q 87 47 2 0,010 0,018 0,030 0,014 0,014
France cd 1 1 q 88 51 3 0,020 0,025 failed 0,017 0,017
Vietnam 1 1 a 94 55 4 0,030 0,026 0,045 0,019 0,020
Algeria 1 1 a 149 85 7 0,028 0,026 0,036 0,023 0,029
Algeria 5 1 a 630 408 23 failed 0,169 0,237 0,145 0,146
Algeria QAM 19 1 a 1361 922 19 failed 0,312 0,433 0,261 0,234
World 1 12 q 1396 1063 71 0,087 0,084 0,082 0,083 0,083
Vietnam 3 3 a 1858 1260 37 2,446 2,008 2,522 failed failed
China 3 4 a 2648 1900 45 failed failed failed 0,349 0,321
One can observe:
• The bad news is that each method fails at least once, the good that at least one works for any model. If this
was true in all cases, convergence would be guaranteed.
• Gauss-Seidel works better for small models. This is quite contrary to expectations, as the cost of inverting a
matrix of growing size should become more and more expensive as size grows. There might be a cost of entry
for Newton.
• Broyden is globally better for large models, for no particular reason. However, it does fail once. But it is the
only one working for the Chinese model.
• For the large Vietnamese model, Broyden fails, and Newton is better than Gauss-Seidel.
Let us now consider the link between model size and computation cost.
We shall present four graphs, all of them considering the logarithm of both variables:
• A global one for all methods
• Three individual ones dedicated to each method.
243
They show that the link is globally log-linear, with a slight downward inflexion for the “inversion” methods. As to the
elasticity itself, a separate estimation gives a value close to 1, and a global one a value of .99 (with a T-stat of 14.7).
2.000
The speed according to model size
miilli seconds per period and replication
Broyden
Gauss-Seidel
0.200
Newton
0.020
0.002
10 20 30 50 70 100 200 400 600 1,000 2,000
Number of core equations
miilli seconds per period and replication
2.000 The speed using Gauss-Seidel
0.200
0.020
0.002
10 20 30 50 70 100 200 400 600 1,000 2,000
Number of core equations
244
3.00
2.00
Milliseconds per period and replication
The speed using Newton
1.00
0.70
0.50
0.30
0.20
0.10
0.07
0.05
0.03
0.02
0.01
10 20 30 50 70 100 200 400 600 1,000 2,000
Number of core equations
.400
milliseconds per period and replication
.200 The speed using Broyden
.140
.100
.060
.040
.020
.014
.010
.006
.004
.002
10 20 30 50 70 100 200 400 600 1,000 2,000
Number of core equations
• The two last graphs compare analytical and numerical methods, for both related algorithms.
• The diagnosis is complemented by an histogram presenting the ratio of costs (when both options led to
convergence).
245
They show clearly :
o That establishing analytical derivatives is quite efficient for the Newton method, with an average gain of
about 40%.
o That for Broyden (where the derivatives are only computed once) there is no gain, but no loss either.
Countries Newton Newton Newton Broyden Broyden Broyden
analytic numerical ratio analytic numerical ratio
France failed failed failed 0,00 0,00 1,31
France cf 0,02 0,03 1,62 0,01 0,01 1,05
France cd 0,02 failed failed 0,02 0,02 0,96
Vietnam 0,03 0,05 1,78 0,02 0,02 1,06
Algeria 0,03 0,04 1,40 0,02 0,03 1,26
Algeria 0,17 0,24 1,40 0,14 0,15 1,01
Algeria QAM 0,31 0,43 1,39 0,26 0,23 0,89
World 0,08 0,08 0,97 0,08 0,08 0,99
Vietnam 2,01 2,52 1,26 failed failed failed
China failed failed failed 0,35 0,32 0,92
246
7.1.7 SOLVING THE MODEL: BASIC EVIEWS TECHNIQUES
The previous program has made model data and equations consistent. The present one will solve it.
7.1.7.1 The syntax
Again, we shall SOLVE the model, but this time the options are more numerous.
Solving the model will generally be done in a program. However, one can use the menus, first by accessing the model
item, then using the “solve” item, which gives access to all the features we are going to present.
In a program, these options can be set before any computation, using the SOLVEOPT statement, or at simulation time
(SOLVE).
The syntax is the following:
model-name.solveopt(options)
Or
solve(options) model-name
Where the options are (taken from the EViews Help file):
m= (default=5000) Maximum number of iterations for solution (maximum 100,000).
c= (default =1e-8) Convergence criterion. Based upon the maximum change in any of the endogenous
variables in the model. You may set a number between 1e-15 and 0.01.
s=arg Solution type: "d" (deterministic, default), "m" (stochastic - collect means only), "s"
(stochastic - collect means and s.d.), "b" (stochastic - collect means and confidence bounds), "a" (stochastic - collect
all; means, s.d. and confidence bounds).
d=arg Model solution dynamics: "d" (dynamic solution, default), "s" (static solution), "f"
(fitted values - single equation solution). Forecasting calls for a dynamic solution, which is the default option. There
are few reasons to choose another (except d=f as above).
n=arg NA behavior: "n" (stop on "NA" values), "t" (do not stop when encountering "NA"
values). Only applies to deterministic solution; EViews will always stop on "NA" values in stochastic solution.
a=arg Alternate scenario solution: "t" (solve both active and alternate scenario and collect
deviations for stochastic), "f" (default, solve only the active scenario).
o=arg Solution method: "g" (Gauss-Seidel), "n" (Newton), "b" (Broyden).
247
i=arg For the initialization values: “a” (actual), “p” (previous period solution). The default
is “a” if the values are available, “p” otherwise. In forecasts, “p” should be applied, but one can initialize the
endogenous which gives the two options (with “a” as the default).
For stochastic simulations:
r=integer (default=1000) Number of stochastic repetitions (used with stochastic "s=" options).
b=number (default=.95) Size of stochastic confidence intervals (used with stochastic "s=" options).
7.1.7.1.1 The suffix
In addition to the above options, EViews allows the user to define the name of solution variables.
Of course, one should not consider using the actual names as such. This will destroy the information on the original
values. However, the connection must be as easy as possible. The obvious solution is to use the original names, but
modify them by some addition: a given prefix or suffix.
This will call for a specific statement:
model_name.assign @all suffix
This statement will add to the text of the model model_name the statement:
assign @all suffix
where suffix is a string.
For instance,
_fra_1.assign @all _b
will give to the variables obtained by the next model solution a name composed of the original variable followed by
the suffix “_b”. In this case, the solution for Q will be called Q_B.
7.1.7.2 Problem processing techniques
Like all packages of this type, EViews does not always reach a solution. We will now provide a typology of techniques
one can use for solving convergence problems, focusing on the ones available under EViews, and the way to apply
them.
248
In all cases one must be certain that the residual check shows no error in the first place. This is essential, and very
cheap (at least in the observation of errors, and the identification of their location).
7.1.7.2.1 The tools provided by EViews
In addition to changing model specifications, EViews provides a number of specific tools. Some of these can only be
specified through menus, but they remain active if one runs a program in the same session.
• Displaying the number of iterations necessary for convergence (only useful of course if the model converges on part
of the sample). But if the model starts diverging after a few periods, it is interesting to observe if the process is
gradual (the number of iterations is growing until breakdown) or sudden (the number is stable until the crash).
This is obtained by
o double-clicking the model item
o using:
solve>diagnostics
The following window opens:
We tick the box for “displaying detailed messages including iteration count by block”.
When we solve the model, a window will open displaying the number of iterations needed for solving each block
(several messages per period if the model contains more than one non-recursive block).
This window will remain open after solving, and one can scroll back and forth, to the first message if needed.
Model: _FRA_1
Date: 03/23/20 Time: 11:57
Sample: 2002Q1 2004Q4
Solve Options:
249
Dynamic-Deterministic Simulation
Solver: Gauss-Seidel
Max iterations = 5000, Convergence = 1e-08
Scenario: Baseline
Solve begin 11:57:47
2004Q1 Block 1 - 3 eqns Solved (recursive block)
2004Q1 Block 2 - 14 eqns Convergence after 12 iterations
2004Q1 Block 3 - 3 eqns Solved (recursive block)
2004Q2 Block 1 - 3 eqns Solved (recursive block)
2004Q2 Block 2 - 14 eqns Convergence after 12 iterations
2004Q2 Block 3 - 3 eqns Solved (recursive block)
2004Q3 Block 1 - 3 eqns Solved (recursive block)
2004Q3 Block 2 - 14 eqns Convergence after 12 iterations
2004Q3 Block 3 - 3 eqns Solved (recursive block)
2004Q4 Block 1 - 3 eqns Solved (recursive block)
2004Q4 Block 2 - 14 eqns Convergence after 12 iterations
2004Q4 Block 3 - 3 eqns Solved (recursive block)
Solve complete 11:57:47
• Displaying the values obtained at each iteration for a set of variables. Of course, following the convergence process
is only practical using the Gauss-Seidel method.
This is again obtained by
o double-clicking the model item
o using:
solve>diagnostics
o Introducing the names of the variables in the window.
250
Alternatively, the same can be obtained in a program, using:
model_name.trace list_of_variables.
This can be useful if you select a given set of important elements, which will then be available at all times after any
simulation.
o After solving, using:
View>trace output
A window will open with the values of the variables at each iteration, for each period (one column of values per
variable).
251
Date Block Iteration Objective CAP Q K
2004Q1 1 1 4.19E+11
2 0 3.54E+11
1 0.022360 3.54E+11
2 0.020004 3.54E+11
3 0.041031 3.54E+11
4 0.004830 3.54E+11
5 0.000891 3.54E+11
6 0.000232 3.54E+11
7 3.16E-05 3.54E+11
8 3.90E-06 3.54E+11
9 4.79E-07 3.54E+11
10 5.87E-08 3.54E+11
11 7.20E-09 3.54E+11
3 1 2.34E+12
2004Q2 1 1 4.18E+11
2 0 3.56E+11
1 0.020415 3.56E+11
2 0.022601 3.55E+11
3 0.097650 3.55E+11
4 0.010504 3.55E+11
5 0.001899 3.55E+11
6 0.000479 3.55E+11
7 6.31E-05 3.55E+11
8 7.52E-06 3.55E+11
9 8.90E-07 3.55E+11
10 1.05E-07 3.55E+11
11 1.25E-08 3.55E+11
12 1.48E-09 3.55E+11
3 1 2.35E+12
2004Q3 1 1 4.20E+11
2 0 3.58E+11
1 0.020342 3.58E+11
2 0.025339 3.57E+11
3 0.137144 3.57E+11
4 0.013647 3.57E+11
5 0.002419 3.57E+11
6 0.000588 3.57E+11
7 7.47E-05 3.57E+11
8 8.53E-06 3.57E+11
9 9.69E-07 3.57E+11
10 1.10E-07 3.57E+11
11 1.25E-08 3.57E+11
12 1.42E-09 3.57E+11
3 1 2.37E+12
2004Q4 1 1 4.24E+11
2 0 3.60E+11
1 0.025946 3.60E+11
2 0.028302 3.59E+11
3 0.153634 3.59E+11
4 0.014515 3.59E+11
5 0.002533 3.59E+11
6 0.000595 3.59E+11
7 7.31E-05 3.59E+11
8 8.05E-06 3.59E+11
9 8.82E-07 3.59E+11
10 9.65E-08 3.59E+11
11 1.06E-08 3.59E+11
12 1.16E-09 3.59E+11
3 1 2.38E+12
If the variable belongs to a recursive block, only one value will be displayed. In our case, the unique values for CAP and
X will preceded the iterative process description, then the unique K value.
The first value (iteration 0) is the initial one.
The columns follow the order used in the solving process, not the one in which the list is specified. This is quite useful
for following the process.
252
But the interpretation of the evolutions as a column of figures is often difficult, in particular if one tries to evaluate the
presence and nature of cycles. It is better to transform these figures into graphs. This can be done in two ways:
• Selecting with the mouse the requested zone (for instance the values for a given period) and copying it into a
graphic software such as Excel.
• The same can be done inside EViews but it is a little more complex. One must:
o Select with the mouse the requested zone
o Create variables associated with the displayed elements (with different names, for instance by inserting a
common prefix).
o Display the variables as a table filled with “na” values.
o Copy the selection into the sheet.
o Apply View>Graph.
• Excluding equations from the solving process.
This is very easy. Using menus, one has to
o access the model,
o use:
Basic options>Edit scenario options
In the “exclude series” block, specify to series to exogenize.
253
Using a program, one needs the statement:
model-name.EXCLUDE list-of-series
The corresponding equations will not be applied, and the variables will retain their initial value.
Note: the list will remain active for all subsequent solutions produced under the same scenario, until it is modified
using a different list. Reverting to the full model calls for a statement with a blank list, which should be stated
immediately.
One can consider eliminating a single variable, like the change in inventories (if it seems to introduce an explosive
process), or a full set, like all the deflators in the model (if the price system seems to explode, or to disturb the
computation of the real sector).
However, one must be aware that if an identity variable is excluded, the identity is no longer enforced. Balancing
variables should never be excluded. For instance, excluding Q makes supply no longer equal to demand (and the
model recursive…).
7.1.7.2.2 Other tools
Some tools can be considered by the user:
7.1.7.2.2.1 USING DAMPING FACTORS
This has been described earlier. Technically, you will have to rewrite the equation. For instance
(4) Q + M = FD + X
254
will become:
(4) Q = rel_q * (FD + X - M) +(1-rel_q) * Q
where damping is enforced by making rel_q different from 1 (lower in principle).
It might be useful to set the factor as a series, even if it will be constant over time. To apply a change to a scalar, the
model needs to be recompiled (by UPDATE).
Also, this method is a little different from the one described earlier. Application of the factors is not done at the end of
the iteration, but at the time the variable is computed. If the variable is used later in the ordering, this affects the
convergence process.
7.1.7.2.2.2 REORDERING THE MODEL
Using Gauss-Seidel, reordering will change the solving process. Not for Newton as the Jacobian is based on derivation
of individual equations.
7.1.7.2.2.3 CHANGING PARAMETER VALUES
Changing the values of parameters will obviously modify the solving process. In particular, a coefficient which looks
too high can be reduced or even eliminated. This could be the case for instance if the intensity for the short-term
accelerator effect in the investment equation.
This technique can be used to eliminate a specific influence of a variable, leaving unchanged its other roles.
This calls for a good knowledge of the properties of the model.
7.1.7.2.2.4 CHANGING EQUATION SPECIFICATIONS
A mechanism which seems to create problems can be replaced by another. For instance, the purchasing power parity
assumption for the exchange rate can be smoothed or replaced by a fixed nominal rate.
7.1.7.2.3 Applying these tools
Let us now propose a methodology in case the model fails in reaching a solution.
In all cases, you should move to menus. This environment will give much more information and provide many more
tools.
7.1.7.2.3.1 THREE SIMPLE CASES
First, let us consider three specific cases, requiring simple treatment.
255
• Some data the process requires is not available.
It should be filled or you should change the period.
One frequent case: you forgot to establish an assumption (compute an exogenous variable) on the future.
If you created a group for the exogenous, you just have to display its values. If there is apparently more than one
series missing, the best strategy is to access the group (perhaps copying the series to a specific page) and use “Edit
Columns” to call for the starting, ending and number of observations of the series.
256
You can see immediately that variables EC_M and EC_X are missing.
If there are too many variables to trust eye control, you can
o Freeze the table.
o Sort it according to the column (D in our case).
The shorter series will appear at the top.
• You made technical mistakes.
o Your model is not consistent with the data. Of course you have performed a residual check, but maybe
something has changed since. One should perform it again.
o The exogenous assumptions have been given abnormal values.
For instance with WD=0 we get:
Solve terminated - Log of non positive number in "DLOG(X) = @COEF(1) * DLOG(WD) + @COEF(2) + @COEF(3) *
RES_X(-1) + EC_X"
o The initial values are abnormal, and the model crashes before it reaches an acceptable region.
To solve the problem, it will be necessary in this case to retrace the causal chaining
For instance if WR is 1000 too high due to wrong units, we get:
Model: _FRA_1
Date: 03/23/20 Time: 15:30
Sample: 2000Q1 2002Q4
Solve Options:
Dynamic-Deterministic Simulation
Solver: Gauss-Seidel
Max iterations = 5000, Convergence = 1e-08
257
Scenario: Baseline
Solve begin 15:30:59
2000Q1 Block 1 - 3 eqns Solved (recursive block)
2000Q1 NA generated for LE - Log of non positive number
2000Q1 Block 2 - 14 eqns No convergence after 4 iterations
Solve terminated - Log of non positive number
And we can follow the evolution of variables in the “Trace output” display:
Date Block Iteration Objective X RHI CO TD M Q
2000Q1 1 1 9.29E+10
2 0 2.01E+11 1.83E+11 6.71E+11 8.78E+10 3.35E+11
1 910.3030 4.33E+13 1.00E+13 1.34E+12 1.23E+11 9.92E+11
2 2929.029 2.22E+14 2.02E+14 2.21E+14 3.59E+14 -1.39E+14
3 330.6255 2.11E+14 1.93E+14 7.34E+13 3.59E+14 -1.73E+14
4 1.452647 2.07E+14 1.89E+14 -1.13E+13 3.59E+14 -1.98E+14
We can observe the following sequence, starting from historical values (Iteration 0):
At iteration 1, household revenue gets much too high, consumption follows, then total demand, imports and GDP.
At iteration 2, imports becomes higher than demand, and GDP becomes negative at the end of the iteration.
At iteration 3, the employment equation uses this negative value in a logarithm, and the process breaks down; as the
first message indicated.
• The maximum number of iterations has been reached.
You should specify a higher number (say 1000, or even 100 000) until convergence. You do not want to miss a late
convergence, which corresponds to a solution. Time is no longer an issue, except If you are producing stochastic or
forward looking estimations, but you can check an other solving algorithm, or a different ordering for .
The absence of convergence with a high number is exceptional. It can only come from oscillations between two very
close values of a single variable (purely numerical problem due to the precision of computer computations) in which
case the chosen threshold is too small.
For instance, a variable which can change signs, which means that it is computed as a difference (like the change in
inventories or the trade balance) takes values much lower than its components, and loses precision digits in the
process. The same is true also for growth rates, when the associated element is stable over time.
In practice, this should happen only with Gauss-Seidel. Again, try another algorithm and consider the result. Either it is
acceptable, or the model crashes. In that case, the next case applies.
7.1.7.2.3.2 THE SOLUTION CRASHES
Let us now consider the main case : the algorithm crashes.
In practice, EVIews will stop computations in only two cases, in addition to reaching the maximum number of
iterations.
258
• The values of a variable become too high.
This will happen essentially with Gauss-Seidel: the spider web process we have presented earlier leads away from the
(existing) solution.
However, it can also happen with Newton (and Broyden less clearly), and much faster. These algorithms produce a
sequence of linearizations and solutions of the linearized model, supposedly more and more precise until it
corresponds to the solution of the model itself. If this linearization does not gain precision it can only lose it, and quite
fast (except in the very unlikely case of oscillations) due to the quadratic nature of the algorithm, which applies both
to convergence and divergence. The associated solution will fast enter a domain with unacceptable values, most likely
as to their sign (negative GDP or unemployment?) which will enter a logarithm when the Jacobian is computed. The
algorithm will crash.
But of course reaching unacceptably high values is also possible, for instance when the Jacobian becomes almost
singular.
Now, the highest value EViews will consider is at this time is more than 10308. Obviously no economic variable can
take this value (even for a country with explosive inflation).
Let us consider the following model (using the names from the small model):
Q + M = FD + X
CO = 0.8 * Q
I = 0.5 * d(Q)
FD = CO + I + g
M = 0.2 * FD
With X and g exogenous.
This model does not seem so crazy. But one can observe (as in a previous version of the example) that increasing g by
1 has a secondary impact of 1.1 (0.8 + 0.5 – 0.2). Using Gauss-Seidel the process will explode. This is what will happen
in practice.
We will ask for the display of iteration numbers, and the sequence of values for the variable q, using:
259
And we will apply the Gauss-Seidel algorithm. We get:
And the sequence of values for Q (obtained by clicking on “Trace output”)
260
As you can see, the high value accepted by EViews calls for a very high number of iterations for the algorithm to
crash.
We have suggested using another algorithm. With Newton we get:
As the model is linear, it converges in one iteration (the second is needed to check convergence).
Solving the model using another algorithm has proven one very important thing: your model has a solution. But this
can have two outcomes, which you will have to separate yourself.
o The solution is acceptable: the algorithm is at fault, or you used it inappropriately. For instance, you used
Gauss-Seidel with an illogical ordering. Switching to a successful method can be done indefinitely for this
particular model.
o The solution is wrong (as in the present case). You have not solved the problem, but you have the elements
to do so, even if some research will be necessary.
Of course, we could have used another of the techniques presented earlier (damping factors, exogeneizing variables)
but changing the algorithm was both faster and more informative
• The second error type(s) will call for a slightly modified model.
1. Q + M = FD + X
2. FD = CO + I + g
3. Log(CO) = Log(Q) + res_co
4. I = 0.50 * d(Q)
5. M / FD = r_m
The use of Log in Equation 3 is the main difference from the previous version.
In the same context as before, we get two different error messages, according to the algorithm.
In the example we are going to develop, we get from Gauss-Seidel.
261
And from Newton:
One can see that Newton take much less iterations, but both crash at the same period (after 6 “successes”).
Now the situation is both better and worse.
• Both algorithms fail very quickly, so switching algorithms is not relevant.
• But they succeed in the beginning, which gives us additional information..
This information will be put to good use.
Let us draw the graph of solution values, first for Gauss-Seidel:
262
Then Newton:
As expected, the two algorithms give the same values (the model solution) except for the erroneous period.
Let us consider the links between the elements, by considering derivative of the system.
1. D(Q) + D(M) = D(FD)
2. D(FD) = D(CO) + D(I)
3. D(CO)/CO= D(Q)/Q or D(CO) = 0.56 D(Q).
4. D(I) = 0.50 D(Q) – 0.50 D(Q(-1))
5. D(M) = r_m D(FD) or D(M) = 0.29 D(FD)
In summary
D(Q) = 0.71 D(FD) = 0.71 . (0.55 D(Q) + 0.50 D(Q)) – 0.50 D(DQ(-1))
D(Q) = - 0.50 / 0.25 D(Q) = - 2 D(Q)
263
We observe the same diverging process as in the graphs. The amplification of the change in Q brings the solution to a
negative value.
We can see that the problem does not lie with the consumption equation, rather with investment. This means that
facing this kind of problem we might have to retrace the causality process, perhaps by several steps.
For Gauss-Seidel, an efficient method (not provided by EViews) to understand the flaws in the convergence process
consists in changing the algorithm to Ritz-Jordan, which takes into account only values of the previous iteration,
eliminating interferences and allowing to improve individualizing divergent processes. At the beginning of a process,
this method individualizes immediately the faulty equation (less efficiently than a residual check, however).
This example has helped us to understand the problem a little better. Let us now treat the general case.
A first statement: to solve the problem, it is essential to solve the model, even by doing it in a non-operational
environment.
Basically, the model builder can face four situations. The model can:
• Never converge, using any algorithm.
• Converge only for a few periods.
• Converge for the whole period considered, but with unacceptable properties.
• Converge for the whole period, with acceptable properties.
The last situation never occurs at the first try. In forty years, it occurred to me once. And I consider myself as a
specialist.
The second and third differ only by the number of periods for which the results provide information.
As to the first case, it should be avoided much as possible, using the approach we propose, as we get no results at all
(except on the way divergence occurs). But this is easier said than done, of course.
Let us state also that when the model fails, there is a good chance the current model has no solution for that period,
as it presents two relatively independent flaws:
o The iterative process is not converging using Gauss-Seidel. If the ordering of the equations reproduces an
economic equilibration, this is not working: starting in the vicinity of a given solution does not lead to
another solution, but away from it.
o The linearization of the model provides the Newton algorithm with an image with very different properties
from what we expect.
Now, to make the model converge, you can apply the following techniques.
o First, you should control extensively the text of your equations (including the parameter values), the status of
your variables (endogenous/exogenous), and the block structure. It is better if somebody outside the
modelling team can contribute.
o You can drop one or several equations, and try solving the model again. The choice depends on the variables
which seem to lead the divergence: either the ones which actually take the worst values, or the elements
which influence them. By forcing a variable to take a given value, one can often measure its contribution to
the problem (especially if the variable goes through unacceptable values) by observing if the problem
disappears. This method has however the same limit as the others: the exogeneity of a variable destroys all
influences forwarded by it, and making the error disappear does not mean we have exogenize the faulty
equation.
264
o You can also change the values of parameters, or drop an individual influence by setting its coefficient to
zero. This is different from dropping the variable altogether: in our model, you can drop the role of the rate
of use in exports but leave it in imports. By applying variations to some strategic coefficients, chosen either
because they play an important role in the model (in the case of a general absence of convergence), or
because they appear linked to the diagnosed error (in case of a local error), one might establish a connection
between these coefficients and the intensity of the problem. However this method, once again, does not
locate necessarily the origin of the problem, which can come from far upstream. And it remains to determine
how to use the information, if one excludes setting arbitrarily the coefficient to the value ensuring
convergence.
o You can try alternate (probably simpler92) formulations. For instance, you could replace a complex CES
function using the relative cost of factors in a sophisticated way, by a very simple complementary factors
function. Of course, you can start by exogeneizing the relative cost of factors.
If the argument of a logarithm takes a negative value, you can introduce a test avoiding this situation (for example
replacing Log(x) by Log (max (x, a)), with a>0. The system might actually converge to an acceptable value (x>0) 93, but
reaching a solution with x<0 will give you information anyway94.
o You can use other starting points. You should try to guess as well as possible the domain in which you
solution will reside, and initialize the variables within it. This is easy on historical periods.
o You can check the derivatives of your equations. You can compute them by hand, or by computations (a little
harder, but EViews provides the elementary tools). If you start from quasi-solution values but the gradient is
wrong, you can go in the wrong direction, perhaps by a wide margin.
o Finally, you can change the simulation period. It is interesting to know if the problem is general, or appears
only for some periods. Examining the particular aspects of the faulty periods (slow growth, strong inflation,
important unemployment...) can give useful information on the origins of the error. Using it can however be
difficult. One specific case is the role of external trade: in forecasts, too fast an expansion compared to local
production and demand can drive the multiplier to abnormal and diverging values, which will disappear if we
simulate the model on the first periods available.
Once the model has converged, at least for a while, you find yourself in the general case, and you can start studying
the results. But the error correction process includes an additional difficulty, as the final version will have to eliminate
the artificial changes you introduced, to make it in fully operational. While an immediate convergence, even for a few
periods, guaranteed that the full model had no technical flaw.
The methods we have presented can be used again, but the goal is now to proceed to the last case. And for this we
have more information. This will of course be easier with a longer simulation, starting from on a regular base
trajectory. In other terms, a forecast, or rather a simulation over the future.
We can also concentrate on:
• As we have done on the small model, observing the way the model is diverging, and which variables (or sets)
appear to lead the way. For instance, variables at constant prices might follow an acceptable path, and prices
92
If the model you want to produce is quite complex, it might be a good idea to start with a simpler version, and
introduce layers of complexity one by one.
93
A rare situation in which the studying the error actually solves the problem.
94
When things go wrong, the potential for solving the problem is of course quite different if you have a (bad) solution
or nothing at all. This means it is generally essential to manage to reach a solution, even in a non-acceptable
framework.
265
explode. Or unemployment might explode while employment is acceptable, driving wages and the partition
of value added between workers and firms....
We can see if the divergence is monotonous or cyclical, regular or erratic, sudden or regular.
• Shocking assumptions and looking at the consequences. They can be acceptable in the beginning, which
means that the problem lies in the dynamics. They can grow wrong in a regular way, or explode suddenly.
The interpretation and solution of a convergence problem are not simple, essentially because of interferences due to
the interactive nature of the system. It will be often necessary to use several methods, and to repeat each of them
several times, which will not dispense with a thorough reflection on the logical structure of the model. It is therefore
essential for the model builder to know his model well, which will be all the more difficult as its size grows.
Your model might describe a sequence of equilibrating processes, which converge initially, but change through
dynamics which makes them diverging after a while. For instance, the expanding effect in the Keynesian multiplier
might get higher than one, making the process explosive (the same with the price-wage loop). Or the error-correction
process can actually become error-increasing.
In an operational model, a growing call for labor in the productive process can make unemployment become negative
in time, and it is used as a logarithm in the wage equation 95. Or as we have seen earlier, an accidentally very low level
of investment can make the associated equation explode suddenly. Or exports growing much slower than imports can
make GDP become negative.
• Finally, when one uses the Gauss - Seidel, the computation of eigenvalues of
can help to understand the convergence process, and evidence the mathematical characteristics of a possible
divergence (intensity and number of divergent processes, presence of cycles). If the highest modulus is higher
than one, this can justify the problem.
Processing this information can be done by a gradual elimination of loop influences, analogous to the one
recommended by Deleau and Malgrange (1978) to study dynamic model properties. In the best of cases, an
association will be obtained between each divergent process and a single loop variable (or two for a couple of
conjugate eigenvalues, producing a cyclical process).
But in any event the change must not reduce the theoretical validity of the model: one does not try to enforce a
solution thanks to an ad hoc formulation, but rather to correct an inconsistency.
7.2 A FIRST VALIDATION
Although the convergence of a model is not without connection to its theoretical validity, it is not enough to ensure it.
For this, a certain number of more or less exact methods must be applied.
7.2.1 EX POST SIMULATIONS
95
Replacing the logarithm by a level does not change the situation, as negative unemployment is of course
unacceptable.
266
Now we want to know if the model is able to perform realistic forecasts. For this, two techniques are available: ex-
post simulations (or ex-post forecasts) over the past, and stochastic simulations, generally over the future. Let us start
with the first case.
We have a problem: we do not know the future (if we did we would not need the model anyway), so we have to rely
on the past, which provides limited information. We are not in the situation of a scientist who can produce relations
as exact as he wants through the required number of experiments.
So we shall perform simulations on the past, but place the model in a situation as close as possible to the one it will
meet in the future. And there is one element it certainly will not know: the residuals in the estimated equations, in
other terms the unexplained part of the behaviors.
What he guesses is their most probable values, the average of their distribution: zero.
So we shall simulate the model on the past, with zero residuals. This is called “ex post simulations”.
They are undertaken on the total set of periods used for the estimation of coefficients: it may seem that the closer
simulation results will be to the observed values, the better the model should be.
One will then compare model results to historical values, through tables of residuals, graphs showing the two curves,
or statistics.
In practice one shall perform:
• Static simulations, which use historical values for lagged variables. We will get in fact a sequence of
independent simulations with a one-period horizon.
• Dynamic simulations, which will use the results from previous periods (except of course for the first l one).
This last criterion should be favored, as it is more representative of the future use of the model. One can also assume
that it is more difficult to meet, but nothing truly proves it.
In addition to the visual observation of errors (made easier by a graphic presentation) information can be synthesized
using to the following criteria:
• The average absolute error
• the average absolute percentage of error.
• the standard error (square root of the average squared error)
• the standard error divided by the average of observations means.
• the bias
One will generally start with one of the first four, which we shall now present with
x t as the simulated value, and x t as the observed value.
The root mean square error:
T
RMSE = 1/ T (x
t=1
t - x t ) 2
The mean absolute error:
267
T
MAE = 1 / T |(x t - x t )|
t=1
The mean absolute percentage error:
T
MAPE = 100 / T |(x t - x t ) /x t |
t=1
The root mean square error, normalized:
T
RMSER = 1/ T (xt=1
t - x t ) 2 x
The choice between a criterion in level or in relative value will come as usual from the nature of the variable. In
general the relative criterion will be more relevant (and more telling): an error of 1% on household consumption (CO)
is more easily interpreted than an error of 3 billion 2005 US Dollars. The absolute criterion will be used for elements
presenting a strong variability, especially if their sign is not determined: the error on the trade balance will be
measured naturally in billions of currents units (Euros or US Dollars). But to better understand the figure, it will be
more appropriate to divide it by a some normalizing element, like GDP. Levels can be used also if the notion of
relativity is already present in the variable itself (growth rates, ratios): the absolute error on the variable «growth rate
of wages» will give the relative precision on the salary itself. The error on the profits ratio will also be better measured
in points96.
As for the choice between the two types of moments: first and second order, it will be often forced, packages
providing in general only one of the two statistics. One should note that:
• The first moment is always lower than the second, when the formula is of the same type (Cauchy-Schwartz
inequality).
• This difference is all the more important as individual errors are of unequal size: a simulation including one or
two particularly badly simulated observations will be particularly hurt by the second criterion.
Finally, another question appears, in the case of dynamic simulations of quantities or especially indexes: should one
consider the error between simulated and observed levels of the variable, or the additional error introduced by the
simulation of the period? In others terms, should one consider the error on the level or on the growth rate? In the
case of a price index, both can be meaningful: in economic forecasts, or when judging the economic efficiency of a
policy, one generally uses the inflation level, but the cumulated evolution on the price levels is the determinant of
price competitiveness and of purchasing power.
And the result can be completely different, as the following example will show.
Let us suppose that the simulation of the price index p, from t = 1 to T (let us suppose T even), uses an econometric
formulation in growth rates, where the formula:
96
We get back to the same notion as usual: to choose the best criterion, one just has to consider the way the variable
(and its changes) is normally presented.
268
𝑝𝑡 = 𝑝𝑡−1 ⋅ (1 + 𝑓𝑡 ) + 𝑒𝑡
can be affected by two sets of errors:
(1) 𝑒1 = 𝑎, 𝑒𝑇 = −𝑎, 𝑒𝑡 = 0 for other values of t.
(2) 𝑒𝑡 = 𝑎 when t is even, 𝑒𝑡 = −𝑎 when t is odd
(the sum of errors is zero in both cases).
We shall not discuss the way these results have been obtained: the first estimation looks skewed, at least for small
samples, while the second should be corrected for autocorrelation. We have to admit the problems we shall meet
come at least partially from faulty estimations.
The criterion "absolute mean relative error" will give the results:
error on the level error on the rate
case (1) a (T - 1 ) / T 2a/T
case (2) a/2 A
The order is completely inverted according to the criterion used.
In more general terms, the issue is the following: is a simulation which presents a lasting gap to observed reality, but a
relatively stable trend parallel to it, better than a simulation which oscillates around true values, with the same error
band as the first97?
More generally, the difference between the two criteria comes from the dynamic character of the model. At a given
period, the error comes from the accumulation of present and past ones, the impact of the latter reduced in principle
with time.
The growth in the variance of the global error will be reduced by the fact that the sum of original residuals is zero, so
positive and negative ones will alternate. However, this condition is only met on the whole sample. It is quite possible,
even with no apparent autocorrelation, that the current sum has a significantly positive or negative value.
97
We prefer the first case, actually both should lead to a change in the formulation itself.
269
So it is not surprising with highly dynamic equations that the smallest simulation error is quite often found at the last
period, and the largest error close to the middle.
But in some occasions, an accumulation of errors can lead the variable away from its observed value. Let us show this
on a very simple example.
Let us suppose that the wage rate depends on the consumption price (in growth rates) and the unemployment rate
(unemployment divided by the total active population).
𝒕𝒙(𝑾𝒕 ) = 𝒂 ⋅ 𝒕𝒙(𝑷𝑪𝒕 ) − 𝒃 ⋅ 𝑼𝑵𝒕 /(𝑳𝒕 + 𝑼𝑵𝒕 ) + 𝒄
Unemployment decreases with job creation, with a reduced intensity:
𝛥𝑈𝑁𝑡 = −𝑑 ⋅ 𝛥𝐿𝑡
Both equations are estimated, and subject to errors. Let us suppose the second error is not auto correlated. On the
series UN the error will be the sum of non-correlated errors: it can very well keep the same sign over the period, for
instance if the few first errors are mostly positive. In the end the error will cancel out, but it will have introduced on W
itself a cumulative positive error which can make its level diverge markedly from its historical value98.
7.2.1.1 Reassessing the criterion
Actually this criterion is not so relevant, and its use is highly questionable, for an evident reason: the numerical history
one seeks to approach was known at estimation. It has been used to determine values of coefficients, but also to
choose among possible formulations. Even a scrupulous model builder will be led sometimes to favor equations that
give results close to reality, to the detriment of theoretical quality.
An extreme case is the introduction of dummy variables, taking the value 1 in one or several periods, and zero
otherwise: they will allow the equation to go perfectly through a point quite badly explained. One does not see why
this introduction, which improves the criterion, can improve the model and its properties.
It is clear in this case that the additional explanation does not add anything to the model, but rather masks a problem
(all the more as these variables are attached to the most badly explained periods). They can be accepted only if the
period presents a specific feature, impossible to translate into an explanatory element or behavior (as in the May
1968 case we have seen earlier, or the present coronavirus pandemy). But even then one runs the risk, by eliminating
the main fluctuations of the explained series, of making the interpretation of its behavior more difficult. And the
problem of forecasting these variables will remain.
More generally, there is no strict connection between the proximity of formula results to the observed reality and its
intrinsic quality (one can observe that very accurate estimations can provide coefficients values entirely unacceptable
by any theory, for example presenting wrong significant signs). A model produced by selecting systematically the
98
Of course, this would not happen with error correction formulations.
270
formulations giving the smallest distance to observed values, to the possible detriment of economic meaning, might
give good ex-post simulations99. By pushing the argument to the extreme, one can end up with a model describing
exactly reality, provided enough variables are added100.
But a model determined according to these principles will probably fail the more difficult test of the analytic shocks,
which we shall present later.
It is nevertheless also clear that a model failing to describe past evolutions will have to be rejected or at least
corrected (but by theoretical rather than technical contributions), as the quality of its forecasts will be even lower.
Passing the present test is therefore necessary, but not sufficient, to guarantee model quality.
Let us consider the formation of simulation errors, originating in the residuals (we shall consider later the
consequences of errors on coefficients and the actual form of the equations). They have three sources:
• The original estimation errors: it is clear that the smaller the estimation errors, the better the expected
adequacy of simulation results to observed reality.
• The expanding effects of the instantaneous links between variables. Taking into account as explanatory
elements variables affected by estimation errors (directly or indirectly) will increase the number of error
sources and in most cases the global error (most equations use sums or products, and in the absence of
correlation, the variance of a sum is the sum of variances).
This is not guaranteed however, as some errors can compensate by pure chance, and some model mechanisms can
play a moderating role. For instance, a positive error on demand will see its effect on production dampened by the
increase in imports it will generate. This will happen in our model:
[14] FD = CO + IP + IC + IH + gd
[15] TD = FD + CI
[16] DLOG(M) = 1.676*DLOG(TD) + 0.2133*DLOG(UR) - 0.1057*DLOG(COMPM)
- 0.1028*RES_M( - 1) - 0.2757+ EC_M
[17] Q + M = FD + X
A positive error on IP and FD will increase M, which will limit the impact on Q.
• The dynamic error (in dynamic simulations), coming from lagged variables which are now the result of a
simulation. The error should grow with each period simulated, as again the number of error sources
increases. However if we impose to all our estimated equations an error correcting structure, the problem is
99
Although it is quite possible that in an incoherent model the instantaneous and dynamic links between variables
exert a strong expanding effect on originally small estimation errors..
100
But the small explanation given by the last variables introduced should give them very low acceptability statistics.
271
partially solved: if a variable moves away from a target value, the behavior of model agent(s) will drive it back
to this target (it is not surprising that an “error correcting” model corrects the errors better than others).
With these models, medium-long-term simulations can actually give better results than the short-medium-
term ones, and ex post simulations can be more accurate at the end than in the middle101.
We shall improve our understanding of the implications of error correction when we address stochastic simulations.
7.2.1.2 Partial tests
To better interpret the simulation errors, some of their sources can be eliminated
• By exogeneizing variables.
The variables will take their exact values, which should reduce the global error. The measure of the decrease gives
some idea as to the contribution of the variable.
However this technique will mix two effects, in the case of a behavioral variable: suppressing the estimation error and
the model errors transiting through this variable. This will make the results difficult to interpret.
In the case of eliminated identities, the change in the error will show clearly the contribution of the variable. But if you
for balancing equations (like the supply-demand equilibrium giving GDP) the constraint will no longer hold true.
One can consider:
o Eliminating a single variable (or a small set) and observing the decrease of the residual compared to the full
model.
o Eliminating a large set of variables (for instance all the deflators) and observing the expansion of the residual
through model mechanisms.
With EViews this is done quite simply through the EXCLUDE statement.
• By re-introducing residuals (starting from a fully null set) :
This is done by setting the estimation residuals to their observed values, starting in two directions:
o Keeping every residual null, except for one equation. The (probable) reduction of the error will show the
gains to be expected from improving the equation.
o Setting every residual to its historical value, except for one equation. The remaining error will show how the
model dynamics expand the original estimation error.
Of course, combinations of variables also can be used: for instance observing the global error coming from external
trade, or the error remaining if all prices showed no residual error....
With EViews, we can simply put the requested residuals to zero (to avoid losing their estimated value, they can be
stored in an alternate variable). However, there is a much better technique, which uses the SCENARIO option. We
shall present it shortly.
• by using static simulations, and comparing them to dynamic ones
101
On the past, this is favored by the fact that the sum of the historical errors is zero.
272
This allows to separate the impact of the error on the first period, and from its dynamic influences.
7.2.2 EX-POST FORECASTS
To make the above criterion lose its artificial character, it looks more appropriate to simulate the model over a period
which has not been used for estimation. This can be done by excluding from the estimation sample the last known
periods, and using them to check the quality.
But this introduces several problems:
• It reduces the size of the sample, and therefore limits the significance of estimated coefficients. To reproduce
the conditions of the test, it is necessary to remove a high number of periods (5 at least in the case of a
medium-term annual model).
• Evaluating the result faces the same subjectivity problems as the previous test.
• But the main problem is that if we select models on the basis of the test, we use actually the same technique
as on the full sample, and the same “ad-hoc” critique applies.
It all depends on the way the test is conducted.
• First, it is always applied to a model (or a set of models) which has been estimated initially on the full sample.
One never restricts the sample during the whole estimation process, then tests the models on the eliminated
periods. This would be quite inconvenient, as these periods are the most representative of the domain on
which the model will be use: the future.
So the search for the best formulas in always conducted on the full sample. When we restrict the period, we get other
equations which might or might not be satisfying from an economic and statistical point of view. If the equations have
remained stable, the residuals are probably of the same size, and it is quite improbable that the multiplying effects of
the model properties will change very much, so the ex post error should not be much larger.
But if the results change, we will question the specifications themselves: the last periods are not consistent with the
global choice.
This means that the only message from the new test is the stability of the set of equations, which could have been
evidenced using a simple Chow breakpoint test, a more objective criterion. What we still get is a more global (and
perhaps utilitarian) measure.
Of course, if the modelling process had been applied to the reduced sample, the process would be much more valid,
provided one uses the result as an absolute measure of model quality, and not as a way to select the best model in a
set. In this case, the process is not much different from estimating on the full sample. The choice relies on a test
(obtaining good results on out of sample simulations) which basically requests equation stability, which means that
the equations obtained are not very far from the ones from the full sample.
In other words, the success of the ex post forecast requires that the estimations work on the full sample, and that they
keep stable if we eliminate the last periods. Of course, stability is a requirement for model reliability. But it is not a
proof: an equation inverting a true and strong link between two variables (making demand depend on imports for
instance) should remain stable over any sample.
7.2.3 SOLVING THE MODEL: SCENARIOS
To apply the above techniques, an EViews feature is quite useful: the scenarios.
When one wants to change the way the model is solved, it can mean:
273
• Solving the same model under
o Different assumptions on exogenous variables
o Different horizons
o Different options for formulations allowing alternate versions (for instance the exchange rate equation can
propose a constant nominal rate, a constant real rate or uncovered interest rate parity).
• Solving different coexisting versions of the same model.
o One can be official, having passed all the requested tests and familiar to the operational forecasters, and the
other can be more experimental, in the process of being tested.
The scenario option provides an efficient way of dealing with these issues.
• It works first through the creation of a new scenario.
In a program, the statement is:
model-name.scenario(n) scenario-name
This will create a blank scenario even if it exists already. If a previously defined scenario should be reused, one should
state:
model-name.scenario scenario-name
Then one should:
• Assign a suffix to the scenario/model solution:
model-name.assign @all suffix
From EViews 8 this can actually be obtained in one step:
model-name.scenario(a=suffix) scenario-name
• Create assumptions associated with the scenario, using the name of the variables followed by the suffix.
274
• State that the solution should look for the alternate versions of these variables by using the override
statement:
model-name.override list-of-overriden-exogenous
For instance to solve our model _fra_1 dynamically, with zero residuals for exports and imports (series called m_ec
and x_ec) and the current suffix “_s”.
' We solve over the past
' with residuals set to zero
' using the suffix « _s »
genr m_ec_s=0
genr x_ec_s=0
_fra_1.scenario "scenario 1"
_fra_1.append assign @all _s
_fra_1.solveopt(n=t m=1000,c=1e-6,o=n,d=d)
_fra_1.override ec_m ec_x
solve _fra_1
To override variables, two operations are needed:
• Stating that the variables have to be overridden.
• Creating the alternate assumptions using the scenario suffix.
If one step is missing, the consequences are different.
If a variable is omitted from the list, the override operation will not be applied, without any adverse consequence (but
no message). This allows to define a pool of potentially overridden variables in which the choice will be made later.
If a variable in the list has not been created, with the adequate suffix, EViews will refuse to solve the model, just as if
the original variable did not exist.
If the program switches to another scenario, the options associated to the previous scenario will be put aside but
remain associated with it, until they are modified under this very scenario. A new override statement cancels any
previous overrides.
To eliminate all overrides, one can just specify a blank list. This should be done as soon as the overrides are no longer
pertinent, as a precaution.
From EViews 8, one can use an alternate option: you can drop overrides individually through the REVERT command:
275
model_name.revert list-of-unoverriden-series
But the most important addition is the possibility to specify directly the changes in the assumptions, using ADJUST.
The syntax is:
model_name.adjust(init=initial_series) series adjustment
For instance the statements:
_mod_1.scenario(a=_2) “scenario 2”
_mod_1.adjust(init=”scenario 1”) gdp =+10000
will create a series called gdp_2 with a value 10000 higher than the value from “scenario 1” (maybe called gdp_1). The
series will be added to the “override” list, and to the “exclude” list as well if the variable is endogenous.
Be careful to introduce a space before the “=” sign, or gdp will take the value 10000!
Introducing a change involving elements, such as series and parameters, is possible but more difficult. One can prefer
using actual expressions, as we have done in our examples.
7.2.3.1 Scenario description
A description can be attached to each scenario, and optionally exported to the series it computes.
276
7.2.3.2 Scenario list view
By using “Scenario view” for a particular model, one will obtain a list of scenarios including any information or
comments, as well as possible overrides and excludes.
7.2.4 OVERRIDING THROUGH MENUS
You have certainly understood that in our opinion menus are not the best way to manage alternate scenarios.
However one might want to use this option to make a quick study, for instance to test a simple assumption.
This is done by accessing the model, then its “Variables” presentation, under a specific scenario (excluding “Baseline”).
Right-clicking on a variable name will open a box allowing you to change its override status, using “Edit Override”.
277
If the variable is not currently overridden, it will be added to the list (endogenous variables will be endogenized first,
by adding them to the EXCLUDE list).
Then a window will open, with the variable values in spreadsheet format, which you can edit.
Just as in programs, clicking on the REVERT item will get the series back to the original values, and the endogenous
back to their original status.
For an endogenous EXCLUDEd variable (through any method), REINCLUDE will also change its status back, but will
create an add factor for the variable, with null values.
7.2.5 OUR EXAMPLE
We shall now simulate the model with residuals set to zero, on the 1980q1 to 2004q4 period, the longest one for
which all estimations could be conducted, considering the availability of exogenous values.
' We solve over the past
' with residuals set to zero
for !i=1 to g_vbeha.@count
%1=g_vbeha.@seriesname(!i)
genr ec_{%1}_s=0
next
_fra_1.scenario "scenario 1"
_fra_1.append assign @all _s
_fra_1.solveopt(n=t m=1000,o=g,d=d)
_fra_1.override. ec_k ec_ic ec_le ec_m ec_x
solve _fra_1
' We compute the residuals and their averages
' - the relative errors
' - the absolute errors on the growth rates
smpl 1980Q1 2004Q2
matrix(20,2) v_psm_b
for !i=1 to g_vendo.@count
%st1=g_vendo.@seriesname(!i)
%st2=g_vendo.@seriesname(!i)+"_s"
series psa_{%st1}=100*({%st2}-{%st1})/({%st1}+({%st1}=0))
series tc_{%st2}=100*({%st2}-{%st2}(-1))/({%st2}(-1)+({%st2}(-1)=0))
278
series tc_{%st1}=100*({%st1}-{%st1}(-1))/({%st1}(-1)+({%st1}(-1)=0))
series psd_{%st1}=tc_{%st2}-tc_{%st1}
v_psm_b(!i,1)=(@mean(@abs(psa_{%st1})))
v_psm_b(!i,2)=(@mean(@abs(psd_{%st1})))
smpl 1980Q1 2004Q4
next
We get the following results
Simulation errors
Percentage
Error on the
error on the
level growth rate
CAP 0,755 0,035
CI 104,041 566,870
CO 1,720 0,210
FD 1,246 0,213
I 2,016 0,465
IC 0,989 0,348
IH 1,720 0,209
K 0,755 0,051
LE 2,431 0,261
LED 2,541 0,384
LT 1,871 0,202
M 2,142 0,837
PRLE_T 1,856 0,038
Q 0,989 0,348
RES_M 0,890 0,747
RES_X 23,156 12,928
RHI 1,720 0,210
TD 1,569 0,363
UR 1,600 1,296
X 2,268 1,195
The quality looks rather acceptable, considering the size of the model 102. The only source or uncertainty comes from
the estimation residuals, which have a zero sum over the period. And our equations, except for the change in
inventories, follow an error correction framework. This means that any residual appearing at a given period is
corrected with time, sometimes very fast. The variance of the global error will grow, but converge in the long run to a
finite value, as we shall observe later.
102
But one could also suggest that the smaller the model, the larger the domain of the assumptions and the smaller
the sources of uncertainty.
279
7.2.6 ANALYTIC SHOCKS
If ex-post simulations give a valued criterion (more or less valid) of the quality of the model in forecasts, analytic
shocks allow judging its properties (so the reliability of policy studies) by studying in terms of quantified economic
logic the results of model simulations. Thus the model will show if its quantitative properties are consistent with the
economic theory that has presided over its construction, and not too atypical as compared to the other models of the
same class (the second condition is not of course necessary, but to ignore it begs for further study and strong
arguments).
For this, a set of simple changes will be undertaken independently, on the main assumptions, trying to encompass
each of the main areas described by the model. Every type of domain must be treated: agents, production function,
supply and demand effects. This means we will study the sensitivity of the economy to a variation of exogenous
demand (by State investments for example), the behavior of households (by a variation of the income tax), of firms (by
a variation of their social contributions), the foreign trade (by the impact of a devaluation or a change in the tariffs
rates), the role of capital productivity, of an exogenous price increase. The quantitative and logical analysis of the
evolutions of the main economic elements will give indications on model quality, by comparison with economic theory
and properties of other models.
If the information gained from this criterion is probably more instructive, it also loses its objective character: a model
which makes imports increase when local firms create new productive capacities will be clearly wrong, but an original
behavior compared to the family of other models (for example if exports show a particularly strong sensitivity to price
competitiveness, or depend on the age of capital) will not be conclusive, in the absence of a general consensus on the
theoretical value. This might even, in some cases, affect the sign of the connection: facing a rise of inflation,
households can increase their savings (to maintain their purchasing power) or decrease them (to buy immediately
what will soon become more expensive).
In practice, to validate an atypical behavior, one will have to identify the source of the originality. Sometimes it will
have been expected before any simulation, as it comes from an original feature introduced voluntarily into the model.
If the model identifies foreign direct investment, a growth in local GDP can attract it and produce an increase in
exports, overcoming the usual substitution effect.
In general, to make the interpretation easier, the change in assumptions will not fluctuate with time. One will
distinguish however three types of shocks, according to the nature of the variation:
• One-time shocks, where the modification disappears after one period. For example the State will increase its
investment for one budgetary year. In this case one will be often interested, not only in the evolution of
effects (which should disappear with time), but also in the cumulation of these effects over time (such as the
total gain on production).
• Sustained shocks, where the modification is maintained for a number of periods, which can be the full sample
used in the simulation. For example a reduction of the income tax rate will be maintained on the period of
simulation.
• Cumulative shocks, where the shock is amplified at each period by an identical amount. Their interpretation
will be often difficult, as the global evolution comes from multiple sources.
It is clear that if one wants to correctly interpret these shocks, they must remain logically acceptable
In summary, one must consider two criteria:
• The acceptability of the shock.
• The readability of results.
This allows discarding one-time shocks in general, especially for policy variables. They fail most of the time on both
grounds. Except for an additional investment or subsidy, policy decisions are generally taken to last for a while,
280
especially if they apply to tax rates or social security benefits. Structural changes (gains on factor productivity) are also
permanent. As to external shocks (like an increase in world demand) they could very well be one-time.
But the most decisive argument is more down to earth: most of the interpretation will be done through graphs, and
one-time shocks leads back all the variables to the baseline scenario. After a few periods, the return to zero of
variations will make it impossible to separate the lines from each other, creating a high confusion close to the time
axis, while in the case of a sustained shock the stabilization of consequences will create a set of more parallel (or
slowly evolving) lines. As to cumulative shocks, they will produce expanding graphs and make the interpretation of the
initial changes (associated with a smaller scale) more difficult.
So we are left most of the time with sustained shocks, which are indeed the easiest to read, provided the shock is
reasonable. This applies first to the intensity: the less reasonable the shock, the more difficult its interpretation. This
means the shock must be small enough (for instance affordable in terms of budgetary cost) and large enough (why
should the government bother with such small decisions?).
But it applies also to the cumulated cost of the shock, in the case of policy decisions. If the period is long enough, and
the shock important enough, one can consider that its cost will become unbearable after a number of periods. And
the economic agents know about this. So not only should we stop the shock in time, but the model should adapt the
behavior of agents to this future event. This is not possible with our present model, in which our agents look only
backwards in time.
This is the most basic principle in the Lucas critique, which we shall approach later, when we address rational
expectations.
Of course, this long-term problem will be more acute when we consider forecasts: we could stress that the sample
period is small enough to allow sustained policy changes.
7.2.6.1 The choice of modified assumptions
Concerning shocked variables, one will look for a set that asks contributions from all the mechanisms of the model: as
the purpose is to validate the whole model, no abnormal behavior should remain. An initial investment in the full
examination of these elements will avoid later problems, the identification of which will prove to be far more
expensive (provided it can be done).
Especially, one will seek to activate the whole set of agents, and the whole set of domains: production, labor market,
prices and wages, external trade, financial mechanisms. Supply and demand oriented influences must all be studied.
However, if there was only one test to realize, the most representative would consist in increasing the exogenous part
of demand, thus determining the Keynesian multiplier. We shall come back to this in the following paragraph.
7.2.6.2 The choice of the results considered.
All variations of endogenous elements may or may not be examined, according to the size of the model. In most cases,
the first analyses will consider aggregated results, resorting only to detailed values if a problem comes to light.
• In the case of a macro-economic model, the main elements will be:
o The supply and demand equilibrium: gross domestic product, demand and its decomposition, exports and
imports (the ratio of the variation of GDP to the increase of exogenous demand defines the Keynesian
multiplier).
o Prices and wages, interest rates.
o Employment and unemployment.
o Foreign and budget balances.
281
o Some ratios: savings ratio, profits, utilization of production capacities, financial balances and debts as a share
of GDP.
• If the model also addresses financial issues, we could add
o The stocks of assets: debts, equities, currency, as a share of GDP.
7.2.7 OUR EXAMPLE
In our example, the number of available assumptions is quite low. The most reasonable choices are:
• Government demand GD
• World demand WD
And maybe also:
• Capital productivity PK
• The savings rate SR
• The depreciation rate DR103
• The share of non-wage household revenue in GDP R_RHIQ
• The share of household housing investment in revenue R_HI
• The residuals in each of the five estimated equations.
The main elements are demand oriented ones. We shall select the most widely accepted: government demand GD.
We shall use the following statements, taken from a program (we have left the comments):
' We produce a shock
103
Shocking DR introduces an interesting problem, as it is de facto forbidden. Indeed, the development of our
investment equation is based on the fact that DR is constant, which allows to integrate it to the constant term, so
changing it violates model specifications. It is the only shock which will actually modify the long-term rate of use, and
this for wrong reasons. To allow it to change, one can simply estimate the change in capital instead of investment (the
assumption on DR stability disappears) and rearrange the present capital equation so that it now determines
investment.
282
' First, an unshocked simulation
_fra_1.append assign @all _b
_fra_1.solveopt(n=t m=1001,o=g,d=d)
_fra_1.override
solve _fra_1
' The shocked simulation
_fra_1.solveopt(n=t m=1002,o=g,d=d)
smpl 1980Q1 2004Q4
genr gd_v=gd+.01*q_b*(t>=1981)
genr dv_gd=+.01*q_b*(t>=1981)
_fra_1.scenario "scenario 1"
_fra_1.append assign @all _v
_fra_1.override gd
solve _fra_1
' We compute the differences
' in absolute and relative terms
for !i=1 to g_vendo.@count
%1=g_vendo.@seriesname(!i)
series dv_{%1}={%1}_v-{%1}_b
series pv_{%1}=100*dv_{%1}/({%1}_b+({%1}_b=0))
next
A few additional comments.
• We have started the shock a little after the start of the simulation. This allows checking that in absence of shock the
two simulations give the same result, which should mean that the difference can be interpreted as coming only
from the shock. Additional differences can come from:
o A different starting date
o A different model.
o Different assumptions.
This is not an absolute proof, however:
o An additional change in the series might start after 1981.
o The specification of the model might contain changes which apply only after 1981 (for instance a new tax) or only
if a condition is met, which will occur later (like a high level of inflation).
To guarantee the absence of such errors, it is advisable to simulate the baseline scenario just before the shock, even it
has been produced earlier. The efficiency of present algorithms and the speed of present computers make the cost
negligible in most cases.
• We produced the alternate assumption in two steps, computing first the shock itself, with the same format (“DV_”
plus the name of the variable) as the absolute endogenous deviations. This will allow us to compare the ex-ante and
ex post changes in GDP and create a graph presenting all changes.
7.2.7.1 The results
We shall base our commentary on three graphs showing: the evolution of the supply-demand equilibrium in relative
terms, the decomposition of final demand in absolute terms (including the initial shock), and the evolution of
elements linked to production.
283
4
3 A shock on Government demand
2
Gross Domestic Product
Imports
1 Final demand
Exports
0
-1
82 84 86 88 90 92 94 96 98 00 02 04
7,000,000,000
6,000,000,000
5,000,000,000 Final demand
The decomposition of demand
Household consumption
4,000,000,000 Productive investment
3,000,000,000 Housing investment
Change in inventories
2,000,000,000 Government demand
1,000,000,000
0
-1,000,000,000
1985 1990 1995 2000
5
Production
4
Percentage variation
3 Gross domestic product
Productive capacity
2 Rate of use
Productive capital
1 Firms employment
Productive investment
0
-1
1985 1990 1995 2000
The properties shown by these graphs are quite usual for this type of model (considering the reduced size of our
version).
284
• The ex-ante demand variation is amplified by the increase in production factors (investment and
employment), the latter affecting household revenue and consumption. These two elements keep increasing
for a while, with some inertia, traditional for an error correcting process.
• Later, both changes become proportional to GDP. But while households obtain as revenue a high share of the
increase, the motivation for investment decreases as capital adapts to GDP. In the long run, the only
incentive left is the replacement of a higher level of discarded capital.
• Capital and capacity adapt slowly, but this inertia makes them overshoot the target. A cyclical process
appears on the rate of use, with a rather fast decreasing intensity.
• External trade plays a simple role.
o The change in exports follows the rate of use, with a decreasing amplitude as capacities adapt.
o Imports take a large initial share in the additional demand, when capacities have not adapted. They
compensate the ex-post increase in demand, reducing the multiplier value. In following periods, capacities
adapt, and the multiplier increases regularly. The overshooting of capital actually favors local producers in
the medium-term.
On the whole, although this model remains overly simplistic, it looks rather sound if we consider its size.
7.2.7.2 The drawbacks of shocks over the past
Although producing historic shocks for a model designed for the future does not face the same criticisms as the ex-
post tests, they still meet several problems.
• They are conducted on a limited period. If one wants to observe the long-term properties of the model, the
available period will generally be too small. The problem will increase if data production has started recently,
or if the country has undergone a recent transformation making the use of older data unacceptable. For
Central and Eastern Europe countries, acceptable series start in 1995 at best.
• They are conducted on the wrong period. If one wants to interpret the dynamic properties of a model over a
sufficiently long-term, it is necessary to start the shocks quite far in the past. At that time, the economy of the
country might have presented different characteristics. Even if the economic behavior was the same, some
structural parameters were quite different, like the importance of external trade, or the relative roles of labor
and capital in the production process
• They are based on a more or less unstable baseline. No model is completely linear (as it will always include
both sums and logarithms). This means that the response to shocks will depend on the baseline, which can
present irregularities. For example, a given increase in government consumption will generate more GDP if
the activity is low, and the country can use more idle capacities to respond to demand.
This means that it will be difficult to decide if irregular variations in the results of the test come from model properties
or from a non-smooth baseline.
Obviously, all these problems disappear when the shock applies to forecasts:
• The period is unlimited, allowing to evidence completely convergence processes and in particular cycles.
• As tests are conducted on the same period as future operational applications, the evaluation will be more
relevant
• The present context and future issues are more familiar to the modeler
• As the baseline is smooth (especially in the medium and long-term when it matters) any irregularity will be
allocated to model properties.
This is why we suggest to limit to a minimum the studies on the past, and proceed to simulations on the future. This is
true even for control of convergence: it is on the future and not on the past that the model will have to converge. But
a limited series of immediate tests can evidence problems right away, which can help to gain time in the global
process by staying on the right path.
285
8 CHAPTER 8: TESTING THE MODEL OVER THE FUTURE
The above arguments lead to the natural idea: test the model on the future.
• We shall have initial information on the reliability of spontaneous forecasts, and of their distance to what we
expect of the near future.
• The results will me more representative of future use.
• The actual results can be interpreted as the actual consequences of present policy decisions.
• The first periods of the shock will be representative of the present efficiency of present policies.
• The tests can be conducted on a period of any length, allowing to observe convergence and cycles.
• With regularly growing assumptions we can test that the simulation is regular, and that it converges to both a
steady state growth and a long-term stable solution.
• Applying to these regular solution constant shocks we can check that we get smooth evolutions, and
interpretation is easier. We also have enough observations to treat the Lucas critique.
There are only two drawbacks:
• We cannot check the simulation results against true values. We shall try to prove this is not so important, and
can be replaced by other tests.
• We do have to produce a simulation over the future, an unknown domain in which convergence might be
more difficult to achieve.
But this convergence will have to be obtained anyway. So our first test will be:
• Can the model converge at any horizon?
In our sense, this is a prerequisite for structural econometric models. This is where this type of model has an
advantage over CGE and VARs: to produce dynamic simulations which show the way the economy converges to a
long-term equilibrium, and allow a full interpretation of the process. If this is missing (if the model converges only over
the medium-term) this advantage is lost.
8.1 MAKING THE MODEL CONVERGE IN THE LONG RUN
First, let us stress that this is not the same as producing a long-term forecast. What we are considering is the very long
run (the first converging version of the model can take several centuries to stabilize). We are not going to suppose
that the model will remain valid for such a period. What we want to achieve is:
A technical reproduction on a long period of the mechanisms associated with the equations we have evidenced, as if
they remained valid forever104. The period is much longer than the longest on which the model will be used.
For instance, many models present a cyclical behavior, linking investment, capital, the output gap and profitability.
The length of this cycle can exceed 30 years. To check that its period is constant, and that it is converging regularly, we
need at least three cycles, which means that our simulations have to last until 2100.
So at the beginning of the tests, one should consider a very long period.
104
But we know this is not the case
286
This could be compared to testing a car (actually a prototype of a car) on an artificial track, for as long as it can run,
exceeding by far its normal life (say several millions of kilometers), at a constant speed. And also changing one of the
car elements, or the driving conditions, and observing the consequences, both immediate and after a long period..
What we want to get is information which will help us to build a better model, which will be used later in quite
different circumstances (a shorter but less regular environment).
We have to keep this in mind when we undertake our simulation.
8.1.1 THE ASSUMPTIONS
To produce a long-term simulation, we need assumptions on the exogenous elements.
The choice of these assumptions can have several goals:
• Producing short-term evolutions which coincide with what we know already of the near future. This can also
be true of the recent past which we have to simulate in the absence of complete information (global GDP but
not its product decomposition).
• Producing reasonable medium-term solutions, consistent with the goals of the Government, or of what we
expect from the local economy (a high but decreasing Budget deficit for instance).
• Allowing the model to converge regularly to a long-term equilibrium, which we shall examine to understand
model properties.
For the time being, we shall concentrate on the last goal. Of course, we shall control that the two first targets are at
least partially met, as an unreasonable simulation will react to shocks in an unreliable way, due to the model non
linearities, making it more difficult to interpret.
This means we shall not differentiate the assumptions which share the same dimension, but apply to them common
growth rates. In practice three rates have to be defined (in addition to a null one):
A rate for populations (including employment).
o A rate for variables at constant prices.
o A rate for deflators.
Of course, the second rate could be replaced by labor productivity. Combining it with employment we would get
quantities.
Actually, we shall see that most model assumptions have no dimension. This should be favored as it suppresses the
need to make normally independent assumptions grow at the same rate (in our model, Government demand GD and
World demand WD).
A tip : in most data sets, the last observed period changes between series. If you want to apply a given growth rate
from the first missing observation, you just have to specify the SMPL statement as:
SMPL period if variable=na
genr variable=variable(-1)*(1 + rate)
287
8.1.1.1 A particular case: the estimation residuals
In general (this is the case for all the equations in our model) the estimated expression has been defined without
dimension, if only to eliminate the risk of heteroscedasticity. And anyway, the most probable value of a residual is
supposed to be zero. But one can suppose that the last value obtained is a better indicator of the future, in particular
if the previous values held the same sign. For the time being, the choice is not so important, as we are not overly
concerned with short-term solutions. But it will become crucial later, when we approach actual forecasts.
8.1.2 ADAPTING THE FORMULATIONS
To make the model converge in the long run, applying any formula to variables which grow at the theoretical rate
should produce a result which grows at its own theoretical rate. This is easily controlled, both visually but also
technically as we shall see later.
Identities will provide this property automatically, if they are specified correctly and if the assumptions are well
defined.
For estimated equations, the error correction format allows to introduce the necessary constraints in the long-term
equation. One does not have to be concerned with the VAR, which includes only growth rates, stable in the long run
by definition. But one has to manage (generally suppress) any additional trend.
Let us explain this on our exports equation, using for now the trend provided by EViews. The basic equation is:
[3] DLOG(x) = 0.4940*DLOG(wd) - 0.09142*res_x( - 1) + 0.01960+ ec_x
[18] res_x = p_x(1) * log(x / wd) + 0.9815 * log(ur) + 0.001441 * @trend(60:1)
Let us suppose the rest of the model (in particular the investment equation) has allowed UR to stabilize in the long
run. Then any growth in res_w will come from:
• X growing at a different rate from WD.
• The trend : @trend(60:1)
To suppress the trend at the end of the historical period, we can replace its expression by105:
@trend(60:1) * (t<=2004) + @elem(@trend(60:1) , "2004q4") * (t>2004)
Starting in the first quarter of 2005 (represented by t>2004) the trend will be replaced by its 2014q4 value.
105
Let us recall the syntax for the @elem function:
@elem(series,”date”) gives a scalar with the value of the series for that particular date.
288
This works, but one can find the expression overly complex. To simplify it, we can apply the technique presented
earlier, estimating the employment equation.
First, let use replace the trend by our time variable t.
Then, we can replace t by (t-2014)106 without changing he actual specification. The constant term will adapt and the
formula will give the same values.
Finally, we can break down the computation into:
• A trend of (t-2014) for the historical period.
• 0 for the forecast period.
Combining both, we get the formula;
(t-2014)*(t<=2014)+0*(t>2014)
Which simplifies of course into:
(t-2014)*(t<=2014)
On this example, let us see how growth rates of endogenous elements converge to that of exogenous.
We have (with log(WD) and ec_x constant)
𝚫𝑳𝒐𝒈(𝑿𝒕 ) = 𝒂 − 𝒃 ⋅ 𝒓𝒆𝒔_𝒙𝒕−𝟏
𝒓𝒆𝒔_x𝒕 = 𝑳𝒐𝒈(𝑿𝒕 /𝑾𝑫𝒕 ) + 𝒄
giving
𝑳𝒐𝒈(𝑿𝒕 ) = 𝑳𝒐𝒈(𝑿𝒕−𝟏 ) + 𝒂 + 𝒄 − 𝒃 ⋅ (𝑳𝒐𝒈(𝑿𝒕−𝟏 ) − 𝑳𝒐𝒈(𝑾𝑫𝒕−𝟏 ))
and
𝑳𝒐𝒈(𝑿𝒕 ) = (𝟏 − 𝒃) ⋅ 𝑳𝒐𝒈(𝑿𝒕−𝟏 ) + 𝒂 + 𝒄 − 𝒃 ⋅ 𝑳𝒐𝒈(𝑾𝑫𝒕−𝟏 )
106
Actually, we it might be better to use 2004.75, associated with the last historical value.
289
The growth rate of X will converge to that of WD.
In very special cases, the trend must not be suppressed but replaced by a given value, as it applies to non-constant
ratios.
For instance in the employment equation, the structural labor productivity follows an estimated trend, estimated by:
𝑳𝒐𝒈(𝑸𝒕 /𝑳𝑬𝒕 ) = 𝒂 + 𝒃 ⋅ 𝒕
The trend of labor productivity will be replaced in forecasts by
𝑳𝒐𝒈(𝑷𝑹𝑳𝒕 ) = 𝑳𝒐𝒈((𝟏 + 𝒕𝒙𝒒)/(𝟏 + 𝒕𝒙𝒏))
Where txq and txn represent (as seen above) the long-term growth rates of quantities and populations.
In the long run, the growth rates of quantities and employment are fixed, which defines the growth rate of labor
productivity. The estimated coefficient has to be replaced by this value.
8.1.3 IMPROVING THE CHANCE (AND SPEED) OF CONVERGENCE
We have presented earlier some techniques. They can be applied indifferently to simulations over the past or the
future. Few new elements will be specific to forecasts.
Actually the only significant one is associated with the starting values of the endogenous, for the computation process
of a given period. EViews allows initializing the process either by “historical” values or those obtained at the previous
period. This is done through the option: i=a (actuals) or i=p(previous) in the SOLVE or SOLVEOPT statement. In
forecasts, only the last one should be available.
But actually one can define values on the future:
• By computing the theoretical values, applying to the endogenous variables their theoretical growth rate.
• By giving to the actual names the value obtained for a previous solution, if it used comparable assumptions to
the current problem.
These values are considered as “actual” by EViews, even if they correspond to future periods 107. They can be used as
starting points, and it is possible (in particular in the second case) that they will improve the speed and probability of
convergence. The issue is whether the growth rate of variables is higher than the difference for the same period
between the two solutions. Both cases are quite possible, depending on the difference between assumptions in the
second case.
107
EViews leaves to the user the management of the present date.
290
In any case, the general principle applies: if one option does not work and another is available, just try it!
8.1.4 SOLVING PARTIAL MODELS
As presented earlier, an efficient way of diagnosing the origin of convergence problems is to exclude endogenous
variable(s) from the solution. This calls for generating values for these variables. Generating starting values for all
variables allows to use this feature at no further cost.
8.1.5 CHECKING THE EXISTENCE OF A LONG-TERM SOLUTION
We shall now present a very efficient way of checking that model specifications allow a long-term solution.
For this solution to exist, the following must be true:
All the exogenous assumptions sharing the same dimension must grow in the long run at a common rate.
Of course, their evolution can differ from this rate for the first periods.
As stated earlier, in practice three basic rates will appear:
• variables at constant prices (let us call it txq)
• deflators (let us call it txp)
• populations (let us call it txn)
The other rates will be combinations of the three (in addition to pure stability). The main ones are labor productivity
and purchasing power of wages: txq-txn or rather (1+txq)/(1+txn)-1
The trends must be stopped, or set to theoretical values (like the trend in labor productivity).
In estimated formulas, the long-term equations (in principle cointegrating relations) should use elements with zero
dimension, and the dynamic equations too. This is not completely necessary, but it makes things much simpler, and is
required anyway if we want the equations to be homoscedastic. Stationarity is more easily checked on elements with
no dimension.
For dynamic equations, using logarithms (in general their variations) will provide this property naturally.
Under the above conditions, every formula must give a solution growing at the rate associated with the element it
defines. For instance, the equation for imports at constant prices must give a result growing at rate q.
It is enough that a single equation does not meet these conditions to forbid the whole model from having a long-term
solution, let alone reach it.
In addition, the estimated formulas must lead to the solution and not away from it. In particular the sign of the
coefficients associated with error correcting terms must be the right one, and the value lower than one to create a
diverging process.
If these conditions are not met, the model will probably explode in the long run, and it is quite possible that this
process affects already the medium-term solutions. To be sure this is not the case, the only option is to check the
presence of a long-term solution.
291
Of course, one can start by simulating the whole model over a long enough period. Unfortunately, even if high care
has been taken, at least one error will generally remain, propagating to the whole system without generally giving
clear clues as to the origin of the problem.
However, a very simple technique can be applied: producing a residual check over the future!
One just has to generate (as proposed earlier) a full set of exogenous and endogenous elements following the above
constraints. Then the model is solved using the “fit” option (separate computation of each equation) and the growth
rate of the endogenous result108 is compared to the growth rate of the variable the equation defines. The two must be
identical. If not, the elements in the equation are not consistent with its formulation.
The origins of problems are somewhat different from the usual residual check. The most frequent are:
• A trend has been forgotten or miscomputed in the equation, for instance the opening of world trade in the
imports equation109.
• The dimension of one or more element on the right hand side is not consistent with its role. This can apply to
identities or estimated formulas.
For instance the exogenous variable associated with social benefits might be computed in purchasing power per head,
while the social benefits equation just multiplies this element by the deflator. Or a linear estimation adds up elements
with various dimensions.
• The dimension of the endogenous variable is inconsistent with the formula. For instance a CES production
function uses the wrong coefficients, used as exponents.
• We have decided to apply different formulas for the past or the future, for instance, blocking the trend after
the estimation period, and the second type contains errors of the above types..
For instance, a trend term such as:
C(1)*((t-2007)*(t<2007)+0.02*t*(t>=2007))
instead of:
C(1)*((t-2007)*(t<2007)+0.02*(t-2007)*(t>=2007))
will make the element jump from 0 to 0.02*2007 when forecasts start in 2007.
108
Not the growth rate of the left hand side expression itself.
109
This trend can be continued in the initial forecasting periods, but it must decrease and disappear after a while. In
this case, one must check the residuals only after this condition is met.
292
The obvious advantage of this technique is to identify most of the erroneous equations in the model (hopefully a
limited se)in one pass. But as in the more usual application, a zero residual does not prove the absence of error, and a
non-zero one can come from several errors. This means that in general reaching an (apparently) cleaned version will
take several iterations.
The programs provided at the end of the book will systematically use this technique.
8.2 CONVERGENCE PROBLEMS IN THE SHORT RUN
Success in the above test does not guarantee reaching a consistent solution, however: now the problem lies in the
initial periods, and particularly in the very first one. 110And the above method cannot be applied, as the lagged
elements on the right hand side do not yet grow at the theoretical rate.
The most immediate danger is an error in the assumptions: the above program, which sets the growth rates to all the
exogenous variables, should avoid this problem. We have checked that their growth rates are correct, and consistent
with the formulas in which they are used. The only error could come from values given explicitly, for instance because
we have advance information on some elements, or because we are trying to specify endogenous target by
manipulating exogenous instruments.
Finally, it is possible that the management of residuals in the first forecast periods brings high variations, and in
particular strong cycles, which
• make one of the elements pass through unacceptable values (such as negative levels for goods, which enter
behavioral equations by their logarithm)
• or produce a model with diverging properties. This can be true for Gauss-Seidel. If the linearized model is
very different from the right one, it might can contain eigenvalues with a modulus higher than one. And the
Newton method can make the solution move in the wrong direction.
As to the remedies, they are not different from the ones developed in the previous chapter..
8.3 CONVERGENCE PROBLEMS IN THE MEDIUM RUN
Now if the model has started converging, and contains a converging error-correcting process, it should give solutions
over the whole period.
Of course this is not true if the second condition is not met: in that case we know nothing about the long run
properties, and divergence can occur at any time. Actually the greatest danger occurs when convergence is still
obtained at the end of the forecasting period, but numerical properties are already wrong, not yet bad enough to be
noticed.
Otherwise, the only danger which remains is if the cycles are too strong, which can only occur if the model is wrong.
But now we have help, just as if we tested a motor which runs for a while before breaking down: it is much easier to
understand the reason than if it does not start at all.
The main solutions have been developed earlier:
• Applying shocks to the main assumptions and observing the consequences. Maybe the diverging process will
start before the breakdown, and its reasons can be made explicit.
110
Just as takeoff is the most dangerous part of a plane flight.
293
• Excluding some variables, or changing the intensity (increasing or decreasing) of local explanations. Now we
can observe how these changes affect the horizon of the crash, and the properties: less cyclical or more
consistent with economic theory. This is much more informative than just trying to get a solution.
8.4 TESTING THE RESULTS
Let us now suppose that our model is converging in the long run. It means that we have efficiently managed to
establish a long-term solution (remember that we are considering a very long period).
8.4.1 OF SIMULATIONS
What we should check at this stage is how the model behaves in the short and medium runs. Of course, we have
introduced assumptions too crude to provide a realistic forecast. In particular, setting from the first the growth of
word demand to the common value txq will stop the expansion of exports (imports should follow). A more realistic
option should be to reduce it gradually. However:
• The impact of this gradual decrease will change as the model evolves, so it will have to be reconsidered with
time.
• This will go against the main present goal in simulating the model over the future, which is obtaining
trajectories which are stable, thus easier to interpret, in a future as near as possible.
In practice the quality of our simulations has just to be reasonable enough not to question the teachings of
sensitivities to shocks. They will represent the most important diagnosis.
8.4.2 OF SHOCKS
For the shocks, we shall use the same « Scenario » feature as in the previous chapter. We have already described the
advantages of waiting for this stage to make the tests. The situation has not changed much.
As we have stated earlier, it is on the future that shocks are better performed:
• They can last longer.
• They start from a smoother baseline.
• The period is the same as for future uses.
In particular, one can check the exact long run values and ratios. Long-term convergence itself is guaranteed by the
previous production of a long-term scenario, provided the shock is not too large.
Concerning the period itself, it is better to start with a very long one (like two or three centuries). This might look
unreasonably high, but it allows:
• To check that indeed the long run is reached numerically.
• To interpret the convergence (or divergence) of potential cycles, which have to be observed in a high enough
number. Cycles with a 50 years period are not uncommon (especially if the model slightly overdoes it) so
observing 4 instances requires 200 years...
Of course once the model is stabilized, the cycles will probably converge faster, and the test period can be reduced,
allowing to concentrate of the short-medium-term properties (which will be much more visible in shortened graphs).
294
8.5 EVIEWS FEATURES
We will now see how we can apply the above principles to EViews.
8.5.1 PRODUCING SIMULATIONS
We shall now describe in turn the phases is the simulation process.
8.5.1.1 Preparing the workfile
Let us see how we can manage this on our simple model. We shall concentrate on the specific features (the full
program is made available later).
The first task is producing a « forecast » file. The period considered should be quite large.
close fra_q.wf1
open fra_q.wf1
wfsave _fra_1.wf1
pageselect fra_m2
pagecopy(page=fra_proj)
pagestruct(end=2100Q4) *
8.5.1.2 Producing the assumptions
We can also compte the endogenous (as justified earlier).
Now we will produce the assumptions.
Fortunately, most of the assumptions have no dimension, and are independent from the growth rates.
The only exceptions are:
• Government and world demand, gd and wd, quantities which grow at txq.
• Government employment lg which grows at txn
scalar txq=exp(0.006)-1
scalar txn=exp(0.0005)-1
smpl 2015Q1 2100Q4
genr T=t(-1)+0.25
' exogenous
for %1 gd wd
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)
next
for %1 r_rhiq dr sr pk compm ec_i ec_le ec_m ec_x ec_ci ct r_ih
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)
next
for %1 lg
295
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txn)
next
for %1 wr
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)/(1+txn)
next
' endogenous
for %1 ci i m x co fd k rhi q cap ih
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)
next
for %1 ur res_m res_x
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)
next
for %1 le lt led
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txn)
next
for %1 prle_t
smpl 2005Q1 2100Q4 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)/(1+txn)
next
Changing coefficients
For our simulation to provide a steady state solution, we only need to change one coefficient: the trend in labor
productivity which must follow txq-txn (in logs) at each period. As it applies to t which grows by 0.25 a each quarter,
we must multiply the coefficient by 4. And as we have seen, our management of trends makes the change in
coefficient apply only from 2004, without calling into question the previous values of trend productivity.
We will replace c_prle(2) by 4*log((1+txq)/(1+txn))
8.5.2 PRODUCING A BASE SOLUTION
To produce a base solution, we use the same technique as usual. We compute the growth rates of the endogenous to
check that they converge to the long-term values (zero, txq, txn or a combination of both).
Two remarks:
• We shall also compute the theoretical growth rates, and the difference between actual and theoretical,
which should decrease to zero in the long run.
Actually we shall compute the logs, for which the variables associated with combinations of growth rates (like labor
productivity) will nevertheless give round values, making a visual check easier.
As stated earlier, this technique represents a trick, and one can rely on actual growth rates. It is in our sense less
efficient, but it is much easier to explain (consider the length of my message, and I am not sure it actually went
through).
• The blank “override” statement insures that only base series are used..
296
• The blank “exclude” statement insures against any previous exclusion associated with the present scenario. It
is free, and one cannot be sure that it is not needed.
smpl 2006Q4 2100Q4
_fra_1.scenario "Scenario 1"
_fra_1.append assign @all _p
_fra_1.override
_fra_1.exclude
_fra_1.solveopt(n=t m=1000,o=g,d=d)
_fra_1.solve
for !i=1 to g_vendo.@count
%st2=g_vendo.@seriesname(!i)
genr tc_{%st2}=100*log(({%st2}_p+({%st2}_p=0))/({%st2}_p(-1)+({%st2}_p(-1)=0)))
genr tc0_{%st2}=100*log(({%st2}+({%st2}=0))/({%st2}(-1)+({%st2}(-1)=0)))
genr dtc_{%st2}=tc_{%st2}-tc0_{%st2}
next
8.5.3 PRODUCING SHOCKS
To produce shocks we use the same technique as before. We apply to gd an increase of 1 GDP point (using q_p, the
GDP level computed by the baseline simulation). We solve the model using the suffix “_v”. Then we compute the
absolute and relative differences (destroying first any variable of this type, to avoid confusions).
Finally we can export the results to Excel (not necessary with the present EViews graphic potential).
A trick: when a simulation breaks down in a program which contains several SOLVE statements for a given model, the
EViews message will not let you identify which of them produced the error. A simple way to make this identification is
to change the maximum number of iterations in the statement (1000 into 1001, 1002,….), as this statement will be
displayed in the message.
genr gd_v=gd+.01*q_p*(t>=2007)
_fra_1.solveopt(n=t m=1002,o=g,d=d)
_fra_1.scenario "Scenario 1"
_fra_1.append assign @all _v
_fra_1.override gd
smpl 2006Q4 2100Q4
_fra_1.solve
for !i=1 to g_vendo.@count
%1=g_vendo.@seriesname(!i)
series dv_{%1}={%1}_v-{%1}_p
series pv_{%1}=100*dv_{%1}/({%1}_p+({%1}_p=0))
next
group chocs pv_* dv_*
write(t=xls) varia_p.xls chocs
' We create groups for the supply-demand equilibrium, and for the production process
group g_v1 pv_x pv_m pv_fd pv_q
group g_v2 pv_cap pv_q pv_le pv_k pv_ur pv_i
297
2.5 The multiplier
2.0
1.5
1.0 Exports
Imports
0.5 Final demand
Gross domestic product
0.0
-0.5
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
2.5 The production process
2.0
Productive capacity
1.5 Gross domestic product
Firms employment
1.0 Productive capital
Rate of use of capacities
0.5 Poductive investment
0.0
-0.5
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
The results look rather similar to the previous shock. We can only observe that with a more recent period, the role of
the trade variables increases, leading to a lower multiplier. The main observation is that now the smooth evolutions
make the results much clearer.
In the long run, the value of the multiplier is about 1.4, a somewhat high value. But we do not take into account the
loss in competitiveness on both external trade elements, which would come through inflation.
We can observe that the rate of use of capacities gets back to its base value. This leads exports to do the same.
The addition of investment and consumption to the ex ante change in Government demand makes the global
variation grow to 2 %. Imports grow less as they depend also on the unchanged exports. And GDP even less, as it
depends even more on exports.
The stability of the rate of use gives investment and capital on one side, and GDP and capacity on the other, a
constant ratio, and thus the same relative change. As productivity of capital is fixed, all these variables change in the
same proportion. And as productivity of labor is also fixed, firms employment follow. The change in inventories also,
although with more sustained fluctuations.
But government employment does not change. Thus total employment and wages grow less, as well as household
income and consumption, which depend also on GDP
8.5.4 CHANGING MODEL SPECIFICATIONS
To test model properties in the long run, it can be interesting
298
• To test alternate estimated formulas, the properties of which have to be ascertained.
• To allow some of its equations to follow formulations different from estimated ones, either through a change
of coefficient values, or even a change in the actual formula.
This can be done of course by producing a new model creation program. But this will have several drawbacks.
• It will take more time and involve more statements.
• It will not single out an “official” version.
• One will have to maintain several model-building programs, making file management less clear.
• The forecasting program will have to be run anyway.
In our sense it is better:
• To maintain an official version of the model, representing the best tested version according to the present
information.
• To introduce the changes in the “forecasting” program (or programs, documenting the differences)..
This can be done in two ways.
• If a new official version has been estimated for some equation, the full set of associated statements should
be specified (residual set to zero, defining the equation and estimating it, storing the estimated residual,
merging the equation).
• If one is just testing an alternate version with at least one fixed coefficient, with or without re-estimating the
rest, we suggest the following procedure, which we shall present using investment as an example:
o Copy into the “forecasting” program the set of statements, as above.
o State a vector of parameters (which will not be estimated).
vector(10) p_i
o Estimate the equation as usual, with a different name (here: the normal name followed by “1”).
coef(10) c_i
smpl 1980Q1 2004Q4
genr ec_i=0
equation eq_i1.ls i/k(-1)=c_i(1)*i(-1)/k(-2)+c_i(2)*(ur_star-ur)/ur+c_i(3)*.125*q/q(-8)+c_i(4)+ec_i
genr ec_i=resid
This step can be bypassed if you are sure that the present equation is correct, and you know the estimation results.
However, we advise it as it displays the situation and helps you control the process.
o Transfer the coefficients into the vector elements, making any desired change in the values. Here we want to
set the tension coefficient to 0.5.
p_i(1)=c_i(1)
p_i(2)=0.50
p_i(3)=c_i(3)
299
o Replace all “c_ “coefficients by “p_” parameters, except for the constant term (this will allow EViews to
estimate the equation, a process which require at least one coefficient). 111
o Estimate the equation using the normal name.
o Store the new residual value.
equation eq_i1.ls i/k(-1)=p_i(1)*i(-1)/k(-2)+p_i(2)*(ur_star-ur)/ur+p_i(3)*.125*q/q(-8)+c_i(4)+ec_i
genr i_ec=resid
_fra_1.merge _eq_i1
Of course, we can modify as many coefficients as we want, but not the constant term if we want the residuals to add
up to zero, and EViews to recognize the equation as econometric.
To apply this technique, we can consider two general strategies.
• Start with no changes, and transfer in succession each equation which requires it.
• Copy at the beginning all equation-building elements into the forecasting program, with no value changes (all
the “c_“ elements are transferred into equivalent “p_”s). The specifications have changed, but not the
numerical properties.
• The requested paragraphs will be changed in turn later, as needed. This calls for an initial investment, but
makes the following changes easier and less error prone.
As you can guess, we favor the second option. But one should consider the share of equations he contemplates
changing112.
Note: when you are testing new values, you should keep the old one as a comment, which will allow you to backtrack
if the situation deteriorates, and also to recall the level of your change.
8.5.5 UPDATING THE MODEL SPECIFICATIONS
The above tests, or other considerations like a desire to introduce a new theoretical feature, can lead the modeler to
update the model. This can represent a very simple change, like replacing the unemployment rate by its logarithm in
the wage equation, or a very complex one, like changing the whole production function.
This process can lead to the introduction of new variables, or completely new behavioral equations.
Once the new set has been established, two options are available:
• If the model specifications are produced by a single program, one can simply edit it. Of course, one should
keep a version of the old program, in case mistakes are made or the new model proves unacceptable, for any
reason.
• One can keep the previous model creation program (or programs) and create an additional updating program
applying the changes.
111
We have no simple solution if the equation has no constant term (as for the change in inventories in our example).
112
Which is difficult to forecast.
300
As you can expect, we favor the first option, which is both safer (recreating the model is completely safe forward) and
clearer (the specifications of the model can be observed by sight, especially if it is not too large).
You can also (if you choose that option113):
• Drop any kind of equation, using :
model_name.drop variable_name (for identities)
model_name.droplink equation_name (for estimated equations)
For instance you can specify:
_fra_1.drop GDP ‘ GDP is made exogenous
_fra_1.droplink eq_i ‘ eq_i, the equation for investment is dropped, and I becomes exogenous
• Replace the formula using:
model_name.replace variable_name (for identities)
model_name.replacelink old_equation_name new_equation_name (for estimated equations)
_fra_1.replace GDP = ……
_fra_1.replacelink eq_i eq_i_new
• Finally (the most important) one can replace one variable name by another in all model occurrences.:
113
Remember that in our opinion, the most efficient method for editing a model is to destroy the previous version and
create a new one from scratch.
301
model_name.replacevar old_variable_name new_variable_name
For instance:
_fra_1.replacevar GDP Q ‘ replaces GDP by Q in the whole model (but not in estimated relations)
There are at least three interesting applications.
o You receive a model from some source and you want to change the names to ones you are familiar to.
o The names follow a language and you want to change it. For instance moving from English to French you
want to replace FD by DF (“demande finale”).
o You obtained a model from the World Bank using its complex names (like “ny_gdp_mktp_kn” for GDP) and
want to change it to plain “Q”.
The good point of this feature is that you can switch between versions, provided you follow the right order. So you can
work with modelers using World Bank names, and exchange the same model, each partner using his own names.
You have to be careful, however. If in your model QV stands for GDP at current prices, and Q at current prices, QV
should be processed first. And if you have specified QV_RES as a residual, it should be treated at the very first.
This means that this method applied essentially to models with names long enough, to avoid mix-ups.
It should be clear however that to follow the process, using programs is mandatory.
8.5.6 MODEL PROTECTION
One can protect a model to avoid modifying or deleting by accident.
This requires access to the Protect / Unprotect item in the Proc menu of a given model.
In case of protection, a password must be specified to update it.
Apparently this feature is only available by menu.
8.6 SOME RECENT USEFUL FEATURES TO BETTER MANAGE SIMULATIONS
Since Eviews 9, features have been added which do not modify the computing process, but make its management
easier. We will present them using the model we have just described.
8.6.1 THE GRAPH OF DEPENDENCIES
You can now asks for a graph showing the links between model elements.
This obtained by
302
Model>View>Dependency Graph
Arrows allow to observe the nature of links (simultaneous, lagged or both) according to a color.
This feature is just right for this size of model. It would have no use for a very small one, and be too complex for a
large one. We shall prove this later.
This graph evidences the density of the links (for instance the role of Q and FD). The only drawback is that it does not
take into account the economic implications. We think the graph we have designed is more helpful in presenting the
model structure (but it is longer to produce).
8.6.2 INTERPRETING THE LIST OF VARIABLES
For a given variable, EViews allows to identify the elements it depends on, and influences (only endogenous of
course). The causality path can then be followed further for one of these elements.
Let us apply this feature to Q (GDP)
303
These variables depend on Q:
These variables determine Q:
8.6.3 THE “PRINT”DISPLAY
To get the text of the model in the best editable form, one should call for the Print View” item in View menu of the
model. You should “include identities” and display coefficients as
“Fixed decimal”.
304
The estimated coefficients will be replaced by their numerical values.
Later you can add equation numbers. For that, we propose is the following sequence, a little long but not error prone.
o You copy the text to Excel (or equivalent).
o Most of the equations have the endogenous at the left. For the others (like Dlogs, maybe 5% of the total) you
type the name of the variable before the expression.
o You insert a column with numbers starting at 1.
o You sort the sheet according to the formulas.
o You delete the blank lines at the end.
o You insert two columns of “[“ before the numbers and “]” after.
o You sort the page according to the (now odd) numbers.
o You delete this first column of numbers and insert a new one between the brackets.
o You delete (by hand) the names you have inserted in the “5%” formulas.
o You select the full zone and “Copy special” it to Word (or equivalent) using the “a” option (the last one).
o You replace the tabulations (hidden but accessible as “^t”) by a null element or a blank . By the way, the item
“insert a jump to the next line” is accessible as “^p”. This can allow you to transform a column of items into a
line, separated by blanks or commas.
Et voilà!
8.6.4 EQUATION SEARCH
One can examine the set of equations, looking for the presence of a given variable (or aa particular equation).
This is done under the “Equations” view using the “find” button at the end of the line.
One can limit the search to endogenous / exogenous, or links.
8.7 RATIONAL EXPECTATIONS
EViews calls them « model consistent expectations » and their rationale is a little different from the usual one.
8.7.1 THEFRAMEWORK
In a rational expectations framework, we suppose that some of the agents (at least) are aware of the future
assumptions, and are able to evaluate their influence on some (at least) of the model variables. This does not need the
305
knowledge of the actual model equations, just the mathematical « application » from the assumptions to the
endogenous.
As they will use this knowledge in their present decisions, some of the present variables will depend on future values,
either of assumptions or endogenous elements depending on these assumptions.
To take into account rational expectations one does not need to believe in them. Interpreting the differences in
economic behavior (and their consequences for the equilibrium) between forward and backward looking agents is
quite interesting if only from a theoretical point of view. The following example will shed some light on this point.
8.7.2 CONSEQUENCES FOR MODEL SIMULATIONS
In this context, we can no longer compute the solution for each period separately, moving from the past to the future.
The solution for a given (future period) will depend on values for which the solution has not been yet obtained.
This introduces two problems:
• Finding a way to take into account future values, belonging to the forecasting period.
• Finding a way to end the simulation, the last simulations taking into account values outside the forecasted
period. This calls for the definition of « terminal conditions ».
8.7.2.1 Inside the forecasting period
The problem is purely technical. Two strategies are considered at present.
• The Fair – Taylor algorithm: it solves the model for each period in succession (using the Gauss-Seidel, Newton
or Broyden algorithm each time), then goes back to the first one and iterates (in a Gauss – Seidel way) until
convergence of the whole system. This is the technique used by EViews.
• The Laffargue (1990) algorithm: basically, it adds a time dimension to the model equations (duplicating them
as many times as there are periods) and solves the associated model as a whole, using the Newton algorithm
with possible improvements (using the fact that the matrix is band diagonal, the number of non-zero
matrixes on a given line depending on the number of backward and forward lags).
Obviously, applying the last algorithm was only made possible by the increase in the computing power of computers,
and more importantly of the memory capacity, as a very large set of values has to be managed at the same time.
8.7.2.2 Outside the forecasting period
EViews provides the usual options for terminal conditions, which are controlled by
t=arg
in the SOLVEOPT statement
arg can take the values:
306
"u" (user supplied – actuals, default)
"l" (constant level)
"d" (constant difference)
"g" (constant growth rate)
8.7.3 TECHNICAL ELEMENTS
EViews will detect automatically the presence of future endogenous variables, and will switch to forward looking
algorithm (actually to Fair-Taylor). One can use the same statements as usual, or change the terminal conditions from
default values, if deemed useful.
8.7.4 OUR EXAMPLE
In our example, we shall limit the case to the investment equation.
In the present model, we use:
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*UR+C_I(3)*.125*Q/Q(-8)+C_I(4)
Dependent Variable: I/K(-1)
Method: Least Squares (Gauss-Newton / Marquard
Date: 03/25/20 Time: 17:42
Sample: 1980Q1 2004Q4
Included observations: 100
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*(UR_STAR-UR)/
+C_I(4)+EC_I
Coefficient Std. Erro
C_I(1) 0.944504 0.013304
C_I(2) -0.002560 0.000898
C_I(3) 0.042822 0.013585
C_I(4) -0.004488 0.001676
R-squared 0.985646 Mean depe
Adjusted R-squared 0.985198 S.D. depen
S.E. of regression 0.000260 Akaike info
Sum squared resid 6.51E-06 Schwarz cr
Log likelihood 685.4474 Hannan-Qu
F-statistic 2197.406 Durbin-Wa
Prob(F-statistic) 0.000000
Now we shall consider four options:
The basic case (full backward looking)
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*UR+C_I(3)*.25*(Q-Q(-4))/Q(-4)+C_I(4)
307
A mixed backward – forward looking case
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*UR+C_I(3)*.25*(Q(2)-Q(-2))/Q(-2)+C_I(4)
A full forward looking case
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*UR+C_I(3)*.25*(Q(4)-Q)/Q+C_I(4)
Another full forward looking case, extending the knowledge to a longer horizon
I/K(-1)=C_I(1)*I(-1)/K(-2)+C_I(2)*UR+C_I(3)*.05*(Q(20)-Q)/Q+C_I(4)
The global influence of Q is the same in each case, only the lags differ.
8.7.5 THE TEST
We shall consider that the State increases its demand by 1% of GDP (as in the previous shocks) but only from 2010 to
2040 (considering perhaps that extending the decision would lead to an unacceptable debt level). The length of this
period might seem too high, but it will be necessary for the analysis114.
Of course, one could also apply the technique to other elements: employment and change in inventories in particular.
One should be able to do it easily.
8.7.6 THE RESULTS
We shall concentrate on a few graphs. All changes are measured in percentage deviations from the baseline.
114
In fact, this decision will lead to an increase in the debt primary of the State by 40 percentage points of GDP
(excluding the additional interest). The level of debt is certainly unbearable, reducing the interest of the study,
particularly concerning the reaction of other agents.
308
2.0
The change in GDP
1.6
1.2 Full backward
Medium forward
in points
Full forward
0.8 Far forward
0.4
0.0
-0.4
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
The change in investment
3
Full backward
2 Medium forward
Full forward
Far forward
1
In points
-1
-2
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
309
1.2
0.8 The change in the rate of use
0.4
In points
0.0
-0.4 Full backward
Medium backward
Full forward
-0.8 Far forward
-1.2
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
The 20 years ahead case
1.6
1.2
Capacity
0.8 Investment
GDP
In points
Rate of use
0.4
0.0
-0.4
-0.8
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
We can see very clearly generally logical elements:
• The more the investment decision is based on future GDP, the more investment adapts in advance.
• This is also true if the horizon increases. In this case, the weight of the periods associated with a change in
decisions goes up.
• This produces a smoother adaptation to a change in Government decisions, in both events. The intensity of
the cycles decreases, and they converge faster.
• With a 40 year horizon (case 4), the output gap does not change too much. Its higher values appear before
the changes, when the investment decisions (and the resulting capital changes) are made according to the
future, not the present, conditions. The negative overshooting of capacities observed in the last period is not
present, and the convergence of the rate of use is monotonous.
310
• The changes appear sooner than the shock than the forward horizon in the equation itself. The early
response of forward looking firms is taken into account even earlier by other firms, introducing a cascading
effect (rather limited, however).
8.8 STOCHASTIC SIMULATIONS
For scientific purposes but also the analysis of a given model, the uncertainty of the model in forecasting may be
directly measured. This error can come from several sources:
8.8.1 PURELY STATISTICAL ERRORS
This means
• The presence of a random residual term in behavioral equations
• The subsequent uncertainty on coefficients, even with an exact specification.
When a model is produced, some of the equations are estimated using an econometric technique, from simple
ordinary least squares to more sophisticated techniques like cointegration. This process provides:
• An estimate of the standard error of the estimation.
• Estimates on the standard errors of the estimated coefficients.
The precision of these two elements depends on sample size, in a different way.
• A larger sample should produce better estimates of coefficients, and reduce their error.
• But for the estimation residual, once the sample is large enough for estimation to be applied, enlarging it
provides essentially a better knowledge of its distribution (in practice its standard error). This is due to the
fact that even with a large sample, introducing new explanations beyond five or six variables leads very
quickly to collinearity problems, and the imagination of the modeler will dry up anyway.
8.8.2 FORMULATION AND ASSUMPTION ERRORS
Actually, these two errors are only half of the types facing model forecasts. One must also consider
• The fact that the estimated formula itself could be wrong. Even if a formula passes all tests, one can always
find another formula (sometimes with very different properties) which will show the same statistical quality,
or better. And even if the formula applied to the past, the agents could very well modify their behavior later.
• The error on the assumptions used for the forecast, for which the model should not be made responsible.
8.8.3 CONSIDERING THE ERRORS
In practice only the last first types can be measured without too much difficulty, using statistics (discarding the last
two): the estimation of coefficients gives under some assumptions an estimation of the distribution of estimation
residuals, as well as of the distribution of coefficients themselves. The moments of forecasted variables then can be
computed, either:
• By drawing at random a sample of residuals, undertaking the associated projections, and observing the
statistical characteristics of results (a « Monte Carlo » technique).
• By determining, analytically or numerically, the linear transformation from residuals to the model solution,
and by applying this transformation directly to the law of residuals, to obtain the law of forecasted variables.
311
• More simply and under clearly less restrictive assumptions, by drawing randomly a sample from the
sequence of observed residuals (with or without reintroducing the elements selected into the pool). This
technique is "bootstrapping".
The same type of method applies to the uncertainty due to coefficients:
• By repeating enough times the following sequence of operations: drawing a vector of residuals on the
estimation period (using the estimated law or the “bootstrap" technique), simulating the model on this
period to obtain a new sample of coherent variables (in the model terms), re-estimating coefficients based on
this sample, and forecasting the re-estimated model. One will then measure the moments of the sample of
projections.
• By using the distribution of coefficients to generate a sample of models, which when forecasted will provide a
sample of variable values.
• By inferring analytically the distribution of the projected variable from the distribution of coefficients, and
using the matrix transformation (Jacobian) to link the variables to the coefficients.
Three main error types should be considered:
• The bias: in the case of a model with non - linear properties (relative to the endogenous, both instantaneous
and lagged), the mean of the solution will not be identical to the deterministic solution (obtained with a null
residual): this introduces a bias.
To improve the measurement of the bias, one can use antithetic residuals, by repeating the simulation using the same
residuals with the opposite sign. This eliminates the first order term, leaving only the second order one, the actual
source of the bias. The negative correlation between the results reduces the error on the sum.
The difference comes of course from the non-linearities of the equations, and any economic model presents non-
linearities. Obvious cases are the presence of variables at current prices, product of a variable at constant prices by a
deflator, or variables computed by applying a growth rate to their past value.
If one is only interested in this accuracy, a single experiment (with a large number of replications) will tell of the
stochastic improvement. If it is small, one can stick with deterministic (and cheaper) simulations. This test can be
made as soon as the model becomes operational, and applied to all subsequent uses.
One must be aware that coefficient uncertainty can introduce non linearities even if the model is linear relative to
coefficients (actually it is not linear to the set: variables + coefficients put together).
Let us take the simplest model (Q=GDP, C= private demand, g=public demand):
𝑸𝒕 = 𝑪𝒕 + 𝒈𝒕
𝑪𝒕 = 𝒂 ⋅ 𝑸𝒕
It can be written as:
𝑸𝒕 = 𝒈𝒕 /(𝟏 − 𝒂)
312
which is not linear relative to a.
• the standard error: this time we consider the distribution of solutions around their average: this criterion
will allow for assessing the reliability of results (starting possibly from the mean of stochastic simulations) by
evaluating a confidence interval, or range of possible values.
• The distribution itself: what we want is a graph of the probability distribution of the random solutions.
Obviously producing a consistent graph calls for more simulations than the previous cases. One interesting
message is its symmetric character Actually EViews takes into account dissymmetry, even without a graph, by
producing a confidence interval eliminating the highest and lowest 2.5% of the values.
The analysis of the uncertainty is not necessarily limited to values of variables. It also can be measured on multipliers
(or derivatives of variables relative to the exogenous), therefore on the efficiency of economic policies, or on
eigenvalues of the structural form, thus on the probability of model convergence (see the following paragraph).
Finally, one could consider the covariance between equations. Cointegration should eliminate the problem, as it
individualizes a group of variables or concepts which evolve independently from the rest of the model.
If one considers also the error on coefficients, the process is a little more complex. Again, the sets of cointegrating
equations are independent, but they always use more than one coefficient (one has to consider the dynamic equation
and the long-term relationship). In this case, it is necessary to take into account the correlation between coefficients,
and the drawing will call for a multivariate normal law, with a non-diagonal covariance matrix.
The two last forms of error, on the functional form and the assumptions, are more difficult to assess. However:
• If we have recent information on some variables, the main aggregates (supply-demand balance, inflation) and
especially on the model assumptions (external environment, public policy), we can simulate the model over
these periods and observe if the evolution of the main aggregates is correct.
A rough evaluation of the functional error can be obtained by reducing the estimation sample, and simulating the
model over the remaining periods (as in an ex-post forecast). Then the size of the residuals can be compared with the
theoretical standard error obtained from the previous method (which supposed the use of the true equations). If the
error is much higher than the standard error (for instance systematically more than twice its value) one can surmise
that the estimated functions are unable to forecast the future, and the functional form must be wrong (or the agent
has changed his behavior, which is in fact the same). This technique is similar to the Chow forecast test.
This technique does not face the critique presented above, as we do not consider getting back to the choice of model.
• As to the error on assumptions, it is outside the responsibility of the model (only of the forecast producers).
But an estimate of it can be obtained by applying an ARMA process to the external assumptions (not the
instruments), and observing the reliability of the associated simulation. The error will then supposed to be
the part of the assumption which cannot be forecasted.
8.8.4 THE INTEREST OF THE TECHNIQUE
Stochastic simulations have a cost, and an advantage. The cost is the increased length of computations.
To get an accurate measure of the distribution, a minimum of 500 replications is considered necessary, 1000 or even
10000 being a better figure. This is clearly the main reason for the delay in the use of this technique: 50 years ago,
solving 1000 times even a small model could take several hours on a mainframe computer. Now the price can be
afforded in most cases: on an average microcomputer, the simulation over 20 periods of the 705 equations model
MacSim-CAN takes less than one second. Multiplied by 1000, we get about 15 minutes, a reasonable figure which
might look large however to contemporaneous modelers, especially if simulations are part of an iterative process
looking for the best forecast.
313
8.8.5 BACK TO OUR EXAMPLE
We shall solve the _fra_1 model over 100 periods, starting from 2005Q1.
The only necessary change concerns the solveopt statement, where « d=d » is followed by « d=a »:
In addition, the solveopt option « r=integer » controls the number of replications (default: r=1000), and « b=number »
controls the confidence interval (default: b=.95)
_fra_1.solveopt(n=t m=1000,c=1e-6,o=g,d=d,s=a)
_fra_1.solveopt(n=t m=1000,c=1e-6,o=g,d=d,s=a,r=500,b=.98)
Actually, the « a » option is not the only one associated with stochastic simulations. It is the one which gives the most
information, including the confidence intervals, while «m» gives only the mean, « s » only the standard deviation and
« b » the means and the standard deviation.
In our case, the solution will create for each variable in the model:
• A series for the mean, adding « m » to the name of the simulated variable. For Q and a current suffix of F, the
name of this variable will be « Q_FM ».
• A series for the standard error, using « S », such as Q_FS
• Series for the higher and lower bounds, using « H » and « L »., such as GDP_FH and GDP_FL
8.8.6 THE RESULTS
We shall perform a stochastic simulation on our small model, under the same conditions as the deterministic forecast,
which we will actually conduct again, to measure the error it introduces.
As the model and the application are not changed from the previous edition, we will be using the same graphs, more
explicit in our opinion.
The program is very similar to the previous one, except for the line:
_fra_1.solveopt(n=t m=1000,c=1e-6,o=g,d=d,s=a)
The following graphs * show:
• For GDP, the evolution of its higher bound, lower bound and deterministic value, as a ratio to the average
value.
• The ratio of the standard error to the average value, for a set of variables. Two graphs are actually provided,
one with 100 replications, the other with 1000.
314
1.06
Gross domestic product
1.04
Ratio to the mean value
1.02
1.00
Low er stochastic value
0.98 Deterministic
Mean stochastic value
Higher stochastic value
0.96
0.94
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
1.04
1.03
1.02 Gross domestic product
Ratio to the mean value
1.01
1.00
Lower stochastic value
0.99
Deterministic value
Mean stochastic value
0.98 Higher stochastic value
0.97
0.96
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
315
1.04
1.03
1.02 Gross domestic product
Ratio to mean GDP
1.01
1.00
0.99 Lower stochastic vallue
Deterministic value
Mean stochastic value
0.98
Higher stochastic value
0.97
0.96
0.95
2010 2020 2030 2040 2050 2060 2070 2080 2090 21004
The standard errors
.035
.030
GDP
Ratio to the mean value
.025 Imports
Final demand
.020 Exports
Capital
.015
.010
.005
.000
2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
In the above graph, one will find from top to bottom: Imports, exports, capital, GDP, final demand.
In addition, the following table gives the average relative difference from the mean, compared with the error obtained
from the ex-post simulation on the past.
316
Stochastic Ex-post
Variable error error
CAP 2.85 1.69
CO 1.31 1.43
FD 1.38 1.36
I 4.36 2.85
K 2.85 2.70
LE 1.68 1.68
LED 1.59 1.36
LT 1.28 1.29
M 3.23 2.81
Q 1.62 1.36
RHI 1.31 1.30
UR 1.35 1.00
X 2.99 1.77
One can observe:
• That the error is relatively small.
• That the error increases at first for all variables, as the sources increase.
• That it converges in the medium-term, faster for the trade elements (which use a cointegrating framework).
This is consistent with stationarity properties.
• That the evolution is much smoother for capital (its values are highly correlated).
• That trade elements show the highest variability (and are correlated). This correlation does not come from Q.
• Comparing the ex post error, we can observe that the stochastic error is generally larger, in particular for the
capacity variables. This can be due to the fact that the sum of the residuals is no longer zero, which helped
the solution to get back to the baseline in the end.
8.9 GOING FURTHER: STUDYING MODEL PROPERTIES
In addition to the use of direct simulations, the mathematical functions provided by EViews allow to apply more
complex techniques.
8.9.1 EIGENVALUE ANALYSIS
This technique can help to understand the dynamics of model convergence or divergence, and interpret them in
economic terms.
Its application calls for the model to be linearized around a reference solution (now stands for the difference
between two simulations):
𝜟𝒚𝒕 = 𝑨𝒕 ⋅ 𝜟𝒚𝒕−𝟏 + 𝑩𝒕 ⋅ 𝜟𝒙𝒕
317
If A is supposed to be constant with time, a shock x undertaken at time t (only) on the vector of the exogenous x will
lead at period t+k to a change in y:
𝚫𝒚𝒕+𝒌 = 𝑨𝒌 ⋅ 𝑩 ⋅ 𝚫𝒙 𝒕
and long-term consequences will depend therefore on the evolution of powers of A k, when k tends to infinite. Three
cases can occur:
Ak decreases to 0.
Ak increases to infinite (or at least one of its terms does).
Ak stabilizes at a finite, non-null limit.
Each of these three cases associates to the measure of the spectral radius of A (larger modulus of its eigenvalues): this
radius will have to be respectively lower, higher or equal to unity. This will lead us to reason in eigenvalues, by stating:
𝐴 = 𝑉 −1 ⋅ 𝛬 ⋅ 𝑉
(𝛬 diagonal, containing eigenvalues of A, and V base of associated eigenvectors).
One obtains then:
𝑉 ⋅ 𝛥𝑦𝑡+𝑘 = 𝑉 ⋅ 𝐴𝑘 ⋅ 𝛥𝑦𝑡 = 𝛬𝑘 ⋅ 𝑉 ⋅ 𝛥𝑦𝑡 = 𝛬𝑘 ⋅ 𝑉 ⋅ 𝐵𝑡 ⋅ 𝛥𝑥𝑡
Therefore, for a given shock on x, its dynamics can be studied by decomposing first period effects 𝐵𝑡 ⋅ 𝛥𝑥𝑡 on the
eigenvector base of A, each element of this decomposition evolving then with time proportionally to the
corresponding eigenvalue. Evolutions can be:
• Divergent, convergent or stationary, according to the modulus of the eigenvalue (higher, lower or equal to
one).
• Monotonous or cyclic, according to the real or complex nature of the eigenvalue.
Actually, the stability of A over time is much more assured if one considers relative variations:
𝜟𝒚𝒕 /𝒚𝒕 = 𝑨𝒕 ⋅ 𝜟𝒚𝒕−𝟏 /𝒚𝒕−𝟏 + 𝑩𝒕 ⋅ 𝜟𝒙𝒕 /𝒙𝒕
318
This property has two explanations
• Most estimated equations are formulated in logarithms.
• By dividing the variation by the value, the result should be stable with time.
Indeed, this modification is equivalent to replacing each variable by its logarithm in model equations.
This does not change fundamentally the above considerations, except that convergence and non-convergence will be
associated now to relative variations: an eigenvalue larger than one will mean that it is the relative difference to the
variable to the base value which will tend to infinite.
As most variables grow with time, this criterion for convergence will be less restrictive than the first. But the following
question must be considered: must one associate to divergence a growing absolute variation or a relative one? For
instance, let us suppose households benefit from a tax exemption, at two different periods in time. To compare the
two, do they consider levels at constant prices, or the ratios to their present income? The truth is certainly between
the two.
8.9.1.1 First example: a very simplified model
As an example, let use an extreme simplification of our usual model.
(1) 𝐼𝑡 = 𝑎 ⋅ (𝑄𝑡 − 𝑄𝑡−1 ) + 𝑏
Firms invest to adapt their productive capacities to the evolution of demand.
(2) 𝐶𝑂𝑡 = 𝑐 ⋅ 𝑄𝑡 + 𝑑 ⋅ 𝑄𝑡−1 + 𝑒
Households consume a share of the present and previous production levels (introducing a lag makes the example
more interesting).
(3) 𝑄𝑡 = 𝐼𝑡 + 𝐶𝑂𝑡 + 𝑔𝑡
Production adapts immediately to demand, composed of investment, consumption, and State demand.
Reasoning in variations, one gets:
319
ΔI 0 0 a ΔI 0 0 −a ΔI 0
(ΔCO) = (0 0 0) (ΔCO) + (0 0 c ) (ΔCO) + (0) Δg t
ΔQ t 1 1 0 ΔQ t 0 0 0 ΔQ t−1 1
or
𝛥𝐼 0 0 (𝑐 − 1) 𝛥𝐼 𝑎
(𝛥𝐶𝑂) = 1/(1 − 𝑎) (0 0 𝑐(𝑐 − 𝑎)) (𝛥𝐶𝑂) + 1/(1 − 𝑎) (0) 𝛥𝑔𝑡
𝛥𝑄 𝑡 0 0 (𝑐 − 𝑎) 𝛥𝑄 𝑡−1 1
This process contains a single lagged, variable, and a single eigenvalue:
𝝀 = (𝒅 − 𝒂)/(𝟏 − 𝒂 − 𝒄)
and the multiplier at period k of a unitary change in State demand at period 0 will be:
𝟏/(𝟏 − 𝒂 − 𝒄) ⋅ 𝝀𝒌
At this level of simplification a formal interpretation is still possible. The eigenvalue adds two lagged dynamic effects:
• The sensitivity of consumption to past households income: d.
• The sensitivity of investments to the previous level of production, representative of capacity levels acquired
earlier: - a.
These two effects are amplified by the multiplier mechanism, due to the present influence of investment and
consumption on present production: 1 / (1 – a -c).
One can therefore logically suppose:
• That a, c and d are positive.
• That 0 < c + d < 1: households consume a part of their additional revenue, itself lower than additional
production.
Our purpose is not to theorize on these formulations. However, we can identify some special cases:
• d + c = 1 (unitary impact of Q on consumption); then the eigenvalue is 1, and while the dynamic process does
not converge, nor does it diverge: the increase in consumption maintains the level of production in the next
period, which stabilizes investment, which has no further impact on growth Q. A demand shock at period 1
will maintain its effects indefinitely.
320
In other words, the multiplying effect: 1/(1-a-c) is compensated in the next period by the reducing effect: d-a or (1-a-c)
– (1- d-c) lower than (1-a-c) if d+c<1.
Or: with c+d<1, the multiplying effect will spread to the two periods with an intensity lower than the initial shock. The
process will converge.
• c = 0 (consumption independent of Q); then the eigenvalue is - a /(1 - a), and the dynamics converge if a < 1 /
2. Otherwise, getting back to the initial level of investment represents more than half the previous increase
of production, and the multiplier more than doubles this change, which leads to an alternate diverging
process.
• a = 0; then the eigenvalue is c, leading to a general convergence if c 0,1
More generally, when 0 < a < 1 and 0 < c +d < 1 the process will converge if:
𝒄 + 𝒅 > 𝟐𝒂 − 𝟏
Which is a less restrictive condition than a > 1 / 2 (if c=0), because the positive effect of the consumption can then
compensate the alternate divergence presented higher.
We can see that even for a very simple example the formal interpretation can present some complexity. For more
realistic cases, one will have to resort to numerical techniques, interpreting results in terms of economic mechanisms
by associating them to variables of the model.
For this one can (following a methodology introduced by Deleau and Malgrange (1978)) eliminate for each variable in
turn its lagged influence (by deleting the associated column and line of A, then by re-computing the eigenvalues). In
the most favorable case, a single eigenvalue will disappear, creating a natural association to the eliminated variable. In
other cases research will be more difficult, and might need the simultaneous elimination of groups of variables, which
one will associate to a group of eigenvalues of the same dimension.
Our example is too simple to present this method, as it contains only a single lagged influence (and therefore a single
eigenvalue). But the analysis we have just made gives a basic illustration of the method. To extend the dynamic
characteristics of the above model, one could have added an equation for capital and productive capacity.
8.9.1.2 Second example: the role of the Taylor rule
We shall now explain, on a very crude example, the role of the real element of the Taylor rule in our type of model.
Let us remind that the Taylor rule represents the behavior of a central bank which uses the interest rate to target
inflation (and also reduce its volatility). See TAYLOR JB. (1993).
It will increase the rate if:
• Inflation is higher than the target (2% for the European Central Bank).
• Tensions appear on productive capacities (the “output gap”) signaling potential inflation.
This action should reduce demand (investment and consumption), activity and therefore inflation.
321
The original formula is:
𝑰𝑹𝒕 = 𝒕𝒙(𝒑𝒕 ) + 𝟎. 𝟓 ⋅ (𝒕𝒙(𝒑𝒕 ) − 𝒕𝒙∗ (𝒑𝒕 )) + 𝟎. 𝟓 ⋅ 𝒈𝒂𝒑
With
• IR the interest rate
• P the price index (deflator of consumption?)
• tx*(p) the inflation target
• gap the output gap indicator, normally the relative difference between the actual and potential values of
output.
Let us consider the following model, measured in real terms, a simplification of our example without external trade
(1) 𝑖𝑡 /𝑘𝑡 = 0.65𝑖𝑡−1 /𝑘𝑡−1 + 0.10[(𝑞𝑡 − 𝑞𝑡−1 )/𝑞𝑡−1 + (𝑢𝑟𝑡 − 𝑢𝑟 ∗)/𝑢𝑟 ∗] + 𝑑 + 𝑎(𝑢𝑡𝑡 − 𝑢𝑟 ∗)
(2) 𝑐𝑡 = 𝑟𝑡 𝑞𝑡 (1 − 𝑏(𝑢𝑟𝑡 − 𝑢𝑟 ∗))
(3) 𝑢𝑟𝑡 = 𝑞𝑡 /(𝑘𝑡 𝑝𝑘𝑡 )
(4) 𝑘𝑡 = 𝑘𝑡−1 (1 − 𝑑𝑑𝑡 ) + 𝑖𝑡−1
(5) 𝑞𝑡 = 𝑐𝑡 + 𝑖𝑡 + 𝑔𝑡
with the endogenous variables:
c household consumption
i productive investment
k capital
q GDP
ur the rate of use of productive capacity
the exogenous variables:
dd the depreciation rate
g government consumption
pk the productivity of capital
r the ratio of consumption to GDP
ur* the target rate of use
322
For the Taylor rule, we have only considered the role of the output gap, which we have represented (this can be
questioned) by the rate of use of capacities.
So we shall use the parameters:
a the sensitivity of investment to the rate of use
b the sensitivity of consumption to the rate of use
If we compute the dynamics of this model by differentiating its equations numerically,
𝛥𝑦𝑡 /𝑦𝑡 = 𝐴𝛥𝑦𝑡 /𝑦𝑡 + 𝐵𝛥𝑦𝑡−1 /𝑦𝑡−1 + 𝐶𝛥𝑥𝑡 /𝑥𝑡
where A, B and C are relative Jacobians supposed constant with time,
we get
𝛥𝑦𝑡 /𝑦𝑡 = (𝐼 − 𝐴)−1 (𝐵𝛥𝑦𝑡−1 /𝑦𝑡−1 + 𝐶 𝛥𝑥𝑡 /𝑥𝑡 )
To describe the dynamic process, we must now compute the eigenvalues of (𝐼 − 𝐴)−1 𝐵, a 6x6 matrix.
But of course, independently from the values of a and b, we should get only two non-zero eigenvalues, as there are
only two equations with lagged elements (and three such elements).
Starting with a=b=0, we get two complex eigenvalues, with
Modulus: .977 period: 18.8
Now if we put the value of a to 0.03, consistent with our Taylor rule (coefficients of -.06 in the investment equation,
and .5 in the Taylor rule itself) we get:
Modulus: .872 period: 20.6
The period not much affected, but the convergence is much faster. After 10 periods an initial shock is reduced by 75%
(instead of 21%).
Considering b (the role of the interest rate in consumption) we get only a small increase in the period, and in the
speed of convergence.
323
In conclusion: the dynamics are almost fully explained by the real elements of capital formation, including the real
interest rate.
The elements allowing to product the above study can be very simply implemented using EViews.
8.9.1.3 Steady state growth paths
Let us go further in determining the conditions for the existence of a path of model solutions, such that the whole set
of variables grows at a constant rate (a « steady state growth path »). The application of these conditions to
formulations of the model will lead, by editing some equations, to the production of a long-term model, with
potentially very different causalities compared to the original model. One will study also the stability of this path, by
introducing an initial gap through a one-time shock, and observing if the long-term dynamics leads back to the path or
away from it (again, reasoning in absolute or relative terms).
These techniques can be clarified through our simple model. Let us recall its formulations (in EViews format):
[1] cap = pk * k(-1)
[2] log(prle_t) = 9.8398+ 0.009382* (t - 2002) + 0.02278* (t - t1) * (t<t1) + 0.01540* (t - t2) * (t<t2)
[3] DLOG(x) = 0.4940*DLOG(wd) - 0.09142*res_x( - 1) + 0.01960+ ec_x
[4] ur = q / cap
[5] i/k( - 1) = 0.944*I( - 1)/K( - 2) - 0.00256*(ur-ur_star)/ur + 0.0428*.125*q/q( - 8) - 0.00448+ ec_i
[6] led = q / prle_t
[7] DLOG(le) = 0.2954*DLOG(led) + 0.1990*LOG(led( - 1)/le( - 1)) + 0.0004748* + 0.01759**((T = 1968.25) - (T =
1968.50)) + ec_le
[8] ic = ct * q
[9] lt = le + lg
[10] rhi = wr * lt + r_rhiq * q
[11] co = rhi * (1 - sr)
[12] ih = r_ih * rhi
[13] ci/q( - 1) = -0.02680*(T = 1968.25) + 0.6128*ci( - 1)/q( - 2) - 0.0021490 + 0.2193*@PCH(q) + 0.1056*@PCH(q( - 1))
+ 0.05492*@PCH(q( - 2)) + 0.03918*@PCH(q( - 3)) + 0.03026*@PCH(q( - 4)) + ec_ci
[14] fd = co + i + gd + ci + ih
[15] td = fd + ic
[16] DLOG(m) = 1.676*DLOG(td) + 0.2133*DLOG(ur) - 0.1057*DLOG(compm) - 0.1028*res_m( - 1) - 0.2757+ EC_M
324
[17] q + m = fd + x
[18] res_x = p_x(1) * log(x / wd) + 0.9815 * log(ur) + 0.001441 * (@trend(60:1) * (t<=2004) + @elem(@trend(60:1) ,
"2004q4") * (t>2004))
[19] res_m = log(m / (fd + ct * q)) ‘ 1.517 * log(ur) +-0.552 * log(compm) + 0.00579 * (@trend(60:1) * (t<=2004) +
@elem(@trend(60:1) , "2004q4") * (t>2004))
[20] k = k(-1) * (1 - dr) +i
If we call
• txq the growth rate of quantities
• txn the growth rate of populations.
We get:
[1] cap = pk * k / (1+txq)
[2] log(prle_t) = 9.8398+ 0.009382* (t – 2002)
[3] log(1+txq) = 0.4940*log(1+txq) - 0.09142*res_x+ 0.01960+ ec_x
[4] ur = q / cap
[5] i/k*(1+txq) = 0.944* i/k*(1+txq) - 0.00256*(ur-ur_star)/ur + 0.0428*log(1+txq)- 0.00448+ ec_i
[6] led = q / prle_t
[7] log(1+txn) = 0.2954*log(1+txn) + 0.1990*log(led /le) + ec_le
[8] ic = ct * q
[9] lt = le + lg
[10] rhi = wr * lt + r_rhiq * q
[11] co = rhi * (1 - sr)
[12] ih = r_ih * rhi
[13] ci/q*(1+txq) = 0.6128* ci/q*(1+txq) - 0.0021490 + 0.2193*txq + 0.1056*txq + 0.05492*txq + 0.03918*txq
+0.03026*txq + ec_ci
[14] fd = co + i + gd + ci + ih
325
[15] td = fd + ic
[16] log(1+txq) = 1.676*log(1+txq) - 0.1028*res_m- 0.2757+ EC_M
[17] q + m = fd + x
[18] res_x = p_x(1) * log(x / wd) + 0.9815 * log(ur)
[19] res_m = log(m / (fd + ct * q)) ‘ 1.517 * log(ur) +-0.552 * log(compm)
[20] k = k / (1+txq) * (1 - dr) +i
First we can observe that in each equation, provided that all the explanatory elements grow at the theoretical rate,
the explained element will also. In particular, all additive terms should grow, taking into account the variables which
enter them, at this rate.
This means that there exists a consistent solution in which all elements grow at the theoretical rate. As the model has
only one solution, this is the solution it should reach, provided the error correction mechanisms lead in the right
direction (from the coefficients, we can suppose they do).
We can also build the long-term model, which follow a logic somewhat different from the dynamic one, as
disequilibria have been stabilized.
• From (20) we get I/K as the rate which allows to adapt capital to both growth and depreciation:
i/k = 1 - 1/ (1+txq) * (1 - dr)
• From (5) we get UR depending on the speed of adaptation. Even if adaptation was immediate (dynamic
homogeneity) UR would still depend on txq, as a higher expected growth calls for an effort in creating
additional capital, for use at the next period:
i/k*(1+txq) = 0.944* i/k*(1+txq) - 0.00256*(ur-ur_star)/ur + 0.0428*log(1+txq)- 0.00448+ ec_i
• From (18) we get RES_X.
• From (16) we get RES_M.
• From (9) we get LED/LE.
If adaptation was immediate the three would not depend on txq or txn. This would also be the case for UR if the
coefficients complied exactly with the error correction framework.
Here the gap will increase with the growth rate, and decrease with the speed of adaptation.
326
• From RES_X, UR and (18) we get X/wd and X:
• From RES_M, UR and (19) we get M/TD
• From UR, (1) and (4) we get Q/K;
• from (5) we get I/Q
• From (13) we get CI/Q
• From (6) and (7) we get LE/Q
• From (9) and (10) we get RHI as a term proportional to Q, plus wr*lg
• From (11 ) and (12) we get CO and IH as proportional to RHI, thus to Q and wr*lg
• From (17) we get FD as proportional to Q plus a term proportional to wr*lg, plus gd
• This gives M/Q
• This gives Q from (2)
• And all the other elements.
Finally, all the terms in the supply – demand equilibrium are defined as proportional to Q, except for:
• Household consumption and investment which are partially proportional to government employment.
• Government demand which is exogenous115.
The sum of these elements makes demand partially exogenous.
• Exports which are proportional to world demand.
We can see that the logic of the model causalities is quite different from the dynamic model. In particular,
o The rate of use depends on the growth rate of the economy and the depreciation rate.
o The gap between targets and actual values too (with UR as an eventual intermediary).
o Knowing this gap we can define all the real elements as a share of GDP, except for exports and an exogenous
element of demand. Let us call it dx.
The supply – demand equilibrium
Q + M = FD + X
becomes:
Q + a . (b . Q + dx) = b . Q + dx + c . wd
Q = ((1 - a) . dx + wd) /(1 – (1 – a) . b)
With a and b constant elements depending on exogenous assumptions and the growth rate txq (and also txn).
This means that if
• gd was defined as a constant share of demand
• lg was defined as a constant share of LE (or of LT = LE + lg)
(which seem logical and even necessary assumptions),
115
Of course, the assumptions will give them values proportional to Q.
327
the exogenous world demand level would define all elements in the local economy (as world inflation will drive
the local one with a more complex model).
8.9.2 THE CASE OF ERROR CORRECTION MODELS: A SIMPLE EXAMPLE
Let us now show how an error correction formulation makes the interpretation of model dynamics both easier and
more meaningful.
Let us introduce the following model
(1) 𝐶𝑂𝑡 = 𝑎 ⋅ 𝑄𝑡
(2) 𝑄𝑡 = 𝐶𝑂𝑡 + 𝐼𝑡 + 𝑔𝑡
(3) 𝐾𝑡 = 𝐾𝑡−1 ⋅ (1 − 𝑑𝑟𝑡 ) + 𝐼𝑡
(4) 𝛥𝐾𝑡 = 𝛼 ⋅ 𝑏 ⋅ 𝛥𝑄𝑡 + 𝛽 ⋅ (𝑏 ⋅ 𝑄𝑡−1 − 𝐾𝑡−1 )
We have defined an error correction model, where the target consists in making K proportional to Q (this equates to a
constant rate of use of capacities if one supposes constant the productivity of capital).
Let us consider the dynamics of this model, along the lines of the eigenvalue analysis presented earlier.
By eliminating CO and I, we can reduce the system into:
(1) 𝑄𝑡 = 𝑎 ⋅ 𝑄𝑡 + 𝐾𝑡 − 𝐾𝑡−1 ⋅ (1 − 𝑑𝑟𝑡 ) + 𝑔𝑡
(2) 𝐾𝑡 = 𝛼 ⋅ 𝑏 ⋅ 𝑄𝑡 + (𝛽 − 𝛼) ⋅ 𝑏 ⋅ 𝑄𝑡−1 + (1 − 𝛽) ⋅ 𝐾𝑡−1
𝛥𝑄 𝑎 1 𝛥𝑄𝑡 0 −(1-d) 𝛥𝑄t-1 1
( 𝑡) = ( )( )+( )( ) + ( )𝑔
𝛥𝐾𝑡 𝛼𝑏 0 𝛥𝐾𝑡 𝑏(𝛽 − 𝛼) 1-𝛽 𝛥𝐾𝑡−1 0
We can apply to this system the same technique as before, with normally two eigenvalues. The process might be a
little too complex for this simple formulation. We shall simplify it by supposing that the speed of adaptation is the
same for a new gap, due to a change in the target, and a pre-existing gap:
=
328
The system becomes
𝛥𝑄 1−𝑎 −1 −1 0 −(1-d) 𝛥𝑄t-1 1
( 𝑡)=( ) ( )( ) + ( ) 𝑔𝑡
𝛥KP𝑡 1 − 𝛼𝑏 1 0 1-𝛼 𝛥𝐾𝑡−1 0
with one eigenvalue:
l = 1 - + (d - ) / (1- a - b )
Let us interpret this element.
If d = the eigenvalue is (1 - ) This means an initial gap between the variable and its target will close with a reason
1 - The reason for this is clear. If we consider the equation system, the capital will depreciate at the speed d, which
is exactly the rate at which the firms « want » it to decrease if a gap has appeared in previous periods, due for instance
to a shock on g. The ex-ante variation will be the same as the desired one, and no additional mechanism will appear.
On the contrary, if d is higher than for instance, the natural decrease in capital will be higher than the desired one.
Firms will have to invest the difference, leading to an increase in capital made larger by the presence of the multiplier.
But for the process to diverge, the difference between d and must be large, or the multiplier effect very important
(which would lead to problems with any model).
It might look strange that an ex-ante decrease in capital (through a higher d) could lead to an ex-post increase; This is
because the behavior of firms as to capital is set by (2), taking into account the variation of production but not the
additional depreciation rate, which is compensated automatically through higher investment (thus production, and
investment again).
Let us now consider the long-term model.
It is now clear that all variables: CO, Q, I, g, CAP grow at the same rate on the steady state growth path. This rate (let
us call it q) is set by « g ».
The long-term model is
(1) 𝐶𝑂 = 𝑎 ⋅ 𝑄
(2) 𝑄 = 𝐶𝑂 + 𝐼 + 𝑔
(3) 𝐾 = 𝐾 ⋅ (1 − 𝑑𝑟)/(1 + 𝑞) + 𝐼
329
(4) 𝐾 ⋅ (𝑞 + 𝛽)/(1 + 𝑞) = (𝛼 ⋅ 𝑞 + 𝛽) ⋅ 𝑏 ⋅ 𝑄/(1 + 𝑞)
or
(1) 𝐶𝑂 = 𝑎 ⋅ 𝑄
(2) 𝑄 = 𝐶𝑂 + 𝐼 + 𝑔
(3) 𝐼/𝐾 = 𝑞 + 𝑑𝑟
(4) 𝐾/𝑄 = 𝑏 ⋅ (𝛼 ⋅ 𝑞 + 𝛽)/(𝑞 + 𝛽)
We observe that none of the causal explanations of the original model has changed. The ratio of production to capital
is fixed, actually based on the target in the original formulation. And investment maintains this constraint, by leading
to an implicit evolution of productive capacity equal to the growth rate (taking into account depreciation).
But one also can see that the target is reached only in special conditions. This is a standard characteristic for an error
correction model.
Equation (4) shows that the cases are:
𝛼 = 1 (dynamic homogeneity, gaps are closed immediately)
𝑞 = 0 (the target does not move)
This is true for any starting values, including those which meet K /Q =b: if q or are not zero, a gap will appear, which
will never close.
We shall stop here, as our purpose was only to illustrate the process, and show that error correction models make the
dynamics more interesting, as well as easier to study.
330
8.10 MISCELLANEOUS MODELLING FUNCTIONS
Versions 12 and 13 allow new features making the modeling process easier.
8.10.1 MODEL DEPENDENCIES
For a given model, it determines the elements it Influences or depends on.
The syntax is
Model_name.@depends(string)
Model_name.@upends(string).
For example, if the model “mod” contains the trade equations:
BAL=XV-MV
RX=XV/MV
Mod.@depends(“XV”) will give the string “BAL RX”
Mod.@upends(“RX”) will give the string “XV MV”
This feature can be useful to check that the requested variables are available, or
8.10.2 DEBUGGING TOOLS FOR EVIEWS PROGRAMS (“PROGRAM DEBUGGING”).
Until version 13, the main debugging possibilities were:
• Introducing “STOP” statements in the program.
• Using the mouse to select part of a displayed program and running it using “Run Selected”.
The new “Debug” option makes the process easier.
First one should set the breakpoints in the program.
click on the “Debug” item (right of “Run”).
Let us use the “pfor_b.prg” program for running the PIC_B model on the future, and concentrate on the lines 57 To 77.
331
Clicking on “79” we introduce a breakpoint (signaled by the red dot).
When we run the program under the “Debug” option it will stop at this point and give access to three items: Watch,
Breakpoints and Callstack.
“Watch” will give the value of the current parameters.
“Breakpoint” will give the list of current breakpoints (only one at present).
332
“Callstack” will give the list of currenly opened programs.
Of course, you can also acceess all the cuerrent elements and display their value.
8.10.3 PROGRAM DEPENDENCY TRACKING (“PROGRAM DEPENDENCY LOGGING”).
For instance, the program pdon_a.prg uses the file fra_xls and creates the file pic_q.wf1.
FILE: c:\eviews\cours_eviews_22\pic_eviews\fra_q.xls
(c:\eviews\cours_eviews_22\pic_eviews\pdon_a.prg: 34)
FILE: c:\eviews\cours_eviews_22\pic_eviews\pic_q.wf1
(c:\eviews\cours_eviews_22\pic_eviews\pdon_a.prg: 187)
9 CHAPTER 9: USING MODELS
In chapter 2 we began to present modelling applications. We are going now to be more precise, concentrating on
practical aspects.
9.1 OPERATIONAL DIAGNOSES
The first application we quoted, and the most natural apparently, is to request from the model a realistic diagnosis of
economic problems. This diagnosis can lean on scenarios or shocks.
The originator of this type of study can be:
• A governmental organization, such as the local Ministry of Finance, which wants to get information on the
future evolution of the economy it manages, or the efficiency of the policy decisions it contemplates.
• An international institution, such as the IMF or the OECD, or a Non-Governmental institute, which wants to
assess the future of the world economy.
• An academic research unit or individual, which wants to study the impact of policy decisions or structural
changes, according to a given theoretical model.
• A private firm dealing in economic services, which provides forecasts and studies to subscribers, either as a
publication or a dedicated study.
9.1.1 SCENARIOS AND THEIR DIFFERENT TYPES
Two types of scenario may be distinguished.
• Trend forecasts, which will use the most probable assumptions (in the case of a unique scenario), or a set of
possible assumptions, covering the scope of foreseeable evolutions.
333
Assumptions will often be developed by experts from outside the area of model building. Let us take as an example:
the forecasts of MESANGE, a French macroeconomic model considering policy decisions as exogenous, managed by
the French Ministry of Finance, in two of its Directorates: at INSEE, the National Institute for Statistics and Economic
Studies at the General Direction of Treasury and Economic Policy (DGTPE).
Variables describing foreign environment are valued through simulations of a set of tools:
o The MZE (Modèle Zone Euro or EuroZone model) model of the CEPII (a French economic institution
specialized in international studies), managed in the same institutions.
o The NIGEM world model, managed and leased by the British NIESR (National Institute for Economics and
Social Research).
o Results from economic forecasts of OECD, the UN and the IMF.
o Numerical evaluations made by experts in international economics and trade, without using any kind of
model.
o The observation of international developments, like the Asian financial crisis, and the assessment of potential
evolutions.
The government decisions are specified following discussions with the DGTPE and especially the people which
produce the numerical evaluation of next year’s budget.
As for assumptions on the evolution of the population and its structure, they will use the results of studies produced
by INSEE or Ministry of Labor demographists.
This type of scenario will give indications of the most probable future evolution, or on the range of foreseeable
evolutions, to the best knowledge of the available experts.
Production of the final forecast will require a large number of model simulations, accompanied by meetings and
discussions. It lasts usually several weeks (let us say 2 to 4). The improved speed of computation helps, but not too
much, as it is mostly used to improve the quality of the results.
• Normative forecasts, in which results are asked to meet a certain number of conditions. These constraints
are set before the development of assumptions.
This type of scenario describes a context in which efforts are made, relative to the foreseeable evolution of the
economy, to reach some targets (such as meeting the criteria for joining the European Monetary Union, usually called
the Maastricht criteria). It will essentially give information on the extent of expected difficulties, and efforts required,
associated to a possible path.
In general the client of this type of study will also be the State, or a public institution. In cooperation with the model
managers, it will set the constraints, and the policy elements used to meet them116. But non-governmental institutions
and researchers can also be interested in this type of issue, perhaps more than the production of a forecast as it
involves more economic reasoning.
If precise goals are set (such as a budget deficit of 3 GDP points), one can work either by a sequence of
approximations on the normal model, or by modifying model specifications by an explicit reversal of causalities. This
last technique introduces two problems:
116
However in some cases forecasters will also include in the potential set of instruments some assumptions which
are not manageable by the State, such as an exogenous change on the behavior of agents, or structural parameters
such as the productivity of factors. This is not unacceptable, provided the option is clearly displayed in the associated
publication. The initial question was then: how should the behavior of agents, or structural parameters, have to
change in order to reach a specific objective?
334
o A formal one: for the model to preserve a unique solution, the number of endogenous elements which are
now given a set value must be equal to the number of instruments that will allow to reach them. For
example, if one wants to balance both the commercial balance and the State budget, it will be necessary to
free two decision variables and two only, to avoid either impossibility or an indetermination 117.
o A technical one: often it will be too complex to produce an identified model which takes into account this
reversal explicitly, and if the solving algorithm requests such a formulation, it will be necessary to revert to
the other technique, determining the solution by a sequence of approximations. Let us suppose for instance
that we want to balance foreign trade by increasing subsidies on investment. To transform the model, one
will have to identify a chain of causalities leading from subsidies to foreign trade (which can be quite long)
and revert each one in turn, through the associated equation. To this obviously difficult if not unfeasible
technique, one can prefer a numerical, iterative one: computing outside the model an approximation of the
necessary subsidy amount, using it to solve the standard model, then correcting the change from the
observation of the remaining error. This sequence of approximations should decrease to an acceptable value.
We will propose soon an alternate method, which allows solving the model for given values of a number of
endogenous variables, provided the same number of exogenous variables are released. This method does not work all
the time, but if it does the solution is quasi-immediate, and can be applied under EViews without accessing the model
code. The program itself is provided, and can be adapted in a few minutes to any existing model managed under
EViews.
But in any case, results cannot be taken for more than what they are, the economic equilibrium associated to
assumptions deviating from normal evolution, whose probability of occurrence cannot be judged. To earn the right to
use this method, one has to display clearly and probably justify the options he has chosen.
Under those restrictions, one is not limited in the elements he can use. For instance one can modify the estimation
residual in the investment equation to measure by how much private firms should increase it (all things being equal)
for the trade balance to get back to equilibrium in the long run. The less controlled the instruments, the less the
simulation can be considered as a policy result. But this does not make the results (and the observation of causal
sequences) less interesting from a theoretical point of view.
9.1.2 MANAGING ACTUAL FORECASTS: MANAGING THE RESIDUALS
On the estimation period, the sensible option for residuals was of course to set them to the difference between the
actual values and the one given by the estimated formula. Now we have to decide on values, starting with the first
forecast period.
Two main (and extreme) options are available.
• Setting them to zero (in principle, the value with the highest probability).
• Keeping the last value in the estimation period (the most recent information we have).
In other words: should we rely on the recent past, or summarize all the past information we have, giving equal weight
to each element, independently from its age?
Each choice can be dangerous:
117
Except if the number of instruments is larger than targets, but limits are set on their variations, as well as a priority
order. Then each instrument will be used in turn, switching to a new one if the constraint has been reached. In
practice we will be left with as many unknowns as constraints. One can also create a synthetic instrument by applying
predefined weights to a larger set.
335
• The first will refuse to take into account recent elements, like a feature which appeared only in the last
periods, and could not be evidenced as too recent (otherwise it should appear in the actual formula). Also, it
will produce shocks on the initial growth rates, which will be difficult to explain. For instance if the last
residual on household consumption was 2%, a structural decrease of 1% (taking only into account the
formula and its variables, and excluding the residual) will be transformed into a growth of 1%.
• The second will violate the zero mean assumption, and suppose that a possibly high shock will be maintained
for the whole period of the forecast.
In our opinion, no option clearly dominates under all circumstances. One should first consider the profile of the last
residuals.
First, their size: obviously if the last residual is close to zero, the problem is minor.
This means of course that one will favor estimations with a low last residual. From a statistical point of view, the
standard error should not depend on the period, but the modeler will concentrate on the last values, the closest to
the computations he will have to manage.
If the residual is large (whatever its sign), one should look at the immediately preceding values. If they are too, there is
clearly a problem with the equation, and something should be done to reduce it. This is also true if they are smaller
but growing: a trend (or a variable with a trend) might be missing.
Maybe the best option is to decide that the last residual is indeed going to last for a while, but with an intensity
decreasing to zero.
But I any case one should first try to understand the reason for the value. This might make things clear as to the way
to manage it.
9.1.3 MANAGING ACTUAL FORECASTS: THE TARGETING OF SIMULATIONS
In the normative case, one has to manipulate the exogenous in order to reach a certain solution. But this happens also
in trend forecasts. One cannot expect the model to give spontaneously an image of the future consistent with what
we expect, especially in the short run where we already have some (partial) information about what is happening.
A targeting procedure has to be applied, which may be automated, by the application of specific techniques.
9.1.3.1 on the past
We might want the base simulation to reproduce exactly the known past. This operation is not futile: it is advisable to
start the analytic shocks from the historical value, eliminating second-order errors.
If all values are known, the technique is simple: we apply to behavioral equations the estimation residual. Each
individual equation will provide historical values, and the global absence of error makes historical values the solution
of the system.
9.1.3.2 in forecasts
In forecasts this method will seek to give some variables values belonging to a certain interval, through the choice of
assumptions and of estimation residuals.
In the short-term the goal will be to adapt results to already known values (or known intervals) of some elements of
the model. At the end of the year, we know already global growth and some of its components, global inflation,
336
unemployment and the trade balance, and the forecast should provide their values, if only for show. Reproducing
wrongly the already known past casts strong doubts on the forecasting quality of the model.
In the medium-term one will seek to keep the general evolution within reasonable limits, in particular meeting some
general conditions. Thus one will not accept too large a gap between the country described by the model and the rest
of the world, as to inflation and growth.
But the fragmentary character of the information will prohibit the use of the method given in the previous paragraph.
The model builder will work in general by iterations, sometimes relaxing some constraints if they prove difficult to
meet. Or he can try the specific algorithm we are proposing.
9.1.4 DIFFERENT TYPES OF SHOCKS
We are going now to categorize the different types of shocks.
9.1.4.1 Analytic and complex shocks
We have already presented analytical shocks, used to interpret and validate model properties, by comparison with
economic theory and properties of the other models. This set of tests also allows the interpretation of model
responses, making the later interpretation of more complex shocks easier. Actually, operational modelling teams
always produce a “shocks catalogue "118 giving the responses of the model to changes in the main assumptions. Given
the numerical quasi-linearity of models, one will get a realistic enough approximation of a complex set of changes by
combining linearly the results of these individual shocks, using a spreadsheet for instance or more simply an EViews
program (combining in one page the contents of several “single shock” pages with the same structure).
Going even further, a few months spent working with a model will give its user the ability to anticipate (at least
roughly) its reactions to given changes, and this without any simulation. He will already have a basic “shocks
catalogue” available in his mind.
But if the intensity of the shock changes with time (for instance to take into account a gradual loosening of quotas in a
trade agreement) we face a larger problem. As we have just stated, most macroeconomic models are roughly linear at
a given period, but they their Jacobian is not stable over time, mostly because of the growing role of external trade. Of
course, the slower the change, the larger the problem. But in the above case, the expected change in the structure of
exports and clients should call for a specific forecast anyway.
One will also use this type of shock for operational studies
• If requested diagnoses are associated to simple decisions.
• If one seeks to decompose the effects of more complex policies.
But to get the exact consequences of complex shocks, one will have to apply them simultaneously to the base
simulation, in a true operational context.
9.1.4.2 External and policy shocks
All these shocks fall also into three categories:
337
• External environment shocks, which seek to measure the consequences of non-controllable events. For a
single country model, the question asked is then: how will a change in the situation in the world affect the
economy of my country?
Two subcategories can be considered.
o Shocks on foreign assumptions, associated for single country models with the situation in other countries, or
elements defined at the world level. A typical shock would represent a depression in Asia, or an increase in
the price of oil, followed possibly by a decrease in world demand and world inflation. Technically, these
elements should not be , as the modelled country or set of countries have some impact on the economy of
the rest of the world. But the cost of modelling this feedback is generally much too high (it could involve the
production and management of a world model) compared to the limited improvement of model properties.
Of course, the approximation (and the associated error) increases with the size of the modelled country. If it
is reasonable to consider the world outside Denmark as exogenous to this country’s events (we have taken
voluntarily a country outside the Euro Zone), this is not so true for Germany (a large country which shares a
currency with a significant part of the world) and even less for the whole Euro Zone itself.
o Shocks on local uncontrolled variables. Examples include technical progress, population, or climate. Again,
one can argue that each of these elements is not completely exogenous: strong economic growth accelerates
technical progress, the birth rate changes with wealth and health of households, growth brings pollution and
climate changes. But the associated mechanisms are difficult to establish, formally complex and limited in
their short and medium-term impact. Therefore the cost of taking them into consideration is generally much
higher than the benefits, except for very specific fields.
The only exception concerns long-term studies. But although it is true that we have solved our model over a very long
period, the goal was only to observe its properties from a technical point of view, and not to produce an actual
forecast. A true long-term model would require a rather different framework, with new formulations relying more on
a formal economic reasoning than on actual estimations, as the sample is generally too small to establish statistically
long-term relationships.
• Economic policy shocks, where interest lies in the consequences of choices on institutional instruments,
decision variables if one considers that the State (or its impersonator) is the client of the study. The question
is now: what will happen if I take the following decisions? And when these shocks are given an objective (for
example: to reduce unemployment by a given margin without eroding too much the other criteria), they will
call for an iterative targeting process analogous to that of normative scenarios.
These elements are not exogenous to the country modeled, as the decisions taken by the Government are largely
based on the local economic situation. But the producer of the study wants to be allowed to manage the decisions,
and not to see the model do it in his place. Moreover, the State has few goals (unemployment, inflation, budget and
foreign deficits, growth) and a lot of potential instruments. Evidencing statistically the link between decisions on the
instruments and the present state of the goals is very difficult if not impossible. The few studies on that subject do not
give in general reliable answers.
• The two approaches may also be combined however, if we look for the decisions which allow to face (or take
advantage of) a given evolution of external conditions. One could for instance determine how the State can
use his own tools to fight losses in trade balance coming from a given depression in world demand. Or an oil
producing country can decide on the use of the additional revenue coming from an increase in the price of a
barrel. This type of exercise is parent again to normative scenarios.
It will be necessary, in all these cases, to care for the presence of formal connections between exogenous variables.
For example, one cannot suppose an increase of the price of oil 119without taking into account its inflationary effects
119
Otherwise, the modeled country will be unduly subject to losses in competitiveness coming from the increase of its
own inflation.
338
on the production price in non-oil-producing countries, or the depression it will cause in world economy. Similarly it
seems difficult to increase the number of civil servants without considering an increase in State consumption (heating,
lighting, and various equipment) and investment (offices, hospitals, schools) according to the type of positions
created.
A very important issue is the impact of public spending (investment, consumption or hiring of civil servants, separately
or together) on local supply. Of course, they modify demand (creating jobs in the public sector increases household
revenue and reduces unemployment). But they should also have a positive effect on potential production, for many
reasons. To give just two examples:
o Creating or improving roads, or the lighting of streets, increases the productivity of the transportation sector,
and of the whole economy (agricultural goods can be delivered fresh over longer distances, attrition from
accidents will diminish…). Some completely new productions can become feasible (access to water can allow
to produce new kinds of vegetables).
o Buying educational books, or hiring more teachers, will improve students’ skills and their productivity when
they join the workforce.
Unfortunately, the influence of these elements is very difficult to quantify. It is left to the discretion of the modeler,
who more often than not keeps it null. A more reasonable option would certainly be to suppose that the effect on
global productive capacity would be the same as actual productive investment, but with a longer delay in its impact.
This would mean that the main difference in behavior between the public and private sectors would be the lower
concern for quick returns, the more widespread effect, and the fact that the Government has the power and
resources to make a decision that others agents cant individually, even if it would be optimal for all of them.
• Finally, one might want to modify structural parameters, such as the productivity of factors, or the
depreciation rate. Even if it does not come from an actual decision, an increase in factor productivity can
come from a new educational policy, or research on the productive process. It can be interesting to observe
its consequences on the whole economic equilibrium.
Falling in the same category will be the residuals on some behavioral equations: the ratio of household
consumption to revenue can be modified ex-ante, as well as the sharing of value added between firms and
workers. According to the formulation, the error-correcting mechanisms and multipliers can lead to increased
or decreased ex-post changes120.
Of course, these shocks fall in the first category concerning their feasibility, as external uncontrolled elements.
9.1.5 THE FORECAST: TECHNICAL ASPECTS
It would appear that once a coherent model has been developed, with its global precision and economic properties
validated by the previous tests, the only efforts remaining for producing a correct forecast would focus on the
definition of assumptions. It is unfortunately not so, and a good deal of know-how will be necessary for the model
builder to produce acceptable forecasts, bearing especially on the following points:
• How to judge the likely character of the forecast.
• How to integrate evolutions of the endogenous elements, known but not yet precisely quantified (concerning
especially the first forecasted periods). This goes from partially known values (as the recent monthly
evolution of unemployment for a quarterly simulation) to vaguer elements, such as the anticipations
provided by non-modelling experts.
120
The price-wage loop will increase the deflationary effect of a wage hike, but the impact of additional demand on
imports will be reduced by capacity adaptation.
339
• Once anomalies concerning the near future are corrected, how these corrections should be taken into
account for next periods. For example, if one decides that estimates of exports have to be decreased in the
first period compared to the value given by the equation, should this gap be maintained in the medium-
term?
• How to interpret results, beyond the simple observation of figures.
In addition, some specific difficulties can appear when generating assumptions:
• In general, they need the intervention of external expertise, as the macro-economic model builder cannot be
fully competent in the whole set of areas concerned. In some cases, it even will be necessary to call for
another model to obtain coherent assumptions (for example a world model will provide assumptions on
international environment).
• External information, even quantified, is often presented neither in the units nor in the categories used by
the model (like trade agreements or the results of wage negotiations).
• The transformation of some qualitative information (climate, strikes, tensions, expectations.....) into
numerical values is even more complex. The present coronavirus outbreak is a huge example of this issue.
Even if some consequences can be foreseen, their intensity and the full set is not conceivable, even when (if?)
the pandemic itself disappears.
9.1.6 CHANGING MODEL SPECIFICATIONS
9.1.6.1 The reasons
Normally, the modeler should make its forecasts on the same model he has estimated. To proceed differently, there
should be some acceptable reason.
One can think of four cases:
1. It is known that a behavior has changed, in a given way: for instance, the Central Bank has decided to change
its method for inflation targeting, or it has just become independent and can decide on the interest rate. Or a
new government has just decided to stabilize the social security deficit, putting the burden on the
contributions of firms and workers, in equal shares.
2. A behavior was present from the first, but we had not enough data, or the variations of the explanatory
element were too small to support it. For instance, the country faced no supply problem, but now capacities
are growing slower than demand and disequilibrium has to be considered121.
3. A behavior appeared late in the sample period, and there were not enough periods to support it. For
instance, a new law introduced part-time work, with a strong impact on the employment process.
4. It is known that the economy of the future will follow different behaviors compared to the past.
In the first case, the behavior is supposed to be known. The associated equation should be introduced, for instance a
new Taylor rule.
121
It was present from the first but had no opportunity to play a role.
340
The second case is more difficult to solve. It applies for instance to a period for which inflation was flat, and its impact
therefore statistically insignificant, while in the future we can expect it to fluctuate much more. Or local inflation was
similar to that of partner countries (stabilizing competitiveness) while differences are starting to appear.
In this case there is no other way than taking values either from economic theory, measurement on other countries
and the observation of resulting properties.
The last two cases can be generally associated with:
o Developing countries
o Countries in transition, in principle from socialist to market economy (some countries can remain in between
like China or Vietnam).
Identifying transition countries
In transition countries, the behavior of all agents has changed as market economy mechanisms have appeared. But
although the technical transition could be quite fast (like German reunification) the transition process has been
generally much slower.
One can separate two cases:
• The Central and Eastern Europe countries, with the addition of Russia and maybe the former USSR republics.
• Asia: mostly China and Vietnam, possibly Laos and Cambodia.
The difference of course is that the later are still under Communist regimes, and that they can be considered as
developing countries, which is also probably the case of the less advanced elements of the first set, like Bulgaria,
Romania and Ukraine122.
Actually in our opinion, if we consider the present situation, the type of regime is not the most important, as for the
most part China and Vietnam behave economically like market economies (perhaps more than present Russia). The
only issue is the date at which the transition can be considered as advanced enough for the economy to be modelled
in this way. Officially, the date was 1978 for China, and 1986 for Vietnam. But the process was quite gradual, and one
should consider delaying the application of a structural model by at least 10 years.
For the Eastern bloc, the transition started at the beginning of the nineties, and can be considered to have proceeded
much faster, in particular because the economy of these countries had been following western standards (in
framework and level) 50 years before . This means modelling could start in 1994-1995, with additional elements:
o The statistical system took some time in changing methods, in particular from the Social Accounting Matrix
system to western-type National Accounts (like the UN defined SCN95, now SCN08).
o Now almost all of these countries provide quarterly accounts. This multiplies the number of observations by
four, which is crucial in this case (although the problem is decreasing with time). Unfortunately, the gain in
information is much lower.
o Some specific events took place during the period, like the partition of Czechoslovakia in 1993, or joining the
European Union (even the EuroZone).
Concerning the data
Now several situations can appear:
122
Although the appreciation is quite dependent on criteria (look at China for instance).
341
• We now have enough data to estimate equations. This is for instance the case of Poland, where data is
available from 1995, using the same methodology, and with behaviors which have converged gradually to
market economy.
But we have also to consider the last two cases:
• In case 4, the equation has to be entirely calibrated, with values taken either from economic theory,
measurement on other countries and the observation of resulting properties.
• In case 3 also, except that some additional information can be obtained from estimations. But it should not
be used alone, even if all statistics look good (unless we get extremely good statistics on a sample not too
small).
• In any case, if the influence of an element (such as profitability on capital or unemployment on the wage
rate) is considered to be growing in the present period, the associated coefficient should also show a growth
in forecasts, probably reduced with time to converge to a constant. This process (initial level of the increase,
speed of convergence) is the responsibility of the modeler, helped by the observation of model properties.
9.1.6.2 The technique
We can use exactly the same technique as for testing model properties over the future, except that the role of
changes in parameter values and formulations will increase.
9.1.7 OPTIMAL CONTROL
This technique consists in computing the values of instruments which gets model variables as close as possible to a
given objective function, under constraint of the model. Compared to the reversal of causalities, the model user will
set here a unique target, normally unreachable, formalized in most cases by a quadratic function of one or several
variables. Then he will apply a maximization algorithm under constraints (equations of the model and bounds on the
evolution of instruments). This will have to be done outside the modelling package (while spreadsheets provide a
simplified version of this technique).
This technique has been quite popular in the fifties and sixties in western countries like France (managed by the
Planning Agency) and remained so in the socialist countries for a longer time123. Now they have been more or less
abandoned, and the remaining applications limited to economic research.
9.2 TEACHING WITH MODELS
The associated models will allow educational or scientific applications. Diagnoses can be observed for themselves, as
the quantified illustration of some theories. But their small size makes them efficient for the application of specific
validation techniques, useful especially if they have been designed to reproduce the properties of a larger operational
model.
The simplicity of these models gives them the natural role of a learning instrument. According to the public and the
goal, different tools may be used.
• Extremely simplified representations, based on formal examples and fixed parameters, clarify theoretical
frameworks (Solow model, IS-LM model, Mundell-Fleming Fair model...). By formalizing them on a numerical
example and solving them under various assumptions, one can better appreciate their properties (the
123
Not surprisingly, Russia has produced the best theoreticians in this field (like Pontryagin) and in the connected field
of matrix theory (like Gantmacher)
342
simplicity of these models often allows a graphical display of their equations as dynamic curves). Courses in
macroeconomics are now often accompanied by computer programs running these representations.
• More applied models, reproducing as faithfully as possible the properties of larger models used for
operational diagnoses. Rather than illustrating an economic theory, the goal here is to put together a limited
number of elementary mechanisms, evidencing the main interactions governing the economic equilibrium.
This will generally be done through shocks, concentrating on the impact of the different economic policy
instruments and the orders of magnitude of their effects.
In this last case, even if the precision of these models remains limited, producing acceptable diagnoses requires to
apply them to real cases. Hence they generally will be based on real data, and often estimated econometrically. And
to make the interpretation easier, one will generally process current issues: for example measuring consequences of a
decrease in work duration, for the French economy of 1998.
9.3 PRESENTATION OF RESULTS
9.3.1 GENERAL ISSUES
The role played in the efficiency of the simulation process by the quality of results presentation is often
underestimated. To researchers, it represents often a minor task, to which they are not very well prepared.
Nevertheless, this element is often a necessary condition to capture the attention of the audience and insure the
success of a study. While a mediocre but well-presented work will sometimes have some success, a work of quality
but badly presented will generally fail in getting any audience, by discouraging its potential listeners or readers,
including the referees124.
A good presentation has to be:
• Error free (this is more difficult than it seems). A single visible error will cast doubt on the whole set of
results.
• Explicit (the meaning of figures presented has to be evident, in particular their classification).
• Not too long.
• Pleasing to the eye (not too dense), but not too flashy (one must not think the author is trying to hide
problems, or has diverted time from his research). PowerPoint slides should not be packed.
• Clearly synthesizing in the right order the main teachings, and only them. In a PowerPoint presentation, one
must not be diverted from listening to the presenter.
• Allowing fast access to a particular element of information.
• Adapted both to the nature of the work and to its public (for different audiences, different presentations can
be produced for the same work125).
Once these principles have been followed, one should not spend to much time sophisticated features, the danger
being criticized for trying to hide the low quality of the results behind the gaudiness of the presentation.
The two main elements of any presentation are tables and graphs, and one must keep in mind that these two types of
presentation, while complementary, must be conceived simultaneously.
9.3.2 TABLES
124
But the truth of this argument depends also heavily on the celebrity of the author.
125
This is made easier by presentation packages.
343
There are two main kinds of graphs, depending to the public:
• Basic tables, designed for the model builder and his working partners, use simple formats to summarize the
whole set of model information. They allow a quick interpretation of the main characteristics of simulations
and shocks. For a macro economic model, the table will generally contain the evolution of the main
components of the supply-demand equilibrium (at constant prices) as well as the evolution of prices,
employment, external and budgetary balances, and some ratios.
The definition of elements presented will be limited, sometimes even to a name representative of the concept. It will
be simpler to implement if adequate procedures have been created by the user inside the model-building software.
For more detailed models, one will not have to consult the whole set of results, except to find reasons for the
abnormal evolution of some aggregated elements.
• More elaborate tables, designed for presentation to a more or less initiated public (the general public, the
clients of the study, partners that have collaborated directly to it, other researchers, non-specialists of the
subject). Even in the case of a working paper, a certain level of quality is needed. This is made easier by the
most sophisticated model-building packages, and their production can be used for some types of publications
not requiring too high a level. The quality of EViews tables has increased dramatically with the last versions,
allowing to choose fonts, symbols and borders. In our opinion, it is no longer necessary to transfer
information to presentation packages, except in very specific cases.
In practice, one observes that, even for if the model is completely assimilated, the production of relatively elaborate
tables makes work easier, and allows to locate, not only the properties, but also the problems. It is a good investment
to produce them as soon as possible, starting with the development phase. Once their framework has been defined,
these tables can be reused indefinitely at no additional cost.
9.3.3 GRAPHS
The advantage of graphs is to clarify series evolutions by a more telling presentation. The disadvantage is the absence
of precise figures. Graphs therefore should be used:
• Along with tables, to complement the detailed results by more synthetic and easily accessible information.
• Alone, if the message to transmit represents a simple synthesis.
The main graph type uses:
• Points, connected or not by segments, or more rarely by curves (differentiated by a color, a type of line, and
often a symbol).
• Histograms, each observation being associated to a set of vertical bars (juxtaposed or piled) each measuring
the size of a variable.
• Pies126, associating to a single observation a circle whose slices correspond to the share of each variable in
the total.
• Two dimensional bubble graphs (since EViews 10), using a third dimension for the diameter of the bubble.
This list is not restrictive, as the imagination can go far in search of visual synthesis. For example, bars in histograms
are often replaced by appropriate drawings (persons, factories, products...).
126
In French, “Camembert”, a regionTapez une équation ici.al cheese with a round shape.
344
Concerning graphs (and presentations as a whole) the modeling packages have made considerable progress in recent
years. Some offer sophisticated graphic features available through a large set of user-friendly functions, and EViews is
not the last in that regard. One will be able for example, immediately after simulation, to display the evolution of
historical and simulated variables or of selected expressions. This display mode will be used in the model development
process, allowing a better interpretation of intermediate results, and revealing problems.
Since EViews 8, you can:
• Decide on the time scope of the graph using a slide line, available at the bottom.
• Add custom arrows.
• Export the graph to a PDF file.
And since EVIews 9:
You can select (as a rectangle) a rectangular part of a graph, which can be useful to evidence a feature of the series,
for a given subsample, without changing the graph itself.
345
9.3.3.1 Graph of model solutions
For graphs, model solutions represent a specific case.
The most interesting feature is the MAKEGRAPH command, specially adapted to model solutions. The syntax is:
model_name.makegraph(options) graph_name model_series_names
If no option is specified only the current scenario solution is displayed.
The main options are:
a : include actuals
c : compare active to baseline
d : include deviations from baseline (as an additional graph in the same frame)
n : do not include active scenario.
The results can be presented raw or transformed, and for stochastic simulations a confidence interval can be
displayed.
Once the graph is created, the usual modifiers can be applied (legend, type of line, colors….).
One has to note that the graph must not preexist. If it does it must be deleted first (using the “noerr” modifier to
avoid a possible error message).
delete(norerr) name_of the graph
9.3.3.2 Multi-graph slideshow
Being allowed to move from graph to graph in a set is very useful if one wants to display the evolution of a group of
important variables in a forecast, or different aspects of the model response to a shock: demand and supply, deflators,
external trade….. This feature might lead to using in presentations an EViews session in parallel to PowerPoint slides.
This done through Zoom +/-
346
9.3.3.3 An example of model graph
We will now present an example of graph corresponding actually to the results of a shock, which we will present later.
We are not going to present the full syntax, just give some basic elements. Editing the statements to adapt to your
case should be rather easy.
The variables represent the relative changes in final demand, GDP,, exports, imports and added value deflator
following a 1% increase in Government demand (considered as the most meaningful shock available.
First we delete any previous version. Using a preexisting name will cause an error.
delete(noerr) gr1
Then we chose the type of graph. We chose lines.
graph gr1.line p_fd pg_gdpm pg_x pg_m pg_pq
Then we decide to place horizontal and vertical grids with given width and color (black). The three figures represent
the amount of the three basic colors : red, green, blue.
_gr_g1.axis(left) grid font(11)
_gr_g1.axis(bottom) grid font(11)
_gr_g1.options gridwidth(0.375) gridcolor(160,0,181)
We add the title at top(t), units at left(l) and size of shock at right( r)
_gr_g1.addtext(t, font=10,x) Graph 1 : Shock on Government demand \r The supply - demand equilibrium
_gr_g1.addtext(l,font=8) in percentage
347
_gr_g1.addtext(r,font=7) Shock : 1% of GDP
The legend will use two columns, with a black frame, positioned at the bottom on the left.
_gr_g1.legend columns(2)
_gr_g1.legend framecolor(160,0,181)
_gr_g1.legend position(botleft) font(10)
Now we define for each of the five lines the symbol, the color, the width, the pattern and the text.
_gr_g1.setelem(1) symbol(1) lcolor(blue) lwidth(0.25) lpat(1) legend(Final demand)
_gr_g1.setelem(2) symbol(2) lcolor(red) lwidth(0.25) lpat(2) legend(GDP)
_gr_g1.setelem(3) symbol(8)lcolor(green) lwidth(0.25) lpat(3) legend(Exports)
_gr_g1.setelem(4) symbol(4) lcolor(orange) lwidth(0.25) lpat(4) legend(Imports)
_gr_g1.setelem(5) symbol(10) lcolor(purple) lwidth(0.25) lpat(5) legend(GDP deflator)
We get the following result, after moving the legend and title to adapt their position to the rest of the graph, the best
we could.
2.4
Graph 1 : Shock on Government demand
2.0
The supply - demand equilibrium
1.6
Shock : +1 point of GDP
1.2
in percentage
0.8
0.4 Final demand GDP
Exports Imports
0.0 GDP deflator
-0.4
-0.8
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
348
10 CHAPTER 10: APPLYING THE ABOVE PRINCIPLES TO OPERATIONAL CASES
In the preceding chapters, we have addressed all the technical aspects of models construction and applications,
illustrating them by a simple example. Even if the techniques proposed have been quite general, we have not
addressed the more economic features of operational models. This is what we are going to do now.
Having mastered the techniques for producing a very small model, the professional will face two kinds of difficulties:
• The structure of the new model has to be more consistent with economic theory (the one he has in mind).
The detail of the specifications must meet his requirements, in terms of assumptions, mechanisms and
results.
• The methodology for producing and managing the model call for slightly different techniques, adapted to a
problem of higher complexity.
In the following pages, we shall address both these questions. As we are not giving a course in macroeconomics127, we
shall only describe the elements which in our opinion are requested for the production of an operational model, the
choices one has to make, and give some directions as to the most promising options. We leave to the reader the
choice of actual decisions and the introduction of additional features he considers adapted to his problem.
So please do not criticize our explanations for being too basic128.They should be considered as a starting point.
If this is your feeling:
• You are a good enough economist to proceed by yourself
• Or you can use any of the more advanced textbooks in the list we provide.
Again, let us stress that you are not reading a book on sophisticated econometrics or economics. Our goal is to show
how the two fields can be merged into a working model, concentrating on the technical aspects of the task.
In any case, the elements we will describe should be interesting, if interpreted as a description of the main options
used by mainstream operational structural models (maybe a little on the old fashioned side).
For this reason, we do not cite a source in which the mechanisms we describe are presented. Rather, we give a list of
books on macroeconomics, or full model presentations, in which most of them will be found in higher (and probably
better) detail.
We will be more directive on the EViews aspects, of course.
10.1 THE ACCOUNTING FRAMEWORK
Before we start dealing with macroeconomic theory, we must describe the framework and the concepts we shall use.
We have already started, of course, but in operational cases we must be much more precise and consistent.
127
Neither did we give a course in econometrics.
128
Some people will certainly decide they are wrong. This is another problem, but most of the features we present
have been used in several models (this is not a proof of their validity of course).
349
As for econometrics, we do not have the ambition of giving a course in national accounting. In some cases, we will use
simplifications which will make the presentation not entirely rigorous.
At the end of the book one will find a list of associated publications, from simple courses to official manuals stating
the official concepts used at present by local or international organizations.
In our opinion, the best reference in that field is still:
LEQUILLER F, BLADES D, (2006), Understanding National Accounts, OECD Publishing, downloadable at:
https://www.oecd-ilibrary.org/economics/understanding-national-accounts_9789264027657-en
For data and methodology, one can refer to:
http://www.oecd.org/sdd/na/
To define the framework of a macroeconomic model, one must define:
• The agents involved
• The operations they share
• The products associated with these operations.
According to the problem the model is designed to solve, these elements will be more or less detailed.
We shall start with a single good, single country model, then extend the classification. The specifics of the building
process will also be dealt with separately.
In a special chapter, we shall address Stock-Flow Consistent modelling, a field which is much more rigorous in defining
agents and operations, requiring the production of fully balanced accounting tables.
10.1.1 THE AGENTS: A FIRST DEFINITION
The operations described in a model will be managed by agents, characterized by their role. We shall separate:
• The firms, which buy and produce goods and market services, distribute revenue.
• The households, who receive revenue and consume part of it.
• The financial institutions, which manage financial transactions: lending, borrowing, insuring.
• The administrations, which gather taxes from the above agents, and address social needs through
redistribution and production of associated elements.
• The rest of the world, representing the foreign agents trading with the first four.
10.1.2 THE OPERATIONS
They are divided into:
• Operations on goods and services: a good (material, such as a bottle of wine or a DVD player) or a service
(immaterial, such as transportation from one place to another or a music lesson) is traded between two
agents, generally in exchange for money. But if can be bartered (a kingdom for a horse) or given for free,
most of the time by some administration (free school or free medicine). In the last case the transaction is
considered as “non-market”.
• Transfers from one agent to the other, which can come in exchange for something, separately from the
above transaction. For instance, wages are paid to households by firms, but they do not buy the actual good,
350
just the contribution to its production. Alternatively, pensions are financed by contributions, but not
necessarily paid by the beneficiaries, and in any case with some delay. And income taxes are not linked with
any service (although they will be used in part for the good of the taxpayer, at least in principle).
Things are not always so simple: an independent taxi driver will sell a service (and use the money as revenue). If
he belongs to a company he will earn wages (a transfer). The company will sell the service and transfer part of the
revenue to him.
10.1.3 THE INTEGRATED ECONOMIC ACCOUNTS
Once these elements have been defined, they can be presented in a table, allowing one line per operation, and one
column per agent. Actually, an additional column should be introduced for the first “goods and services” part: for each
operation, on the left for total expenditures, on the right for total revenue.
This is particularly useful for transfers: in general, wages will be an expenditure for firms and a revenue for workers.
But all agents pay wages, including households: this technique allows to separate the two types: paid and received.
• In addition, the table will describe the steps in the economic process: the operations will be distributed in
sequential sub-tables, each presenting a balance as the last line.
We shall find in succession:
• The “production” account, describing all operations linked to production, but also imports and exports. It
gives value added. This part will call for an additional column, as lines are not balanced (exports represents an
expenditure for the Rest of the world, but the revenue is not distributed among agents).
• The “primary distribution of income” account, introducing the immaterial expenses: wages, direct taxes,
subsidies (-). It gives the operating surplus.
• The “secondary distribution of income” account, further distributing revenue: dividends, revenue of property,
and also social transfers from the administration. The balance is disposable income.
• The “use of disposable income” account, describing essentially household consumption. The balance is
savings.
• The “capital” account, describing the use of savings to acquire goods. It includes various forms of investment
and the change in inventories. Its balance is the financial capacity.
• The “financial” account, which explains how the financial capacity is used(if positive) or met (if negative). It
describes money, loans, credits and equities. Its balance is zero.
10.1.4 SECTORS, BRANCHES AND PRODUCTS,
Now we can separate the “firms” or rather “non-financial firms” into categories, according to its type of production.
Three notions appear here:
• The sector: it categorizes a firm according to its main activity (car making for the Peugeot car manufacturer).
• The branch: it separates the firm into each of its activities (spare parts production and distribution for the
Peugeot car manufacturer).
• The product: it represents the good or service actually produced by the firm. It is normally equivalent to the
branch, but not in all countries: the difference can come from “fatal products” coming automatically from a
process designed to create another good (such as hydrogen coming from oil refining).
351
As we shall see later, this decomposition calls for the definition of intermediate consumption, a two-dimensional
variable associated to the consumption of one good to produce another. This can be true also for investment (in one
product by one branch) or inventories129.
10.1.5 AGENTS SUBDIVISIONS
National accounts can further subdivide the agents. For instance one can separate:
• Households into individual (a family) and collective (a convent).
• Financial institutions into the Central Bank, other banks, and insurance companies.
• Firms into State owned, individual firms and companies, local or foreign owned.
• The administration into central government, local agencies and specific agencies, such as the social security
agency.
• The rest of the world into countries and zones (even for a single country model). It will allow to make the
model assumptions more explicit.
Of course, other more statistical classifications can be used, like the revenue level of the household or the occupation
of its head.
10.1.6 A MULTI COUNTRY MODEL
We shall develop later the specifics of a multi-country model. Let us just say for now:
• Of course, the system must be duplicated for each of the countries described. Specific elements can be
introduced in some cases (like individualizing oil for producing countries).
• Consistency must be enforced between the exports of each country and the imports of others. This can be
done either by identifying individual trade flows, or by creating a specific export demand variable for each
country, based on the global demand of each of its clients.
• Finally, the categories can be countries, associations (like the European Union) or related zones (like Sub-
Saharan Africa).
10.2 A SINGLE COUNTRY, SINGLE PRODUCT MODEL
We shall start with the economic developments needed by our model to reach operational status.
10.2.1 THE ECONOMIC ASPECTS
The very small model we have built as an example is of course too limited to be used for operational studies. It
presents nevertheless:
• A link between production and revenue.
• External trade, depending on available productive capacity.
• Production defined as local demand plus net exports.
• A simple production function.
129
Even though in this case some elements are identically null, such as investment in the energy product or
inventories in financial services.
352
• A description of the output gap, its consequences and the way it is closed.
What do we need in addition, at the least?
• A price system, and its links with the real sector (in both directions).
• A better description of the behavior of firms and households.
• A financial sector.
• A full description of the State budget and its instruments.
This will also call for a redefinition of the scope of assumptions.
10.2.1.1 The productive process
This part of the model (one speaks often of “blocks”) will not define production, but rather potential production (or
productive capacity), as a function of available factors.
Why not actual production itself? There are two ways to consider production:
• Actual local production, contributing along with foreign exporters to the satisfaction of demand (both local
and foreign), in a share depending on relative prices and available capacities).
• Potential production, given by the production function, taking into account technology and the level of
factors (capital and labor), themselves chosen by firms according to their relative costs, expected demand,
and profits conditions.
We want our model to follow the most logical causal sequence, which is:
1. Firms form expectations on demand.
2. They decide to try adapting their capacity.
3. They decide on the optimal level of factors allowing this capacity.
4. Technical difficulties, and concerns in adapting the factors too fast will create inertia in the definition of
actual factors.
5. Actual factors will define actual capacity.
6. A gap appears with target capacity.
7. Firms will try to close this gap by increasing factor levels, correcting the decision in 4.
Alternate version:
2. Firms decide on the production target taking into account profitability.
The comparison between actual and potential production will play an important role in some behaviors.
This is the sequence that the model will describe, actual production being obtained late in the process, once demand
is known (as in the small model).
This capacity for production will be measured:
• For employment, in man - years or man - quarters according to model periodicity
• For capital, at constant prices, in the currency of the country.
The function can also include:
• Energy consumption
• Intermediate goods (like raw materials).
353
Actually, capacities are generally defined in terms of value added, a more reliable notion as we have explained earlier.
This means the two last elements are not considered, or rather their level will come automatically from value added
itself.
The first issue concerns the logical link between capacity and factors. We have already seen:
• Complementary factors. For a given capacity, there is a single optimal process requiring a fixed combination
of labor and capital. Starting from an optimal combination, adding to the process a quantity of one factor
does not increase capacity, or allow using less of the other factor. This capacity is obviously optimal
regardless of the relative costs. Actually labor productivity has generally some flexibility, and capital is the
truly constraining factor, as temporary and limited increases in labor productivity can be achieved (for
instance by increasing the number of hours worked).
This is the simplest option, in its formulation, estimation and understanding of properties. Operational models use
generally more sophisticated frameworks:
• Cobb-Douglas. The elasticity of substitution is unitary. This means that if the ratio of the cost of labor to
capital changes by 1%, the optimal ratio of capital to labor will change by 1% too, for a given capacity
requirement. So the sum of the elasticities of capacity to labor and capital is one.
• CES (Constant elasticity of substitution). Now the elasticity can take any fixed value (with the right sign).
Of course, the CES option covers both others (with fixed elasticities of 0 and 1 respectively).
The framework calls also for:
• A definition of the relative cost.
The relative cost of labor and capital is not just measured by the ratio of the wage rate to the investment deflator.
One has to take also into account:
o Social contributions of firms: they contribute to the cost of labor.
o The interest rate: while capital is bought immediately130, labor can be bought (rented) when the time comes
(slavery has been abolished for some time now). So a firm which has money can save it, and one which has
not does not have to borrow. If capital itself can be rented, the distinction disappears.
o The depreciation rate: capital wears out, while when a worker “wears out” through old age or sickness, he
will leave and can be replaced by a new worker at no cost except training (pensions have already been saved
as a share of wages).
o The future evolution of wages: if wages are currently growing faster than inflation, firms can expect labor to
become less competitive in the future. The gain from having output transferred to fast developing countries
becomes lower as they reduce the gap with developed ones. This applies in particular to present Asian
developing countries like Vietnam or Cambodia.
• The possible changes in technology.
The issue here is to decide if the technology decided at investment time (which defines the roles of labor and capital)
can change later.
130
Actually, some forms of capital (like buildings, computers or patents) can be rented or leased.
354
Basically, the options are:
o A single available technology (Clay-Clay).
o A technology chosen at installation time, with no possible later change (Putty-Clay). This means basically that
the “complementary factors” option applies to factors once they are installed.
o A technology with a permanent possibility of change (Putty - Putty). The same substitution option applies to
factors at any period.
The reader fill find easily examples for the three options.
10.2.1.1.1 A specific problem: the statistical determination of productive capacity
To determine the actual value of capacity, we have several options, depending on the available information
• In some countries (such as France), a survey asks a sample of firms by how much they could increase their
production using the present factors (hiring more people if necessary). This gives the firm’s capacity. Then a
synthetic measure of capacity will be obtained using the same weights as for computing actual production, ,
and the rate of use as a ratio of the global values.
Then we shall use the capacity series to estimate its equation. For this, we can specify the actual behavior of firms,
and optimize their profits under a capacity constraint using the formula we want to estimate. This applies when the
factors are substitutable (otherwise the optimum solution is set from the start, and does not depend on relative
costs). Taking the derivative of the function according to both labor and capital will give a set of equations with
common coefficients, which one can estimate as a system. This method takes into account fully and explicitly the role
of the relative costs.
• If we know only the level of factors (capital is sometimes missing in the country’s statistics), we can specify
the production function, and estimate its parameters over the actual values of production. We can suppose
that the estimated formula gives normal production, and the residual is the output gap. Again, the ratio of
actual to “normal” production gives the rate of use, but this time to a constant factor (the average rate of
use).
We can also (a better solution in our opinion) apply the first method, using actual production instead of capacity.
Again, the estimated capacity (reconstructed by applying the production function to the estimated factors, considered
as optimal) will give a normal level of production, and the difference to actual production the output gap. However,
we have to consider a delay before this level is reached, through the same inertia we have described earlier.
• If we do not have this information, we can always smooth production, and use the result as a “normal
production” level (at a normal rate of use of capacities). For this, applying to actual data a Hodrick-Prescott
filter is the most usual technique. If we suppose the “normal” rate of use of capacities constant over time, we
get capacity at an unknown multiplicative factor.
This technique does not require a choice of production function, or the availability of a series for capital (which is
often absent or unreliable). Neither does it provides it, which will be a problem for model specification and
interactions.
10.2.1.2 The change in inventories
We see no specific reason to modify the framework used by the small model. More sophisticated formulations could
use:
• A full error-correction framework, provided we knew the level of inventories.
355
• An influence of demand: if it goes up suddenly, some of it can be met by relying more on inventories. This
element will be difficult to introduce, as it calls for a negative influence, while value added has a positive one,
and both elements are positively correlated. This means the over estimation of one coefficient can be
compensated by over estimating the second.
• An influence of prices: the more expensive the inventories, the shorter the time they will be stored.
Measuring the change in inventories at current prices introduces a specific problem, as one must also consider the
change in the price of previously stored inventories, especially for a quarterly model.
10.2.1.3 Unemployment
This is a new concept, which would have fitted easily into the small model (increasing its size a little131)
We shall consider that the variations of employment do not transfer fully to unemployment. Job creation will attract
to the labor market previously inactive persons, who shall take some of the jobs offered: the work force (employed +
unemployed) will increase.
For instance, creating a small firm in the Vietnamese mountains will allow housewives to combine employment with
domestic work.132 Or employees of a closing down factory will not necessarily remain in the labor market if their
qualification is not required elsewhere.
But the level of unemployment should also influence its dynamics. If it is high, the incentive to join the work force will
be lower. Favorable employment prospects will lead young people living with their parents to start their working life.
On the contrary, a depressed labor market will persuade aged workers to retire earlier (and they will be incited to).
And some of the unemployed will stop looking for a job, and leave the work force.
Also, the higher the unemployment level, the higher the quality of the best unemployed. Observing the situation, the
average unemployed people will lower their expectations for getting a job, leading them to leave the work force.
On the contrary, at a low level of unemployment, the unemployed will feel that they stand a good chance over their
competitors, most of them being either inefficient or not really looking for employment.
This obviously corresponds to an error correction framework, leading to a target rate of unemployment (and also of
participation of potential workers to the labor force, as we shall see).
10.2.1.4 The price system
The role of prices in a model is of course essential. But it is not so simple to introduce, even for a minimal model like
the one we presented above. In this case, several deflators have to be introduced simultaneously, associated with the
elements in the supply - demand equilibrium:
• GDP
• Final demand
• Exports
131
We have tested it, and it works. You can try too. And you can also link the real wage rate, making it endogenous.
132
Which is not considered as employment (maybe because it is not paid, and does not affect GDP, even if paid
housework does).
356
• Imports.
And in addition:
• Wages (possibly including social security contributions)
• Possibly, deflators for each element in the decomposition of demand: consumption, investment, government
demand....
• The price of foreign currency (the exchange rate)
• The prices of lending and borrowing (the interest rates)
Moreover, trade prices have to be defined including and excluding taxes. This distinction applies to external trade (for
defining competitiveness and trade balance) and local demand (for defining final and intermediate consumptions).
And the price of labor (the wage rate) must take into account social contributions.
Not all these elements have to be estimated. Behaviors should be associated with:
• GDP (firms decide on the price at which they sell, once they take into account the cost of input).
• Exports (local exporters do the same).
• Imports (now we consider foreign exporters)133.
• Wages (the result of a negotiation between workers and firm managers).
The demand price is obviously missing. Its role will be to balance the supply-demand equilibrium at current prices:
𝑃𝑓𝑑 ⋅ 𝐹𝐷 + 𝑃𝑥 ⋅ 𝑋 = 𝑃𝑞 ⋅ 𝑄 + 𝑃𝑚 ⋅ 𝑀
Or
𝑃𝑓𝑑 . 𝐹𝐷 = (𝑃𝑞 ⋅ 𝑄 − 𝑃𝑥 ⋅ 𝑋) + 𝑃𝑚 ⋅ 𝑀
In which demand at current prices appears as the sum of two demands : supplied by local firms (production – exports)
and foreign firms (imports).
• In the system, the deflators will depend on each other. For the time being, we will only give indications. A
more detailed reasoning will come with actual estimations.
o The GDP deflator depends on the wage rate, or rather the wage cost per unit produced.
If the wage cost goes up, firms will have to increase prices to keep their margins. If the slice of the pie decreases, the
size of the pie can compensate.
133
Remember we are building a single country model. The description of trade will be different with several
connected countries.
357
They do not have to do it immediately, and they have also to consider their competitiveness on the local and foreign
markets (for exporting firms).
And if the firm maximizes its margin, it will be more accurate to consider the global cost, including amortization of
capital.
o The wage rate depends on the consumption price, but maybe also on the value added price.
If prices go up, workers ask for a raise in wages to sustain their purchasing power. But again, firms are more likely to
accept raises if they were able to increase their own price.
o Trade prices depend on the cost supported by the exporter, and on the price set by its competitors. This
means they have to maintain their margins and their competitiveness at the same time.
This behavior is obviously based on production prices, the price at which they sell. This means the cost of intermediate
consumptions has to be taken into account. For instance, a country having access to cheap oil will be able to export at
lower prices, even at the same cost in value added (and the same margins). But this introduces a problem, as until
now the single product feature allowed us to neglect intermediate consumption, a variable difficult to manage as its
value depends on the classification and the length of the production process.
The behavior also has to apply to the same currency. If the export price uses the currency of the exporter, the price of
its competitors measured in foreign currency has to be corrected by the exchange rate.
o The price of demand depends on the price at which local producers and foreign exporters sell on the local
market.
This uses the identity above.
Another important issue concerns the behavior of firms selling on both on local and foreign markets.
Two behaviors can be considered:
o The firms define both selling prices separately. Local firms start by defining a price for selling on the local
market, using the above behavior. Then their export price will average this price and that of competitors.
o The firms define first a global selling price, allowing to reach a global margin rate, then they chose a
combination of the two prices which meets this target. This means that a decrease in the export price
(possibly designed to stay competitive facing foreign deflation) will have to be compensated by an increase in
the local selling price.
o Of course, this behavior is completely justified for firms which sell only on one market, like producers of Nike
shoes in Vietnam or private teachers in France.
The choice will have a strong impact on the price system. Actually the second option will increase the intensity of the
price-wage loop: if local costs go up, firms refuse to apply completely these costs to exports (as they do not want to
lose their competitiveness), and maintaining global margins calls for a larger increase in local selling prices (which does
not happen if targets are defined separately).
• This equilibrium is subject to external influences, either endogenous or exogenous.
o Endogenous
If labor productivity goes up, firms need fewer workers and can pay them more. They can also lower their
prices.
358
If output is too low compared to capacities, firms can first lower prices to sell more (later they can adapt their
capacities).
If unemployment goes down, workers can increase their demands without the risk of firm managers to look
elsewhere.
o Exogenous.
The concerned variables are the indirect tax rates.
One will generally consider:
VAT.
The other indirect tax rates, such as the tax on tobacco, gas, alcohol.
Tariffs.
And also
The rate of social security contributions by firms.
If indirect134 tax rates (such as VAT, tax on gas, cigarettes, social contributions paid by firms...) go up then firms should
adapt their price if they want to keep their margins.
Two elements:
o First, it is quite important to separate these taxes in a model, for the usual reason: their base is different, and
their impact on the economy also.
VAT applies only to demand. But the most important feature is that it does not apply to exports (exporters
can deduct it before they sell abroad), and they apply to imports. VAT on foreign cars is the same as on local
ones, and applies to the total value. And when the car producer looks for electronic equipment, increasing
VAT on this good will not change its decision on its origin as it can deduct VAT anyway135.
On the contrary the other indirect taxes apply only to local productions, even though the impact of this
difference is not so high, as imported goods are often taxed at the moment they are sold. For instance, the
tax on tobacco applies also to imported cigarettes, and the tax on alcohol to imported whisky.
Concerning local tariffs, they are not deductible in the general case. This means that a change in their rate
will affect directly the competitiveness of imported goods, unless the importer decides to compensate the
effect by adapting its margins.
As to tariffs applied to local products by foreign countries, they affect directly the competitiveness of exports.
This means they have to be considered, even though their statistical value is not directly obtained from the
national accounts of the country.
o Second, formalizing the role of taxes relies obviously on the rates, the variables decided by the state (or
foreign states for tariffs on local exports). These rates will affect deflators, and allow computing the amount
of the tax once the base is known.
134
These taxes are called indirect because they are not paid directly by the ultimate payer, contrarily to income tax,
corporate tax....
135
It might not apply it fully, however, if it fears a drop in its sales.
359
It should be clear that the right way to formalize these taxes is to set the rate as an exogenous decision variable, and
not to estimate the amount as some modelers might be tempted to do. This allows handling the decision easily, both
in forecasts and shock analysis. And on the past, the technique is quite simple: the tax amount and the base are
known, and this allows computing the rate, used as an exogenous ratio. The associated identity (tax = rate x base) will
hold exactly true. We shall see later how to handle these rates on the future.
Obviously, the rate obtained will be different from the legal one (generally lower!).
This technique is consistent with the general approach: make the government decisions exogenous, but identify first
what is the true decision variable.
• Prices can also influence real elements
o The selling price of local producers determines the quantities they will sell. This is also true of exporters,
through the comparison between their export price and the price on the market on which they sell
o The relative costs of labor and capital influence the choice of the factors in the productive process.
o More generally, ratios of prices affect the ratios of elements (or the shares in a total). For a given global
consumption level, reducing the price of one good will increase its share.
o A higher inflation reduces the purchasing power of previous savings, calling for a larger reconstitution effort.
o And of course prices enter the definition of variables at current prices, when they are separated into volume
and deflator (elements in the trade balance, wages...). For the elements in a sum, a different evolution of
deflators will change the shares at current prices. And some ratios are computed at current prices, with a
different deflator on the numerator and denominator.
All this is described by the following graph.
10.2.1.5 The account of firms
We have already dealt with the supply side, defining the adaptation of production factors: employment and capital, to
target capacity, as well as the decision on prices, based generally on the short-term maximization of profits.
This means that most of the remaining equations will be definitions, describing the firms account without calling for
any theoretical elements.
360
There are three exceptions:
• The tax on profits, which should be again computed by applying a rate to a base. This is more complex than
usual, however, as:
o Computing profits in a model is quite complex, and not all models are able to do it. Sometimes it is necessary
to use a proxy, making the apparent rate more difficult to interpret.
o The timetable for the tax calls for a dynamic equation, as the tax is not generally paid in the same period as
the associated profits (but there can be a provision to pay immediately). So a formula describing the
mechanism must be established.
o The tax on negative profits is not negative, but null, introducing a bias on the apparent rate.
• The income tax, which can be subtracted from wages, or paid later, with a possible advance system.
• Dividends paid by firms, which can be estimated or constructed through an identity (using a rate in the same
manner as taxes). Again, one must decide on the dynamics, as dividends follow the profits. Also, the
beneficiary of dividends has to be identified (sharing must be done between the five usual agents).
Of course, the complexity of formulations (and even the identification of elements such as dividends) depends on the
role of the model, whether it is used by researchers trying to answer global theoretical issues, or by policy advisers
addressing in detail the evolution of the next State budget.
10.2.1.6 The behavior of households
On the contrary, we have not yet addressed the decisions of households.
Basically:
• Households obtain revenue from several sources, the main ones being:
o Wages
o The revenue of individual workers
o Remittances
o Social benefits of various kinds
o Interests from loans in local or foreign currency
o Interests from bills and bonds
o Dividends from equities
o Rent transfers with other households (a service)136
• They use this revenue:
o To pay the income tax
o To consume various goods and services
o To save, and in particular in housing, but also in deposits, bonds, stocks and goods (such as art).
To be considered operational, even a single product model must use some detail, as the economic processes through
which these revenues are obtained, and the consequences of spending decisions, are quite different from each other.
136
Actually, it would be strange to consider that if a household buys the apartment it was renting, the service
disappears and GDP decreases. For that reason, housing owners are considered by National Accounts as paying to
themselves a fictitious rent.
361
Another principle of modelling: favor the detail which allows separating behaviors.
Basically
• For revenue:
o Wages paid by firms should be the product of an average wage rate (coming from the price block) by the
number of workers (from the production block).
o The number of civil servants will generally be exogenous, but not the wage rate, which can be different from
firms’.
o Wages paid by households (mainly for housekeeping) can be identified or not, according to the type of
model.
o Remittances will depend on the wage rate in foreign countries, corrected by the exchange rate.137
o Social benefits are generally separated in five types: sickness, family subsidies, unemployment benefits,
invalidity from working accidents, pensions. It is clear that:
Each of these elements depends on inflation, but at different degrees.
Most of them depend on population, and often a given type of population. For instance, the number of
children, the number of people having reached retirement age, or of unemployed.
All of them depend on economic activity, again in a variable way. For instance, unemployment benefits
decrease with GDP, working accidents increase, and pensions should increase (in principle) with the revenue
from the contributions which finance them.
They also depend on a decision made by the State (for instance the purchasing power is maintained).
This means an operational model should try to separate these items, to take into account their differences in
behavior.
In this way the model will show naturally:
The change in benefits with the number of beneficiaries.
The change in benefits with the government policy.
o The interests will be described globally, in a subsequent paragraph (and in more detail in the chapter on SFC
models). Let us only stress that for households the interest rates (lending and borrowing) can be deviate from
market values through state intervention. In France, a limited number of savings benefits from a higher
guaranteed rate, and borrowing to buy housing can be done at a lower rate (0% in some cases).
o Dividends will be treated later with the firms’ account.
o As to the revenue from housing (rents), its role in a model is limited, as it mostly represents a transfer from
households to other households. For owners of real estate, it is even a transfer within the same household.
There are reasons to consider it, however: it can be subject to taxation, and it enters GDP.
It is not essential to consider marginal elements, such as lottery winnings, inheritance, donations, fines...
o Finally, one can formalize the transfers from abroad (or to abroad). For developing countries remittances can
represent a sizable share of household revenue (more than one third of GDP for Tajikistan). For a single
country model, they should be exogenous, perhaps even in current terms (a notable exception to the general
principle, with particular consequences).
• For expenditures:
137
Modelling the number of expatriates is an interesting issue.
362
o The income tax should be computed as a rate applied to revenue before tax, obtaining the historical values of
the apparent rate by dividing the amount by the base. The model will then get the tax by applying the
exogenous rate to the base. This base poses the usual timing problem: the tax can be paid after the revenue
is obtained (with a provision mechanism).
Also, applying an average rate to all households can be acceptable for forecasts (which allow this rate to change with
time), but less for the shocks addressed to a category of households at one extremity of the spectrum: in a traditional
macroeconomic model, a decrease in the tax on large incomes or an increase in benefits for the poor, of the same ex
ante size, will have the same ex post consequences 138. To eliminate this error, an ad hoc correction has to be made on
the savings rate itself.
This problem appears in most models, coming from the fact is that the tools to solve it are not available. National
Accounts separate firms using the goods they produce, but not households according to any categorization, including
the level of revenue. Some surveys address the problem, and their teachings could be used to create specific data.
This means some solution might be found, but without doubt at a high cost. Actually, the same problem arises if one
wants to separate firms not according to sectors, but to size, considering that small firms act differently from large
ones.
o Once the disposable income is known, the first task is to separate it into consumption and savings,
considered as whole in most models (for multi-product models the situation will be more complex).
The most common technique is to compute consumption first, as a ratio to revenue, then savings as a residual. We
shall develop this with estimations.
Consumption is generally determined at constant prices (which means in purchasing power). The usual determinants
are:
o The level of revenue (measured also in purchasing power). The higher the revenue, the higher the
consumption level, but the lower the share of consumption (the poor do not save, and remember that buying
a house is considered as savings).
o The recent evolution of revenue. Households take some time in adapting their behavior to an increase (or
decrease) in revenue. And a sudden hike (especially if it is destined to be permanent, like a promotion) can
lead them to invest in housing, which can actually decrease consumption for a while.
o Inflation (the “real holdings” or Pigou effect). Present savings contain a large share of monetary elements
(deposits, bonds with fixed rates...). Current inflation reduces their purchasing power, which has to be
complemented by additional savings. The effort is proportional to the inflation level.
o The unemployment rate. For employed workers, an increase in the chance of losing their jobs (measured
more by the change in the rate than its value139) leads them to save a larger share of their present revenue, if
they want to optimize their utility across time.
o The (short-term) interest rate: in general, people prefer satisfying a given need now than later. But this has a
cost, the interest they have to pay. The lower the rate, the more they will indulge in immediate consumption.
This is particularly true for durable goods: if a household wants to watch TV on a set using the latest technology (and
thinks that after its purchase, in its whole life it will have enough resources to afford this kind of set) the only reason
for not buying one right now (and increasing its satisfaction for a period) is the actualized cost, which is lowered with a
decrease in interest rates. What it has to consider is not the cost of the good, but the cost of making its acquisition
earlier.
138
Of course, the impact on consumption will be higher if the increase concerns the poor.
139
Although the actual rate plays also a role: a higher value implies a higher turnover, and a high risk of participating in
the turnover.
363
If the good is perfectly durable, and can be sold back at its original value at constant prices, things are just as if it was
renting the good. If the interest rate is divided by 2, the “price” of the increase in satisfaction is divided by 2.
For non-durable goods, the situation is different. The household has already optimized its consumption over time. If
the interest rate changes, it might be tempted to consume earlier, but if the marginal utility of the good is decreasing
fast, the pattern of consumption will not be much affected. A person dreaming of visiting the pyramids, and saving for
that purpose, might make the trip earlier but will not do it again.
What matters is the real rate:
o They allow comparing goods at constant prices.
o If households assume their revenue will grow with inflation, they will optimize in real terms.
Once consumption is determined, savings are computed as a residual, and generally as a global element. This option
can be discussed, as different kinds of savings can be assumed to follow different behaviors.
In particular, housing investment is negatively affected by interest rates (a specific rate, but one can assume it follows
the global rate) while financial savings are positively so. Buying a house calls for obtaining a given good and asking
another agent to provide the collateral, in return for interests. Buying a bond means lending collateral to another
agent to use it as a spending tool (maybe to buy a durable good), in return for interests but this time in the other
direction.
We will develop this feature when we address SFC models.
10.2.1.7 External trade
In a single country model, the rest of the world is exogenous.
This means that we consider only influences from the world to the country, and not the other way around.
Of course, this is not really exact, even for the smallest of countries (or in that regard for a country region, a town or
an individual): by increasing your consumption and so local production, you create a fraction of a job, a small amount
of household revenue, and again more consumption.
What we consider is that the influence is too small to have a sizable effect, and that the cost of producing (and
running) a model describing it is too high compared to the gain in the accuracy of results. This is essentially true for
lower or medium sized countries like Latvia or Bolivia, much less so for larger countries like France, and quite untrue
for the USA or the European Union considered as a whole. For instance, when we used the MacSim world model for a
shock analysis, the French Keynesian multiplier for 2000 was 1.3 if we ran the full model, but only 1.1 if we ran the
French model by itself. The iterative feedbacks of German imports from France, coming from the increase of German
exports, will have the largest share in the difference. Considering the evolution of world trade, the present difference
should be even wider.
This means that the exchanges of the country have to be considered from the point of the country:
• Exports are the share of production of goods and services which is sold by the country to the rest of the
world.
• Imports are the share of local demand for goods and services which is not produced in the country but
bought from the rest of the world.
364
Both elements will be computed using the currency of the country. However, using elements at constant prices will
mean using the exchange rate of the base year, so the currency issue is not relevant, introducing only a scaling by a
constant factor140.
However, the trade elements having the same nature, their logical determinants will be the same. The main difference
will come only from the relative size of the two markets (buyer and seller) in the trading operation: the single
country’s importance (or GDP) will always be much lower than that of the rest of the world, although this is less
obvious again if we model the USA or the European Union as a whole.
These elements will be:
• Demand: for a country to sell a given good to a partner country, demand for this good must be present, part
of this demand must be addressed to the world market, and the quality of local products must appeal to the
importing country. For instance, French exports of wine will depend on the world demand for wine, and the
natural preference of importing countries for foreign wine (starting with their status as wine producers).and
French wine in particular. And at present Vietnam will not export so much rice and coffee to markets asking
for high quality products. However, one has to consider evolutions of this criterion, as the limits on land
availability and technical progress is changing the nature of supply.
To define world demand, one can:
o Use a global index, provided for instance by an international institution like the OECD or the World Bank.
o Weigh roughly the growth of GDP or better demand in the main trading partners.
o Use the precise weighting of partners and the evolution of their imports, to build an accurate index.
The best but rather complicated option (which we have actually implemented) is to establish a recursive prologue, in
which foreign demand to the country is computed by weighting the global demand from each partner by the share of
this partner in exports. The data requirements are higher, but a combination of local trade statistics and international
data (from the World Bank WDIs?) should provide the information.
The same technique can be used to compute competitors’ deflators, using the same weightings (and different ones for
imports).
This framework allows to study the consequences of detailed shocks (like a decrease in Chinese growth for a
Vietnamese model).
Defining demand introduces a problem.
For imports, we have already seen that including intermediate consumption in the supply-demand equilibrium (thus
considering production on one side and total local demand on the other) is quite a problem for models, as the level of
intermediate consumption depends on the number of steps in the production process. The single product feature has
until now eliminated the need for considering intermediate consumption. But imports do contain intermediate goods,
whether they represent energy (oil, gas, even electricity) or primary goods (from untreated wood to electronic
components). And these intermediate goods are necessary to export.
A simple solution is to consider the ratio of intermediary consumption to value added. Looking at the figures, we can
indeed observe that the “technical coefficients”, the number of units needed to produce a unit of value added or GDP,
is rather constant.
140
This is only true if we consider a single rest of the world, or we measure it in a single currency. More on this later.
365
So we have just to consider a composite demand, as the sum of final demand itself, and intermediate consumption as
a function of GDP (or rather value added, as intermediate consumption excludes VAT).
• Price competitiveness: to decide whether to buy a good from a local or foreign producer, a country will
compare the local price with the foreign exporters’ price. And to choose among potential sellers, the
importing country will consider their relative price at a given quality (remember that the deflators consider
goods at the same quality level, an increase in quality improving the value at constant prices instead of the
deflator).
We generally observe that the relative price is less of an issue when the buyer contemplates buying local or foreign
goods, than when he has to choose between foreign sellers. This follows economic logic: local goods are supposedly
designed for the local market, and some goods are not or hardly substitutable (local bus tickets, meals in restaurants
or newspapers).
This means in our case that the sensitivity of exports to price competitiveness should be higher than for imports.
Exports depend on world demand to the world market, and once a country has decided to import, the price will play a
more important role than in the import decision itself.
Finally, of course measuring competitiveness must use deflators defined in the same currency. It can be any currency,
as applying the same exchange rate to both elements of the ratio will not change its value. In the case of exports, this
means that measuring their deflator in local currency calls for a foreign price measured in the same units. As the
exchange rate is identified, this foreign price will be endogenous, as the product of two assumptions: the foreign price
in foreign currency, and the correcting exchange rate (a deflator). It is perhaps more logical (and equivalent in
practice) to consider both prices in foreign currency, the local one being corrected by the symmetric exchange rate.
To compute precisely this element, one can use the same technique as for foreign demand: weighting foreign
production prices by the share in the country’s imports and exports (giving two separate indexes).
• The available capacities
The third element is the potential to supply additional demand, which means the presence of available productive
capacities. The relevant variable is naturally the rate of use of capacities, independent from the size of the economy.
The choice of this option is not so straightforward, however. One could argue that as long as the rate is lower than
one, additional demand can be satisfied. We have already shown that this is not true: demand concerns a wide range
of products, and one cannot generally be substituted for another, in particular in the short-term. And some products
may see their capacity completely saturated.
Let us explain the process again, this time in an import-oriented way.
Actually, the average rate of use is based on a distribution of rate values, from zero (hopefully a few cases) to one
(probably a sizable number). When global demand increases, it addresses a range of firms, in variable intensity. Some
of these demands will be addressed to firms already unable to provide more, and some others will make them reach
that level. The proportion of firms working at full capacity will grow. Of course, in some cases, another available
product can represent a substitute. But the simplest option is to import the same product, as the missing product
should be available somewhere in the world (maybe at a higher price, but this should be treated by the price
competitiveness).
The “missing” demand increases with the share of firms which cannot increase their production, having reached
capacity.
Of course, this phenomenon applies essentially in the short-term, as firms will react by investing, which will increase
capacity and close the output gap with time. But this process can be slow, even if full adaptation should be obtained in
the very long run.
366
But if we follow the above reasoning, we observe:
o That the larger the country, the lower the probability that a given absolute, but also relative increase in
demand will face local supply problems. This increase in demand will be more diversified, and the available
capacities will be more diversified too141. At the world level, the issue disappears.
o That in our case, the rest of the world should not face any supply problem, which means that for both our
country’s imports and exports, only the local rate of use should be taken into account.
Of course, this is not true for a multi-country model. This issue will be addressed later.
• And a last condition can appear for the exporting country. If the world requires a given good, the
characteristics of that good produced in the country must also be adapted to the particular demand, which
can be quite specific. For instance, facing an increase in the foreign demand for cars, a country might be able
to supply them at a competitive price, but the type of cars they propose might be unsuitable. This might
concern the size, the metrics used, the sophistication, the powering energy, the security features... Of course,
one cannot consider going in such a detail in a model, even if the data was available (which it is not).
Unfortunately, finding an element describing this feature is less straightforward than above, especially for a single
product model. The simplest idea is to use the age of capital, assuming that a younger productive process will be
better adapted to present demand142. For instance, a recently built car factory might follow market trends in
producing smaller cars, or more energy efficient ones. The age of capital can be derived simply from the chronology of
investment and depreciation, if we consider that this depreciation applies equally to all generations of present capital,
or that capital disappears brutally a given number of years after its implementation. Another assumption leads to
more complex, but manageable, formulas.
• A last element (linked to the previous one) is the share of foreign investment in the productive capacity. The
product from foreign firms can be better adapted to foreign demand (and they might have been designed for
the foreign market in the first place). This feature is implemented in some of or models
10.2.1.8 The Budget
In operational models, describing fully and consistently the Government budget is an absolute requirement.
This is true even if the model is not going to be used by Government advisers, but rather by experts in applied
economics. A crucial goal of these researchers is to assess the consequences for the economy of Government
decisions, external events, or structural changes, considering the most probable impact or the range of possibilities,
possibly under different model formulations (like different options on the interest rate). The approach might be more
or less applied (the advisers might try to produce an image of the next budget, to be presented to the Parliament, and
the scientists will try to see how the adoption by the Central Bank of a Taylor rule will stabilize the economy), but the
tool required is quite similar.
As we have stated above, the best way of defining the associated equations is to build identities, computing an
endogenous revenue or expenditure as the product of an endogenous base by an exogenous rate. The equations will
hold true over the past, and the modeler will be responsible for (and allowed to) establishing future assumptions on
the rate. Of course, he does not have to keep this rate constant, and can rely on an estimated trend as a starting base.
But the final decision will be his.
141
This would not happen if the additional demand was in a specific good.
142
Especially foreign demand as its role is increasing with time, so its influence on the nature of investment will be
higher in later periods.
367
This technique answers to the following objection: if we consider VAT, even with constant legal rates, the apparent
rate will change (grow) with the affluence of households, able to increase the share of highly taxed products in their
consumption. One solution is to establish a trend, used as a base value, and to deviate from this trend as a policy
decision.
If these principles are followed, it will be possible to produce a table showing the evolution of all budget elements, in
current terms and in GDP points, both features being obviously required for presentations.
Another important principle of modelling: if you cannot choose between the possible presentations for a given
concept (value at constant prices, at current prices, growth rate, and ratio to another variable) just look at how this
concept is presented in economic publications (focusing on the ones designed for the general public). Or wait until you
will have to use the figures in your own presentations, then measure your reaction and that of the public.
10.2.1.9 Financial and monetary elements (basic version)
This feature will be much improved in the chapter on Stock-Flow Consistent models. For now, we give only some pre-
SFC elements, for modelers with limited interest in financial features.
In any model, this represents the most variable and controversial part. The first models had little or no financial
equations. Even at this stage, the financial block can be limited to the definition of a few rates, and their impact on the
real sector (these rates can even be exogenous, generally in real terms). On the contrary, this block can be so
developed that the purpose of defining a real sector can be considered as a way to complete the links between
financial elements, for instance describing the creation and origin of additional lending if a decrease in interest rates
draws investment upward.
In our opinion, even a real side oriented model should include:
• A base interest rate set by the Central Bank of the country.
• A short and a long-term rates in the currency of the country.
• An average rate on current net borrowings.
• A rate on the present debt, being computed from the chronology of past rates, perhaps as an autoregressive
function.
• One or several foreign rates, applied to borrowings in foreign currency both in the country and in the rest of
the world.
• The net interests paid by all (five) agents, considering two currencies for the interests paid to the Rest of the
World.
From this basic option, developments can consider:
• Identifying the debt of agents (or their financial holdings).
• Separating it into currencies (local, US Dollars, maybe Euros for non EMU countries).
• Separating it into short-term and long-term.
• In addition, one or more forms of money supply can be formalized.
Most of these equations should be established as identities, based on available data or assumptions. Exceptions can
concern:
• The Central Bank rate, following perhaps a Taylor rule, but not necessarily. Actually, the same model should
allow several options (using a separating parameter).
• The short-term and long-term rates could include a risk premium, depending for instance on the current
budget deficit or its most recent values.
368
• The spread between long and short-term could depend on growth expectations (truer if they are partly or
totally rational) and the health of the local economy.
We will develop all these aspects later.
369
10.2.2 THE EVIEWS PROGRAMS
We shall now present EViews programs producing the data and the framework for a model of the type we have just
described.
Note: all the programs presented here are available on the EViews site, or on request from the author (who shall
provide an access through Google Drive)...
Just as in the previous example, we shall start by stating completely the identities, but limiting the definition of the
behavioral equations to a declaration of the type of function we intend to estimate (this will be done later).
As we have stated before, producing the data and stating the equations can be done in any order, but creating the
groups and predicting the residual check must wait for both tasks to be completed.
However, we shall separate the program which creates the data, and the one which creates the equations (and
produces the residual check).
The reasons for this shall be made clear later, but we can already state the main issue: during the model building
process, both tasks will evolve, but separately most of the time. Changing the model specifications will be much more
frequent than changing the data set, and we want to avoid running a task without reason.
It seemed to us that our presentation would be clearer if we started with model production. It might be also the more
natural sequence for the model builder: first considering the scope of data available (without studying individual
series in detail143) then writing down the equations in sequence, observing more precisely if each series is present
(directly or through a transformation), then creating the data building program.
10.2.2.1 General elements on running programs
In the general case, you just have to access the program and run it.
But you can also run part of a program, by selecting it with the mouse (in the usual Windows way), clicking on the
right button, and choosing “Run Selected”.
This is generally more efficient than the previous method of copying the selected part into a blank program, and
running it. However, the new method does not allow editing, useful when one wants to run a selected AND modified
set.
Symmetrically one can exclude temporarily from execution part of a program, by “commenting it out”. To do this, one
should select the relevant part, click on the right button, and choose “Comment Selection”. To reactivate the
statements, one should select them again and use “Uncomment Selection”.
This can be a little dangerous, especially if you (like myself) have the reflex of saving the program before each
execution. To avoid destroying the original, one can save first the modified program under another name 144.
Finally, one can ask a column of numbers to be displayed left of the program lines, switching the “LineNum” option to
“+”.
143
For instance one can start from the list of variables presented at the end of the book.
144
Only once of course.
370
This is particularly efficient if you use the “Go To Line” statement 145, but also to locate the error position when the
error message provides the information.
Unmatched parenthesis in "_GR_G1.LEGEND COLUMNS(2" on line 10.
The only drawback is that the statement has to be displayed in a single line, which requires activating the “Wrap+”
option. This means long statements will be cut to the screen size.
10.2.2.2 Producing the model
Let us now see how the economic elements we have described can be combined into an EViews program, along the
lines of the simpler case.
In this program, we shall expand the principles of the method we have outlined earlier, to create the framework of a
small but operational model. As in the example, we shall stop before any estimation. What we should get is a set of
equations in which:
• Identities are fully defined.
• Behavioral equations are defined as identities presenting, using specific notations, the explained variable and
its explanatory elements.
The following text will essentially contain the EViews statements, and comments present in the file itself (a quote
mark appears at the start of the line). Additional comments will appear without quote mark, and use a different font.
They are generally associated to features requested longer explanations, which would have taken too much place in
the program.
‘ We start with the usual definition of the directory
cd "c:\program files\eviews11\book"
‘ We state that any test results will be exported to a Rich Text Format file named _mod_1.rtf
‘ output(r) _mod_1 would create a basic text file.
output(r,o) mod_1
‘ We close the input and output file
145
However, you must be careful to update the numbers when the program changes.
371
‘ in case a version is already open
close data_1
close mod_1
open data_1
wfsave mod_1
‘ We define the sample period as the maximum available
‘ restrictions will appear with actual estimations
smpl 1962Q1 2017Q4
' We now need to create a new (blank) model
‘
‘ Trick
‘ We give the model a name starting with « _ »
‘ « _ » is the first element in any alphabetical ordering
‘ As the workfile elements are displayed in alphabetical order
‘ this means it will appear first in the display
‘ (of course among elements using the same trick).
‘ This is quite useful for frequently accessed elements
‘ In case the model already exists, we delete it (the “noerr” option allows failure if the model does not
exist).
delete(noerr) _mod_1
‘ So this means we start afresh
If we do not do this, the previous equations will be retained, even if the new equations redefine the same variables
(even using identical formulations). This means variables would be defined twice, which EViews will obviously refuse,
producing an erroneous model. This happens because identities are « appended » as text to the model specification.
We will not present them, as we consider much more efficient to state the full model in the same program. We have
already presented the reasons for this method: clarity of the model text, better error identification, easier transfers,
and easier management of programs.
This is not true for estimated equations, as we shall see later.
model _mod_1
372
‘ We define the f scalar, for the production of the behavioral « identities »
scalar f
We start with the production block
‘------------------------------------------------
‘ The production block
‘------------------------------------------------
‘ The rate of use of capacities is the ratio of actual GDP to potential GDP
_mod_1.append UR=Q/CAP
‘ GDP balances the supply – demand equilibrium
_mod_1.append GDP+M=FD+X
‘ To get value added we substract VAT at constant prices, the product of the base year VAT rate by final
demand excluding VAT, at constant prices
Identifying value added is necessary to compute firm’s margins and deflators excluding VAT.
In a model at constant prices, the VAT rate would be more or less constant, and VAT proportional to GDP. But here the
use of VAT at current prices, and a deflator excluding VAT, calls for excluding VAT from a denominator measured at
constant prices. We shall use value added Q in the production function (even though GDP would provide a similar
explanation). .
_mod_1.append Q=GDPM-r_vat0*FD/(1+r_vat0)
‘ Investment depends on GDP, the rate of use, profitability, previous capital and the long-term real interest
rate. If we consider capital – labor substitution, it will depend on the relative cost of factors.
373
We see that we can already introduce lags (for the previous level of capital) and elements of formulations (for the real
interest rate).
_mod_1.append I=f*(Q+UR+RPROF+K(-1)+(IRL-100*@pchy(PC))+RELC)
We have chosen to measure capital at the end of the period.
It is the sum of the non-discarded previous capital
and the investment implemented during the period.
Actually what we consider here is the investment purchased during the period. There is no guarantee that it can be
used readily for production.
_mod_1.append K=K(-1)*(1-rdep)+I
‘ The change in inventories will depend on GDP
_mod_1.append IC=f*(Q)
‘ Employment by firms too, and also possibly on the relative cost of factors.
_mod_1.append LF=f*(Q+RELC)
‘ Productivity of labor is the ratio of value added to firms’ employment
_mod_1.append PL=Q/LF
‘ Productive capacity depends on firms’ employment and the initial level of capital)
_mod_1.append CAP=f*(LE+K(-1))
374
‘ Total employment introduces civil servants
‘ actually we should also identify households’ employees
‘ such as maids and janitors
_mod_1.append LT=LF+lg
‘ The actual work force depends on employment and the potential work force,
‘ in practice the population in age of working.
_mod_1.append POPAC=f*(LT+pop65)
‘ The unemployed are the jobless inside the work force
_mod_1.append UN=POPAC-LT
‘ The unemployment rate
_mod_1.append UNR=UN/POPAC
‘----------------------------------------------------------
‘ The price block
‘----------------------------------------------------------
‘ The unitary wage cost is the wage
‘ (including contributions by firms)
‘ necessary to produce one unit of value added
_mod_1.append UWC=WR*(1+r_scf)/PL
‘ The value added deflator will depend on UWC
375
‘ and the rate of use
_mod_1.append PQ=f*(UWC+UR)
‘ The production price aggregates the value added deflator
‘ and the price of intermediary consumption
‘ weighted by its influence :
‘ the demand price excluding VAT
‘ It will be needed for defining the trade prices.
_mod_1.append PP=(PQ+tc*PFDXT)/(1+tc)
‘ The deflator of final demand is the ratio of demand
‘ at current and constant prices.
‘ It balances the supply – demand equilibrium at current prices
‘ All other elements are or will be computed elsewhere
_mod_1.append PFD=(GDPMVAL+MVAL-XVAL)/(GDPM+M-X)
‘ We compute also the deflator excluding VAT
‘ by inverting the relation with PFD:
_mod_1.append PFDXT=PFD*(1+r_vat0)/(1+r_vat)
‘The deflators for demand elements depend on the global one.
‘The relation can be estimated, or we can apply an exogenous ratio
_mod_1.append PC=f*(PFD)
_mod_1.append PI=f*(PFD)
_mod_1.append PIG=f*(PFD)
‘The wage rate depends on deflators:
‘PC from the point of view of workers
‘PQ from the point of views of firm managers
‘And also on labor productivity and unemployment
376
_mod_1.append WR=f*(PC+LP+UNR+PQ)
‘The trade deflators in local currency depend on the local and foreign production prices
‘which have to be converted in local currency through the exchange rate
_mod_1.append PX=f*(PP+ppx+ER)
_mod_1.append PM=f*(PP+ppx+ER)
‘ The exchange rate can be exogenous
‘ or depend on the inflation differential (PPP assumption)
_mod_1.append ER=f*(PP+ppx+ERX)
‘ The short-term interest rate can be exogenous
‘ in nominal or real terms
‘ or depend on inflation and the output gap (Taylor formula)
_mod_1.append IRS=f*(IRSX+(IRSR+100*@pchy(PC))+ (150*@pchy(PC)+50*(UR-urd)/urd))
‘ The long-term interest rate depends on a lag structure of short-term rates
‘ possibly with a spread
_mod_1.append IRL=f*(IRS+spread)
‘ The rate on new borrowings is an average of short and long-term rates
_mod_1.append IR=f*(IRS+IRL)
‘ The rate on previous borrowings depends on the previous rate
‘ and the current rate
‘ according to the reimbursement speed
377
_mod_1.append IRM=f*(IRM(-1)+IR)
‘----------------------------------------------------------
The households block
‘----------------------------------------------------------
‘ The firms’ wages
_mod_1.append WF=WR*LF
‘ The civil servants wages
‘ could be different on average from firms’
_mod_1.append WG=WR*lg
‘ Total wages
_mod_1.append W=WF+WG
‘ Social benefits are exogenous in purchasing power per head
‘ They could be separated into risks
_mod_1.append SOCB=socbr*PC*popt
‘ Additional revenue can be linked to value added
_mod_1.append REVQ=r_revq*QVAL
‘ or be exogenous in purchasing power
378
_mod_1.append REVX=r_revx*PFD
‘ More complex formulations can be used
‘ Social security contributions use an exogenous rate
_mod_1.append SCW=r_scw*W
‘ Household income is defined as a sum of its elements
_mod_1.append HI=W-SCW+REVQ+REVX+SOCB
‘ Income tax applies to the revenue of the previous year
‘ again, this can be made more complex
‘ or use the present value of revenue
‘ according to the country’s rules
_mod_1.append ICT=r_ict*HI(-1)
‘ Now we compute disposable income
_mod_1.append HDI=HI-ICT
‘ also in purchasing power
_mod_1.append HRDI=HDI/PC
‘ Household consumption will depend on:
‘ * real disposable income
‘ * inflation
‘ * unemployment
379
‘ * the short-term interest rate
_mod_1.append HCO=f*(HRDI+PC+UNR+(IRS-100*@pchy(PC)))
‘ * a more explicit relation could be used
‘ _mod_1.append HCO=f*(HRDI+HCO(-1)/HRDI(-1)+PC+d(UNR)+(IRS-100*@pch(PC)))
‘ or even:
‘ _mod_1.append Dlog(HCO) =f*(Dlog(HRDI)+Log(HCO(-1)/HRDI(-1)+Dlog(PC)+d(UNR)
‘ +(IRS-100*@pch(PC)))
‘ This will not change the diagnosis on model structure
‘ But will state the type of relationship one wants to estimate
‘ As a personal reminder or a presentation to partners
‘-------------------------------------------------------------
' The firms block
‘-------------------------------------------------------------
‘ Value added at current terms
_mod_1.append QVAL=PQ*Q
‘ The value added tax
_mod_1.append VAT=r_vat*PFDXT*FD/(1+r_vat0)
‘ Gross Domestic Product
_mod_1.append GDPMVAL=QVAL+VAT
‘ And its deflator
380
_mod_1.append PGDPM=GDPMVAL/GDPM
‘ Subsidies are proportional to value added
_mod_1.append SUBS=r_subs*QVAL
‘ Margins are computed as:
‘ value added plus subsidies minus « other indirect taxes »
‘ minus wages including firms’ social security contributions
_mod_1.append MARG=QVAL*(1+r_subs-r_oit)-WR*LF*(1+r_scf)
‘ The margins rate
_mod_1.append RMARG=MARG/QVAL
‘ The tax on profits is based on past profits
‘ A more complex rule can be applied
_mod_1.append IFP=(PROF(-1)+IFP(-1))*r_ifp
‘ Profits exclude:
‘ * household revenue from production
‘ * the tax on profits
‘ * net interests paid
_mod_1.append PROF=MARG-REVQ-IFP-NIF
‘ The profits rate applies to capital
‘ at the cost of renewal
381
_mod_1.append RPROF=PROF/(PI*K(-1))
‘ We can also compute the ratio of margins to capital
‘ a more stable and reliable concept
_mod_1.append RPROB=MARG/(PI*K(-1))
‘ The balance of firms excludes spending on
‘ * productive investment
‘ * the change in inventories
‘ as margins includes yet unsold value added
_mod_1.append FCAPF=PROF-PI*I-PFD*IC
‘ Net interests paid depend on
‘ * their past value
‘ * the rate on past debts
‘ * the rate on new debts
‘ * the balance
_mod_1.append NIF=f*(NIF(-1)+IRM+IR+FCAPF)
‘ -------------------------------------------
' The external trade block
‘ -------------------------------------------
‘
The import price including tariffs
_mod_1.append PMT=PM*(1+r_tar)/(1+r_tar0)
‘ Import price competitiveness compares
‘ * the local production price
‘ * the import price including tariffs
_mod_1.append COMPM=PMT/PP
382
‘ Final demand at constant prices is the sum of its components
‘ including a residual proportional to GDP
‘ (a reasonable assumption)
‘ Housing investment could be identified
_mod_1.append FD=COH+I+IC+CG+IG+fdxr*Q
‘ Imports depend on
‘ * Final demand and intermediate demand
‘ (proportional to value added)
‘ * The rate of use of capacities
‘ * Price competitiveness
_mod_1.append M=f*(FD+Q+UR+COMPM)
‘ Export price competitiveness compares
‘ * the foreign production price
‘ * the local export price including tariffs
‘ The exchange rate corrects the currency difference
_mod_1.append COMPX=PX*(1+r_tarx)/(1+r_tarx0)/(PPX*ER)
‘ Exports depend on
‘ * World demand (both final and intermediate)
‘ * The rate of use of capacities
‘ * Price competitiveness
_mod_1.append X=f*(WD+UR+COMPX)
‘ Trade flows are computed in current terms
_mod_1.append MVAL=PM*M
_mod_1.append XVAL=PX*X
383
‘ The export import ratios are computed:
‘ * at current prices
‘ * at constant prices
‘ * for the deflators
_mod_1.append RCVAL=XVAL/MVAL
_mod_1.append RCVOL=X/M
_mod_1.append TTRAD=PX/PM
‘ The trade balance
_mod_1.append TRB=XVAL-MVAL
‘ Net interests paid to the Rest of the world
‘ are separated depend on currency.
‘ They depend on:
‘ * their past value
‘ * the rate on past debts
‘ * the rate on new debts
‘ the international rate is used for interests
‘ in foreign currency
‘ * the balance
‘ * the exchange rate for the debt in foreign currency
_mod_1.append NIXL=f*(NIXL(-1)+IRM+IR+TRB)
_mod_1.append NIXX=f*(NIXL(-1)+IRMX+IRX+TRB+ER)
_mod_1.append NIX=NIXL+NIXX
‘ The financing capacity
_mod_1.append FCAPX=TRB-NIX
‘------------------------------------------------------
' The State budget block
‘------------------------------------------------------
‘ Most of its elements have been computed already
384
‘ as transfers between the State and another agent
‘ Social contributions paid by firms
_mod_1.append SCF=r_scf*Wf
‘ Other indirect taxes
_mod_1.append OIT=r_oit*QVAL
‘ Tariffs
_mod_1.append TAR=r_tar*MVAL
‘ Social contributions paid by the State
_mod_1.append SCG=R_SCG*WG
‘ Total revenue
_mod_1.append REVG=SCF+SCG+SCW+OIT+IFP+ICT+VAT+TAR+r_revg*QVAL
‘ Government investment at current prices
_mod_1.append IGV=IG*PIG
‘ Government consumption at current prices
_mod_1.append CGV=CG*PFD
385
‘ Government demand at current prices
_mod_1.append FDGV=CGV+IGV
‘ Net interests paid by the State depend on
‘ * their past value
‘ * the rate on past debts
‘ * the rate on new debts
‘ * the balance
_mod_1.append NIG=NIG(-1)*IRM/IRM(-1)-IR/100*FCAPG
‘ Total expenditures
_mod_1.append EXPG=FDGV+WG+SUBS+SOCB+NIG+SCG+r_expg*QVAL
‘ Government balance
_mod_1.append FCAPG=REVG-EXPG
‘ Government balance in GDP points
_mod_1.append FCAPGP=100*FCAPG/GDPMVAL
‘ Total GDP
‘ The sum of market and non-market GDP
‘ equated to the total wage cost of the State
_mod_1.append GDPVAL = GDPMVAL+WG+SCG
386
‘--------------------------------------------------------------
‘ End of model specifications
‘--------------------------------------------------------------
387
10.2.2.3 Producing the data: an OECD example
Now we will present a program creating the data for the above model. We have used a very simple case, in which the
model builder has access to the OECD “Economic Perspectives”, a file containing about 5000 quarterly series
describing the world economy. Each OECD country is described individually using the same concepts (with a few
exceptions), and other important countries (like China) or zones (like Latin America) using less detail. However, the
definitions vary slightly across countries, and although the same set of series is always technically present, some of
them do not contain any value, or very few. For instance, the notion of savings is not always the same, and capital
stock is not always available.
As we shall see, our base contains all the data we need for France, with one or two minor exceptions.
Obviously, this program is quite case dependent, and a user starting from another base will use a rather different one.
A program with the same purpose is available for the World Bank World Development indicators, with a lot more
countries and series, but an annual periodicity.
However:
• The user might actually start from this OECD base, a quite popular one.
• Some of the tasks (creating the workfile, producing a trend...) will appear in all cases.
• And actually the program will not be so different. Almost of the variables accessed in the OECD base will be
available in any base of this type, and one will just have just to replace the “OECD” names, changing some of
the concepts if needed.
The only additional problems for the prospective user of the program will come from:
• Unavailable data which has to be estimated, or guessed.
• Additional data required by the introduction of new elements in the model.
‘==================================
' An example of data transfer
'==================================
' This program will start from the original French data
' provided by OECD Economic perspectives and named fra_*,
' the prefix used by OECD to identify French statistics
' We decide on the directory
' This is not generally necessary
' except if one works on several projects
' or maintains several directories for the same project
' It guarantees trial versions do not destroy official ones
cd " d:\eviews\_pic_2020\pic_eviewsf"
' It will create the data for our model
388
' with the prefix f_*
'-------------------------------------------------------------------------
' This technique can be used with any source
' where the original series use the same prefix
' the results will be created with any different prefix
' In the best case, if the set available is the same (or larger) than the OECD set
' one has just to replace the OECD names in the following statements
'-------------------------------------------------------------------------
' we close the original file fra_1
' containing the sole fra_* data
' and some global OECD series named OECD_*
' we close also the file which will receive the French data
' in case it is already open
' having two versions of the same file open in memory is quite dangerous...
close fra_1
close data_1
' We open the original file (presently closed)
' and save it under the name data_1 for the French data
open fra_1
save data_1
' Now the file should contain only original data
' called fra_*
' we delete any existing f_* series
' just in case, this should not happen...
delete f_*
' We have to make an assumption on the sharing of indirect taxes
389
' into VAT and other indirect taxes
' as OECD provides only a global variable
' p_oit = assumption on the share of oit in indirect taxes
'
scalar f_p_oit=0.2
' we create a time trend
' with the value of the year for the first quarter
' to which we add 0.25 for each following quarter of the year
' 1994 : 1994Q1
' 1994.25: 1994Q2
' 1994.50: 1994Q3
' 1994.75: 1994Q4
' 1995.00: 1995Q1 ....
' This will be quite useful to:
' create yearly time trends
' replace actual dummy variables by expressions using logical conditions
' much easier to manage
smpl 1962Q1 1962Q1
genr f_t=1962
smpl 1962Q2 2017Q4
genr f_t=f_t(-1)+0.25
' Now we start with the supply - demand equilibrium
smpl 1962Q1 2004Q4
genr f_gdpval=fra_gdp
‘ Gross domestic product at current prices
genr f_gdp=fra_GDPV
‘ Gross domestic product at constant prices
'--------------------------------------------------------------
390
' Supply
'--------------------------------------------------------------
' The model separates market GDP
genr f_gdpmval=fra_gdp-fra_cgw
‘ = gross domestic product at current prices – Government wage consumption at current prices
genr f_pcog=fra_cg/fra_cgv
‘ = Government total consumption at current prices / = Government total consumption at constant prices
genr f_gdpm=fra_GDPV-fra_CGw/f_pcog
genr f_pgdpm=f_gdpmval/f_gdpm
' Trade at current and constant prices, deflators
genr f_m=fra_MGSV
genr f_x=fra_XGSV
genr f_pm=fra_PMGS
genr f_px=fra_PXGS
genr f_xval=f_px*f_x
genr f_mval=f_pm*f_m
' The deflator of final demand (including VAT and OIT)
genr f_pfd=(f_GDPmVAL+f_MVAL-f_XVAL)/(f_GDPm+f_M-f_X)
genr f_fd=fra_TDDV-fra_CGw/f_pcog
' Identifying the indirect taxes This separation is important as VAT applies to final demand
genr f_oit= f_p_oit* fra_TIND
genr f_vat = (1-f_p_oit)* fra_TIND
genr f_r_vat =f_vat/(f_fd*f_pfd-f_vat)
smpl 2014q1 2014q4
391
scalar f_r_vat0=@mean(f_r_vat)
scalar f_r_oit0=@mean(f_r_oit
smpl @all
genr f_r_oit=f_oit/(f_gdpmval-f_vat-f_oit)
genr f_pfdxt=f_pfd*(1+f_r_vat0)/(1+f_r_vat)
' Value added excluding VAT (but including OIT)
genr f_qval=f_gdpmval-f_vat
genr f_q = f_gdpm-f_r_vat0*f_fd/(1+f_r_vat0)
genr f_pq=f_qval/f_q
' Capital, rate of use, capacity
genr f_k=fra_KTPV
genr f_ur=fra_GDPV/fra_GDPVTR
genr f_cap=f_q/f_ur
genr f_pk=f_cap/f_k(-1)
genr f_urd=1
'-------------------------------------------------------------
' Demand
'-------------------------------------------------------------
' Intermediate consumption
' Not identified in the OECD data base
‘ We specify the ratio as 1 (a common value)
genr f_tc=1
genr f_ic=f_tc*f_q
genr f_td=f_fd+f_ic
' Production price
genr f_pp=(f_qval+f_ic*f_pfdxt)/f_td
392
' Elements of demand and their deflators
' Household consumption
genr f_coh=fra_CPV
genr f_pcoh=fra_cp/fra_cpv
' Investment, depreciation rate
genr f_i=fra_IBV
genr f_pi=fra_ib/fra_ib
genr f_rdep = (f_k(-1) + f_i - f_k) / f_k(-1)
' Housing investment
genr f_hih=fra_IHV
' Government investment (excluding State firms)
genr f_ig=fra_IGV
genr f_igv=fra_IG
genr f_pig=f_igv/f_ig
' Government consumption (excluding State firms)
genr f_cogv=fra_cg-fra_cgw
genr f_cog=(fra_cg-fra_cgw)/f_pcog
' Government demand
genr f_fdgv=f_cogv+f_igv
genr f_fdg=f_cog+f_ig
' Change in inventories
genr f_ci=fra_iskv
393
' Individual demand deflator ratios
genr f_r_pi=f_pi/f_pfd
genr f_r_pcoh=f_pcoh/f_pfd
genr f_r_pig=f_pig/f_pfd
genr f_r_pcog=f_pcog/f_pfd
' Wage rate
genr f_wr=fra_WAGE/(fra_ET-fra_ES)
genr f_r_scf = fra_WSSS/fra_WAGE -1
' Checking that the decomposition of final demand is correct
genr f_fdxr=(f_fd-f_coh-f_i-f_hih-f_ic-f_fdg)/f_q
'-----------------------------------------------------------------
' employment, unemployment, population
'-----------------------------------------------------------------
' Employment
genr f_lt=fra_ET
genr f_lg=fra_EG
genr f_lf=fra_ETB
' Labour productivity
genr f_pl=f_q / f_lf
' Wages and wage cost
genr f_w=f_wr*f_lt
genr f_wf=f_lf*f_wr
genr f_uwc=f_wr*(1+f_r_scf)/f_pl
' Unemployment
394
genr f_un=fra_UN
genr f_unr=fra_UN/(fra_ET+fra_UN)
' Population
genr f_pop=fra_POPT
genr f_pop65=fra_POPT
genr f_popac=f_lt+f_un
'-------------------------------------------------------------------------------
' Financial elements
'-------------------------------------------------------------------------------
' Exchange rate
genr f_er=1/(fra_EXCHEB/@elem(fra_EXCHEB,"1995"))
genr f_erx=f_er
' Nominal interest rates
genr f_irs=fra_IRS
genr f_irl=fra_IRL
genr f_ir = fra_irwyp
genr f_irm=f_ir
genr f_irsx=fra_irsaf
genr f_irlx=fra_irlaf
genr f_irmx=fra_irfor
genr f_irx=fra_irfor
‘ Real interest rates
‘ Here we assume the formulas used for the short-term
‘ and the average interest rate on past debts
‘ their logic will be explained later
genr f_irr=f_ir-100*@pchy(f_pcoh)
395
genr f_irl_ec=0
genr f_irsr=f_irs-100*@pchy(f_pcoh)
genr f_irst=f_irs-(150*@pchy(f_PCoh)+50*(f_UR-f_urd)/f_urd)
scalar p_f_irm=0.8
genr f_irm_er=f_IRM-(p_f_irm*f_IRM(-1)+(1-p_f_irm)*f_IR)
' relative cost (for substitution between labor and capital)
genr f_relc=f_wr*(1+f_r_scf)/f_pi/(f_ir/100-@pchy(f_pcoh)-4*log(1-f_rdep))
genr f_spread=3
'---------------------------------------------------
' Households
'---------------------------------------------------
' Global elements
genr f_hi =fra_YRH-fra_TRPH
genr f_ict=fra_tyh
genr f_r_ict = fra_TYH / f_hi(-1)
genr f_hdi=f_hi-fra_tyh
genr f_hrdi=f_hdi/f_pcoh
genr f_sr=1-f_coh/(f_hrdi)
' Individual elements
genr f_socb=fra_trrh
genr f_scw= fra_TRPH - f_r_scf*f_w
genr f_r_scw=f_scw/f_w
genr f_rpro=f_hi-(f_w-f_scw+f_socb)
genr f_revx=(1-0.5)*f_rpro
genr f_revq=0.5*f_rpro
genr f_r_revx=f_revx/f_pfd
genr f_r_revq=f_revq/f_qval
genr f_wg=f_wr*f_lg
genr f_hdir=f_hdi/f_pcoh
396
'--------------------------------------------------------
' external trade
'--------------------------------------------------------
' Tariffs
genr f_tar=0
genr f_r_tar=f_tar/f_mval
genr f_r_tarx=0
scalar f_r_tarx0=@elem(f_r_tarx,"1995")
scalar f_r_tar0=@elem(f_r_tar,"1995")
' Foreign elements
genr f_wd = fra_XMVMKT
genr f_ppx=OECD_PGDP
' Competitiveness
genr f_PMT=f_PM*(1+f_r_tar)/(1+f_r_tar0)
genr f_compm=f_pmt/f_pp
genr f_COMPX=f_PX*(1+f_r_tarx)/(1+f_r_tarx0)/(f_PPX*f_ER)
' Ratios and balances
genr f_rcvol=f_x/f_m
genr f_rcval=f_xval/f_mval
genr f_ttrad=f_px/f_pm
genr f_trb=f_xval-f_mval
genr f_FCAPX=fra_FBGSV
' Interests paid
‘ In the OECD data set, the interests paid to the RoW are not identified
‘ We apply the formulas used by the model
‘ They will be explained later
397
smpl 1990Q1 2004Q4
scalar p_nix=0.5
genr f_nixl=-f_ir*f_trb/40*p_nix
genr f_nixx=-f_irx*f_trb/40*(1-p_nix)
smpl 1990Q2 2004Q4
genr f_nixl=(f_nixl(-1)*f_irm/f_irm(-1)-f_ir*f_trb/400*p_nix)/(1-f_ir/400)
genr f_nixx=(f_nixx(-1)*f_irmx/f_irmx(-1)*f_er/f_er(-1)-f_irx*f_trb/400*(1-p_nix))/(1-f_irx/400)
genr f_nix=f_nixl+f_nixx
genr f_fcapx=f_trb-f_nix
smpl 1962Q1 2004Q4
'---------------------------------------------------------
' Government budget
'---------------------------------------------------------
' Revenue
genr f_scf=f_r_scf*f_wf
genr f_r_scg= f_r_scf
genr f_scg=f_wg*f_r_scf
genr f_ifp=fra_tyb
genr f_r_oit=f_oit/(f_qval)
genr f_revg=f_ict+f_oit+f_vat+f_scf+f_scg+f_tar+f_scw+f_ifp
genr f_r_revg=0
genr f_recg=0
genr f_r_recg=0
' Expenditures
genr f_socbr = fra_TRRH /f_pcoh/f_pop
genr f_subs=fra_TSUB
genr f_r_subs = fra_TSUB / (f_qval)
' Interest rates and balance
398
genr f_nig=fra_gnintp
genr f_fcapg=-FRA_NLG
genr f_nig_er=(f_NIG-(f_NIG(-1)*f_IRM/f_IRM(-1)-f_IR/400*f_FCAPG))/f_qval
genr f_expg=f_revg-f_fcapg
genr f_r_expg=(f_expg-(f_fdgv+f_wg+f_scg+f_nig+f_socb+f_subs))/f_qval
' Balance in GDP points
genr f_fcapgp=100*f_fcapg/f_gdpval
'---------------------------------------------------------------
' Firms account
'---------------------------------------------------------------
' Margins
genr f_marg=f_qval*(1+f_r_subs-f_r_oit)-f_wf*(1+f_r_scf)
genr f_rmarg =f_marg / f_qval
' Interests paid, profits and balance
genr f_prof=fra_PROF
smpl 1978Q1 2004Q4
genr f_fcapf=-fra_NLB
smpl 1978Q1 1978Q1
genr f_nif=f_ir*f_fcapf/11
smpl 1978Q2 2004Q4
genr f_nif=f_nif(-1)*f_irm/f_irm(-1)-f_ir/400*f_fcapf
smpl 1962Q1 2004Q4
genr f_rprob = f_marg/(f_pfd*f_k(-1))
genr f_prof1=f_marg-f_revq-f_ifp-f_nif
genr f_prof_er=(f_prof-f_prof1)/f_qval
genr f_r_IFP=f_ifp/(f_PROF(-1)+f_IFP(-1))
genr f_rprof=f_prof/(f_pfd*f_k(-1))
genr f_fcapf_er=(f_fcapf-(f_prof-f_pi*f_i-f_pfd*f_ci))/f_qval
save data_1
399
10.2.2.4 Creating the model groups and checking the data – equations consistency
These two tasks can be started only when the model has been defined, and the data created.
Creating the groups needs not only the model, but also the data, as EViews groups can only be built from (existing)
series.
This program can be run after the two tasks, in the same session. This is why we did not introduce the usual
statements specifying the directory and opening the workfile.
The program is model independent. It:
• Creates a list of endogenous and exogenous variables.
• Solves the model equations separately from each other twice, for different values of the scalar « f ».
• Separates the endogenous into identity and behavioral using the fact that the equations for the latter contain
the scalar « f »
• Computes the level and relative differences between historical values and the results of one of the
simulations above.
‘ ==========================================
‘ creating the groups
‘===========================================
‘ We create groups for the endogenous and exogenous
‘ The statement « makegroup »
‘ * applies to the model « _mod_1 »
‘ * creates a group of endogenous (@endog) or exogenous (@exog)
‘ * with a name _g_vendo or _g_vexog
‘ The modifiers specify the elements in the group
‘ « n » means we do not want the suffixed elements
‘ (suffixed by the current suffix, actually « _c »)
‘ « a » means we want the actual names (no suffix)
‘ This overrides strange EViews default options
_mod_1.makegroup(a,n) _g_vendo @endog
_mod_1.makegroup(a,n) _g_vexog @exog
' Now we want to separate identity and behavioral variables
‘ We shall use a trick
‘ The future behavioral equations use the scalar "f"
' We solve the model twice for different values of "f"
' using different suffixes for the solution
' We need to "update" the model
400
' for the parameter change to be taken into account
smpl 1995Q1 20017Q4
f=1
_mod_1.append assign @all _d
solve(d=f) _mod_1
f=2
_mod_1.update
_mod_1.append assign @all _c
solve(d=f) _mod_1
' We create two empty groups
group _g_vbeha
group _g_viden
' Then we consider in turn each endogenous variable
' _g_vendo.@count is the number of elements in _g_vendo
' _g_vendo.@seriesname(!i) is the name of the ith element in the group
' We consider the variable built from this name and the two suffixes "_c" and "_d"
' If the values are different then the variable is behavioral
' We test the condition on a given period
' (but not the base year as there is a chance the right hand side gives zero
' in which case the method does not work)
' If it is true we add the element to the list of behavioral elements
' otherwise to the list of identity elements
' « endif » close the condition
' « next » closes the loop
' the brackets « {» and « }» delimitate the parameter
' they are dropped after the replacement
for !i=1 to _g_vendo.@count
%1=_g_vendo.@seriesname(!i)
scalar {%1}_eq=(@elem({%1}_c,"2014Q1")<>@elem({%1}_d,"2014Q1"))
if {%1}_eq=1 then
_g_vbeha.add {%1}
401
else
_g_viden.add {%1}
endif
next
‘ ==========================================
‘ computing the residuals
‘===========================================
' Now we check the residuals
‘ We define the suffix as « _c »
_mod_1.append assign @all _c
‘ We reduce the sample to a period
‘ for which each equation can be computed
‘ taking into account the presence of lags
smpl 1986Q2 2017Q4
‘ But only for identities
‘ However, we were able to compute the « behavioral »
‘ over the given period
‘ which means that we will be able to estimate
‘ The dc_{%1} are the absolute differences
‘ The pc_{%1} the relative ones
‘ To avoid dividing by zero, we use a trick
‘ We add to the variable a boolean one
‘ Testing if the variable is zero
‘ If it is true we get 0 divided by 1 = 0
‘ If not the computation is not affected
‘
‘ This means we have just modified the outcome of divisions.
‘ so that dividing zero by zero now gives zero
for !i=1 to _g_viden.@count
%1=_g_viden.@seriesname(!i)
402
genr dc_{%1}={%1}-{%1}_c
genr pc_{%1}=100*dc_{%1}/({%1}+({%1}=0))
next
10.2.2.5 Estimating the equations: examples
We shall now apply the above principles to the estimation of equations. Basically, this means replacing in the previous
program the identities declaring the modeler’s intentions by actual equations.
We shall only present the new program sequences, associated with this process. The full model building program is
available as an annex, and includes a summary of the comments we are going to make. It is also available as a .prg file
on the model site.
Of course, the results we shall obtain are specific to the French case. They should only be considered as a (working)
example, and will probably prove much less directly useful than the previous programs.
One will observe that we are not always following a pure and clean methodology. Our opinion is that we behave
better than most. And anyway rather than a perfect element, what we are proposing here an example of the general
methodology used by model builders, including deviations from the “politically correct” path.
For instance, we shall use cointegration only once, and build most error correction models in one step. You can guess
(rightly) that we could not find a formulation fitting the required conditions, either because one of the elements was
stationary, cointegration was not present, or the coefficients in the cointegrating equation were not significant or
satisfactory.
We shall deal with the equations in sequence, considering two cases for the production function: complementary
factors and Cobb-Douglas. The choice will affect other elements, as estimating a different capacity will affect the rate
of use, an explanatory variable.
10.2.2.5.1 The complementary factors case
Let us first deal with the complementary factors case. The framework is very similar to the one we have seen earlier.
Investment and employment are estimated separately.
10.2.2.5.2 Estimating investment
As in the small model, capacity will be defined by capital.
As to investment, it is given by:
𝐼𝑡 = 𝐾𝑡 − 𝐾𝑡−1 ⋅ (1 − 𝑑𝑟𝑡 )
The investment equation will be:
403
𝐼𝑡−1 𝑄𝑡
𝐼𝑡 /𝐾𝑡−1 = 𝑎 ⋅ 𝐾 𝑡−2
+𝑏⋅𝑄 − 𝑐. (𝑈𝑅 ∗ − 𝑈𝑅𝑡 )/𝑈𝑅𝑡 ) + 𝑑
𝑡−1
or in EViews terms:
equation _eq_i.ls(p) i/k(-1)=c_i(1)*i(-1)/k(-2)+c_i(2)*(q-q(-1))/q(-1)-c_i(3)*(@mean(ur)-ur)/ur+c_i(4) +i_ec
In our equation, accumulation rate will come from:
• Its previous value.
• The change of value added over the last period, representing the firms’ expectations for future growth.
• The target change in the rate of use.
The results are rather satisfactory even though the lagged influence looks quite high, which can lead to problems in
the period following a shock..
10.2.2.5.2.1 ESTIMATING EMPLOYMENT (COMPLEMENTARY FACTORS CASE)
Again, we shall use the same framework as in the small model. We start by defining the labor productivity trend. The
same breakpoint tests applied to labor productivity leads to the two following breakpoint periods: 1972Q4 and
1992Q4.
Chow Breakpoint Test: 1972Q4 1992Q4
Null Hypothesis: No breaks at specified breakpoints
Varying regressors: All equation variables
Equation Sample: 1963Q1 2005Q4
F-statistic 1491.695 Prob. F(4,166) 0.0000
Log likelihood ratio 620.8195 Prob. Chi-Square(4) 0.0000
Wald Statistic 5966.779 Prob. Chi-Square(4) 0.0000
Taking into account these breaks, estimating the trend in labor productivity gives:
404
Dependent Variable: LOG(Q/LF)
Method: Least Squares
Date: 04/02/20 Time: 17:58
Sample (adjusted): 1963Q1 2017Q4
Included observations: 220 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 11.46786 0.003141 3651.269 0.0000
T-2017 0.009231 0.000195 47.42078 0.0000
(T-1972.75)*(T<=1972.75) 0.019739 0.000826 23.89066 0.0000
(T-1992.75)*(T<=1992.75) 0.017956 0.000378 47.51469 0.0000
R-squared 0.997510 Mean dependent var 11.05890
Adjusted R-squared 0.997475 S.D. dependent var 0.348982
S.E. of regression 0.017536 Akaike info criterion -5.231159
Sum squared resid 0.066419 Schwarz criterion -5.169457
Log likelihood 579.4275 Hannan-Quinn criter. -5.206242
F-statistic 28840.75 Durbin-Watson stat 0.232734
Prob(F-statistic) 0.000000
11.6
11.2
10.8
.06
.04 10.4
.02
10.0
.00
-.02
-.04
-.06
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
smpl 1962Q1 2017Q4
coef(10) c_plt
_eq_plt.ls(p)LOG(Q/LF)= c_plt(1) + c_plt(2)*(t-2018)*(t<=2018)
+4*log((1+txq)/(1+txn))*(t-2018)*(t>=2018)+c_plt(3)*(t-1972.75)*(t<=1972.75)
+c_plt(4)*((T<=1992.75)*(T-1992.75))+c_plt(5)*(t-2008.25)*(t<=2008.25)+c_plt(6)*(t>=2008.50)
_eq_plt.resids(p)
close _eq_plt
genr log(lpt)=log(q/lf)-resid
genr lfd=q/plt
To the two breaks which take place in 1972 and 1992 (last quarters in each case), we needed to add a third break in
2008Q2, and a dummy variable starting in the third quarter. All these additions are validated by the relevant tests.
As in the previous case, we observe that actual labor productivity is stationary around the trend.
405
This gives us target employment, from which we build the error-correction model:
coef(10) c_lf
smpl 1962Q1 2017Q1
genr lf_ec=0
equation _eq_lf.ls(p) dlog(lf)=c_lf(1)*dlog(lfd)
+c_lf(2)*log(lfd(-1)/lf(-1))+c_lf(3)*(t=1968.25)+c_lf(4)*(t=1968.50)+c_lf(5)*(t=1968)+lf_ec
_eq_lf.resids(p)
close _eq_lf
genr lf_ec=resid
Dependent Variable: DLOG(LF)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/02/20 Time: 22:37
Sample (adjusted): 1963Q2 2017Q1
Included observations: 216 after adjustments
DLOG(LF)=C_LF(1)*DLOG(LFD)+C_LF(2)*LOG(LFD(-1)/LF(-1))+C_LF(3)
*((T=1968.25)-(T=1968.50))
Coefficient Std. Error t-Statistic Prob.
C_LF(1) 0.211368 0.027205 7.769419 0.0000
C_LF(2) 0.151266 0.018871 8.015851 0.0000
C_LF(3) 0.007869 0.002699 2.915794 0.0039
R-squared 0.306256 Mean dependent var 0.000853
Adjusted R-squared 0.299742 S.D. dependent var 0.003294
S.E. of regression 0.002757 Akaike info criterion -8.935719
Sum squared resid 0.001619 Schwarz criterion -8.888840
Log likelihood 968.0577 Hannan-Quinn criter. -8.916780
Durbin-Watson stat 1.056421
.010
.005
.000
.010
-.005
.005
-.010
.000
-.005
-.010
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
As for the small model, we had to introduce a dummy variable, with opposite signs, for the middle quarters of 1968.
During the whole second quarter, a so-called “student revolution” actually blocked the economy, in particular
transportation. GDP went down sharply, but employment did not really follow, as managers guessed (rightly) that this
406
situation was transitory. This calls for a dummy for that quarter (with a positive coefficient). In the next quarter, things
got back to normal, calling for a high growth in GDP, but employment was there to meet it.
One will also observe that we ended the estimation period in the first quarter of 2017. The evolution of employment
was difficult to explain for the second and fourth quarters. And rather than introducing dummy variables (a difficult
task at the end of the sample), we have elected to leave the management of residuals to the forecast step.
10.2.2.5.3 The Cobb-Douglas case
In the framework of a policy model, the complementary factors option is quite questionable and limited. But it is also
quite simple to implement, and leads to easier interpretation of properties. Moving to a Cobb-Douglas specification
might be simple at first sight. For instance we could consider modifying slightly the complementary factors
framework, defining employment and capital separately based on a capacity target, and simply establishing the
capacity using the Cobb-Douglas function.
But then we lose the most interesting property of this framework: taking into account the (endogenous) sensitivity of
labor and capital factors to the ratio of their relative costs.
This means that a shock producing a 1% increase in target capacity would lead in the long run to the same relative
increase in both factors, even if the same shock has modified the ratio of labor and capital costs.
We think this is too simplistic, and our formulas will take into account this effect. This will show for instance that a
decrease in the social security contributions of firms will especially favor employment, and that the effect of a demand
shock on unemployment and wages will favor capital.
The Cobb-Douglas assumption supposes a unitary elasticity of the share of factors to the relative cost.
However, using this assumption calls for developing a much more complex framework. Let us consider its elements in
turn.
10.2.2.5.3.1 THE FRAMEWORK
We shall now describe the framework in detail.
10.2.2.5.3.1.1 MARGINS MAXIMIZATION
In this framework, firms will try to maximize their margins
(1) 𝑝𝑞𝑡 ⋅ 𝑄𝑡 − 𝑤 𝑐𝑜𝑠 𝑡𝑡 ⋅ 𝐿𝐸𝑡 − 𝑘 𝑐𝑜𝑠 𝑡𝑡 ⋅ 𝐾𝑡−1
under constraint of the production function:
407
(2) 𝐿𝑜𝑔(𝐶𝐴𝑃𝑡 ) = 𝛼 ⋅ 𝐿𝑜𝑔(𝐿𝐸𝑡 ) + (1 − 𝛼) ⋅ 𝐿𝑜𝑔(𝐾𝑡−1 ) + 𝛽 ⋅ 𝑡 + 𝛾
which leads us to maximizing
(3)𝑝𝑞𝑡 ⋅ 𝑄𝑡 − 𝑤 𝑐𝑜𝑠 𝑡𝑡 ⋅ 𝐿𝐸𝑡 − 𝑘 𝑐𝑜𝑠 𝑡𝑡 ⋅ 𝐾𝑡−1 − 𝜆 ⋅ (𝐶𝐴𝑃𝑡 − 𝑒𝑥𝑝(𝛽 ⋅ 𝑡 + 𝛾) ⋅ 𝐿𝐸𝛼𝑡 𝐾1−𝛼
𝑡−1 )
relative to both LEt and Kt-1.
Derivation of (3) gives:
−𝑤 𝑐𝑜𝑠 𝑡𝑡 − 𝜆 ⋅ 𝛼 ⋅ 𝑒𝑥𝑝(𝛽 ⋅ 𝑡 + 𝛾) ⋅ 𝐿𝐸𝑡𝛼−1 𝐾𝑡−1
1−𝛼
=0
Or
(4) −𝑤 𝑐𝑜𝑠 𝑡𝑡 − 𝜆 ⋅ 𝛼 ⋅ 𝑒𝑥𝑝(𝛽 ⋅ 𝑡 + 𝛾) ⋅ (𝐾𝑡−1 /𝐿𝐸𝑡 )1−𝛼 = 0
and equivalently
(5) −𝑘 𝑐𝑜𝑠 𝑡𝑡 − 𝜆 ⋅ (1 − 𝛼) ⋅ 𝑒𝑥𝑝(𝛽 ⋅ 𝑡 + 𝛾) ⋅ (𝐾𝑡−1 /𝐿𝐸𝑡 )𝛼 = 0
Dividing (4) by (5) on both sides gives:
(6)𝑤 𝑐𝑜𝑠 𝑡𝑡 /𝑘 𝑐𝑜𝑠 𝑡𝑡 = 𝛼/(1 − 𝛼) ⋅ (𝐾𝑡−1 /𝐿𝐸𝑡 )
which shows indeed the unitary elasticity of the ratio of factors to the ratio of costs.
From (2) and (6) we get:
408
(7) 𝐿𝑜𝑔(𝐿𝐸𝑡 /𝐶𝐴𝑃𝑡 ) = (1 − 𝛼) ⋅ 𝐿𝑜𝑔(𝐿𝐸𝑡 /𝐾𝑡−1 ) − 𝛽 ⋅ 𝑡 − 𝛾
= (1 − 𝛼) ⋅ 𝐿𝑜𝑔(𝛼/(1 − 𝛼)𝑘 𝑐𝑜𝑠 𝑡𝑡 /𝑤 𝑐𝑜𝑠 𝑡) − 𝛽 ⋅ 𝑡 − 𝛾
(8) 𝐿𝑜𝑔(𝐾𝑡 /𝐶𝐴𝑃𝑡 ) = −𝛼 ⋅ 𝐿𝑜𝑔(𝐿𝐸𝑡 /𝐾𝑡−1 ) − 𝛽 ⋅ 𝑡 − 𝛾
= −𝛼 ⋅ 𝐿𝑜𝑔(𝛼/(1 − 𝛼)𝑘 𝑐𝑜𝑠 𝑡𝑡 /𝑤 𝑐𝑜𝑠 𝑡) − 𝛽 ⋅ 𝑡 − 𝛾
But to apply this framework to a full model, we have to take into account several elements.
10.2.2.5.3.1.2 TARGETS AND ACTUAL VALUES
The above presentation applies to targets: knowing the relative costs, firms will estimate a target level of capacity,
then the target levels of factors which allow reaching this capacity.
This means that we have to define this capacity, then compute the target level for both factors, then define the
process leading from target values to actual values.
10.2.2.5.3.1.3 CAPACITY
For defining capacity, we shall suppose that firms have a target rate of use, constant over the period. This means that
target capacity is proportional to production (or rather value added). In the estimation of equations (7) and (8), we will
replace the ratio of factors to capacity by the ratio of factors to production, without loss of generality (the target rate
of use will be absorbed by the constant term).
10.2.2.5.3.1.4 THE TIME FACTOR
In the above framework, we have used only instantaneous elements. But we have to consider the nature of our
variables.
• Capital and capacity are measured at a given point in time
• Employment, in our definition, is an average level across one period (one quarter).
We shall suppose that target capital, employment (and implicit target capacity) are given by the system, using as
target the actual level of production, but that actual capacity for the period is given by actual employment and the
initial level of capital. As capital is measured as end-of-period, we shall use the lagged value.
10.2.2.5.3.1.5 THE INERTIA OF FACTORS
We shall also assess that optimal decisions are not implemented immediately. As explained before, the reasons are
both technical (the length of the investing process) and psychological (risk aversion). So firms go only part of the way
to the target, starting from the previous decision level. We shall try to estimate the inertia factor, allowing different
values for labor and capital.
409
The first factor should be less inert, as the penalty for errors is lower (we have to consider annual wages compared to
full cost of capital), and managing their consequences is also easier (laying down workers is easier than selling back
unneeded equipment).
10.2.2.5.3.1.6 THE RELATIVE COST
Of course, it should compare the price of capital (actually investment 146) to the price of labor (the wage rate). Actually,
things are a little more complex.
• The wage rate should include social contributions.
• Once purchased, capital can be used as long as it is not destroyed or obsolete147, whereas employment is
bought for a single period148.
• The price of capital should take into account the fact that it has to be purchased at once, whereas the
alternate factor, labor, is paid for at the moment it is used, or even later. This delay should call for the
introduction of the interest rate.
• The price of labor is measured in time spent. This means it is expected to increase in real terms with labor
productivity.
• Capital depreciates over time. For workers, the efficiency generally increases then decreases, depending on
the product. But firms can always replace older workers by new ones, at minimal cost (retirement financing is
generally included in the wage cost).
We shall actually compare the yearly wage cost with the price of investment, equivalent to the price of capital at the
cost of renewal. One has to consider that the increase in the efficiency of capital is included in the variable at constant
prices, not in the deflator (this is called the « quality effect »). To spread the cost of capital over its period of use, we
shall divide its deflator by an estimated factor, which should associate more or less with the number of periods of its
productive life.
•
𝑟𝑒𝑙𝑐 = (𝑤𝑓 ⋅ (1 + 𝑟_𝑠𝑐𝑓)/𝑝𝑖) ⋅ 1/((𝑖𝑟 − 𝑝𝑐) − 4 ⋅ 𝐿𝑜𝑔(1 − 𝑟𝑑𝑒𝑝))
Some explanations are needed for the last term. The increase in the cost of capital for a given period is the sum of the
interest rate and the annual depreciation:
•
𝑧 = (𝑖𝑟 − 𝑝𝑐) + 4 ⋅ 𝐿𝑜𝑔(1 − 𝑟𝑑𝑒𝑝)
146
Remember that the evolution of capital quality changes its value at constant prices, not its deflator, measured at a
given quality (or efficiency). This means that the present deflator applies also to past values of capital, as the efficiency
of a quantity is independent from the time at which it has been purchased. Applied to a given capital level, it will give
the value of investment necessary to replace this capital, with the same efficiency.
147
Maybe with a decreasing efficiency.
148
This was different in the Roman Empire, when you could buy slaves. Then labor and capital had closer properties.
410
This means the ratio between the cost of labor and capital will give to the first an advantage of:
1/(1 + 𝑧 𝑘 )
Or over the full period of
1/(1 + 𝑧) + 1/(1 + 𝑧)2 + 1/(1 + 𝑧)3 +. . . . = 1/𝑧
10.2.2.5.3.2 THE ESTIMATION
Let us see how we can apply this technique in our model, under EViews. We have seen that we have to estimate two
equations, with common coefficients. This will be done through a system, a feature which we will have now the
opportunity to present.
First let us compute the relative cost, as explained above:
smpl 1962Q1 2017Q4
genr relc=wr*(1+r_scf)/pi/(ir/100-@pchy(pc)-4*log(1-rdep))
genr k_ec=0
genr lf_ec=0
Now we have to estimate a system of two equations with common coefficients.
• As usual we create two vectors: coefficients and parameters.
• We destroy any preexisting system of the same name.
• We initialize the values with a guess.
• (this is important as convergence of this system is a little tricky under EViews)
coef(10) c_cd
vector(10) p_cd
delete cd
smpl 1962Q1 2017Q4
system cd
411
c_cd(4)=0.1
c_cd(1)=100
c_cd(3)=0.02 ‘ the yearly growth of total factor productivity: guessed at 2%
c_cd(2)=0.65 ‘ the share of employment in the process, guessed at 0.65 149
p_cd(8)=0.0
p_cd(9)=0.0
cd.append log(k/q(-1))=-c_cd(3)*(t-2017)-c_cd(7)+c_cd(2)*log(relc)
cd.append log(lf/q(-1))=-c_cd(3)*(t-2017)-c_cd(1)+(c_cd(2)-1)*log(relc)
Two methods are available; Full Information Maximum Likelihood (FIML), and Seemingly Unrelated Regression (see
Zellner (1962)). We test both but we chose the latter.
System: __CD
Estimation Method: Seemingly Unrelated Regression
Date: 04/03/20 Time: 11:59
Sample: 1980Q4 2017Q4
Included observations: 149
Total system (balanced) observations 298
Linear estimation after one-step weighting matrix
Coefficient Std. Error t-Statistic Prob.
C_CD(3) 0.003820 0.000672 5.683930 0.0000
C_CD(7) 4.539931 0.214654 21.14995 0.0000
C_CD(2) 0.462223 0.017133 26.97902 0.0000
C_CD(1) 4.962403 0.214680 23.11530 0.0000
Determinant residual covariance 4.62E-05
Equation: LOG(K/Q)=-C_CD(3)*(T-2017)-C_CD(7)+C_CD(2)*LOG(RELC)
Observations: 149
R-squared -4.729373 Mean dependent var 1.005121
Adjusted R-squared -4.807857 S.D. dependent var 0.051824
S.E. of regression 0.124893 Sum squared resid 2.277333
Durbin-Watson stat 0.066958
Equation: LOG(LF/Q)=-C_CD(3)*(T-2017)-C_CD(1)+(C_CD(2)-1)
*LOG(RELC)
Observations: 149
R-squared 0.260933 Mean dependent var -11.26714
Adjusted R-squared 0.250809 S.D. dependent var 0.158847
S.E. of regression 0.137491 Sum squared resid 2.759949
Durbin-Watson stat 0.074370
The estimation shows a 0.46 coefficient for labor (rather low compared to usual values) and a significant positive
trend in global factor productivity.
We generate the estimated (“desired”) series as target values.
149
A figure also close to the share of the wage cost in the total.
412
smpl 1963Q4 2017Q4
genr log(kd/q(-1))=-c_cd(3)*(t-2017)-c_cd(7)+c_cd(2)*log(relc)
genr log(lfd/q(-1))=-c_cd(3)*(t-2017)-c_cd(1)+(c_cd(2)-1)*log(relc)
We estimate the actual results as weighted averages of the estimated result and the previous actual value, with an
additional trend.
Actually we have fixed the inertia factors, at levels which give to the model reasonable properties.150
smpl 1970Q4 2017Q4
coef(10) c_k
vector(10) p_k
genr k_ec=0
coef(10) c_lf
vector(10) p_lf
genr lf_ec=0
p_lf(1)=0.20
p_k(1)=1.00
equation _eq_k.ls(p) log(k/q(-1))=p_k(1)*log(kd/q)+(1-p_k(1))*log(k(-1)/q(-1))+c_k(2)
+c_k(3)*(t-2017)*(t<=2017)+k_ec
genr k_ec=resid
equation _eq_lf.ls(p) log(lf/q(-1))=p_lf(1)*log(lfd/q)+(1-p_lf(1))*log(lf(-1)/q(-1))+c_lf(2)
+c_lf(3)*(t-2017)*(t<=2017)+lf_ec
genr lf_ec=resid
150
OK, we do a little cheating here, please forgive us.
413
Finally we generate capacity applying the estimated formula to actual factor values, and we compute the rate of use
of capacities.
genr log(cap)=c_cd(7)*(1-c_cd(2))+c_cd(1)*c_cd(2)
+c_cd(3)*(t-2017)*(t<=2017)+c_cd(2)*4*log((1+txq)/(1+txn))*(t-2017)*(t>2017)
+c_cd(2)*log(lf)+(1-c_cd(2))*log(k(-1))
genr ur=q/cap
We have to observe that this rate of use is different from the one we got with the complementary factors function.
This means that the equations in which it enters will have to be estimated again: value added deflator, exports and
imports.
10.2.2.5.4 The change in inventories
We shall use the same explanation as the simple model, simplifying the formula by using the present growth rate of
value added.
The unusual profile of the dependent variable is due to the use of an original yearly series.
coef(10) c_ci
smpl 1962Q1 2017Q4
genr ci_ec=0
smpl 1963Q1 2017Q4
equation _eq_ic.ls(p) ic/q(-1)=c_ic(1)*c_ic(2)*@pchy(q)+c_ic(2)+(1-c_ic(1))*ic(-1)/q(-2)+c_ic(3)+ic_ec
_eq_ci.resids(p)
close _eq_ci
genr ci_ec=resid
414
Dependent Variable: IC/Q(-1)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 16:18
Sample (adjusted): 1964Q1 2017Q4
Included observations: 216 after adjustments
Convergence achieved after 5 iterations
Coefficient covariance computed using outer product of gradients
IC/Q(-1)=C_IC(1)*C_IC(2)*@PCHY(Q)+C_IC(2)+(1-C_IC(1))*IC(-1)/Q(-2)
+C_IC(3)+IC_EC
Coefficient Std. Error t-Statistic Prob.
C_IC(1) 0.281439 0.041693 6.750247 0.0000
C_IC(2) 0.065153 0.011786 5.528009 0.0000
C_IC(3) -0.065597 0.011851 -5.535271 0.0000
R-squared 0.706556 Mean dependent var 0.000118
Adjusted R-squared 0.703801 S.D. dependent var 0.001890
S.E. of regression 0.001029 Akaike info criterion -10.90758
Sum squared resid 0.000225 Schwarz criterion -10.86070
Log likelihood 1181.019 Hannan-Quinn criter. -10.88864
F-statistic 256.4312 Durbin-Watson stat 1.105253
Prob(F-statistic) 0.000000
.010
.005
.000
.008 -.005
.004 -.010
-.015
.000
-.004
-.008
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
10.2.2.5.5 Unemployment
This is a new equation compared to the small model. Actually it seemed clearer to us to model the work force
(employment + unemployment). As it depends on employment, the quality of estimation will be exactly the same
(only the R-squared will change). The coefficient of the work force POPAC will be higher by 1 compared to a
formulation using unemployment UN.
Otherwise, the equation follows fully the framework defined above.
The short-term sensitivities are reasonable, and the long-term ones too, with a slightly lower value. We had to decide
on the rate of convergence to the long-term relation.
It is logical to expect that a permanent improvement of the labor market will attract more and more job seekers.
However, the erratic evolution of residuals in the later period is rather disquieting.
415
coef(10) c_popac
p_popac(3)=0.50
smpl 1962Q1 2017Q4
genr popac_ec=0
d(popac)/pop65(-1)=c_popac(1)*d(lt)/pop65(-1)+c_popac(2)*d(pop65)/pop65(-1)
-p_popac(3)*(popac(-1)/pop65(-1)-c_popac(4)*lt(-1)/pop65(-1)-c_popac(5))+[ar(1)=c_popac(6)])
+popac_ec
_eq_popac.resids(p)
genr popac_ec=resid
close _eq_popac
Dependent Variable: D(POPAC)/POP65(-1)
Method: ARMA Conditional Least Squares (Marquardt - EViews legacy)
Date: 04/03/20 Time: 16:55
Sample (adjusted): 1962Q3 2017Q4
Included observations: 222 after adjustments
Convergence achieved after 8 iterations
D(POPAC)/POP65(-1)=C_POPAC(1)*D(LT)/POP65(-1)+C_POPAC(2)
*D(POP65)/POP65(-1)-0.5*(POPAC(-1)/POP65(-1)-C_POPAC(4)*LT(
-1)/POP65(-1)-C_POPAC(5))+[AR(1)=C_POPAC(6)]
Coefficient Std. Error t-Statistic Prob.
C_POPAC(1) 0.608286 0.048134 12.63745 0.0000
C_POPAC(2) 0.183219 0.086401 2.120565 0.0351
C_POPAC(4) 0.446176 0.093014 4.796890 0.0000
C_POPAC(5) 0.357420 0.051223 6.977662 0.0000
C_POPAC(6) 0.977937 0.014378 68.01539 0.0000
R-squared 0.423523 Mean dependent var 0.001048
Adjusted R-squared 0.412896 S.D. dependent var 0.001146
S.E. of regression 0.000878 Akaike info criterion -11.21481
Sum squared resid 0.000167 Schwarz criterion -11.13817
Log likelihood 1249.844 Hannan-Quinn criter. -11.18387
F-statistic 39.85605 Durbin-Watson stat 2.427417
Prob(F-statistic) 0.000000
Inverted AR Roots .98
416
.006
.004
.002
.000
.004 -.002
.002 -.004
.000
-.002
-.004
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
10.2.2.5.6 The value added deflator
The formula will change according to the type of production function: the measure of the rate of use will be different,
and in the Cobb Douglas case the cost will include amortization of capital.
In both cases the equation follows an error correction format: a long-term relationship between the rate of use and
the margins rate (or rather the share of the cost in value added), and a dynamic equation freeing the elasticity of the
deflator to the cost.
In the Cobb-Douglas case, we have to consider the total cost: wage and capital.
genr cost=(wr*lf*(1+r_scf)+0.05*pi*k(-1))/q
We distribute the cost of capital over 20 periods (5 years). Reducing this value would not change much the properties.
The coefficients are rather different from the previous case: stronger dynamic effect, slower correction.
417
Dependent Variable: DLOG(PQ)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 17:57
Sample: 1974Q1 2017Q4
Included observations: 176
Convergence achieved after 5 iterations
Coefficient covariance computed using outer product of gradients
DLOG(PQ)=C_PQ(1)*DLOG(COST)+C_PQ(2)*DLOG(UR)+C_PQ(3)
*(LOG(PQ(-1)/COST(-1))-C_PQ(2)*LOG(UR(-1)))+C_PQ(5)+PQ_EC
Coefficient Std. Error t-Statistic Prob.
C_PQ(1) 0.652817 0.023028 28.34928 0.0000
C_PQ(2) 0.365519 0.061221 5.970488 0.0000
C_PQ(3) -0.018863 0.006386 -2.953605 0.0036
C_PQ(5) 0.026904 0.008276 3.250874 0.0014
R-squared 0.869624 Mean dependent var 0.009297
Adjusted R-squared 0.867350 S.D. dependent var 0.010069
S.E. of regression 0.003667 Akaike info criterion -8.356375
Sum squared resid 0.002313 Schwarz criterion -8.284318
Log likelihood 739.3610 Hannan-Quinn criter. -8.327149
F-statistic 382.4193 Durbin-Watson stat 1.976983
Prob(F-statistic) 0.000000
.05
.04
.03
.02
.01
.012 .00
.008 -.01
.004
.000
-.004
-.008
-.012
1975 1980 1985 1990 1995 2000 2005 2010 2015
Residual Actual Fitted
An alternate version uses cointegration.
We test the cointegration between the rate of use and the margins (as the ratio between the value added deflator and
the cost). It reflects the fact that facing a low rate of use firms will reduce their margins to increase demand.
It works.
418
Date: 04/03/20 Time: 17:23
Sample (adjusted): 1972Q2 2017Q4
Included observations: 183 after adjustments
Trend assumption: Linear deterministic trend (restricted)
Series: LOG(PQ/COST) LOG(UR)
Lags interval (in first differences): 1 to 4
Unrestricted Cointegration Rank Test (Trace)
Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.100608 29.59230 25.87211 0.0164
At most 1 0.054149 10.18772 12.51798 0.1189
Trace test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.100608 19.40458 19.38704 0.0497
At most 1 0.054149 10.18772 12.51798 0.1189
Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I):
LOG(PQ/COST) LOG(UR) @TREND(62Q2)
19.03573 -32.15614 -0.003083
-0.295312 66.21976 -0.001362
Unrestricted Adjustment Coefficients (alpha):
D(LOG(PQ/... -0.001096 -0.001306
D(LOG(UR)) 0.000840 -0.000750
1 Cointegrating Equation(s): Log likelihood 1422.333
Normalized cointegrating coefficients (standard error in parentheses)
LOG(PQ/COST) LOG(UR) @TREND(62Q2)
1.000000 -1.689252 -0.000162
(0.79135) (0.00025)
Adjustment coefficients (standard error in parentheses)
D(LOG(PQ/... -0.020863
(0.00940)
D(LOG(UR)) 0.015996
(0.00592)
The dynamic equation works also, with a low correcting coefficient.
419
Dependent Variable: DLOG(PQ)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 17:38
Sample (adjusted): 1971Q2 2017Q4
Included observations: 187 after adjustments
DLOG(PQ)=C_PQ(1)*DLOG(COST)+C_PQ(2)*DLOG(UR)+C_PQ(3)
*RES(-1)+C_PQ(5)+C_PQ(6)*(T-2017)*(T<=2017)
Coefficient Std. Error t-Statistic Prob.
C_PQ(1) 0.462592 0.036840 12.55669 0.0000
C_PQ(2) 0.227275 0.080422 2.826041 0.0052
C_PQ(3) -0.022939 0.006275 -3.655477 0.0003
C_PQ(5) 0.028252 0.008094 3.490630 0.0006
C_PQ(6) -0.000304 3.77E-05 -8.067195 0.0000
R-squared 0.796039 Mean dependent var 0.009865
Adjusted R-squared 0.791557 S.D. dependent var 0.010248
S.E. of regression 0.004679 Akaike info criterion -7.865245
Sum squared resid 0.003984 Schwarz criterion -7.778851
Log likelihood 740.4004 Hannan-Quinn criter. -7.830238
F-statistic 177.5822 Durbin-Watson stat 2.087218
Prob(F-statistic) 0.000000
.05
.04
.03
.03 .02
.02 .01
.00
.01
-.01
.00
-.01
-.02
1975 1980 1985 1990 1995 2000 2005 2010 2015
Residual Actual Fitted
10.2.2.5.7 The wage rate
The wage rate estimation is the only occasion which allows us to use cointegration, and thus to conform to the
accepted principles of econometrics.
The estimation process is exactly the same as for the small model.
We suppose that the wage cost is indexed in the long run: for 50% on the value added deflator (the firms want to
stabilize the share of wages in value added) and for 50% on the consumption deflator (workers want to their
purchasing power to follow the gains in productivity). The weighting could be adapted through a parameter.
So first we test that the stationarity of _luwc = log(uwc/(0.5*pq+0.5*pc))
420
genr _luwc=log(uwc)-0.5*log(pq)-(1-0.5)*log(pcoh)
uroot(p) _luwc
Then that of the unemployment rate itself.
uroot(p) unr
Both fail.
Then we test the cointegration of the two elements, which works.
coint(b,p) _luwc unr
421
Date: 04/03/20 Time: 18:08
Sample (adjusted): 1973Q4 2017Q4
Included observations: 177 after adjustments
Trend assumption: No deterministic trend (restricted constant)
Series: _LUWC UNR
Lags interval (in first differences): 1 to 6
Unrestricted Cointegration Rank Test (Trace)
Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.100436 23.92640 20.26184 0.0149
At most 1 0.028906 5.191742 9.164546 0.2631
Trace test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**
None * 0.100436 18.73466 15.89210 0.0174
At most 1 0.028906 5.191742 9.164546 0.2631
Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I):
_LUWC UNR C
-12.04745 -55.71775 8.626833
15.27940 -4.114727 -4.265647
Unrestricted Adjustment Coefficients (alpha):
D(_LUWC) 0.002981 -0.000523
D(UNR) 0.000222 0.000250
1 Cointegrating Equation(s): Log likelihood 1461.761
Normalized cointegrating coefficients (standard error in parentheses)
_LUWC UNR C
1.000000 4.624858 -0.716071
(0.96386) (0.08071)
Adjustment coefficients (standard error in parentheses)
D(_LUWC) -0.035914
(0.00888)
D(UNR) -0.002676
(0.00152)
The coefficient for UNR in the cointegrating equation is significant with the right sign. It remains to be seen if this
equation gives good properties to the full model.
For the dynamic equation, we had to set the indexation as globally unitary (dynamic homogeneity) with a lag
422
structure. All the elements are quite significant, but the error correction is quite slow (a result we found in other
circumstances for the same equation).
Dependent Variable: DLOG(WR)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/04/20 Time: 20:24
Sample (adjusted): 1972Q2 2017Q4
Included observations: 183 after adjustments
DLOG(WR)=C_WR(1)*DLOG(PC)+C_WR(2)*DLOG(LPROD)+C_WR(4)
*RES_WR(-1)+C_WR(5)*(T-2018)*(T-2018)+C_WR(6)
Coefficient Std. Error t-Statistic Prob.
C_WR(1) 0.598216 0.055015 10.87363 0.0000
C_WR(2) 0.151471 0.062906 2.407877 0.0171
C_WR(4) -0.014572 0.003991 -3.650994 0.0003
C_WR(5) 7.51E-06 9.67E-07 7.764355 0.0000
C_WR(6) -0.001007 0.000909 -1.108496 0.2691
R-squared 0.865434 Mean dependent var 0.013277
Adjusted R-squared 0.862410 S.D. dependent var 0.011659
S.E. of regression 0.004325 Akaike info criterion -8.021948
Sum squared resid 0.003329 Schwarz criterion -7.934257
Log likelihood 739.0082 Hannan-Quinn criter. -7.986402
F-statistic 286.1925 Durbin-Watson stat 1.269681
Prob(F-statistic) 0.000000
.08
.06
.04
.02 .02
.01 .00
-.02
.00
-.01
-.02
1975 1980 1985 1990 1995 2000 2005 2010 2015
Residual Actual Fitted
10.2.2.5.8 The trade deflators
For this equation we will not use an error correction format, rather a homogenous indexation. This means we shall not
separate the short and long-term sensitivities.
As usual for this type of equation:
• Exporters show a higher attention to their costs than to the price of their competitors. This is particular true
for the exporters to France, as the role of French prices is not completely significant.
This will dampen the dynamics of the price-wage loop. The impact of trade on the price of demand can be explained in
the following way.
423
o Imports are a share of global demand. They are bought at the import price. The higher its sensitivity to
foreign costs, the higher the difference to the local production price, and the higher the reducing impact of
imports on the global demand price.
o If local producers decided on their selling price on the local and foreign markets independently (a possible
behavior that we did not consider), the sensitivity of the demand price to local costs would clearly be less
than one.
o But in our framework, the production price is decided globally, and the lower sensitivity of the export price
has to be balanced by a higher sensitivity of the price at which they sell on the local market. The higher the
impact of local costs on the export price, the lower the necessary compensation.
One can see that in the transition from production to demand price, the higher the role of the production cost in the
price set by the exporter, the higher the first (negative) effect and the lower the second (positive) one. 151
In the extreme, if all exporters take only into account their costs, the import price will not be affected, and as the
export price will change just as the global production price, no compensation will be needed. The damping effect will
be maximal.
• The additional trend is negative and quite significant. It probably represents a structural but permanent shift
in traded goods to the ones which present the lowest price.
To reach a steady state in the long run, these trends will have to be suppressed after a while. Here we did it
immediately, but true forecasts should call for a gradual decrease.
coef(10) c_px
smpl 1962Q1 2017Q4
genr px_ec=0
smpl 1962Q1 2017Q4
equation _eq_px.ls(p) log(px)=c_px(4)*log(pp)+(1-c_px(4))*log(ppx*er)
+c_px(6)+c_px(5)*(t-2017)*(t<=2017)+[ar(1)=c_px(7)]+px_ec
_eq_px.resids(p)
close _eq_px
coef(10) c_pm
smpl 1962Q1 2017Q4
genr pm_ec=0
smpl 1962Q1 2017Q4
equation _eq_pm.ls(p) log(pm)=c_pm(4)*log(pp)+(1-c_pm(4))*log(ppx*er)
+c_pm(6)+c_pm(5)*(t-2017)*(t<=2017)+[ar(1)=c_pm(7)]+pm_ec
_eq_pm.resids(p)
close _eq_pm
151
This could be formalized easily, but we hope the message is already clear.
424
Dependent Variable: DLOG(PX)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 19:40
Sample (adjusted): 1975Q2 2017Q4
Included observations: 171 after adjustments
Convergence achieved after 3 iterations
Coefficient covariance computed using outer product of gradients
DLOG(PX)=0.7*DLOG(PP)+C_PX(2)*DLOG(PPX*ER)+C_PX(3)
*(LOG(PX(-1))-C_PX(4)*LOG(PP(-1))-(1-C_PX(4))*LOG(PPX(-1)*ER(
-1)))+C_PX(6)+C_PX(5)*(T-2017)*(T<=2017)
Coefficient Std. Error t-Statistic Prob.
C_PX(2) 0.203108 0.038859 5.226755 0.0000
C_PX(3) -0.089478 0.030611 -2.923041 0.0040
C_PX(4) 0.635238 0.126448 5.023700 0.0000
C_PX(6) -0.006800 0.003914 -1.737285 0.0842
C_PX(5) -0.000839 0.000399 -2.106586 0.0367
R-squared 0.195857 Mean dependent var 0.005023
Adjusted R-squared 0.176480 S.D. dependent var 0.022168
S.E. of regression 0.020117 Akaike info criterion -4.945705
Sum squared resid 0.067179 Schwarz criterion -4.853843
Log likelihood 427.8577 Hannan-Quinn criter. -4.908431
F-statistic 10.10771 Durbin-Watson stat 1.816121
Prob(F-statistic) 0.000000
Dependent Variable: DLOG(PM)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 19:43
Sample (adjusted): 1975Q2 2017Q4
Included observations: 171 after adjustments
Convergence achieved after 4 iterations
Coefficient covariance computed using outer product of gradients
DLOG(PM)=C_PM(1)*DLOG(PP)+C_PM(2)*DLOG(PPX*ER)+C_PM(3)
*(LOG(PM(-1))-C_PM(4)*LOG(PP(-1))-(1-C_PM(4))*LOG(PPX(-1)
*ER(-1)))+C_PM(6)+C_PM(5)*(T-2017)*(T<=2017)
Coefficient Std. Error t-Statistic Prob.
C_PM(1) 0.726492 0.187994 3.864435 0.0002
C_PM(2) 0.266741 0.030791 8.663057 0.0000
C_PM(3) -0.078071 0.037872 -2.061446 0.0408
C_PM(4) 0.625968 0.113616 5.509510 0.0000
C_PM(6) -0.005655 0.002856 -1.980227 0.0493
C_PM(5) -0.000673 0.000315 -2.136540 0.0341
R-squared 0.465638 Mean dependent var 0.005416
Adjusted R-squared 0.449445 S.D. dependent var 0.020259
S.E. of regression 0.015032 Akaike info criterion -5.522834
Sum squared resid 0.037283 Schwarz criterion -5.412601
Log likelihood 478.2023 Hannan-Quinn criter. -5.478106
F-statistic 28.75589 Durbin-Watson stat 1.562780
Prob(F-statistic) 0.000000
ls dlog(px)=c_px(1)*dlog(pp)+c_px(2)*dlog(ppx*er)+c_px(3)*(log(px(-1))-c_px(4)*log(pp(-1))-(1-
c_px(4))*log(ppx(-1)*er(-1)))+c_px(6)+c_px(5)*(t-2017)*(t<=2017)
ls dlog(pm)=c_pm(1)*dlog(pp)+c_pm(2)*dlog(ppx*er)+c_pm(3)*(log(pm(-1))-c_pm(4)*log(pp(-1))-
(1-c_pm(4))*log(ppx(-1)*er(-1)))+c_pm(6)+c_pm(5)*(t-2017)*(t<=2017)
425
10.2.2.5.9 Household consumption
Our equation follows as usual an error correction specification (estimated in one step!) following almost completely
the framework presented earlier.
The change in consumption depends on:
• The change in real income (over the last year).
• The change in unemployment.
• Inflation (the « real holdings » effect).
• An error correction term.
• A negative time trend, representing the increase in the wealth of households and their saving potential (in
particular their accession to housing ownership).
The only influence we could not evidence is that of the real short-term interest rate.
coef(10) c_coh
smpl 1962Q1 2022Q4
genr coh_ec=0
smpl 1962Q1 2017Q4
equation _eq_coh.ls(p) dlog(coh)=c_coh(1)*.25*log(hrdi/hrdi(-4))+c_coh(2)*dlog(unr)
+c_coh(3)*log(pc/pc(-4))+c_coh(5)*dlog(coh(-1))+c_coh(6)+c_coh(7)*log(coh(-1)/hrdi(-1))
+c_coh(8)*(t-2017)*(t<=2017)+coh_ec
genr coh_ec=resid
426
Dependent Variable: DLOG(COH)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 12:39
Sample (adjusted): 1963Q1 2017Q4
Included observations: 220 after adjustments
DLOG(COH)=C_COH(1)*.25*LOG(HRDI/HRDI(-4))+C_COH(2)
*DLOG(UNR)+C_COH(3)*LOG(PC/PC(-4))+C_COH(5)*DLOG(COH(
-1))+C_COH(6)+C_COH(7)*LOG(COH(-1)/HRDI(-1))+C_COH(8)*(T
-2017)*(T<=2017)+COH_EC
Coefficient Std. Error t-Statistic Prob.
C_COH(1) 0.417154 0.128560 3.244825 0.0014
C_COH(2) -0.048968 0.011362 -4.309732 0.0000
C_COH(3) -0.035504 0.017628 -2.014029 0.0453
C_COH(5) -0.213361 0.066355 -3.215459 0.0015
C_COH(6) -0.304138 0.098582 -3.085123 0.0023
C_COH(7) -0.064356 0.020801 -3.093930 0.0022
C_COH(8) -0.000153 4.79E-05 -3.195567 0.0016
R-squared 0.342567 Mean dependent var 0.006449
Adjusted R-squared 0.324048 S.D. dependent var 0.007509
S.E. of regression 0.006173 Akaike info criterion -7.305809
Sum squared resid 0.008118 Schwarz criterion -7.197830
Log likelihood 810.6390 Hannan-Quinn criter. -7.262204
F-statistic 18.49791 Durbin-Watson stat 2.091141
Prob(F-statistic) 0.000000
.06
.04
.02
.03 .00
.02 -.02
.01
.00
-.01
-.02
-.03
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
10.2.2.5.10 Exports
For exports we shall use again an error correction framework, estimated in one pass.
• The substitution effect appears through the average of the rate of use over the last two quarters (the
dynamic and long-term effects are not otherwise separated). Price competitiveness uses the same technique.
• A significant trend had to be introduced and will be stopped in the future. It is possible that the measure of
world demand is biased.
427
Dependent Variable: DLOG(X)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 20:48
Sample (adjusted): 1974Q2 2005Q4
Included observations: 127 after adjustments
Convergence achieved after 4 iterations
Coefficient covariance computed using outer product of gradients
DLOG(X)=C_X(1)*DLOG(WD)+C_X(2)*LOG(X(-1)/WD(-1))+C_X(5)
+C_X(6)*T+[C_X(3)=AR(1)]
Coefficient Std. Error t-Statistic Prob.
C_X(1) 0.421293 0.084847 4.965310 0.0000
C_X(2) -0.158407 0.041881 -3.782334 0.0002
C_X(5) 2.257588 0.689654 3.273506 0.0014
C_X(6) -0.001118 0.000343 -3.256103 0.0015
R-squared 0.247388 Mean dependent var 0.011818
Adjusted R-squared 0.229031 S.D. dependent var 0.018045
S.E. of regression 0.015844 Akaike info criterion -5.421050
Sum squared resid 0.030877 Schwarz criterion -5.331469
Log likelihood 348.2367 Hannan-Quinn criter. -5.384654
F-statistic 13.47692 Durbin-Watson stat 2.002881
Prob(F-statistic) 0.000000
.08
.06
.04
.02
.04 .00
-.02
.02 -.04
.00
-.02
-.04
1975 1980 1985 1990 1995 2000 2005
Residual Actual Fitted
genr x_ec=0
smpl 1962Q1 2003Q4
equation _eq_x.ls(p) dlog(x)=c_x(1)*dlog(wd)+c_x(2)*log(x(-1)/wd(-1))
+c_x(3)*0.5*(log(ur)+log(ur(-1)))+c_x(4)*0.5*(log(compx(-1))+log(compx))+c_x(5)
+c_x(6)*(t-2005)*(t<=2005)+x_ec
_eq_x.resids(p)
close _eq_x
428
genr x_ec=resid
Dependent Variable: DLOG(X)
Method: Least Squares (Gauss-Newton / Marquardt steps)
Date: 04/03/20 Time: 20:51
Sample (adjusted): 1974Q2 2003Q4
Included observations: 119 after adjustments
DLOG(X)=C_X(1)*DLOG(WD)+C_X(2)*LOG(X(-1)/WD(-1))+C_X(3)*0.5
*(LOG(UR)+LOG(UR(-1)))+C_X(4)*0.5*(LOG(COMPX(-1))
+LOG(COMPX))+C_X(5)+C_X(6)*(T-2005)*(T<=2005)
Coefficient Std. Error t-Statistic Prob.
C_X(1) 0.415130 0.083749 4.956817 0.0000
C_X(2) -0.201800 0.048941 -4.123297 0.0001
C_X(3) -0.254627 0.121300 -2.099154 0.0380
C_X(4) -0.202750 0.060155 -3.370481 0.0010
C_X(5) -0.038971 0.018205 -2.140639 0.0345
C_X(6) -0.007775 0.001966 -3.953931 0.0001
R-squared 0.326431 Mean dependent var 0.011474
Adjusted R-squared 0.296627 S.D. dependent var 0.018593
S.E. of regression 0.015593 Akaike info criterion -5.434828
Sum squared resid 0.027477 Schwarz criterion -5.294704
Log likelihood 329.3723 Hannan-Quinn criter. -5.377928
F-statistic 10.95263 Durbin-Watson stat 2.071445
Prob(F-statistic) 0.000000
.08
.06
.04
.02
.00
.04 -.02
.02 -.04
.00
-.02
-.04
74 76 78 80 82 84 86 88 90 92 94 96 98 00 02
Residual Actual Fitted
10.2.2.5.11 Imports
For imports we use the same framework, but:
• We had to set the dynamic demand coefficient to unity.
• The rate of use is not lagged.
• Price competitiveness is measured over 6 periods.
429
' estimating imports
'show err_m
coef(10) c_m
smpl 1962Q1 2022Q4
genr m_ec=0
smpl 1962Q1 2017Q4
equation _eq_m.ls(p) dlog(m)=dlog(fd+tc*q)+c_m(2)*log(ur)+c_m(3)*dlog(compm)
+c_m(5)+c_m(6)*(t-2017)*(t<=2017)+[ar(1)=c_m(7)]+c_m(8)*log(m(-1)/(fd(-1)+tc*q(-1)))+m_ec
genr m_ec=resid
Dependent Variable: DLOG(M)
Method: ARMA Conditional Least Squares (Marquardt - EViews legacy)
Date: 04/03/20 Time: 21:54
Sample (adjusted): 1963Q3 2017Q4
Included observations: 218 after adjustments
Convergence achieved after 8 iterations
DLOG(M)=C_M(1)*DLOG(FD+TC*Q)+C_M(2)*LOG(UR)+C_M(3)
*DLOG(COMPM)+C_M(5)+C_M(6)*(T-2017)*(T<=2017)
+[AR(1)=C_M(7)]+C_M(8)*LOG(M(-1)/(FD(-1)+TC*Q(-1)))
Coefficient Std. Error t-Statistic Prob.
C_M(1) 2.154307 0.116071 18.56025 0.0000
C_M(2) -0.020114 0.045942 -0.437819 0.6620
C_M(3) -0.103709 0.042707 -2.428405 0.0160
C_M(5) -0.046274 0.029940 -1.545553 0.1237
C_M(6) 0.000941 0.000504 1.865146 0.0635
C_M(8) -0.029705 0.018188 -1.633248 0.1039
C_M(7) 0.096754 0.073690 1.312989 0.1906
R-squared 0.684084 Mean dependent var 0.013658
Adjusted R-squared 0.675101 S.D. dependent var 0.024959
S.E. of regression 0.014227 Akaike info criterion -5.635817
Sum squared resid 0.042706 Schwarz criterion -5.527140
Log likelihood 621.3040 Hannan-Quinn criter. -5.591921
Durbin-Watson stat 1.961930
Inverted AR Roots .10
430
.2
.1
.0
.06
.04 -.1
.02 -.2
.00
-.02
-.04
-.06
65 70 75 80 85 90 95 00 05 10 15
Residual Actual Fitted
10.2.2.6 Checking again model – data consistency
Now that equations have been estimated, we should check again data-equations consistency, this time on all
equations.
The identities should still hold true, and adding the estimation residual at the end of the behavioral equations should
make them hold true, too.
So this process could look unnecessary. This will generally be true if all estimations have been repeated immediately
before the check, which means that they have been grouped, along with the definition of identities, in the same
program. This is actually the best technique, as it allows producing the full model through a single program, and not
through a sequence (which needs some organization and can lead to errors).
But if the estimates somewhat dated, and if one is not sure that the data have not been modified since, errors might
have appeared.
And in any case, this check is almost free (the control that all the residuals are negligible can be done with a program,
with a single answer) so there is no reason not to do it.
Of course, our example gives acceptable results152.
152
But achieving this result took some time…
431
smpl 1995Q2 2017Q4
‘ Now we compute the residuals
‘ for all the endogenous
‘
‘ The process is the same
‘
‘ The dc_{%1} are the absolute differences
‘ The pc_{%1} the relative ones
‘ To avoid dividing by zero, we use trick 1
‘ We add to the variable a boolean one
‘ Testing if the variable is zero
‘ If it is true we get 0 divided by 1 = 0
‘ If not the computation is not affected
‘
‘ This means we have just modified the outcome of divisions.
‘ Dividing zero by zero now gives zero
for !i=1 to _g_vendo.@count
%1=_g_viden.@seriesname(!i)
genr dc_{%1}={%1}-{%1}_c
genr pc_{%1}=100*dc_{%1}/({%1}+({%1}=0))
next
10.2.2.7 Solving the model on the future
We are coming now to the crucial part of model testing, observing how it performs on the field where it will be used:
the future.
Globally, the technique is the same as explained earlier. The differences are minimal.
We start with the usual statements: changing the directory, creating a special expanded file called “proj_1”.
cd "d:\eviews\__fra_2020"
' we set the workfile to proj_1, quarterly from 1962 to 2100
'
close proj_1.wf1
close mod_1.wf1
open mod_1.wf1
save proj_1
pagestruct(end=2100Q4) *
432
Now we are using deflators. For the long-term assumptions, we need to include their growth rate:
' we define long-term growth rates
scalar txq=exp(0.005)-1
scalar txn=exp(0.002)-1
scalar txp=exp(0.006)-1
Obviously, the set of exogenous variables, and of assumptions to define, is larger.
This program simulates the model over the future
' we expand series for projection: EXOGENOUS variables
' constant expansion
for %1 COH_EC ERX FDXR K_EC CI_EC IR_ER IRL_ER IRMX IRSR IRST IRSX IRX LF_EC M_EC
NIF_ER NIG_ER PK PM_EC POPAC_EC PQ_EC PX_EC R_EXPG R_ICT R_IFP R_OIT R_PCOH
R_PCOG R_PI R_PIG R_REVG R_REVQ R_SCF R_SCG R_SCW R_SUBS R_TAR R_TARX R_VAT
RDEP TC URD WR_EC X_EC IRM_ER relax_q relax_pfd fcapf_er prof_er smpl 2005Q1 2100Q4
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)
next
' expansion with GDP growth rate
for %1 COG IG WD R_REVX HIH
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)
next
' expansion with population growth rate
for %1 LG POP65 POP
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txn)
next
' expansion with deflators growth rate
for %1 PPX
smpl 2004 2100 if {%1}=na
433
genr {%1}={%1}(-1)*(1+txp)
next
' expansion with special growth rate
for %1 SOCBR
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)/(1+txn)
next
genr t=t(-1)+0.25
It will be also useful (but not necessary) to create values for the endogenous:
smpl 2005Q1 2100Q4
' we expand series for projection: we initialize ENDOGENOUS variables
‘ constant expansion
for %1 COMPM COMPX ER FCAPGP IR IRL IRM IRS RCVAL RCVOL RES_WR RMARG RPROB
RPROF TTRAD UNR UR
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)
next
‘ expansion with long-term GDP growth rate
for %1 CAP COH FD HRDI I CI K M Q X GDPM IC
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)
next
' expansion with population growth rate
for %1 LFD LF LT POPAC UN popt
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txn)
next
' expansion with deflators growth rate
for %1 PCOH PFD PFDXT PI PIG PM PMT PP PQ PX UWC
434
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txp)
next
next
' expansion with variables at current prices growth rate
for %1 CGV EXPG FCAPF FCAPG FCAPX FDGV HDI HI ICT IFP IGV MARG MVAL NIF NIG NIX NIXL
NIXX OIT PROF GDPMVAL QVAL REVG REVQ REVX SCF SCG SCW SOCB SUBS TAR TRB VAT W
WF WG XVAL
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txp)*(1+txq)
next
' expansion with labour productivity growth rate
for %1 PL PLT
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txq)/(1+txn)
next
' expansion with labor productivity at current prices growth rate
for %1 WR
smpl 2004 2100 if {%1}=na
genr {%1}={%1}(-1)*(1+txp)*(1+txq)/(1+txn)
next
But the essential issue is the guarantee that a stationary path will be obtained.
10.2.2.7.1 General elements
The same elements as above apply to this case.
But the higher complexity of the problem makes the methods proposed earlier more efficient if not necessary:
• Making shocks on the main assumptions and observing the consequences.
• Excluding some variables, or changing (increasing, decreasing or suppressing) local explanations.
10.2.2.7.2 Ensuring convergence in the long run: further elements
Let us return to the constraints the model must follow, and see how the above principles apply to a more complex
case.
• First, care must be taken to avoid any remaining trend.
435
As we have said earlier, all elasticities should be unitary in the estimated equations, except of course for elements
without dimension (ratios). The easiest method is obviously to use an error correction framework: the constraints can
be set in the cointegrating equations (estimating them in one pass does not change this property, just the validity of
the method….). They can be released in the VAR (the dynamic equation) at no cost to convergence, in principle.
One will observe that almost all our estimated equations, even if cointegration was applied only twice, contain a long-
term expression linking ratios without dimension. The only exceptions are:
• The trade prices where no difference is made between short and long-term, but homogeneity is obtained by
constraints on coefficients.
• For investment:
o In the complementary factors case, most elements represent derivatives (with no dimension). The only
exceptions are the rate of use and the profits rate, both ratios which should stabilize in the long run.
However, the rate of use is no longer fixed as in the simpler case.
o In the Cobb-Douglas case, the capital – output ratio is linked to a ratio of prices (labor to capital).
The only real problem lies with employment, which has a specific dimension.
o We have already treated the complementary factors case. The trend in labor productivity is estimated. In the
long run, both trends in employment (populations) and value added (a variable at constant prices) are fixed,
and so is the trend in labor productivity, with a value which must replace the estimated one 153.
o For the Cobb-Douglas case, three elements are taken into account to define productive capacity:
The trend in global factor productivity, at present estimated.
The contribution of labor: txn
The contribution of capital: (1- ) txq
Basically, the total growth rate of capacity should be the same as all variables at constant prices, which gives:
𝑡𝑥𝑞 = 𝑎 + 𝛼 ⋅ 𝑡𝑥𝑛 + (1 − 𝛼) ⋅ 𝑡𝑥𝑞
𝑎 = 𝛼 ⋅ (𝑡𝑥𝑞 − 𝑡𝑥𝑛)
This formula is easy to interpret: capital units will have a constant productivity (remember that the value at constant
prices includes the increase in quality). Labor units do not, so to achieve their share of the increase in capacity their
productivity must increase as the difference between growth of quantities and growth of labor.
In this way the total growth of capacity will meet the constraint.
153
Or one of trends must be set to agree with the estimated value.
436
To present the issue in a clearer way, it might be better to consider, not employment, but employment “efficient
units” in which each worker is value at his potential contribution to production, just like capital. The value of these
units at constant prices will grow as production, and there will be no need for an additional productivity trend.
Once these conditions are set, the model should have a long-term solution, to which the formulations (in particular
error-correction behaviors) should make the model converge.
10.2.2.7.3 Making the model converge in the short run
The problems are the same as in the simple case, but
• The danger is higher as more elements and mechanisms are concerned. The probability of spurious cycles
grows.
• Deflators are generally more volatile than quantities (being less “real”).
• The risk of making an error on assumptions is higher, as the process of definition is more complex: trying to
produce an accurate forecast on a partially known future will lead us to define some of them explicitly.
• The trends which have to be blocked are more numerous, and they might appear in more complex equations.
• Finally, it is possible that the management of residuals in the first forecast periods brings high variations, and
in particular strong cycles.
10.2.2.7.4 Making the model converge in the medium run
Again, the main danger lies in the cycles, with a higher probability.
But the fact that the model has converged at least for some periods will provide us with important information, and
the possibility to conduct additional tests, as we have already seen.
10.2.2.7.5 Specific forecast elements
Now that we have made the model converge over the whole period, we can test its properties through responses to
shocks. Then we can move to actual forecasts.
10.2.2.8 Producing a forecast
The technique is the same as usual. We can use any algorithm.
' we solve the model in the future (no shock)
‘ The letter for the base simulation is b
_mod_1.append assign @all _b
smpl 2020Q1 2100Q4
_mod_1.scenario "Scenario 1"
‘ We make sure that the full model is solved
‘ and that we use the normal values of exogenous (with no suffix)
‘ We use the Gauss-Seidel method
_mod_1.override
_mod_1.exclude
437
_mod_1.solveopt(n=t m=1000,o=g,d=d)
_mod_1.solve
‘ We compute the growth rates
‘ with an additional term in the denominator
‘ it avoids dividing by zero (which gives an error message)
‘ if the value is zero the expression gives 0 / (0 + 1) = 0
‘ if not zero it doean not change the result as the condition gives zero
for !i=1 to _g_vendo.@count
%1=_g_vendo.@seriesname(!i)
genr td_{%1}=100*d({%1}_b)/({%1}_b(-1)+({%1}_b(-1)=0))
genr td0_{%1}=100*d({%1})/({%1}(-1)+({%1}(-1)=0))
genr dtd_{%1}=td_{%1}-td0_{%1}
next
10.2.2.9 Producing shocks on assumptions
Compared to the previous example, the process will be a little more complex. In particular, we shall consider eight
shocks.
Comments are imbedded in the program.
' Now we shall produce a set of shocks (in the present case 7)
smpl 2000 2100
' The group called shocks_v will contain the list of 7 shocked variables
group shocks_v ig erx r_vat r_tar r_tarx wd m_ec
' The group called shocks_l will contain 7 letters associated to each shock
' but the letters must be known as series
' as in EViews groups can only contain series
' We create the artificial series
' but only for the non-existing ones of course
for %z g r t f y w n
if @isobject(%z) =0 then
438
genr {%z}=na
endif
next
' We create the group
group shocks_l g r t f y w n
' Now we compute the additional change for each assumption,
' using the name of the variable
' The shock will only start in 2021
' leaving one unshocked period
‘ to check that the difference comes only from the shock
‘ The name of the assumption will combine the shocked variable with the letter
' 1 - Shock g: +1 GDP point on Government demand
genr ig_g=ig+.01*gdpm_b*(t>=2021)
' 2 - Shock r: 1% devaluation of the Euro
genr erx_r=erx*(1+.01*(t>=2021))
' 3 - Shock t: -1 point on the VAT rate
genr r_vat_t=r_vat-.01*(t>=2021)
' 4 - Shock f: -1 point on the local tariffs rate
genr r_tar_f=r_tar-.01*(t>=2021)
439
' 5 - Shock y: -1 point on the foreign tariffs rate
genr r_tarx_y=r_tarx-.01*(t>=2021)
' 6 - Shock w: +1% on World demand addressed to France
genr wd_w=wd*(1+.01*(t>=2021))
' 7 - Shock n: +1% on ex-ante imports
genr m_ec_n=m_ec+.01*(t=2021) -.01*c_m(8)*(t>2021)
The loop on the shocks
for !j=1 to shocks_v.@count
smpl 2005Q1 2100Q4
' We get the name of the shocked variable and the associated letter
%2=shocks_v.@seriesname(!j)
%3=shocks_l.@seriesname(!j)
' We set the solution suffix using the letter
' We control no endogenous is excluded from the simulation
' We override the variable associated to the current shock
' We solve the model
_mod_1.scenario "scenario 1"
_mod_1.append assign @all _{%3}
440
_mod_1.exclude
_mod_1.override {%2}
_mod_1.solve(n=t m=1002 c=1e-6 o=g d=d)
' We compute the difference to the base simulation, in absolute and relative terms
for !i=1 to _g_vendo.@count
%1=_g_vendo.@seriesname(!i)
series d{%3}_{%1}={%1}_{%3}-{%1}_b
series p{%3}_{%1}=100* d{%3}_{%1}/({%1}_b+ (%1}_b=0))
series dv_{%1}=d{%3}_{%1}
series pv_{%1}=p{%3}_{%1}
next
' We create groups for variations for a specific list of important variables
‘ the number is restricted to make a table legible
group g_v{%3} P{%3}_PFD P{%3}_PM P{%3}_PX P{%3}_PQ P{%3}_WR P{%3}_FD P{%3}_gdpm
P{%3}_X P{%3}_M P{%3}_I P{%3}_COH P{%3}_UR P{%3}_CAP P{%3}_LF P{%3}_K P{%3}_UNR
D{%3}_FCAPGP P{%3}_RCVAL P{%3}_RCVOL P{%3}_TTRAD
' or very important ones
‘ displayed in a graph
‘ the number is further restricted to make the graph legible
group g_w{%3} P{%3}_FD P{%3}_gdpm P{%3}_X P{%3}_M P{%3}_PQ
' We create a group for variations for all variables
group g_v2 pV_* dV_*
' We store all shocks in Excel files (2025 and 2100)
441
' using the letter associated to the shock
smpl 2020Q1 2025Q4
write(t=xls) v2_{%3}.xls g_v2
smpl 2020Q1 2100Q4
write(t=xls) w2_{%3}.xls g_v2
next
10.2.2.10 Applying the programs
We shall now present the results obtained from the previous programs.
10.2.2.10.1 Producing the model framework
Basically, the program creates an item called _mod_1 in the workfile. Of course, this model cannot be solved at this
time. However, accessing this item gives alternately
• the list of variables,
• the list of equations,
• the original source code,
• the block structure.
The list of equations and the code repeat the model creating statements, and are of little interest. The list of variables,
as stated earlier, helps locate problems such as logically endogenous variables with no equation (they appear as
exogenous) or typing errors in variable names (they create also an exogenous item).
The last element is the most interesting. Let us show the result for our model.
Number of equations: 88
Number of independent blocks: 7
Number of simultaneous blocks: 3
Number of recursive blocks: 4
Largest simultaneous block: 52 Equations (3 feedback vars)
Block 1: 1 Recursive Equations
ict(43)
Block 2: 52 Simultaneous Equations (3 feedback vars)
lfd(5) lf(8) lt(11) popac(14) un(15)
unr(16) pl(9) pc(22) wr(26) wf(50)
wg(39) w(40) scw(41) socb(38) revx(37)
pi(23) cost(27) cap(13) ur(3) pq(18)
qv(47) revq(36) hi(42) hdi(44) hrdi(45)
coh(46) ic(10) irl(32) irs(31) ir(33)
er(30) pfdxt(21) pp(19) px(28)
compx(64) x(65) xv(67) pm(29) pmt(60)
compm(61) kd(4) k(6) i(7) fd(62)
m(63) mv(66) vat(76) gdpmv(48) gdpm(1)
relc(35) q(2) pfd(20)
442
Block 3: 10 Recursive Equations
id(12) uwc(17) pig(24) res_wr(25) irm(34)
pgdpm(49) subs(51) marg(52) rmarg(53) ifp(54)
Block 4: 3 Simultaneous Equations (1 feedback var)
prof(56) fcapf(59) nif(55)
Block 5: 18 Recursive Equations
rprof(57) rprob(58) rcval(68) rcvol(69) ttrad(70)
trb(71) nixl(72) nixx(73) nix(74) fcapx(75)
scf(77) oit(78) tar(79) scg(80) revg(81)
igv(82) cogv(83) fdgv(84)
Block 6: 3 Simultaneous Equations (1 feedback var)
expg(86) fcapg(87) nig(85)
Block 7: 1 Recursive Equations
fcapgp(88)
What the above shows is that the 88 equations in our model can be decomposed into a sequence of seven « blocks ».
Three of them are defined as recursive, two as simultaneous.
Let us define these notions, which are associated to the model solving process.
• Recursive means that the set can be arranged in such a way that no variable appears in any equation before
the one in which it is computed. Obviously, this means that computations will give to each variable its exact
value, and after each variable has been computed once no further computation is needed.
• Simultaneous means that each variable in the set depends on a variable computed later, either directly or
through a sequence of (possibly recursive) influences. Therefore, this variable will not take its exact value
immediately, except if the variable(s) responsible for the non-recursivity was given its exact value in the
beginning.
Starting from any order, EViews is able to build an initial recursive set, as variables which depend only on , lagged, or
variables which have been included earlier in the set. We will call this set a « prologue ».
It is also able to build a final recursive set, as variables which influence no variable in the rest of the set, or only
variables which have already been included in the set. We will call this set an « epilogue ».
Once this is done, the rest of the model could be considered simultaneous. However, this set can eventually be
separated again in two consecutive ones, if no variable in the second set influences any variable in the first. The first
set can have an epilogue, the second one a prologue, which can be joined into an intermediate recursive block. And
the process can be applied again to the two simultaneous blocks, until partition is no longer possible.
In our case, we observe:
• A small initial recursive block with ICT (Income tax) and IFP (Tax on firms’ profits). They obviously depend on
an exogenous tax rate and revenue from the previous period.
• A large (51 equations) block containing two interconnected loops: the Keynesian supply – demand – supply
equilibrating process, and the wage - price loop. Taking any couple of variables in the set, one can link the
first to the second, and the second to the first.
• A large (11 equations) recursive block, containing mostly descriptive elements (the terms of trade), items in
the State budget (amount of tariffs) and variables influencing only the future (the capital stock)
• A small non-recursive 3 equations block, linking Firms profits, Firms Financial capacity, and net interests paid
(depending on the capacity and entering profits These elements depend on the rest of the model, but have
no influence.
443
• A large (18 equations) recursive block, containing mostly descriptive elements (the terms of trade), items in
the State budget (number of tariffs) and variables influencing only the future (the capital stock)
• A small non-recursive 3 equations block linking Government balance (revenue – expenditures) interests paid
(depending on the balance) and expenditures (including interests).
• A 1 equation final block for the balance in GDP points.
10.2.2.10.2 Producing the data
We hope the comments included in this program make it self-explanatory. Basically, the OECD data is used to create
the model series in sequence, and the results are saved. Of course, elements used in any computation must have been
created in a previous statement (if the sequence defined a model it would have to be recursive).
To adapt this program to another model using the same concepts, we advise to replace all references to the original
“FRA_” by the specific name in the original file. This should produce a nearly correct version, with some exceptions:
• Unavailable variables which will have to be created (a common example is capital)
• Variables known using a different definition. This will change the logic of the statements. For instance OECD
provides wages including and not including contributions, but not the contributions themselves, which have
to be computed as a difference. If they are available, a direct transfer is possible.
10.2.2.10.3 Creating the model groups and checking data – equations
This program is also self-explanatory. The elements produced are:
• Groups for the endogenous (separated automatically into identity and behavioral) and .
As explained in the program, the fact that the behavioral equations, and those only, contain the element « f » is used
to separate them from the identities.
On has just to run two « residual check » simulations with different values of “f”, and identify as behavioral the
variables for which the results are different.
• A set of absolute and relative differences between the historical series and the result given by the associated
equation.
Let us make a comment on checking that the residual is null.
Actually, the value of the residual is seldom exactly null. Due to the limited precision of EViews, we generally get a
value like 10-6 (percent, so zero to 8 digits precision). Once the model reaches a certain size, it becomes difficult to
check by sight that the value is sufficiently small (the error can concern a limited range of periods, even a single one).
There are at least two ways to treat the problem, one immediate but limited, the other harder to implement but more
efficient and exact.
o The first is to produce an EViews graph. If all residuals are negligible, we should get a set of Brownian
motions, with a very low higher and lower maximum values. So we know immediately if there are no errors.
But it is difficult to identify the culprits (and an EViews graph is limited in series display).
o The second is to export the relative errors to an Excel sheet, sort the file for a given year, and concentrate on
the top and bottom values. The process is repeated until only negligible errors remain.
444
10.2.2.10.4 Estimating the equations
We have already presented its elements in detail.
10.2.2.10.5 Solving the model on the future
The solving process gives the same results as usual. If convergence is achieved (our case), no message is produced. If
not, we have described earlier the techniques one can use to solve the problem.
10.2.2.10.6 Producing a forecast
In our case, the goal is not to produce an actual forecast, but a simulation giving results acceptable enough to be used
as a base for the shocks which will follow.
This property can be controlled by displaying the growth rates of the main variables in the short and medium runs.
In our case, the model converges normally under Gauss-Seidel, to values not very different from history for the first
periods (we have only presented the convergence of the rate of use UR and the value-added deflator PQ).
Model: _MOD_1
Date: 04/14/20 Time: 17:41
Sample: 2017Q1 2020Q4
Solve Options:
Dynamic-Deterministic Simulation
Solver: Gauss-Seidel
Max iterations = 5000, Convergence = 1e-06
Scenario: Scenario 1
Solve begin 17:41:45
2017Q1 Block 1 - 1 eqns Solved (recursive block)
2017Q1 Block 2 - 51 eqns Convergence after 24 iterations
2017Q1 Block 3 - 11 eqns Solved (recursive block)
2017Q1 Block 4 - 3 eqns Convergence after 6 iterations
2017Q1 Block 5 - 18 eqns Solved (recursive block)
2017Q1 Block 6 - 3 eqns Convergence after 7 iterations
2017Q1 Block 7 - 1 eqns Solved (recursive block)
2017Q2 Block 1 - 1 eqns Solved (recursive block)
2017Q2 Block 2 - 51 eqns Convergence after 29 iterations
2017Q2 Block 3 - 11 eqns Solved (recursive block)
2017Q2 Block 4 - 3 eqns Convergence after 6 iterations
2017Q2 Block 5 - 18 eqns Solved (recursive block)
2017Q2 Block 6 - 3 eqns Convergence after 7 iterations
2017Q2 Block 7 - 1 eqns Solved (recursive block)
2017Q3 Block 1 - 1 eqns Solved (recursive block)
2017Q3 Block 2 - 51 eqns Convergence after 26 iterations
2017Q3 Block 3 - 11 eqns Solved (recursive block)
2017Q3 Block 4 - 3 eqns Convergence after 7 iterations
2017Q3 Block 5 - 18 eqns Solved (recursive block)
2017Q3 Block 6 - 3 eqns Convergence after 7 iterations
445
2017Q3 Block 7 - 1 eqns Solved (recursive block)
2017Q4 Block 1 - 1 eqns Solved (recursive block)
2017Q4 Block 2 - 51 eqns Convergence after 28 iterations
2017Q4 Block 3 - 11 eqns Solved (recursive block)
2017Q4 Block 4 - 3 eqns Convergence after 7 iterations
2017Q4 Block 5 - 18 eqns Solved (recursive block)
2017Q4 Block 6 - 3 eqns Convergence after 8 iterations
2017Q4 Block 7 - 1 eqns Solved (recursive block)
2020Q1 Block 1 - 1 eqns Solved (recursive block)
2020Q1 Block 2 - 51 eqns Convergence after 22 iterations
2020Q1 Block 3 - 11 eqns Solved (recursive block)
2020Q1 Block 4 - 3 eqns Convergence after 8 iterations
2020Q1 Block 5 - 18 eqns Solved (recursive block)
2020Q1 Block 6 - 3 eqns Convergence after 11 iterations
2020Q1 Block 7 - 1 eqns Solved (recursive block)
2020Q2 Block 1 - 1 eqns Solved (recursive block)
2020Q2 Block 2 - 51 eqns Convergence after 22 iterations
2020Q2 Block 3 - 11 eqns Solved (recursive block)
2020Q2 Block 4 - 3 eqns Convergence after 8 iterations
2020Q2 Block 5 - 18 eqns Solved (recursive block)
2020Q2 Block 6 - 3 eqns Convergence after 8 iterations
2020Q2 Block 7 - 1 eqns Solved (recursive block)
2020Q3 Block 1 - 1 eqns Solved (recursive block)
2020Q3 Block 2 - 51 eqns Convergence after 25 iterations
2020Q3 Block 3 - 11 eqns Solved (recursive block)
2020Q3 Block 4 - 3 eqns Convergence after 8 iterations
2020Q3 Block 5 - 18 eqns Solved (recursive block)
2020Q3 Block 6 - 3 eqns Convergence after 8 iterations
2020Q3 Block 7 - 1 eqns Solved (recursive block)
2020Q4 Block 1 - 1 eqns Solved (recursive block)
2020Q4 Block 2 - 51 eqns Convergence after 22 iterations
2020Q4 Block 3 - 11 eqns Solved (recursive block)
2020Q4 Block 4 - 3 eqns Convergence after 8 iterations
2020Q4 Block 5 - 18 eqns Solved (recursive block)
2020Q4 Block 6 - 3 eqns Convergence after 8 iterations
2020Q4 Block 7 - 1 eqns Solved (recursive block)
Solve complete 17:41:45
446
1.40
1.39
1.38
Convergence in 2020 1.37
0.990 1.36
0.988 1.35
0.986
0.984
Rate of use
0.982
GDP defator
0.980
0.978
2 4 6 8 10 12 14 16 18 20 22
.030
The quarterly growth rates
.025
Value added
.020 Vaue added deflator
Rate of use
.015
.010
.005
.000
-.005
2025 2030 2035 2040 2045 2050 2055 2060
If we consider now the evolution with time, we can see that all growth rates stabilize after 40 years. But the quantities
and the output gap do it rather quickly, as well, while the deflator takes more time. Of course, this depends a lot on
the original distance to the solution.
10.2.2.10.7 Producing shocks on assumptions
Let us now see how our model answers to shocks on its assumptions. We shall be brief, and only present and
comment the results for four of them: our goal is mostly to show that the present framework and the associated
estimations can provide a model with consistent properties. With a little bit of luck, you should be able to do the
same, or probably better. The model(s) we are presenting are far from perfect, and we did not try to make them so154.
All shocks will start in the first quarter of 2010 and will be sustained for the whole period. To make interpretation
easier, we shall limit the periods to the first fifteen years (for the short and medium-term dynamics) and the last three
(for the long-term impacts).
10.2.2.10.7.1 AN INCREASE IN GOVERNMENT DEMAND
154
Actually, we can identify a few coefficients which, with different values, would improve slightly model properties.
Of course, we will not apply these changes.
447
Starting in the year 2021 we increase Government demand by 1% of the GDP forecasted for the same period. This will
make interpretation easier than for a constant shock; in particular concerning the “Keynesian multiplier”.
The definition of this element, in practice the most frequently used to characterize a model, is quite simple. In the
presence of a given shock on demand, coming for instance from Government consumption, it measures the ratio
between the change in GDP produced by the shock and its initial size. If the ratio is higher than one, it means that the
economic effects of the shock on the economy are higher than the negative ones. Of course, this figure has also to be
interpreted.
In our case, the “Keynesian multiplier” will be obtained naturally by considering the relative percent change of GDP
itself. If GDP increases by more than 1%, it means that the response is higher than the shock.
We observe in Graph 1 that the multiplier grows then stabilizes to an average value of 0.9.
Final demand presents the same evolution, around a higher level, stabilizing at 1.6% in the long run, and imports at
1.2%. The initial decrease in exports is amplified with time, reaching 0.7%.
These elements can be easily explained, considering the model as a whole.
As local demand and GDP are of the same order of magnitude (their difference is net exports, relatively small
compared to GDP), the initial (or ex-ante) shock increased demand by roughly 1%. This means that the impact of
model mechanisms reduces the ex-post value of GDP growth. Let us detail this aspect.
2.4
Graph 1 : Shock on Government demand
2.0
The supply - demand equilibrium
1.6
Shock : +1 point of GDP
1.2
in percentage
0.8
0.4 Final demand GDP
Exports Imports
0.0 GDP deflator
-0.4
-0.8
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
The second graph shows that productive factors (capital and employment) grow significantly, requiring a strong effort
on investment in the short run. As explained earlier, firms expect production to represent a “normal” share of
capacities, defining a target on their rate of use. If production increases, a gap will appear, which firms will try to close
by increasing capacities through investment and labor. However, this process is slowed by the cautiousness of firms,
and the technical difficulty in implementing their decisions. This inertia applies especially to capital, for which the
consequences of wrong decisions are more costly. It is only in the last period of our sample that capacities get close
to the target, with a possible overshooting in the next years.
448
1.4
Graph 2 : Shock on Government demand
1.2 The production elements
1.0
Shock : +1 point of GDP
0.8
in percentage
0.6
Value added Capacity
0.4 Capital Employment
Rate of use Relative cost
0.2
0.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
However, capital and employment evolutions diverge from capacity. In the beginning, employment grows more, as
adaptation to the target is faster (the change in capital comes only from investment, itself quite inert). In the medium
run, the order is reversed by the increase in the relative cost of labor, coming from the decrease in unemployment
and the partial indexation on the value-added deflator, the element with the highest increase among prices. In the
end the increase in employment is quite limited, and capital increases by 1.3% (representing an increase in investment
of the same amount).
Consumption grows even slower. There are several reasons for this inertia.
• Employment adapts gradually to the production increase.
• The purchasing power of wages is affected initially by non-unitary indexation on prices, the gap building up
with additional inflation. This loss will disappear as the price increase stabilizes.
• The short-term sensitivity of consumption to revenue is lower than one.
• Inflation reduces the purchasing power of current savings.
• A large share of household revenue is not indexed on activity.
The real wage rate profits from the decrease in unemployment, but this has an adverse moderating effect on
employment itself, as we have seen.
The initial decrease in unemployment is the main explanation for the evolution of prices (Graph 6). In the short and
medium terms, the higher inflation comes from tensions on the rate of use (with higher sales perspectives, firms feel
that they can apply higher prices), and on the labor market through the unemployment rate.
449
3.5 Graph 6 : Shock on Government demand
The prices
3.0
Wage rate Consumption
2.5 Value added Exports
Imports
2.0
Shock : +1 point of GDP
in percentage
1.5
1.0
0.5
0.0
-0.5
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
One moderating element is the short-term increase in labor productivity, due to the inertia facing an increase in
production. But we have seen that the adaptation of labor to production was rather fast. 155 This element creates very
temporary deflation.
Let us now consider the trade balance (Graphs 4 and 5). We know that the trade prices have a different sensitivity to
local inflation (0.2 for imports, 0.6 for exports). And the short and long-term competitiveness elasticities are lower
than one. But the change in local prices is not negligible, especially in the long-term.
2.0
1.5
1.0
Graph 4 : Shock on Government demand
0.5
Shock : +1 point of GDP
External trade at current prices
in percentage
0.0
-0.5 Exports Imports
Exports-Imports ratio
-1.0
-1.5
-2.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
155
As to capital, the effect is more complex: the investment effort, destined to generate capacities for the future, is
higher in relative terms than the present increase in GDP. But in the cost we only consider the share of the present
period, through a correcting coefficient.
450
Graph 5 : Shock on Government demand
1.0 The exports - import ratios
0.5
0.0
Shock : +1 point of GDP
At constant prices Terms of trade
-0.5
in percentage
At constant prices
-1.0
-1.5
-2.0
-2.5
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
The evolution of exports is the simplest to explain: they are essentially sensitive to the loss in price competitiveness,
with the highest elasticity. They follow also the rate of use of capacities, but the effect is limited by the coefficient and
the evolution of the rate itself, which returns to the base value at the end of the sample. Graph 1 shows that exports
decrease for the whole period, due to the high increase in inflation.
Imports are more affected: they follow local demand, with the addition of losses in market shares, coming first from
the limits on available capacity (decreasing with time) then from the gradual loss in price competitiveness, which
compares the local production price to rather stable import prices. Both effects roughly offset each other during the
period, and the profile of imports is quite stable. The growing gap with demand comes essentially from intermediary
consumption goods, a large share of which follows the decrease in exports.
Concerning the balances, the “gain” on the terms of trade is limited, and the loss terms of trade are little affected, and
the loss is not so different on both nominal and real balances.
Now, as to the Government budget, the additional expenditures lead to some revenue, coming mostly from the taxes
on production. But as borrowing finances this measure, the interests paid will also increase.
There are two ways to present the changes in the budget. Its variables are measured at current prices. But:
Measured in levels, each variable (including the original Government demand) increases with economic growth and
inflation, which makes it difficult to interpret its evolution.
The immediate idea is to eliminate this time effect by dividing the change by a variable representative of the growth,
the most neutral idea being to use GDP at current prices.
We get the following table, in which we present the changes for the first year, then for the last quarter of selected
years.
451
2021Q1 2021Q2 2021Q3 2021Q4 2022Q4 2023Q4 2024Q4 2025Q4 2026Q4 2030Q4 2050Q4 2100Q4
Expenditures
Total expenditures 0.97 0.93 0.90 0.91 1.67 2.13 2.21 2.31 2.47 3.02 4.44 6.43
Demand 1.00 1.00 1.00 1.01 1.05 1.07 1.09 1.11 1.14 1.23 1.53 1.62
Wages -0.00 0.02 0.04 0.05 0.09 0.10 0.11 0.12 0.13 0.18 0.30 0.33
Social contributions -0.00 0.01 0.02 0.03 0.05 0.05 0.05 0.06 0.07 0.09 0.15 0.17
Social benefits -0.00 -0.00 -0.00 -0.00 0.00 0.01 0.01 0.02 0.02 0.04 0.09 0.10
Subsidies 0.02 0.02 0.02 0.02 0.02 0.03 0.03 0.03 0.04 0.05 0.09 0.10
Interest paid -0.05 -0.12 -0.18 -0.19 0.46 0.87 0.91 0.97 1.08 1.44 2.28 4.12
Revenue
Total revenue 0.07 0.20 0.28 0.34 0.46 0.49 0.53 0.59 0.65 0.87 1.48 1.66
Value added tax 0.04 0.04 0.04 0.05 0.05 0.06 0.06 0.06 0.07 0.08 0.12 0.13
Other taxes on production 0.00 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.02 0.03 0.03
Tariffs 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Social contributions employers 0.01 0.05 0.08 0.10 0.14 0.14 0.15 0.17 0.18 0.24 0.40 0.44
Social contributions workers 0.01 0.06 0.09 0.11 0.15 0.15 0.17 0.18 0.20 0.27 0.45 0.50
Income tax 0.00 0.02 0.03 0.04 0.07 0.08 0.08 0.09 0.11 0.15 0.27 0.30
Tax on profits 0.00 0.02 0.01 0.01 0.00 0.01 0.01 0.02 0.02 0.03 0.06 0.08
Government balance -0.90 -0.72 -0.62 -0.57 -1.21 -1.63 -1.67 -1.72 -1.82 -2.15 -2.96 -4.78
We observe a relative stability of the elements, with two opposite evolutions:
• An increase in revenue with wages (both through social security payments and household consumption).
• A gradual increase of interests paid, with the accumulation of debt.
The first effect wins in the short run, but the second reverses the order quickly.
Of course, it is interesting to obtain a picture of the evolution of the Government budget, measured in currency. But
we have presented part of the issue
Even with a shock on real demand, the values in the table combine both a change at constant prices and an effect of
inflation, and the interpretation is still difficult, in terms of gains and losses. The growth of a revenue with time might
mask a decrease in real terms.
Eliminating both effects makes no sense (the multiplier disappears…), but eliminating inflation allows to reason in
constant terms.
We will do this in the second table by computing the difference between elements divided by their solution deflator
𝑥𝑡𝑠 𝑥𝑡𝑏
𝑒𝑡 = −
𝑝𝑞𝑡𝑠 . 𝑞𝑡𝑏 𝑝𝑞𝑡𝑏 . 𝑞𝑡𝑏
eliminating inflation but keeping the change in the real value in the numerator.
We get quite a different picture as to the evolution of variations (of course, for a given period, the relative evolution
remains the same). As could be expected, the figures are much reduced and more stable.
452
2021Q1 2021Q2 2021Q3 2021Q4 2022Q4 2023Q4 2024Q4 2025Q4 2026Q4 2030Q4 2050Q4 2100Q4
Expenditures
Total expenditures 0.98 0.95 0.91 0.91 1.57 1.94 1.96 2.00 2.10 2.35 2.78 4.26
Demand 1.01 1.01 1.01 1.01 1.00 1.00 1.00 0.99 0.99 0.97 0.92 0.90
Wages 0.00 0.02 0.04 0.05 0.07 0.07 0.07 0.08 0.08 0.09 0.08 0.08
Social contributions 0.00 0.01 0.02 0.03 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04
Social benefits -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.01 -0.01 -0.02 -0.02
Subsidies 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.03 0.03 0.03
Interest paid -0.05 -0.12 -0.18 -0.19 0.44 0.82 0.84 0.88 0.97 1.24 1.73 3.24
Revenue
Total revenue 0.07 0.22 0.29 0.34 0.39 0.37 0.37 0.39 0.40 0.44 0.44 0.42
Value added tax 0.04 0.04 0.04 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05
Other taxes on production 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01
Tariffs 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Social contributions employers 0.02 0.06 0.09 0.10 0.12 0.11 0.11 0.11 0.12 0.13 0.13 0.12
Social contributions workers 0.01 0.06 0.09 0.11 0.13 0.12 0.12 0.12 0.13 0.14 0.14 0.13
Income tax 0.00 0.02 0.03 0.04 0.05 0.05 0.05 0.05 0.05 0.06 0.05 0.05
Tax on profits 0.00 0.02 0.01 0.01 0.00 0.01 0.01 0.01 0.01 0.02 0.02 0.02
Government balance -0.90 -0.73 -0.62 -0.57 -1.18 -1.57 -1.59 -1.61 -1.69 -1.92 -2.34 -3.84
In GDP points
But the relative evolution of the ex-ante and ex-post Government balance remains the same.
For Government demand, the change keeps mostly at its original level, but not exactly. This relatively small
discrepancy is of due to the deflator (GDP for the computation, final demand for the variable). If we use the final
demand deflator in the above formula, we get:
2021Q1 2021Q2 2021Q3 2021Q4 2022Q4 2023Q4 2024Q4 2025Q4 2026Q4 2030Q4 2050Q4 2100Q4
Expenditures
Total expenditures 0.97 0.94 0.91 0.91 1.58 1.96 1.98 2.03 2.13 2.43 3.00 4.52
Demand 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
Wages 0.00 0.03 0.04 0.05 0.08 0.08 0.08 0.08 0.09 0.10 0.12 0.12
Social contributions 0.00 0.01 0.02 0.03 0.04 0.04 0.04 0.04 0.04 0.05 0.06 0.06
Social benefits 0.00 -0.00 0.00 -0.00 -0.00 0.00 -0.00 0.00 -0.00 0.00 -0.00 -0.00
Subsidies 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.03 0.03 0.04 0.04
Interest paid -0.04 -0.11 -0.17 -0.19 0.44 0.82 0.84 0.88 0.98 1.25 1.78 3.31
Revenue
Total revenue 0.09 0.22 0.30 0.35 0.41 0.40 0.40 0.43 0.45 0.51 0.60 0.61
Value added tax 0.04 0.04 0.04 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.06 0.06
Other taxes on production 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01
Tariffs 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Social contributions employers 0.02 0.06 0.09 0.11 0.12 0.11 0.11 0.12 0.13 0.14 0.17 0.17
Social contributions workers 0.02 0.06 0.09 0.11 0.13 0.13 0.13 0.13 0.14 0.16 0.19 0.19
Income tax 0.00 0.02 0.04 0.04 0.06 0.06 0.06 0.06 0.06 0.07 0.09 0.09
Tax on profits 0.00 0.02 0.01 0.01 0.00 0.01 0.01 0.01 0.01 0.02 0.03 0.03
Government balance -0.89 -0.72 -0.61 -0.57 -1.16 -1.56 -1.58 -1.60 -1.68 -1.92 -2.39 -3.91
In GDP points
But of course, using different deflators in the same table is not an option, as the balances would no longer hold true.
Running the associated programs, the reader will be able to observe other elements. But he is also expected to
produce his own model, and to observe how his changes (limited or extensive) modify model properties.
10.2.2.10.7.2 2 A SUPPLY SHOCK: DECREASING THE VAT RATE BY ONE POINT
We suppose that the Government decreases the VAT rate by 1 point.
453
Graph 6 : Shock on the value added tax Wage rate Consumption
The prices Value added Exports
0
Imports
-1
Shock : -1 point of the rate
-2
in percentage
-3
-4
-5
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
Graph 6 shows that all prices decrease significantly. Of course, the highest impact is on the consumption price (fully
depending on VAT). The other prices follow, at various levels depending on their sensitivity.
The wage rate is a special case. Representing initially an average between the consumption and value-added deflators.
It diverges gradually upwards, with the gains in purchasing power due to the lower unemployment.
454
One can observe the loop between the wage rate and the value-added deflator, moderated by trade deflators (the
import price is less sensitive to the GDP deflator). Compared to a complementary factors framework, the margins
target includes also the capital cost, moderating the role of wages but introducing an additional loop.
The evolution of the supply-demand elements is presented in graph 1 (we have dropped inflation for clarity).
1.2
Graph 1 : Shock on the value added tax
1.0 The supply - demand equilibrium
0.8
Shock : -1 point of the rate
in percentage
0.6
0.4
Final demand GDP
0.2
Exports Imports
0.0
-0.2
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
The gain in competitiveness favors exports, once capacities have started to adapt, creating growth through demand in
investment and consumption goods. The subsequent increase in imports (coming from local demand but also from
intermediate goods and energy requested by exports) is accentuated at first by the initial bottlenecks.
1.4
1.2 Value added Capacity Graph 2 : Shock on the value added tax
Capital Employment The production elements
1.0 Rate of use
0.8 Shock : -1 point of the rate
in percentage
0.6
0.4
0.2
0.0
-0.2
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
Then competitiveness, exports and GDP keep increasing. The output gap will close, but very slowly as the target is
moving, with a combination of factors favoring capital due to the increase in the real wage rate, as in the previous
case. As employment adapts faster than capital (which closes slowly with its target), capacity takes a little time in
increasing.
455
The impact on the real trade balance (Graph 5) is negative at first, as the increase in GDP is followed by demand, both
final and intermediate. In the medium-long-term, the gains in competitiveness on both trade elements reverses the
evolution. But the loss on the current balance remains large.
Graph 5 : Shock on the value added tax
The exports - import ratios
0.50
0.25
0.00
Shock : -1 point of the rate
-0.25
in percentage
-0.50
-0.75
-1.00
-1.25
-1.50
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
At constant prices Terms of trade
At current prices
As to the Government budget, its evolution is similar to the previous case, except for the interests with a growing
importance of interests paid, which decrease at first with the rate itself (remember it is indexed on the evolution of
CPI, which decreases for the first two years). However, the cost is lower, and the gain on GDP is higher: this
instrument seems more efficient on that aspect.
10.2.2.10.7.3 A DEVALUATION (LEAVING THE EUROZONE)
Our shock supposes that France leaves the Eurozone, and suffers a devaluation as the consequence.
We will depreciate the new currency by 1% compared to the Euro.
Of course, this shock is purely technical, as many of its consequences are not made explicit.
Our first graph presents prices, the second quantities.
Concerning the latter, we see that in the long run they all increase by 1%, with a hierarchy consistent with the
formulas. The highest growth is for the wage rate, as the shock presents a temporary increase in GDP. As it
disappears, the gain disappears too.
456
1.2 Graph 6 : Shock on the depreciation rate
The prices
1.0
0.8
in percentage
Shock : +1%
0.6
Wage rate Consumption
0.4 Value added Exports
Imports
0.2
0.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
Concerning quantities, exports profit the most from the shock, improving GDP. But imports actually increase a little,
the gains in competitiveness being more than balanced by the additional demand (essentially from intermediate
consumption required by exports).
Graph 2 shows that the relative cost favors employment initially, as the wage rate indexation takes a few periods,
while the price of investment follows the general demand deflator. The real wage rate decreases a little, and
consumption follows. Final demand actually decreases.
Graph 2 : Shock on the exchange rate
.06 The production elements
.05 Value added Capacity
Capital Employment
Rate of use
.04
.03
in percentage
Shock : +1%
.02
.01
.00
-.01
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
457
.12 Graph 8 : Shock on the exchange rate
Demand in GDP points
.10
Final demand Household consumption
.08 Investment Total demand
GDP
.06
in percentage
Shock : +1%
.04
.02
.00
-.02
-.04
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
The last graph presents the evolution of trade. We see that the gains in competitiveness (associated with a loss on the
terms of trade) are reduced gradually from the first, but that the gains in real trade take some time in reaching the
highest level (due to the initial bottlenecks and increase in demand). A very limited gain at current prices appears only
in the medium run.
.2
.1
.0
-.1
in percentage
Shock : +1%
Graph 5 : Shock on the depreciation rate
-.2 The exports - import ratios
At constant prices Terms of trade
-.3
At current prices
-.4
-.5
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
10.2.2.10.7.4 A DECREASE IN THE FOREIGN TARIFFS RATE
As you have certainly guessed, this shock is essentially demand-oriented:
The improved competitiveness will increase the demand addressed to France, with basically the same effects on GDP
as Government demand, from the supply – demand equilibrium.
458
1.4
1.2 Graph 1 : Shock on foreign tariffs Final demand GDP
The supply - demand equilibrium Exports Imports
1.0 GDP deflator
Shock : -1 point of the rate
0.8
in percentage
0.6
0.4
0.2
0.0
-0.2
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
Of course, one can argue that lower import prices from France will reduce inflation in the rest of the world, which
should affect French import prices. World GDP should also change (in a dubious way, positive through disinflation,
negative through the higher share of France in world trade). But both these effects can be considered negligible,
especially if we consider the cost of the alternative: building a reliable world model.
Of course, a model considering the whole European Union would represent a different issue.
.8
Graph 5 : Shock on foreign tariffs
.7 The exports - import ratios
.6
.5 Shock : -1 point of the rate
in percentage
.4
.3
.2
At constant prices Terms of trade
.1
A current prices
.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
On the whole, the current gain stabilizes at constant prices after a while, when capacities have adapted to the new
demand level. But the growing inflation brings a gain on the terms of trade, and the balance in current terms improves
regularly.
459
460
10.2.2.10.7.5 A DECREASE IN THE LOCAL TARIFFS RATE
This shock brings traditionally the most complex mechanisms, and its consequences are quite volatile from one model
to the other. Two main channels have to be considered:
• The improved imports competitiveness increases their share in local demand, reducing local output, with the
traditional consequences of a demand shock, only negative.
• The lower import prices bring global disinflation, especially for demand but also for value added through
lower wages (indexed on the consumer price). This disinflation helps local firms to compete with foreign
producers on the foreign and local markets, reducing the initial gap on the first.
• The cost of capital (part of it imported) decreases more than the value-added price. This increases
profitability and creates investment and productive capacity per se, independently from demand. Firms
working at full capacity will see the constraint released156, and local producers as a whole will gain market
share on the local and foreign markets. At the same time, to improve demand for these new capacities, they
will decrease their prices.
First let us consider prices, the field impacted directly:
Graph 6 : Shock on local tariffs Imports Exports
The prices Wage rate Imports including tariffs
Value added Production
0.0
Final demand
-0.4
Shock : -1 point of the rate
-0.8
in percentage
-1.2
-1.6
-2.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
To understand the order for the evolutions, we had to introduce additional lines, hoping this did not make the graph
too crowded.
We observe in sequence a set of proportional evolutions:
o The import price excluding tariffs, linked essentially to the world price.
o The export price, less sensitive to the world price.
o The value-added price, the most subject to deflation, as in addition to the lower consumption price wages are
sensitive to the higher unemployment.
156
And possibly reappear at a higher production level.
461
o The production price, a weighted average of value added and demand (intermediate consumption).
o The demand price, a weighted average of imports and local prices (with a small impact from exports).
And two special cases:
o The import price including tariffs, with a one-point difference to the base evolution.
o The wage rate, which should follow an average of value added and demand, but is also subject to inertia, and
to unemployment which increases then decreases, for a reason we will explain.
Graph 6 : Shock on local tariffs Wage rate Consumption
The prices Value added Exports
Imports
0.0
-0.4
Shock : -1 point of the rate
-0.8
in percentage
-1.2
-1.6
-2.0
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
Things should be clearer on a specific graph.
• The real wage rate increases a little.
• Exports price competitiveness grows regularly then stabilizes.
• Import price competitiveness decreases and changes sign when the local production price overcomes the
import price including tariffs.
So, in the long run the global outcome of the shock is a gain in competitiveness on both trade elements.
462
0.8 Graph 6 : Shock on local tariffs
Competitiveness
0.4
0.0
Shock : -1 point
Exports price (= -competitiveness) Imports price including tariffs
in percentage
-0.4 Production price Import competitiveness
Real wage rate
-0.8
-1.2
-1.6
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
Let us now study the evolution of quantities.
.5 Graph 1 : Shock on local tariffs
The supply - demand equilibrium
.4
.3
Shock : -1 point of the rate
in percentage
.2
.1
Final demand GDP
Exports Imports
.0
-.1
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 98 99 00
In the short and medium runs, imports show the highest increase, as the only ex ante positively influenced element.
But the improvement of local competitiveness limits that effect. At the same time, disinflation profits to exports, and
also to demand: profitability favors investment, and the gain in purchasing power of the wage rate favors
consumption and imports in turn.
After 15 years, both trade elements gain competitiveness, and the real trade balance starts improving. Demand and
GDP too. The only negative evolutions are for the Government budget, but also for the trade balance: the limited gain
in real terms is obtained through a large decrease in the export price, and a loss in the actual terms of trade (as seen
initially on Graph 6). The local consumers do not actually buy more goods, and the Government will have to increase
taxes if it wants to balance the budget.
463
.4
Graph 5 : Shock on local tariffs
The exports - import ratios
.2
.0
Shock : -1 point of the rate
in percentage
-.2 At constant prices Terms of trade
At current prices
-.4
-.6
-.8
2020 2022 2024 2026 2028 2030 2032 2034 2098 2100
10.3 A SINGLE COUNTRY, MULTI PRODUCT MODEL
Let us now consider separating the products of our model into several categories. We shall address in turn:
• The reasons for product decomposition
• The specific features introduced by product decomposition, and the way it can be treated.
10.3.1 THE MAIN REASONS FOR PRODUCT DECOMPOSITION
There are three main reasons for product decomposition:
• The decomposition allows to take into account differences in the values of structural elements.
This appears when structural parameters take different values from one product to the other, which implies that they
will react differently to external influences (exogenous or endogenous), or exogenous assumptions concerning the
product.
• The decomposition allows to evidence differences in the sensitivity to some explanatory elements.
• For some economic element or field, the links between variables follow different mechanisms, calling for
different specifications which cannot be summarized by mathematical aggregation.
10.3.1.1 Structural differences
They can be identified easily within the framework of our model. The main elements are:
464
• Factor productivity: the quantity of labor and / or capital necessary to produce one unit of good can be
different. This is particularly true for agriculture, in which labor productivity is particularly low 157. And also,
for services, where the need for capital is generally limited.
• The wage rate: the average wage obtained by workers can be quite different (linked often to labor
productivity, which can offset the effect when we consider the unit cost).
• The different shares in each demand item, allowing to present different sensitivities to a given increase in
consumption or in investment.
• The amount of intermediate consumption of each good needed to produce one unit of a given good.
• The sharing of production destination between the local and foreign markets (exports).
• The sharing of local demand between the local and foreign producers (imports).
• The separation between firms, households, the state and foreign agents in the local production process, with
different consequences on revenue and its use.
• The taxation of goods (VAT, other indirect taxes, tariffs on local imports and exports).
All these elements will be taken into account at no cost, once you have separated the goods in the model. The only
problems come from:
• Data availability, which can be partial. For instance, the separation into goods could be known, but not the
full input-output table, or the account of firms.
• Model complexity, both for writing down, checking and understanding the results.
10.3.1.2 Differences in sensitivity
This can apply to all parameters in estimated equations. Of course, one should be concerned with the most important
ones, and economic theory can point out for which elements significant differences can be expected. For instance, the
need for inventories might be more important in the manufacturing sector, or the role of unemployment limited in
defining the agricultural wage, or the sensitivity of external trade to price competitiveness can be low for services.
This does not call for specific formulas. The data should dictate the differences in properties.
10.3.1.3 Differences in logical behavior
This time the very formulation of a behavior, or its causal logic, will be different.
This applies essentially to two domains.
10.3.1.3.1 The production processes
• In the agricultural sector, one can assume that producers maximize their output. For this, they can use
plantations (fields and trees), and animals, or a combination of both (like plowing, or raising trout around rice
fields). They need also a minimum amount of labor, tools and intermediate consumption goods (like fodder).
Increasing these levels can improve output, as well as better infrastructures (irrigation, storage and
transportation facilities)158. But once these are defined, production follows, corrected heavily by
157
Especially in family managed units.
158
Remember that production includes only the goods actually bought (consider the supply-demand equilibrium) or
entering inventories, which do not apply to agricultural goods (processed goods are considered industry).This means
spoiled products are not considered, and improving transportation will increase agricultural production, all things
being equal.
465
unforecasted climatic conditions. This is also mostly true of fishing. The impact of an increase in demand is
limited at least in the short run (even if perhaps cattle will be slaughtered earlier, more fish will be caught, or
more trees will be cut).
• In the manufacturing sector, we can assume the availability of several processes, among which the producer
will chose according to the relative costs of capital and labor. Assuming a constant and unitary elasticity leads
to a Cobb-Douglas formulation. We can also choose a simple complementary factors specification, as well as
a more general CES function, or even a very specific set of formulas, describing for instance the production of
electricity using dams, coal, oil or nuclear power, according to the level of instantaneous demand.
Actually, one of the advantages of product identification is to associate a sophisticated framework to the goods which
actually deserve it. In the single good case, the elements of the alternative: using a simple function or associating a
Cobb-Douglas to the whole economic spectrum are both harmful to model quality and results.
But the main advantage of this identification is to define a productive capacity, and the associated rate of use. This
element will play an important role (as we have seen already) in three elements: the role of local producers in the
supplying of local and foreign goods, the investment behavior, and the short-term choice by firms of the mark-up
applied to the production cost.
• Finally, in the services sector, the choice of factors can depend on the relative cost (computers versus
people), but the role of capacity is less clear, as in many cases producers have a considerable leverage on
output, independently of installed capital and even employment: a sudden and high increase in the demand
for touristic travel can generally be met by travel agencies, if customers are willing to change their plans. The
quality of the service will probably decrease, but not its cost, the element by which the output is measured.
Dissatisfaction of customers does not reduce the value of a purchase, except if some reimbursement is due.
10.3.1.3.2 External trade
The trade in agricultural goods, manufactured goods and services follows obviously different rules.
• For agriculture, the export price is generally set to the world price for a given quality, and a given share of the
local production is proposed on the foreign market, according to the type and quality of goods, local policy,
pre-established contracts.
• For manufactured goods, additionally to demand, price competitiveness is essential, as well as available
capacity which can create temporary bottlenecks on the supply of specific elements.
• For services, the determinants are completely different and much less clear, but the amounts traded are less
important (except for tourism which is quite difficult to model). Financial services can be formalized
considering deposits and loans, rates and maturities (look at the chapter on Stock-Flow Consistent models).
10.3.1.3.3 Consequences for product decomposition
The above remarks should have made clear that the minimal decomposition should introduce three categories:
primary, secondary and tertiary products.
Going further, one can consider
• Using more products (more on this later).
o The most immediate idea concerns energy, especially for oil or gas producing countries. It is clear that the
level of production can change widely according to the will of the producer, that the variable production
costs are relatively small, and that the whole production should have no problem being sold, at a price highly
variable but defined on the world market.
466
o This is also true in general of mining, which should therefore be counted with the secondary goods, in its own
category, or together with energy.
o Construction can also be identified, as it follows the other products, and household revenue. The production
process is relatively straightforward for a given type, and it is neither imported nor exported, although its
inputs can be (like wood and steel).
o Services can be separated into Financial (managing loans) and non-Financial (transportation or meals in
restaurants).
• Identifying the type of ownership. Firms can be divided in state and private, and the latter into individual
firms and companies. Foreign owned firms can be identified (using FDI, repatriating profits). And this comes
naturally in the above case, particularly for Stock-Flow Consistent models, for with the agents column must
separate the two productions. And even if the production of individual owners is merged with firms’,
identifying households as an agent calls for separating the housing service.
10.3.2 INTRODUCING INTERMEDIATE CONSUMPTION
A very important development corresponds to intermediate consumption.
Any model, for any product, has to follow the equilibrium:
Production + imports = Total demand + exports
However, at the global level, subtracting total intermediate consumption on both sides give:
GDP + imports = Final demand + exports
This almost frees the model from defining intermediate consumption, a concept which is not easy to handle, as it
depends on the production process itself, more precisely on its number of stages.
Its only remaining role is the fact that in trade equations, price competitiveness is based on production prices:
preference for goods from a country is obviously based on the price at which they sell.
This means that the trade prices (exports and imports) must consider production prices, and that competitiveness
must compare the global cost of the exporter to the price at which its competitors sell (both including intermediate
consumption). This means for example that a country which has access to cheap oil or gas (like the US, Russia or
Algeria) has a comparative advantage over Germany or China, both against exporters to its market and against other
competitors on foreign markets159.
159
However, as we are using deflators, this means the trade prices for these countries will be less sensitive to the cost
of energy, which represents a lower share of the total cost. They will gain competitiveness if the price goes up, but
lose if it goes down (as they do not profit from the shock).
467
The above equation
Production + Imports = Total demand + Exports
Can be written as:
Value added + intermediate consumption by the product + imports
= Total demand + intermediate consumption of the product + exports
This calls for the identification of a matrix of intermediary consumptions, a square one with the dimension of the
decomposition.
Typically, the definition of each of these elements at constant prices will suppose that producing each unit of a given
good requires a given amount of each good (including itself).
At current prices, one could apply the deflator of the product used (excluding VAT, as it does not apply here) at the
global demand level, or at the global intermediate consumption level. The availability of the matrix at current prices
improves the situation by allowing to define an individual deflator for each cell.
Sometimes, the statisticians provide these tables only for some years, which calls for an interpolation.
10.3.3 SPECIFIC SECTORAL ISSUES
Now that we know the reasons for product decomposition, let us see how this affects model structure.
The most immediate change is of course the multiplication of categories. According to the item, the extension can
apply to a different criterion, or even two criteria at the same time.
A precise list of the items belonging to each category will be proposed as an annex.
First we can consider the products:
• Product 1: agricultural products (such as unprocessed food)
• Product 2: manufactured goods (such as household appliances)
• Product 3: services (such as transportation or banking).
This decomposition applies naturally to traded elements, such as consumption or exports.
But the same decomposition will apply to productive units, according to their output. Farms, industrial plants and
travel agencies will be classified in each of our three categories, in order.
And here a distinction must be made, as a given unit can produce different goods. In that case:
468
• The whole unit will be classified in the branch of its main activity.
• It will be fractioned into products, each share being allocated to the sector it represents.
In our model, we shall not separate branches and sectors, and we shall use alternately the two terms. But branches
and products can play different roles, and appear sometimes for the same variable, introducing a double indexation.
In addition, it might look interesting to proceed further in the decomposition, in two cases:
• For agriculture (and fishing) the model would benefit from a separation into artisanal and industrial units:
o Artisanal units use less capital (machinery) and much more labor, with a lower productivity.
o Artisanal employment will not be very sensitive to output (family units will often include a generally inactive
work force, which can be called for if needed).
o They also use fewer intermediate goods (such as fertilizer and fodder).
o Their revenue is entirely appropriated by households.
o They export a lower share of their production.
This is particularly interesting in developing countries, in which the role of agriculture is still important, especially for
employment (less for output). In 2015 Vietnam, it represented 41% and 15% respectively.
• The manufacturing sector includes energy and mining. Oil, gas and ore are traded through a process different
from other manufacturing products:
o The international price is fixed, and competitiveness plays no role (even for imports).
o Capacity also plays no role in the short run: quantities exported are decided by the exporter, and imports are
directly linked to demand160.
This means we shall separate in our example:
• Agriculture and fishing production into artisanal and industrial
• Manufacturing exports and imports into energy and non-energy products. We could also include mining; as it
plays an important role in the economy of some countries (Chile, Morocco, Australia…).
This is an example of the (frequent) case where decompositions can lead to further decompositions:
• it is only for agriculture that artisanal/industrial decomposition is really efficient,
• And excluding energy from manufactured trade comes from the fact that in that regard, it behaves more like
the primary product. This calls for a local correction of the general decomposition, if we want the model to
present reliable trade equations.
• Finally, the evolution to the integration of financial elements calls more and more to the identification of
financial firms.
10.3.4 THE DATA
For detailed data, very often the information gets scarcer. The following problems can occur:
• Detail on goods and services at current prices is known only from the production side, not the demand side.
160
But if local production goes down, imports can increase.
469
• For the detailed series which are available, deflators are known only at a more aggregate level, making the
computation of variables at constant prices approximate.
• Investment (and capital) is not available by product, or by investing branch, or (most often) in two-
dimensional detail. The same problem can appear for the change in inventories.
Fortunately, investment is made essentially in manufacturing goods and construction (plus livestock and plants for
agriculture). If the model uses a global secondary product, the dimension of the products is not required 161. In any
case, many cells in the matrix will contain null values.
Intermediate consumption by product and use is available more frequently, but not always (this means the input-
output matrix is not entirely known, perhaps only for some periods).
• Employment is not detailed by branch, or by categories of firm’s ownership.
• The part of the account of firms giving transfers (wages, subsidies...) is not detailed by branch.
• The period for which data is known is shorter, or more difficult to collect. For instance, full tables might be
produced yearly for the current year.
10.3.5 CONSEQUENCES FOR ESTIMATIONS
If the sample gets too short, we can still try to estimate. But we should not rely on results, however favorable. We
should just use them as another indication on the associated behavior, along with economic theory, observations on
the way the local economy works, and estimations at the global level.
But very often the above problems will not allow to estimate individual equations at the detailed level.
One can:
• Calibrate the equations using theoretical coefficients.
• Use the results estimated at the aggregate level.
• Take the coefficients from other models, describing a similar country (Thailand for Vietnam for instance, or
Austria for the Czech Republic). This option presents a surprising good point: the country used will generally
be more advanced (as it has a better and more developed data system) which means that its estimations are
based on a period which should represent the future of our modeled country, the actual period on which
economic studies will be performed.
• Estimate the equations as a system, in which some coefficients will be forced to take the same value. The size
of the “sample” is multiplied by the number of categories.
And actually, this difficulty has another silver lining: partially free from econometric restrictions, we can apply the
formulations which conform the most to economic theory, in specification and value of coefficients. In particular, we
can apply error-correction frameworks in each formula, leading to a model which will:
• Provide a long-term solution on any future period, without any change to specifications (with a few
restrictions on assumptions).
• Separate completely the long-term specification from the dynamics leading to it.
• Allow to interpret completely and easily the coefficients (using values which conform to theory).
161
Although investment contains a share of services (architects, patents).
470
10.3.6 THE PRODUCTION FUNCTION
As we have stated above, we can use different solutions for each branch.
10.3.6.1 Primary product
This is a field is quite difficult to manage. Capacity can depend on:
• Capital in product 1: Land made ready for use, plantations, and cattle.
• Capital in product 2: Machinery, more or less sophisticated.
• Public capital: availability of water and electricity, road and telecommunication network.
• Services capital: transportation, storage.
• Intermediate inputs: fodder, fertilizer.
• Climatic conditions (exogenous of course).
Formulations can be more or less complex, from a simple production function (complementary or Cobb-Douglas) to a
logical system including conditions and non-continuous functions. Output could also be decomposed into more
categories (agriculture/fishing/forestry) or into products (rice and coffee for Vietnam, fruits for Central America…). In
the last case, physical quantities could be used (tons, liters, numbers).
But basically, the short run definition of quantities produced should not depend so much on demand, but rather on
potential production including climatic factors (in the long run of course, profitability and market size will affect the
creation of capacities).
This has an important impact on model properties. As the supply – demand equilibrium still has to apply, we need a
new balancing element, which can be imports, exports, demand or one of its components. The simplest solution is to
use imports, which means that:
• Local demand has to be satisfied162
• Exports are determined by world demand and competitiveness (with a limited impact as agricultural export
prices are mostly determined at the world level).
• The share of global demand which cannot be satisfied locally has to be imported 163.
Other frameworks can be considered, such as:
• Local demand has to be satisfied
• Imports are controlled by local agents or the State.
• What is left from local production is exported.
Or:
• Exports are controlled by the state or local agents, according to market conditions.
• Local agents are allowed to buy the non-exported share of local production.
• Imports are also controlled.
162
But remember it is endogenous, so it follows global activity.
163
But if the price of imported goods is too high, consumption will decrease and will also move to other goods.
471
• This gives total local demand.
It is quite possible to consider several of these frameworks simultaneously, but only if two or three categories are
defined, each with its four elements which verify the particular equilibrium.
We can expect the consequences for model properties of the actual choice to be quite important.
10.3.6.2 Industrial product
One of the main purposes of the decomposition is to identify a sophisticated production function for the
manufacturing sector, leaving out the other branches for which this choice is much less natural, and would have a
negative impact on the quality of a global estimation.
A moderately complex option is the Cobb-Douglas function, which we presented earlier.
10.3.7 UNEMPLOYMENT
Unemployment follows the same logic as the simple model. However, one could define different sensitivities to
employment according to the branch. For instance, in construction, the lower share of qualified workers could lead
job creation to hire immediately productive unemployed.
10.3.8 CHANGE IN INVENTORIES
The new element is that we should develop the equation in two directions: the good in the inventory, and the branch
using it. This means that in principle we will have at least 9 variables. However, services are not stored as inventories,
and most agricultural products cannot be either (in particular for a yearly model).
To set the coefficients, we can consider theory, expert advice or the ratio of total inventory change to the change in
value added.
10.3.9 HOUSEHOLD CONSUMPTION
The decomposition brings a new feature: consumption has to be divided into products.
The most common option is to compute global consumption according to the single product option.
Then we separate consumption into products, using an error correction framework, with a target sharing depending
on the relative prices of consumption goods. After the decomposition has been applied, we need to ensure that the
sum is consistent with the total (this can be done using a system or applying a correcting factor).
Depending on the product, a high level of consumption can be maintained (representing habit forming) or negatively
correlated (if the good lasts several periods). For instance, a household could get used to eating meat, or wait several
years to buy a new car.
The situation is simpler if we consider two products, as we only need a single parting ratio, once global consumption is
known.
10.3.10 EXTERNAL TRADE
472
This is the second field in which product decomposition allows to introduce differences in the formulas across
products.
This originality will come from:
• Different weights: the share of exports in production, and imports in demand, is different from one product
to the other (and of course between imports and exports).
• Different sensitivities: estimating the role of price competitiveness can give different answers; if the equation
has to be calibrated, economic theory could lead to use different values.
• Different formulations: the role of tensions on capacities can be limited to the manufacturing product.
10.3.11 WAGES
For wages, we should use the same (theoretical) framework as in the simple model.
However, the indexation process is a little more complex. We still have a choice (in particular in the long run) between
an indexation on the consumption price and the value added deflator. But if the first element is measured at the
global level, the second corresponds to the branch. This means we can observe the consequences of dissymetric
shocks (such as an increase in subsidies to one sector), or shocks having different sectoral impacts (like an increase in
the margins of exporting firms, mostly industrial).
Also, the sensitivity to unemployment can be quite different, leading to dissymetric inflationary properties.
Finally, identifying the artisanal and industrial branches of agriculture will allow considering different productivities
and wage levels.
10.3.12 PRICES
10.3.12.1 The value-added deflator
With different production frameworks, the cost variable used in the long-term target will differ. And the role of the
rate of use should be less intense (or even absent) outside of manufacturing.
10.3.12.2 The production price
This is a domain in which the complexity increases: the intermediate consumption of a branch is the sum of individual
two-dimensional elements, each valued at its own price (which has to be defined).
10.3.12.3 The trade prices
Once production prices are known, we can move to the trade prices. To define the price of competitors, we need to
identify an average foreign production price for each of our products, which will depend on the structure of traded
goods, and of customers and suppliers. We have already addressed this issue.
If we only have global production prices, we can at least use them for individual equations.
10.3.12.4 The demand prices
The above deflators allow us to compute value added and trade at current prices. As we know already the values at
constant prices, the demand deflators can be defined by identities.
473
However, one must check that, at current prices, the data still verifies the equilibrium between demand and supply. A
condition difficult to enforce in practice, and even more in simulations, in which convergence can be achieved at the
global level while individual elements will diverge. Here the error-correction format can help.
A small error can be accepted, and either
• Forgotten: this is dangerous as the model will never get a perfect historical fit, once estimated equations
have been fed their residuals.
• Treated as a (multiplicative) residual. We have seen that it is a much better option.
10.3.12.5 The government budget
As in the simple model, we have to define a detailed government budget. The elements associated with goods and
services (demand and taxes) will have to be separated. Others will not (transfers from and to households in
particular).
10.3.12.6 The EViews program
We can now build a decomposed model, based on the same framework as our single product version, and using the
same concepts.
We have decided (and it should be obvious) that the problem was too complex for a full model to be presented here.
We shall only give a table presenting the decomposition level applied to these concepts. It can add no, one or two
dimensions to the single product case.
10.3.12.1 A list of elements
We present here a list of elements, coming actually from a 3-product operational model for Vietnam. It applies the
artisanal – industrial decomposition of agriculture we presented earlier.
In the “categories” column,
T stands for total,
1, 2 and 3 for the associated products (or branches, or sectors),
1a and 1i for the decomposition of agriculture into artisanal and industrial.
2e and 2m for the decomposition of product 2 into energy and the rest (mostly associated to manufacturing).
474
Name Type Description Units Type categories
CAP Endog Productive capacity Constant prices Product 1,2,3
CI Endog Change in inventories Constant prices Product T, 1,1a,1i,2,3
COG Exog Consumption of Government Constant prices Product 1,2,3
COGV Endog Consumption of Government Current prices Product T
COH Endog Household consumption Constant prices Product T,1,2,3
COMPM Endog Imports price competitiveness (import / local production price) Product 1,2i,3
COMPX Endog Exports price competitiveness (export/ foreign production price) Product 1,2e,2i,3
COST Endog Cost of wages and capital Constant prices Branch 2
CPI Endog Consumption deflator Deflator base year=2014 Product T
CRV Endog Export - import ratio at current prices Ratio Product T
CRVOL Endog Export - import ratio at constant prices Ratio Product T
ER Endog Exchange rate (value of the local currency) Deflator base year=2014 T
EXPG Endog Expenditures of Government Current prices T
FBAL Endog Financing Capacity of Firms Current prices Product 1i,2,3
FD Endog Final local demand Constant prices Product T,1,2,3
FDGV Endog Final demand of Government at current prices Current prices Total T
FDI Endog Foreign direct investment at constant prices Current prices Product 1,2,3
FDV Endog Final local demand Current prices Product T,1,2,3
FOBAL Endog National commercial balance Current prices Product T
475
GBAL Endog Financing Capacity of Government Current prices Product T
GDPM Endog Market Gross Domestic Product Constant prices Product T,1,2,3
GDPMV Endog Market Gross Domestic Product at current prices Current prices Product T,1,2,3
GDPV Endog Total Gross Domestic Product at current prices Current prices Product T
HDI Endog Households disposable income Current prices Product T
HI Endog Household revenue Current prices Product T
HRDI Endog Household disposable income in purchasing power Constant prices Product T
IC Endog Intermediate consumption at constant prices Constant prices Product x Branch (1,2,3)*(1,2,3)
ICV Endog Intermediate consumption at current prices Current prices Product x Branch (1,2,3)*(1,2,3)
IG Exog Government investment at constant prices Constant prices Product 1,2,3
IGV Endog Government investment at current prices Current prices Total T,1,2,3
IHH Endog Housing households investment Constant prices Total T,2,3
INCT Endog Income tax Current prices Total T
(1,2,3)*(T,1a,1i,1,2,3),T
IP Endog Productive investment Constant prices Total
*T
IPD Endog Target productive investment Constant prices Product by Branch 2*2
IR Endog Current interest rate on new loans Points T
IRL Endog Long-term interest rate on new loans Points T
IRM Endog Average interest rate on current debts Points T
IRS Endog Short-term interest rate on new loans Points T
K Endog Productive capital Constant prices Product x Branch (1,2,3)*(T,1a,1i,1,2,3),T
476
*T
KD Endog Target productive capital Constant prices Product x Branch 2*2
KV Endog Productive capital at current prices Current prices Branch 1a,1i,1,2,3
L Endog Total employment Thsds persons Branch T,1a,1i,1,2,3
L_G Exog Employment Thsds persons
L_M Endog Employment Thsds persons
LD Endog Target Firms employment Thousand persons Branch 1a,1i,2,3
LF Endog Firms employment Thousand persons Branch 1a,1i,2,3
LH Exog Household productive employment Thousand persons Branch 1a,1i,2,3
LP Endog Labour productivity Thousand persons Branch T,1,2,3
LPTA Exog Labour productivity trend Thousand persons Branch 1a,1i,3
LS Endog Wage earners Thousand persons Total 1,1a,1i,2,3
LS_F Endog Wage earners firms Thousand persons Total T,1a,1i,2,3
LS_H Endog Wage earners households Thousand persons Total T,1a,1i,2,3
M Endog Imports Constant prices Total T,1,2e,2i,3
MARG Endog Margins of firms Current prices Branch 1,1a,1i,2,3
MV Endog Imports at current prices Current prices Total T,1,2e,2i,3
NIF Endog Net interests paid by Firms Current prices Branch 1i,2,3
NIG Endog Net interests paid by Government Current prices
NIX Endog Net interests paid to the Rest of the World Current prices
477
OIT Endog Other indirect taxes Current prices Product T,1,1a,1i,1,2,3
PCOG Endog Government consumption deflator Deflator base year=2014 Product 1,2,3
PENS Endog Pensions Current prices
PENSR Exog Pensions rate Constant prices
PEX Endog Export deflator Deflator base year=2014 Product T,1,2,2e,2i,3
PEXT Endog Export deflator including tariffs Deflator base year=2014, USD Product 1,2e,2i,3
PFD Endog Local demand deflator Deflator base year=2014 Product T,1,2,3
PFDG Endog Government demand deflator Deflator base year=2014 Product T
PFDXT Endog Local demand deflator excluding VAT Deflator base year=2014 Product 1,2,3
PGDPM Endog Market GDP deflator Deflator base year=2014 Product T
PIC Endog Price of intermediate consumption Deflator base year=2014 Product x Branch (1,2,3)*(1,2,3)
PIG Endog Government investment deflator Deflator base year=2014 Product 2,3
PIHH Endog Household housing investment deflator Deflator base year=2014 Product T
PIM Endog Import deflator Deflator base year=2014 Product T,1,2e,2i,3
PIMT Endog Import deflator including tariffs Deflator base year=2014 Product 1,2e,2i,3
PIMU Exog Energy import price in US Dollars Deflator base year=2014 Product 2e
PIP Endog Productive investment deflator Deflator base year=2014 Product T,1,2,3
POP Exog Population Thousand persons
POPAG Exog Population in age of working Thousand persons
PP Endog Production deflator Deflator base year=2014 Product 1,2,3
478
PQ Endog Value added deflator Deflator base year=2014 Product 1,2,3
PROF Endog Profits of Firms Current prices Branch 1i,2,3
PT Endog Tax on profits Current prices Branch T,1i,2,3
PWM Exog World price imports weighting Deflator base year=2014, USD Product 2i,3
PWX Endog World price exports weighting Deflator base year=2014 Product T,1,2e,2i,3
Q Endog Value added Constant prices Product T,1,1a,1i,2,3
QA Endog Target value added Constant prices Product 2
QV Endog Value added at current prices Current prices Product T,1,1a,1i,2,3
R_COGV Exog Ratio of non-wage Govt consumption to GDP
R_CPI Exog Ratio of CPI to Final demand deflator
R_EXPG Exog Ratio of non-identified Administrations demand to GDP
R_IHH Exog Ratio of Housing investment to household revenue Branch 2,3
R_INCT Exog Income tax rate
R_IP Exog Ratio of productive investment to GDP Product x Branch 3*(1,1a,1i,2,3)
R_LFA Exog Share of Firms employment in total Branch 1a,1i,2,3
R_OIT Endog Rate of Other Indirect taxes (than VAT) Branch 1,1a,1i,2,3
R_PCOG Exog Ratio of Govt consumption to final demand deflator Branch 1,2,3
R_PFDG Exog Ratio of Govt demand to final demand deflator
R_PIC Exog Ratio of Interm. consumption to Demand (excluding taxes) deflator Product x Branch (1,2,3)*(1,2,3)
R_PIG Exog Ratio of Govt investment to Final Demand deflator Product 2,3
479
R_PIHH Exog Ratio of Housing investment to final demand deflator
R_PIP Exog Ratio of Productive investment to final demand deflator Product 1,2,3
R_PTI Exog Rate of the tax on profits Branch 1i,2,3
R_REVG Exog Ratio of other Government revenue to GDP
R_REVQ Exog Ratio to production of other Household revenue from production Branch 1i,2,3
R_REVX Exog Non wage non production Household revenue, in purchasing power
R_SCF Endog Rate of social contributions by Firms Branch 1,1a,1i,2,3
R_SCF_G Exog Rate of social contributions by Government
R_SCF_H Exog Rate of social contributions by Households
R_SCW_G Exog Rate of social contributions by wage earners Government
R_SCWA Exog Rate of social contributions by wage earners Branch 1a,1i,2,3
R_SUBS Exog Rate of subsidies (to value added at current prices) Branch 1,1a,1i,2,3
R_TARIFF Exog Rate of Local tariffs Branch 1,2,2e,2i,3
R_TARX Exog Rate of Foreign tariffs Branch 1,2e,2i,3
R_UN Endog Unemployment rate
R_VAT Exog VAT rate Product 1,2,3
R_WG Exog Ratio of Government to market wage rate
R1_A Exog Share of artisanal in Agriculture
R1_I Exog Share of industry in Agriculture
RDEP Exog Depreciation rate of capital Product x Branch (1,2,3)*(1a,1i,2,3)
480
RELC Endog Labour - Capital relative cost Branch 2
REVG Endog Revenue of Government Eq364
REVQ Endog Revenue of Households from production, other than wages. Branch T,1i,2,3
REVX Endog Non wage non production Household revenue Constant prices
RFDI Exog Share of FDI in investment Product 1,2,3
RMARG Endog Margins ratio Branch 1,1a,1i,2,3
RPROB Endog Margins ratio including capital amortization Branch 1i,2,3
RPRO Endog Profits ratio Branch 1i,2,3
RS_F Exog Share of wage earners in Firms Branch 1a,1i,2,3
RS_H Exog Share of wage earners in Households Branch 1a,1i,2,3
SCF Endog Social contributions by Firms Current prices Branch T,1,1a,1i,2,3
SCF_G Endog Social contributions by Government Current prices
SCF_H Endog Social contributions by Households Current prices
SCW Endog Social contributions by Wage earners Current prices
SOCB Endog Social benefits Current prices
SOCBR Exog Social benefits in purchasing power Constant prices
SUBS Endog Subsidies to firms Current prices Branch T,1,1a,1i,2,3
TARIFF Endog Local tariffs Current prices Product T,1,2,2e,2i,3
TC Exog Technical coefficient Product x Branch (1,2,3)*(1,2,3)
TRBAL Endog Trade balance Current prices Product T
481
TTRAD Endog Terms of trade (export - import price ratio) Thousand persons
UN Endog Unemployment Thousand persons
UN_R Exog Unemployment Thousand persons
UN_U Endog Unemployment
UR Endog Rate of use of capacities Branch 1,2,3
VAT Endog Value Added Tax Current prices Total T,1,2,3
W Endog Market wage rate At constant prices Branch 1,1a,1i,2,3
W_G Endog Government wage rate Thousands at constant prices
WAGE Endog Total wages Current prices Branch T,1,1a,1i,2,3
WAGE_F Endog Wages paid by Firms Current prices Total T,2,3
WAGE_G Endog Wages paid by Government Current prices T
WAGE_H Endog Wages paid by Households Current prices Total T,2,3
WCOST Endog Unitary wage cost Thousands at constant prices Branch 1,2,3
WD Exog World demand at constant prices Index at constant prices Product 1,2e,2i,3
WF Endog Wage rate paid by Firms Thousands at current prices Branch 2,3
WH Endog Wage rate paid by Households Thousands at constant prices Branch 2,3
X Endog Exports Constant prices Total T,1,2,2e,2i,3
XV Endog Exports at current prices Current prices Total T,1,2,2e,2i,3
482
483
10.4 A MULTI COUNTRY, SINGLE PRODUCT MODEL
We shall now address the issue of building, maintaining and using a model describing the economy of more
than one country. We shall also consider the similar issue of a single country but multi-region model.
10.4.1 FIRST ISSUE: THE MODELS
To produce a multi – country model, we first need single country ones. Two issues have to be considered:
• The degree of similitude of the country models
• The description of the Rest of the World
10.4.1.1 The single country models
We shall suppose the country models follow the lines used above. What matters now is the differences allowed
in the specifications, from one country to the other. Several cases can be considered:
A - The models are allowed to be completely different. This even applies the concepts used, such as the
product decomposition. The only condition is that, once identified the elements linking the models (basically
the trade variables at constant and current prices) they can be converted in such a way that they can be
transferred as input to the other models. But for instance, a single product and multi product models can
coexist. The transfer process will apply aggregation or disaggregation as required. In this case the names of the
variables will probably be different too.
B - The models use the same concepts but the economic framework is different. For instance, a model can
use a Cobb-Douglas production function, the other a CES.
C - The models use the same equations, but the coefficients are different. Some exceptions can appear, for
instance an additional lagged variable, but the logic remains the same. In general, these differences should be
due to estimation, which means that a common data set must be available.
D - The models use the same equations and the same coefficients. This means that either the formulas have
been estimated as a panel (by polling all the countries data together) or calibrated.
The choice of option depends on the case:
• Option A can be considered if the models already exist, which simplifies the process. The case is even
stronger if the models have to keep on living as independent versions.
This is the case for instance for the LINK project, which links into a multi-country version, under the
management of the United Nations’ Department of Economic and Social Affairs, and the University of Toronto,
models maintained as independent forecasting tools by country institutes.
Another example is the grafting of existing a national model in an international environment, also existing. For
example, the MODUX model managed by the Statistical Institute of Luxemburg (STATEC) has been integrated
into the MacSim environment (see below) in order to enrich its projections and especially its impact analyzes.
Such a possibility is provided by the managers of MacSim (which basically means the author of this book).
484
The drawback is of course that the models can follow different economic properties, and the differences in
mechanisms, leading to different sensitivities to shocks, are not necessarily justified by reliable country
specifics (if they are, then option A is clearly optimal). As stated earlier, econometrics based on the same data
can justify formulations with quite different properties, even by pure chance (two economists with the same
economic philosophy can reach different formulations through the meanders of successive estimations).
• Option B is option A with no initial models. Then the models will be easier to manage, but the above
problems remain.
• Option C is even simpler to manage and its properties easier to interpret. The only drawback is the
constraints on model formulations, which forbids introducing country-specific formulations. But of
course, some minimal differences can be accepted, such as a longer lagged structure, the long run
indexation of wages on different deflators.
• Option D limits even more country specifics. What only remains is the effect of the size of the country,
the structure of its partners (imports and exports) and structural parameters like labor and capital
productivity, the weight of the public sector or the tax structure, or the sharing of production revenue
between the firms and the workers.
The only justification is just this: that the only difference in properties comes from undisputable elements, the
consequences of which they describe faithfully.
From the above elements we can derive the following suggestion: if the models do not preexist, option C is
clearly preferable, with some elements from option B if they are really justified, in particular by the observation
of country characteristics.
For instance, for an oil producing country the trade in energy elements could be individualized. For the interest
rate, identifying the policy of the central bank should lead to the introduction of the associated rule. And
maybe if there is sufficient proof that factor substitution follows a zero or unitary elasticity, the associated
production function (complementary factors or Cobb-Douglas) should be used.
10.4.2 SECOND ISSUE: INTER-COUNTRY TRADE
Once the single country models have been defined, they have to be merged into a single entity.
The main issue is to make consistent the imports and exports of each country, at constant and current prices.
The obvious option is to associate behaviors with the actual decision processes:
• Global imports are decided by the importing countries.
• Export prices are decided by the exporters, and import prices are computed by weighting identities,
taking into account the relative exchange rate. The decision can depend on the targeted market.
Once imports of a country are decided, they must be shared between the countries which export to it. There
are two options:
• Identifying the individual imports by their source and converting them into individual exports from the
supplying country. This is obviously the most logical solution. It has the essential advantage of making
exports and imports consistent at the global level.
• However, some models use a more devious technique: they compute global potential exports of a
country by weighting by constant shares the imports of its partners.
Of course, this option gives to total exports and exports different values, and some correction must be
applied. The main advantage of this method is to allow estimation of exports based on actual global values.
485
Also, it avoids identifying a large number of trade flows (growing as the square of the number of
countries).
We favor strongly the first option.
• It identifies more variables, but the associated information is important. The interpretation of the
consequences of shocks (on domestic or trade related variables) is much clearer and more
informative.
• It does not call for corrections, giving automatically consistent results.
• It does not require more information: the weights used by the other method can be used to compute
the trade flows.
• But the most important in our sense is that it locates the decision process where it belongs: exports
are not decided by exporters, but by importing countries which, once they have decided to import,
chose among the potential suppliers, essentially by considering relative prices.
Of course, it does not allow global estimation. The individual trade flow equations will have to be either
calibrated (perhaps using values from global estimations) or estimated using panel techniques.
From now on, we shall concentrate on this technique.
10.4.3 A CONSISTENT METHOD: MACSIM
We shall present now a consistent method for defining in sequence the above elements. It is used by the
MacSim system (2001, 2012).
We shall start, not with imports but with the price system, as prices depend only on local elements, while
imports depend on price competitiveness.
10.4.3.1 The price system
We start by establishing a coherent system for trade prices. For this we shall start from the export price
formulations (estimated or set). For clarity we shall use calibrated equations, but the method applies to
estimated ones also.
We shall suppose that the traditional equation applies to each client country. But this time the exports
destination will be identified.
We shall consider I as the exporting country, j as the destination.
To get individual export prices, we correct by the exchange rate:
𝐿𝑜𝑔(𝑝𝑒𝑥𝑖,𝑗 ) = 𝑎 ⋅ 𝐿𝑜𝑔(𝑝𝑝𝑖 ) + (1 − 𝑎) ⋅ (𝐿𝑜𝑔(𝑝𝑝𝑗 𝑐ℎ𝑖 /𝑐ℎ𝑗 ) + 𝑒 ⋅ 𝑡 + 𝑓
Where i is the exporter and j the client, and chi is an index, representing the evolution (from the base year) of
the value of a common currency (in practice the US dollar) compared to the currency of country i. If the
currency of i depreciates, the index grows.
486
𝑝𝑖𝑚𝑗,𝑖 = 𝑝𝑒𝑥𝑖,𝑗 . 𝑐ℎ𝑗 /𝑐ℎ𝑖
and we get the global import price through an average:
pim j = m j ,i pim j ,i m j ,i
i i
(We shall see later how we compute the trade flows at constant prices).
10.4.3.2 The foreign rates of use
In the single country models, we have seen the fundamental role of the rate of use of capacities for the local
country. But we supposed the foreign capacities to be infinite.
Now we can consider, to define the imports of a country, the rate of use for providing countries. If the rate of
use increases in Germany, it will lose market shares in France, relative to other exporters but also to French
manufacturers.
To get the average, we shall use the same method as above (without currency corrections).
𝑢𝑟𝑥𝑗 = ∑𝑖 𝑚𝑗,𝑖 𝑢𝑟𝑖 ⁄∑𝑖 𝑚𝑗,𝑖
10.4.3.3 The global imports
We can now determine the global imports of country i, by modifying slightly the equation from our single
country model, to consider the capacity of exporters:
𝐿𝑜𝑔(𝑚𝑖 /𝑡𝑑𝑖 ) = 𝑏 ⋅ [𝐿𝑜𝑔(𝑢𝑟𝑖 ) − 𝑐 ⋅ 𝐿𝑜𝑔(𝑢𝑟𝑥𝑖 )] + 𝑑 ⋅ 𝐿𝑜𝑔(𝑝𝑚𝑖 /𝑝𝑝𝑖 ) + 𝑒
To simplify the explanation, we have presented only the long-term part. The dynamic equation should contain
the same change.
This means that, as in the single country models, a general decrease in the available capacity of exporters will
reduce their exports, through a substitution effect. The coefficient c, lower than one, takes into account the
larger size of this set of countries.
487
10.4.3.4 The trade flows
Finally, we have to separate imports into individual exports. Once again, we shall take into account relative
competitiveness, and fluctuations in available capacities, relative to the above average. Actually, rather than
estimating an equation (we do not have the associated data) we shall use:
mi , j = mi .bi , j 1 − i ( pex j chi / ch j − pimi ) − i (ut j − utxi )
which means that (as for the single country models) exporters to one country will increase a « natural » share
with competitiveness and available capacity, this time relative to their competitors. More precisely, the A and B
matrixes (one derived from the other) represent the sharing of bilateral trade associated with identical rates of
use and competitiveness for all countries (not necessarily the most natural assumption, nor the present
situation). Scalars and are fixed, and will be set at a higher value than for the import equation (between 0.5
and 1.0). This should represent the higher versatility of the choice of its providers by a country, once the
decision has been made not to buy local products.
One will observe that this technique guarantees the identity of the sum of individual exports with its global
value, without any correction. The weighted sum of deviations from averages, computed using the same
weightings, is of course zero.
Of course, the coefficients can be different from one market to another, but not within one market.
The system can be summarized with this graph:
The trade block
Rate of Imports of
Production Demand
Price i
use country i
i
i
Import Rate of
Price Imports Capacity exporters
price
compet i i compet i to i
i
Export Price Cap
Exports Rate of
price of j compet of j compet of j
of j use
market i market i on market i
to i j
Produc Total
tion price Exports
Exports of
j of j country j
In addition to the above, we can now introduce accounting equations:
488
Exports from country i to j:
𝑥𝑖,𝑗 = 𝑚𝑗,𝑖 ⋅ 𝑐ℎ0𝑗 /𝑐ℎ0𝑖
Where usd0i represents the base year value of the currency of country i, in US Dollars.
Total exports are computed as a sum
xi = x i, j
The average export price of country i:
𝑝𝑒𝑥𝑖 = ∑ 𝑝𝑒𝑥𝑖,𝑗 ⋅ 𝑥𝑖,𝑗 /𝑥𝑖
10.4.3.5 Third issue: common behaviors
But we have also to consider behaviors common to a set of countries, which take into account global variables.
The most obvious case is the presence of a global formal agreement, such as the use of a common currency by
the European Monetary Union.
This option is very easy to implement: one just has to create a new “country” using aggregated concepts,
considering only the relevant variables.
The identification of the countries in the union can be done by equations considering all countries, and
applying a Boolean variable (1 = yes, 0 = no) to each concept.
10.4.3.6 The Rest of the World
If the scope of the countries described is large enough, we cannot expect that shocks applied to a large subset
or the whole model will have no feedback through the economy of the remaining countries. This is true in our
MacSim system, which models most of Europe (the Euro zone + the United Kingdom) and also the US and
Japan. Moreover, with the technique we are proposing (computing the shares of each country in each
country’s imports) endogenizing the Rest of the World allows its exports to conform to the global framework,
avoiding any unbalance.
But of course, we are not going to estimate a ROW model: we do not have the data, and the heterogeneity is
too high.
489
What we propose is:
• To define the trade elements as usual: imports are shared according to the above framework, and
exports participate to the above competition.
• In addition, to create a set of multipliers, defining the feedback of the Rest of the World exports on its
GDP and imports, and linking its imports to its final demand.
It should be quantified to give a response similar to the single country models, with the exception of infinite (or
quasi-infinite) capacity and high sluggishness of prices.
This can be (and actually has been) done quite simply.
10.4.3.7 Estimating with common values, using the SYSTEM feature
If we use option C above, we do not have to give to all coefficients the same value. Of course, we have to
estimate the constant term, if we want the mean residual to be zero for each country (and EViews not to refuse
a system without any coefficient to estimate).
We should probably estimate also the possible trends, which should be country specific.
Let us show an example of application, based on our MacSim model. We will consider the consumption
equation, using the same explanations as in the single country model. As our purpose is not studying the
economic implications, our comments will be minimal.
We want the change in consumption to depend on:
• The change in real income (over the last year).
• The change in unemployment.
• Inflation (the « real holdings » or Pigou effect).
• An error correction term.
• A supposedly negative time trend, representing the increase in the wealth of households and their
saving potential (in particular their accession to housing ownership).
Our system will consider 14 countries: Belgium, Germany, France, Great Britain, Italy, Japan, Luxemburg;
Netherlands, USA, Spain, Finland, Greece, Ireland, Portugal (the MacSim system aggregates the last four, but
we can separate the estimations). We will give to each country the same weight, without considering their size.
Using different coefficients for each equation, we will define the following system:
for %1 BE DE FR GB IT JP LX NL US ES FI GC IR PT
genr {%1}_coh_ec=0
delete c_{%1}_coh
coef(10) c_{%1}_coh
__coh.append dlog({%1}_coh)=c_{%1}_coh(1)*.25*log({%1}_hrdi/{%1}_hrdi(-4))
+c_{%1}_coh(2)*d({%1}_unr)+c_{%1_coh(3)*log({%1}_pc/{%1}_pc(-4))
+c_(%1}_coh(4)*log({%1}_coh(-1)/{%1}_hrdi(-1))+c_{%1}_coh(5)
+c_{%1}_coh(6)*({%1}_t-2008)*({%1}_t<=2008)+{%1}_coh_ec
next
490
_coh.sur(p)
And get the following results: a table (some programming is involved).
Dymamic Unempl Inflation Error Cor. Constant Trend See/DW
Belgium 0.243 -0.517 -0.049 -0.018 -0.001 -0.000 0.004
3.491 -3.776 -2.349 -1.609 -0.904 -4.022 0.808
Germany 1.479 0.158 -0.315 -0.347 -0.027 0.000 0.007
4.576 0.269 -2.496 -2.400 -2.449 0.989 2.336
France 0.550 -0.188 -0.006 -0.039 -0.002 -0.000 0.006
3.383 -0.613 -0.159 -1.067 -0.460 -0.306 2.554
Great Brtain 0.299 -0.595 -0.092 -0.047 0.002 -0.000 0.010
2.070 -1.531 -3.006 -1.430 0.823 -2.309 2.286
Italy 0.043 -0.398 -0.021 -0.019 -0.002 -0.000 0.007
0.495 -2.037 -1.410 -2.756 -0.961 -2.759 1.464
Japan 0.391 -1.056 -0.107 -0.081 -0.004 -0.000 0.011
2.689 -1.351 -3.162 -2.104 -1.401 -0.143 2.580
Luxemburg 0.781 0.073 -0.258 -0.118 -0.038 -0.000 0.008
3.250 0.067 -2.529 -1.768 -1.502 -1.143 1.555
Netherlands 0.213 -0.826 0.064 -0.102 0.001 0.000 0.010
1.247 -1.924 1.422 -3.105 0.435 0.901 2.392
USA 0.185 -0.507 -0.124 -0.035 0.005 -0.000 0.006
1.523 -3.040 -3.917 -1.188 2.435 -1.294 1.951
Spain 0.220 -0.463 -0.018 0.006 0.008 0.000 0.006
1.770 -2.949 -0.718 0.427 2.889 0.108 1.725
Finland 0.238 -0.573 -0.033 -0.024 0.004 -0.000 0.011
2.274 -2.984 -0.996 -0.708 1.736 -1.208 2.172
Greece 0.274 0.252 0.058 -0.007 0.011 0.000 0.019
1.415 0.617 0.731 -1.094 1.123 0.603 2.399
Ireland 0.794 0.912 -0.032 -0.022 0.001 -0.000 0.010
4.256 4.061 -0.683 -1.810 0.110 -0.957 2.039
Portugal 0.350 0.284 -0.471 0.000 0.010 -0.001 0.007
1.243 0.947 -2.449 0.008 1.170 -2.687 2.797
The estimations:
491
Equation: DLOG(BE_COH)=C_BE_COH(1)*.25*LOG(BE_HRDI/BE_HRDI(
-4))+C_BE_COH(2)*D(BE_UNR)+C_BE_COH(3)*LOG(BE_PC
/BE_PC(-4))+C_BE_COH(4)*LOG(BE_COH(-1)/BE_HRDI(-1))
+C_BE_COH(5)+C_BE_COH(6)*(BE_T-2008)*(BE_T<=2008)
+BE_COH_EC
Observations: 136
R-squared 0.432945 Mean dependent var 0.005667
Adjusted R-squared 0.411135 S.D. dependent var 0.005789
S.E. of regression 0.004443 Sum squared resid 0.002566
Durbin-Watson stat 0.831542
Equation: DLOG(DE_COH)=C_DE_COH(1)*.25*LOG(DE_HRDI
/DE_HRDI(-4))+C_DE_COH(2)*D(DE_UNR)+C_DE_COH(3)
*LOG(DE_PC/DE_PC(-4))+C_DE_COH(4)*LOG(DE_COH(-1)
/DE_HRDI(-1))+C_DE_COH(5)+C_DE_COH(6)*(DE_T-2008)
*(DE_T<=2008)+DE_COH_EC
Observations: 52
R-squared 0.382124 Mean dependent var 0.003249
Adjusted R-squared 0.314963 S.D. dependent var 0.008780
S.E. of regression 0.007267 Sum squared resid 0.002429
Durbin-Watson stat 2.389686
Equation: DLOG(FR_COH)=C_FR_COH(1)*.25*LOG(FR_HRDI
/FR_HRDI(-4))+C_FR_COH(2)*D(FR_UNR)+C_FR_COH(3)
*LOG(FR_PC/FR_PC(-4))+C_FR_COH(4)*LOG(FR_COH(-1)
/FR_HRDI(-1))+C_FR_COH(5)+C_FR_COH(6)*(FR_T-2008)
*(FR_T<=2008)+FR_COH_EC
Observations: 104
R-squared 0.138231 Mean dependent var 0.004755
Adjusted R-squared 0.094264 S.D. dependent var 0.005932
S.E. of regression 0.005646 Sum squared resid 0.003123
Durbin-Watson stat 2.559371
and the residuals (we can see that for Greece the data is not seasonally adjusted):
492
DLOG(BE_COH) Residuals DLOG(DE_COH) Residuals DLOG(FR_COH) Residuals DLOG(GB_COH) Residuals
.015 .02 .02 .06
.010 .01 .04
.01
.005 .02
.00
.000 .00 .00
-.01
-.005 -.02
-.01
-.010 -.02 -.04
-.015 -.03 -.02 -.06
1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000
DLOG(IT_COH) Residuals DLOG(JP_COH) Residuals DLOG(LX_COH) Residuals DLOG(NL_COH) Residuals
.03 .04 .03 .03
.02 .02 .02 .02
.01 .00 .01 .01
.00 -.02 .00 .00
-.01 -.04 -.01 -.01
-.02 -.06 -.02 -.02
-.03 -.08 -.03 -.03
1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000
DLOG(US_COH) Residuals DLOG(ES_COH) Residuals DLOG(FI_COH) Residuals DLOG(GC_COH) Residuals
.02 .02 .06 .08
.04 .06
.01 .01
.02 .04
.00 .00 .00 .02
-.02 .00
-.01 -.01
-.04 -.02
-.02 -.02 -.06 -.04
1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000
DLOG(IR_COH) Residuals DLOG(PT_COH) Residuals
.06 .015
.04 .010
.005
.02
.000
.00
-.005
-.02 -.010
-.04 -.015
1975 1980 1985 1990 1995 2000 1975 1980 1985 1990 1995 2000
We can observe that most coefficients are significant with the right sign.
Now we can decide to force the dynamic coefficient to take the same value across countries; we just have to
modify the equation, replacing the country code by “to” for this coefficient:
coef(10) c_to_coh
_coh.append dlog({%p}_coh)=c_to_coh(1)*.25*log({%p}_hrdi/{%p}_hrdi(-4))+c_{%p}_coh(2)*d({%p}_unr)
+c_{%p}_coh(3)*log({%p}_pc/{%p}_pc(-4))+c_{%p}_coh(4)*log({%p}_coh(-1)/{%p}_hrdi(-1))+c_{%p}_coh(5)
+c_{%p}_coh(6)*({%p}_t-2008)*({%p}_t<=2008)+{%p}_coh_ec
We get:
493
Dymamic Unempl Inflation Error Cor. Constant Trend See/DW
Belgium 0.274 -0.523 -0.044 -0.018 -0.001 -0.000 0.004
7.472 -3.794 -2.238 -1.574 -0.879 -4.074 0.816
Germany 0.274 -0.170 -0.360 -0.338 -0.025 0.000 0.008
7.472 -0.263 -2.585 -2.106 -2.070 0.159 2.386
France 0.274 -0.390 -0.003 -0.036 -0.000 -0.000 0.006
7.472 -1.359 -0.089 -0.983 -0.001 -0.015 2.549
Great Brtain 0.274 -0.636 -0.094 -0.047 0.002 -0.000 0.010
7.472 -1.680 -3.217 -1.453 0.890 -2.480 2.292
Italy 0.274 -0.322 -0.018 -0.017 -0.001 -0.000 0.007
7.472 -1.614 -1.223 -2.511 -0.313 -1.625 1.438
Japan 0.274 -1.173 -0.108 -0.076 -0.004 -0.000 0.011
7.472 -1.546 -3.201 -1.966 -1.529 -0.509 2.593
Luxemburg 0.274 -0.895 -0.264 -0.130 -0.037 -0.000 0.009
7.472 -0.881 -2.527 -1.886 -1.405 -1.094 1.550
Netherlands 0.274 -0.723 0.059 -0.095 0.001 0.000 0.010
7.472 -2.162 1.357 -3.110 0.472 0.899 2.401
USA 0.274 -0.472 -0.115 -0.035 0.005 -0.000 0.006
7.472 -2.968 -3.858 -1.158 2.341 -1.159 1.946
Spain 0.274 -0.428 -0.012 0.007 0.007 0.000 0.006
7.472 -3.085 -0.502 0.493 3.004 0.276 1.725
Finland 0.274 -0.567 -0.031 -0.023 0.004 -0.000 0.011
7.472 -2.998 -0.944 -0.701 1.686 -1.150 2.178
Greece 0.274 0.238 0.058 -0.007 0.011 0.001 0.019
7.472 0.588 1.439 -1.172 1.178 0.652 2.399
Ireland 0.274 0.548 -0.127 -0.038 0.012 0.000 0.010
7.472 2.784 -3.674 -3.358 4.739 0.989 1.973
Portugal 0.274 0.218 -0.465 0.008 0.011 -0.001 0.007
7.472 0.774 -2.412 0.157 1.452 -2.744 2.808
with an average value for this particular coefficient, and similar results for the others.
Finally, we can apply the constraint to all four economic elements:
494
Dymamic Unempl Inflation Error Cor. Constant Trend See/DW
Belgium 0.268 -0.379 -0.031 -0.011 -0.000 -0.000 0.005
7.758 -5.964 -3.539 -3.403 -0.462 -4.997 0.767
Germany 0.268 -0.379 -0.031 -0.011 -0.002 -0.000 0.009
7.758 -5.964 -3.539 -3.403 -0.523 -1.335 2.581
France 0.268 -0.379 -0.031 -0.011 0.002 -0.000 0.006
7.758 -5.964 -3.539 -3.403 1.243 -1.671 2.556
Great Brtain 0.268 -0.379 -0.031 -0.011 0.005 -0.000 0.010
7.758 -5.964 -3.539 -3.403 2.623 -0.778 2.287
Italy 0.268 -0.379 -0.031 -0.011 0.000 -0.000 0.007
7.758 -5.964 -3.539 -3.403 0.260 -2.986 1.458
Japan 0.268 -0.379 -0.031 -0.011 -0.001 -0.000 0.011
7.758 -5.964 -3.539 -3.403 -0.245 -2.826 2.619
Luxemburg 0.268 -0.379 -0.031 -0.011 0.007 0.000 0.009
7.758 -5.964 -3.539 -3.403 1.987 1.267 1.635
Netherlands 0.268 -0.379 -0.031 -0.011 0.003 -0.000 0.010
7.758 -5.964 -3.539 -3.403 1.637 -1.044 2.433
USA 0.268 -0.379 -0.031 -0.011 0.006 -0.000 0.006
7.758 -5.964 -3.539 -3.403 4.956 -0.535 1.831
Spain 0.268 -0.379 -0.031 -0.011 0.005 -0.000 0.006
7.758 -5.964 -3.539 -3.403 3.441 -0.577 1.671
Finland 0.268 -0.379 -0.031 -0.011 0.004 -0.000 0.011
7.758 -5.964 -3.539 -3.403 1.758 -1.686 2.121
Greece 0.268 -0.379 -0.031 -0.011 0.021 0.001 0.020
7.758 -5.964 -3.539 -3.403 3.398 1.530 2.218
Ireland 0.268 -0.379 -0.031 -0.011 0.008 0.000 0.011
7.758 -5.964 -3.539 -3.403 3.155 0.172 1.990
Portugal 0.268 -0.379 -0.031 -0.011 -0.001 -0.001 0.007
7.758 -5.964 -3.539 -3.403 -0.364 -2.128 2.642
Now all economic coefficients are considered significant, profiting from the number of observations attributed
to them.
The associated programs, including table production, are available from the author.
10.4.3.8 The sequence of production
To produce a multi-country model, one should proceed through the following sequence:
10.4.3.8.1 Deciding on the model framework
We shall suppose that we build our model from scratch, which in this case should lead to a set of country
models using the same economic framework but estimated separately, with limited variations in the estimated
formulas.
In this case one should consider:
495
• The data available.
• The type of behaviors the model should describe.
• The expected degree of detail.
This should allow deciding:
• The variables used by the model (they have to be useful and available).
• The full form of the identities.
• The type of behaviors which should be estimated, including the explanatory elements.
10.4.3.8.2 Producing a set of country models
Then one should build a set of country models, using the technique described earlier. Of course, a global data
set has to be produced, containing series for each single model, with the same name except for an identifier.
For instance, French exports can be called FR_X.
As usual, concerning behavioral equations, one should start with declaration of intent before moving to
estimations.
Several sequences can be considered. One can build models one by one, or apply simultaneously each phase
(data production, framework, estimation, testing) to the whole set. Probably the best solution is to start with
one country, go through the whole process, and apply it to the set of remaining countries, this time phase by
phase simultaneously.
This allows measuring the difficulties, to solve the general problem on a single case, and defining at a limited
cost the full procedure, which will then be applied to the global case. In particular, if problems were to be met
at a later stage, this would probably call for reconsidering the initial ones. Having to modify the full set of
programs can then become quite costly.
The choice is not so simple, however, as:
• Running estimations for the same equations on a set of countries can be organized quite easily using
EViews loops.
But the estimation processes must succeed together, as the equations are not allowed to differ too much. An
interesting direction found for a given country, but not associated to its specific features, can be applied only if
it works also for other countries.
This means more iterations will be needed than for a single country model. The only issue is to limit their
number.
Two additional elements:
• After the estimation process has been applied to the whole set, it is possible that results proved
acceptable for a sizable share of countries (perhaps above 50%). Then if the remaining countries do
not present too specific characteristics (like having a different political system, or a different structure
of production), their formulas can be calibrated using the average of the successful results. This is
what we have done for the first version of the MacSim system.
• If it proves difficult to individualize a given equation, one can resort to panel estimation, estimating
the equation for all countries at the same time, and forcing some coefficients (at least one) to take the
same value for all countries. This technique has been presented earlier, including a practical example.
This is what we have done for the second version of the MacSim system.
496
10.4.3.8.3 Assembling the models
Once satisfying single country versions have been produced, assembling them is essentially technical.
10.4.3.8.3.1 COMPUTING THE TRADE FLOWS
If formalizing the trade between countries uses explicit individual flows (our preferred option) one needs a
matrix of transfers. It can represent either the share of exporters in each country’s imports, or the share of
clients in each country’s exports. Both matrixes are needed, but one can be built from the other in an exact
way. The matrix elements should be considered as series. If it is only available for some periods, it should be
extrapolated, but in any case, the modeler should be allowed to make it change with time in the future. This is
particularly essential for transition economies: the share of Russia in Polish exports has decreased following the
breakup of the Soviet Union, but it is not stabilized and could increase again in the future. Also, establishing a
trade agreement between countries will increase the share of their mutual trade.
From these matrixes the individual trade flows can be computed. If the matrix is known only at current (or
constant) prices, a global assumption has to be made, the simplest being that the same export (or import) price
applies to one country whatever its destination (or origin).
Of course, if we use official statistics to add up the result in one dimension, the other dimension will not give
the actual value. For instance, if we start from imports and share them between exporters, the sum of
individual exports of one country will show a difference with the national accounts. The issue is the size of this
difference: it should not be higher than the residual of the equation if it had been estimated, otherwise a
correction could be made (but this would bring back an inconsistency at the world level).
Of course, the model simulations and forecast results will be consistent, as the model system is consistent by
construction. And the consequences of shocks too, as the difference between two consistent systems. We are
using the error observed on the past as an indication of the global precision of the model on the future, not of
its internal consistency.
10.4.3.8.3.2 FORMALIZING THE TRADE FLOWS AT CONSTANT PRICES
We can do it using the framework defined earlier. We have only to consider the following technical issues:
• The export equations of single models will disappear.
• The trade flows have two dimensions and call for an additional loop.
• If the trade flows are not available as series, the coefficients will have to be calibrated. They must be
common for a given market, and we cannot expect the choice between import origins to use the same
price elasticity as the choice between imports and local products. We can suppose that the role of
relative prices us more important for deciding between producing locally and importing and choosing
between different providing countries.
Again, one advantage of our proposal is that trade flows will add up to imports without any correction, as the
weighted differences to the weighted mean cancel out naturally.
10.4.3.8.3.3 FORMALIZING THE TRADE DEFLATORS
It imports are decided by importers, the transaction price is decided by the exporter.
This time we can estimate the global export price of each country (depending on local costs and the price
applied by competitors). Then we can apply the results to each export market, using as competitors’ price the
one specific of this market.
497
Apart from this, we have to suppose that for each exporter the same formula applies to all its clients, including
coefficient values.
10.4.3.8.3.4 INTRODUCING IDENTITIES
Finally, we need to compute identities, like global country imports and exports at current prices.
10.4.3.8.3.5 PRODUCING COMMON ELEMENTS
The other issue is the presence of an agreement between countries, defining a variable common to each of
them, based on one or several common elements.
For instance, under purchasing power parity, the exchange rate of the Euro will apply to all Euro zone
countries, based on Euro zone inflation.
Equations will have to be built:
• Summarizing the explanatory indicators in the group.
• Computing the common element, generally giving a choice between several options (real interest rate
or Taylor rule, fixed real or nominal exchange rate).
The formula must allow changing the list of countries concerned. This is very easy:
• For computing the indicators, the element of each country in the model will be multiplied by an
indicator (1=in, 0=out) to indicate if it participates in the weighted sum.
• For the use of the common result, an option must be present in each country’s equation, using the
same indicator (if 1, use the common value, if 0, use a country specific formula).
Even in the absence of agreement, it can be useful to compute common indicators (for instance the GDP of the
full European Union, or of ASEAN).
10.4.3.8.3.6 THE REST OF THE WORLD
The Rest of the World “model” will need the usual trade flows equations, and in addition:
• An equation linking GDP and local demand to exports.
• A global imports equation linked to RoW demand and price competitiveness (its prices can be affected
by the other countries inflation). Introducing an influence of the rate of use should depend on the size
of the RoW.
10.4.3.8.3.7 SOLVING THE MODEL
The procedure for solving the model is the same. The number of equations is just larger, and the testing
procedure more complex.
Another problem is the convergence: with less exogenous variables the solution is less anchored on
assumptions, and the probability of divergence might increase. But one can also consider that an individual
problem can become diluted in the global equilibrating process. Nothing is clear on that issue.
The most important issue is perhaps the price system: with less anchoring, even if each country’s inflation
converges in the long run to a common value, this can happen at highly heterogeneous levels. If a country’s
price indexes are higher than the others, the inflation created normally by a demand shock can be offset by the
498
increase in the share of its lower priced imports. The overall effect can be deflationary, introducing properties
which should be considered abnormal, especially as they are due to the distance to the base year.
This can be solved by managing the system of residuals in such a way that the long-term prices remain rather
close to each other. The targeting technique we will present later can be used here164.
10.4.3.8.3.8 USING THE MODEL
Of course, if this model is more expensive to produce, it also allows a wider range of studies, and makes them
more reliable compared to the single country case. As we have said earlier, the demand multiplier for the first
MacSim version is 20% higher for France using the whole system than if we run the French sub-model by itself.
In addition to a better study of changes in the local assumptions of a single country, for this country and
abroad, this type of model allows in particular to study consequences of:
• Global events such as a slowing down of the world growth, or a change in the price of oil, taking into
account its consequences for all countries. We no longer need to make assumptions consistent for the
rest of the world. This adds feedback to the method we have presented earlier.
• Global trade agreements.
• Global policies (for instance a decrease in VAT rates in the European Union).
These elements can be assessed as a whole, and by comparing their consequences for each country.
All this can be done taking into account the role of common rules, such as the establishment of the Euro Zone,
which can be changed at will.
10.4.4 THE EVIEWS PROGRAM
Again, the sequence of EViews programs is too complicated and too problem-specific to be presented here.
However, a model is working, and is used in operational studies. Its elements are available and can be provided
through a special agreement.
10.5 A REGIONAL MODEL
Building a regional model presents the same features as in the multi-country case, with some differences:
• There is no currency issue, nor any difference in the definition of monetary rules. But the level of
interest rates can vary.
• Variables such as various deflators are not independent from one region to the other.
• Variables such as unemployment have to be computed at the country level, or at least regional
variables will influence strongly other regions, especially if we consider migrations.
• “Importing” from other regions is much easier. This means in our sense that instead of a sequence:
definition of imports then separation into providers, all forms of demand from one region should be
separated immediately into providers, the region itself being treated like the others. But of course this
region must be given a higher potential share, at the same level of competitiveness. Some products
will have to come from the region, in particular services, even if intermediate consumption can be
”imported”. For instance, restaurants in Paris can propose seafood from Brittany.
164
We use it quite often in practice.
499
• But this information has less chance of being available, in the absence of regional trade
measurements.
• Migrations are much easier, and should be taken into account, through both their determinants and
their effects. Transfer of household revenue is also easier and more natural.
• One should also consider the difficulty in transporting goods from one region to the other.
All these elements have been taken into account in one of our projects.
10.6 A MULTI COUNTRY, MULTI PRODUCT MODEL
Basically, the issues combine the features (and the difficulties) of both dimensions: countries and products.
• In addition to combining the features described earlier, the only problem lies in our opinion in the
much higher complexity of the problem. Even if the EViews programs can be compacted through the
use of loops, the actual number of variables is much higher, which means that the control of model
properties will be more complex. In practice it will have to be conducted first at a global level, then as
a detailed one.
• Also, the assumptions for forecasts and response to shocks will be more difficult to produce, especially
if one wants to profit from the additional detail (which should be expected).
• Also, the requirements in terms of data are of course more important, maybe even unrealistic.
Nevertheless, we have produced a working version for 3 regions and 3 products.
10.6.1 THE EVIEWS PROGRAM
Although such a program has actually been created, its size is too important to present if here.
The technology is not too complex, as it just combines the elements from its two dimensions.
500
11 CHAPTER 10: QUASI ACCOUNTING MODELS
Until now, our proposals on model building supposed
• The availability of a large set of numerical information both in scope and length, with a short
periodicity if possible.
• A good insight in the way the local economy is working, both on the whole and for each specific
behavior.
In that situation, one can:
• Devise a global framework for the model.
• Establish a consistent set of formulas which describe accurately the local behaviors, possible
considering several options.
• Test each of these formulas using currently accepted methods, based on enough information to make
the results conclusive one way or the other.
Once this is done, the model can be tested, and options reconsidered. Testing is applied again, until
convergence is obtained to a satisfying version, in both economic and statistical terms.
Unfortunately, these conditions are not always present, in particular for developing countries for which:
• The statistical system is not efficient, and has not been set in place for a long time.
• The national accounts system is not precisely established.
• There is no consensus on the way the economy is working.
• Anyway, the behaviors evolve very quickly.
• The unmeasured informal sector plays an important role.
Even when enough information is available, one might want to get results in a detail for which information is
too limited and statistical testing cannot be considered.
One can resort to using a Quasi-Accounting model.
The basic idea is replacing behavioral equations by “Quasi-Accounting” ones, in which the behavior is still
present but simplified, to a given degree.
11.1 THE ORIGINALITY OF THE APPROACH
In terms of complexity, in particular the number of equations, a QAM is not necessarily less developed. The
lower sophistication can be used to develop two aspects: the number of products can be increased, and the
number of operations too.
11.1.1 THE PRODUCT DETAIL
As long as econometrics are not involved, the number of products treated is only limited by data availability.
501
For instance, employment by product can be defined as an ratio to its value added, and imports a share of its
total demand.
11.1.2 THE OPERATIONS DETAIL
To complete the specifications, one clearly needs to define an input-output table, around which the other
operations will be defined, through the following framework.
This framework ensures that supply and demand are balanced for each product. Of course this calls for a
complete Input-Output Table for each period (a sparse availability will introduce residuals). TThis description
502
can be completed by a full economic table, with agents in columns and operations in lines, balanced also in
transfers by identifying for each of them the emitter and the receiver.
Finally, as with Stock-Flow Consistent models, we can consider developing the financial account, both in flows
and in stocks.
Two issues can be raised:
• The above table is defined at current prices. It looks difficult to consider only current elements. The
economic diagnosis relies for a large part on real elements, if only the specific supply – demand
balance and the productive process (investment – capital – GDP).
One simple solution would be to identify a system of exogenous prices, and use it to compute the real values.
But this poses two problems: the supply and demand balances would no longer hold true, and the
consequences of shocks would not take into account the difference in price evolutions (for instance the import
prices are not so sensitive to local inflation).
Some behaviors apply to elements at constant prices, like the quantity of one product needed to provide a unit
of another.
• These tables do not consider supply, and there is no control over the feasibility of production, and the
sensitivity of trade values to available productive capacities, let alone the impact on prices.
However, once capital and employment are defined, a kind of production function can be introduced. For
instance, the calibrated investment equation can stabilize the ratio of GDP to capital.
11.2 THE LEVEL OF SIMPLIFICATION
The simplification can be applied at several levels.
The simplest solution is to consider only ratios. For instance imports can represent an exogenous share of total
demand, and consumption a share of household revenue One has just to specify correctly the base. Imports
should be a ratio to total demand (including intermediate consumption) and VAT a ratio to final demand
(including imports and excluding exports). For employment, the model should divide output by an exogenous
labor productivity. In a model freed from econometric constraints, it is essential to define causalities using the
relevant elements.
Proceeding a little further, one could define investment, not as a share of GDP, but as a ratio to capital,
depending on the growth of GDP. Coefficients could be calibrated, with capital defined as the sum of
investment and the non-discarded share of the previous value.
Even further, capacity could be defined by applying a ratio to capital (a complementary factors option) or even
a function of capital and labor (à la Cobb-Douglas). Then the output gap could enter the investment formula,
ensuring its disappearance in the long run.
This output gap could also influence both exports and imports.
We can also define the change in inventories as proportional to the change in GDP of the same product. This
option corresponds to yet unsold products. For the intermediary goods not yet proceeded, a specific variable
should be constructed, using as weights the vector of technical coefficients associated with the product.
As to endogenizing the prices, this is a much more complex problem…
503
11.3 THE GOOD POINTS
The Quasi Accounting model is not statistically validated, and its level of simplification will be generally very
high. But is has its own advantages.
• No econometric constraint has to be applied. This frees the economist from a long, often tedious and
risky process (remember David Hendry). It leaves him a choice, which can come from:
o Theoretical considerations
o Its own experience in building structural econometric versions.
o The literature.
For a multi-product model, coefficients can be set at the same value, or be modulated according to the
product. For instance, the production of minerals can be independent from capacity. Estimating
coefficients consistent with the nature of the product is quite an impossible task, even if the data is
available.
Formulations can even be different. Production can be exogenous, or follow a complementary factors or
Cobb-Douglas (even CES?) formula.
• An error-correction specification can be applied, without going through the testing process of
stationarity and cointegration.
• With no econometrics, one can use shorter series.
• The data needed is clearly identified, plus capital and labor. If one uses the framework described
above, the data should be directly available from statisticians working on national accounts and their
tables, and the model results presented in a the very format the users are accustomed to, including
advisors to decision makers.
• The chosen specifications are easier to understand, and to explain. Summarizing the choices can be
done using a single equation in most cases. In EViews, the equation can be specified once in a loop
over products.
• Properties are also easier to control and modify if needed.
• It is probable (but not provable) that the simplification will limit the probability of convergence
problems. The model should be more linear in practice, and the Newton algorithm should prove more
efficient (Gauss-Seidel too, actually).
• Once you have built a successful model, you can export the programs to another case. The main
difference should relate to the number of products (a single parameter) or maybe the detail of
operations.
11.4 TWO EXAMPLES
A little less than 10 years ago, we built our first two Quasi Accounting models.
The first one was part of a statistical project improving the Algerian economic information system (project
AMECO). The statistical office provided a series of annual tables, according to the above framework, with a
detail of 19 products. This led to the production of a Quasi-Accounting model, along the lines presented above.
In the following tables, we present the consequences of a 1% shock on demand, separately for each product
(we use the exogenous Government demand, so the shock should be considered technical rather than realistic
in terms of policy).
504
Of course, these results should call for a specific study. We will content ourselves with showing the interest of
using a Quasi-Accounting model, compared to a structural version. The elements we will present are rather
straightforward.
The first table shows the consequences for GDP itself, after 4 years, measured in millions of constant dirhams.
The black line represents the ex-ante shock, which allows us to observe the Keynesian multiplier.
Agriculture sylviculture fishing
Water and energy
Hydrocarbons
Services and oil related construction Shock in Government demand The change in Gross Domestic Product
Mines and quarries
160,000
Metals chemicals electricity electronics
Construction material 140,000
Construction
120,000
Chemicals plastic rubber
Food industry 100,000
Textiles material
80,000
Leather and shoes
Wood and pepper 60,000
Various industries
40,000
Transportation and telecom
Commerce 20,000
Hotels and restaurants
0
Services to firms
In millions of2012
constant Dirhams
Services to households
Not surprisingly, the highest values correspond to elements produced locally, and which call for additional local
production, either as intermediate goods or investment.
This corresponds to primary goods procuced locally (on the left) or tertiary services (on the right). Services
present the highest values, especially “Hotels, cafes and restaurants” a product which, in addition to being
produced locally, consumes a lot of intermediary goods and also calls for construction (which itself is produced
locally, even though some of its intermediary consumption is imported…). We could probably travel along the
chain further, but the model computation does the whole job.
The second table presents the consequences for imports. The values present as expected a negative
correlation to the first table (a statistical test would certainly confirm it). But the correlation is not always
strong: hydrocarbons call for little imports but also little intermediate consumption, investment and labor. And
the food industry is largely imported in Algeria but calls for high local inputs (first of all agricultural products).
Again, the black line stands for the ex-ante increase in demand.
505
Agriculture sylviculture fishing
Water and energy Shock in Government demand The change in Imports
Hydrocarbons
Services and oil related construction
Mines and quarries 120,000
Metals chemicals electricity electronics
Construction material 100,000
Construction
Chemicals plastic rubber 80,000
Food industry
Textiles material 60,000
Leather and shoes
Wood and pepper 40,000
Various industries
Transportation and telecom 20,000
Commerce
Hotels and restaurants 0
Services to firms 2012 Dirhams
In millions of constant
Services to households
Finally, we present the decomposition of the supply-demand equilibrium for specific products, used in the
above arguments. They confirm the conclusions.
Government demand in product 01 : "Agriculture, sylviculture, fishing"
Government demand in product 03 : "Hydrocarbons"
2.0
1.4
1.6 1.2
1.0
1.2
0.8
0.8 0.6
0.4
0.4
0.2
0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00
+ 1 point of GDP
Government demand in product 06 : "Metals, chemicals, electricity, electronics" Government demand in product 07 : "Construction material"
G1 : Balance in GDP points
1.4
1.6
1.2 1.4
Final demand
Gross domestic product 1.0 1.2
Exports
0.8 1.0
Imports
Production 0.8
0.6
Intermediary consumption 0.6
0.4
0.4
0.2
0.2
0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 08 : "Construction" Government demand in product 17 : "Hotels and restaurants"
2.0 2.4
1.6 2.0
1.6
1.2
1.2
0.8
0.8
0.4
0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
506
Now, we shall consider a Vietnamese model built for the Ministry of Agriculture and Rural Development, in
2012.
This model starts again from full tables (including an I/O matrix) identifying this time 138 products and
produced by the General Statistical Office.
Due to the purpose of the model (and limitations on memory size of computers 165), we have compacted the
tables into 52 products, with a focus on agriculture and related categories (like packaged food, leather
garments, and veterinary services). Considering an I/O matrix alone (at constant and current prices, with the
associated deflators) would call for 3 x 139 x 139 = 57963 equations.
At present, the model lacks some of the possible improvements
• Prices are still exogenous.
• The output gap has an impact on investment but not external trade.
This means that:
• Supply oriented policies have no impact.
• The multiplier is higher than usual.
Of course, this could be improved.
Let us shock the demand for product 23: processed rice (patty is product 1). At the global level, we get the
following results (in percentage).
We can observe the increase in global demand and the decrease in the output gap.
IC by intermediate product of product 23 : Rice
G1 : Supply - demand equilibrium (global) G3 : Production factors
Final dem and
.008 .014
Intermediate consumption
.007 Exports
.012
GrossDomestic Product
Added value
.006 Im ports
.010 Productive capital
Em ploym ent
.005
.008 Rate of use
.004 Investm ent
.006
.003
.004
.002
.002
.001
.000 .000
2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025
Now let us compare the consequences of a demand shock for two primary products: paddy (with low
intermediate consumption and almost no imports) and raw rubber (even if Vietnam has planted hoveas, it
isalmost entirely imported, with very low intermediate consumption). We separate the ex-post impact on the
product itself and the others. We use the same scale for both graphs.
165
This would probably no longer be true.
507
We can observe that only the first shock extends to local production and demand (intermediate but also final),
with a very high multiplicative effect. Of course, other products would give fewer extreme results.
Government demand in product 01 : Paddy (all kinds) Government demand in product 04 : Raw rubber
4 4
All products The product itself Other products All products The product itself Other products
3 3
Final demand
Intermediate demand
Gross Domestic Product
2 2 Intermediate use
Imports
1 1
0 0
2017 2017
To complete the diagnosis, let us observe the consequences of an increase in exogenous demand by 1% of
GDP, for different products. To introduce this decision, we used Government demand, even though in real life
Government does not necessarily have a use for the particular product.
We shall present the relative change in supply-demand elements, and (more explicitly in our opinion) the
changes in GDP points of the same elements, and of a decomposition of local demand (total and final).
We have concentrated on products showing a panel of reactions as wide as possible.
Differences can come from:
• Most importantly, the share of imports in the original demand.
• The quantity of intermediary goods needed, and their own import share.
• The investment needed to produce it and the intermediary goods.
• The number of jobs created, and the consumption associated to the additional household revenue.
All the subsequent waves of influence are considered, as with the multiplier effect.
Although our purpose is not to produce a study on the subject, we can observe that:
The diagnosis calls for relative variations, but the decomposition in absolute terms can provide more
information. It can concern the share on imports in demand, the importance of intermediary consumption and
the decomposition of demand.
For instance:
508
Government demand in product 01 : Paddy (all kinds) Government demand in product 04 : Raw rubber
6 1.4
G1 : Supply - demand equilibrium in GDP points
1.2
5
Final demand 1.0
Gross domestic product 4
Exports 0.8
Imports 3
Production 0.6
Intermediate consumption 2
0.4
Total demand
1 0.2
+1 point of GDP 0.0
0
2016 2017 2018 2019 2020 2021 2022 2023 2024 2016 2017 2018 2019 2020 2021 2022 2023 2024
Government demand in product 01 : "Agriculture, sylviculture, fishing" Government demand in product 02 : "Water and energy"
+ 1 point of GDP + 1 point of GDP
2.0 2.0
G1 : Balance in GDP points G1 : Balance in GDP points
1.6 1.6
Final demand Final demand
Gross domestic product 1.2 Gross domestic product 1.2
Exports Exports
Imports Imports
0.8 0.8
Production Production
Intermediary consumption Intermediary consumption
0.4 0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 03 : "Hydrocarbons" Government demand in product 04 : "Services and oil related construction"
+ 1 point of GDP + 1 point of GDP
1.4 2.4
G1 : Balance in GDP points G1 : Balance in GDP points
1.2 2.0
Final demand
1.0 Final demand 1.6
Gross domestic product Gross domestic product
0.8
Exports Exports 1.2
Imports 0.6 Imports
Production Production 0.8
Intermediary consumption 0.4 Intermediary consumption
0.2 0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 05 : "Mines and quarries" Government demand in product 06 : "Metals, chemicals, electricity, electronics"
+ 1 point of GDP + 1 point of GDP
2.0 1.4
G1 : Balance in GDP points G1 : Balance in GDP points
1.2
1.6
Final demand Final demand
1.0
Gross domestic product 1.2 Gross domestic product
0.8
Exports Exports
Imports
0.8
Imports 0.6
Production Production
Intermediary consumption Intermediary consumption 0.4
0.4
0.2
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
509
Government demand in product 07 : "Construction material" Government demand in product 08 : "Construction"
+ 1 point of GDP + 1 point of GDP
1.6 2.0
G1 : Balance in GDP points G1 : Balance in GDP points
1.4
1.6
1.2
Final demand Final demand
Gross domestic product 1.0 Gross domestic product 1.2
Exports 0.8 Exports
Imports Imports
0.6 0.8
Production Production
Intermediary consumption 0.4 Intermediary consumption
0.4
0.2
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 09 : "Chemicals, plastic, rubber" Government demand in product 10 : "Food industry"
+ 1 point of GDP + 1 point of GDP
2.0 2.4
G1 : Balance in GDP points G1 : Balance in GDP points
2.0
1.6
Final demand Final demand 1.6
Gross domestic product 1.2 Gross domestic product
Exports Exports 1.2
Imports Imports
0.8
Production Production 0.8
Intermediary consumption Intermediary consumption
0.4
0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 11 : "Textiles, material" Government demand in product 12 : "Leather and shoes"
+ 1 point of GDP + 1 point of GDP
1.4 2.4
G1 : Balance in GDP points G1 : Balance in GDP points
1.2 2.0
Final demand 1.0 Final demand 1.6
Gross domestic product Gross domestic product
0.8
Exports Exports 1.2
Imports 0.6 Imports
Production Production 0.8
Intermediary consumption 0.4 Intermediary consumption
0.2 0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 13 : "Wood and pepper" Government demand in product 15 : "Transportation and telecom"
+ 1 point of GDP + 1 point of GDP
1.4 2.0
G1 : Balance in GDP points G1 : Balance in GDP points
1.2
1.6
Final demand 1.0 Final demand
Gross domestic product
0.8
Gross domestic product 1.2
Exports Exports
Imports 0.6 Imports
0.8
Production Production
Intermediary consumption 0.4 Intermediary consumption
0.4
0.2
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 16 : "Commerce" Government demand in product 17 : "Hotels and restaurants"
+ 1 point of GDP + 1 point of GDP
2.0 2.4
G1 : Balance in GDP points G1 : Balance in GDP points
2.0
1.6
Final demand Final demand 1.6
Gross domestic product 1.2 Gross domestic product
Exports Exports 1.2
Imports Imports
0.8
Production Production 0.8
Intermediary consumption Intermediary consumption
0.4
0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
Government demand in product 18 : "Services to firms" Government demand in product 19 : "Services to households"
+ 1 point of GDP + 1 point of GDP
2.0 2.0
G1 : Balance in GDP points G1 : Balance in GDP points
1.6 1.6
Final demand Final demand
Gross domestic product 1.2 Gross domestic product 1.2
Exports Exports
Imports Imports
0.8 0.8
Production Production
Intermediary consumption Intermediary consumption
0.4 0.4
0.0 0.0
10 12 14 16 18 20 22 24 26 28 30 98 00 10 12 14 16 18 20 22 24 26 28 30 98 00
510
11.4.1 WORKING WITH BOTH TYPES
We have already stated that in some cases, using a SEM (Structural Econometric Model) is impossible, and a
Quasi Accounting model is de facto the only option available.
It is also possible for some economists or institutions to decide that both options are possible, but a QAM gives
the most interesting results (detail versus sophistication) or the cheapest (this quite clear, provided the data us
available). But if a structural model is already operational, would a QAM help?
We can consider two options.
• The first is to use the two models separately in the same study.
o The structural model will give more sophisticated information, in particular on the short and medium-
term dynamics.
o The QAM will give more detailed information, and assert better the structural implications, as we have
just seen.
The main problem lies in the differences in the global diagnostic. One can force the two solutions to give the
same global results, but this is not feasible for the response to shocks.
• The second is to use the QAM as help to improve the results of the SEM. Many options can be
considered.
o One can use the QAM as a prologue, aggregating the detailed assumptions as global elements. This
can produce new ex-ante assumptions. But one can also run the QAM and aggregate the results to
provide the same assumptions, now ex-post according to the QAM. This can concern:
* The external trade: World demand by product / importer, the foreign deflators
* Production: capital productivity, labor productivity, the technical coefficients
* The tax rates: VAT, other indirect taxes, social security, subsidies
o But one can also consider running both models simultaneously.
Let us take employment as an example. One can:
* Estimate detailed equations for employment, formalizing labor productivity and using an error
correction framework giving the employment levels.
* Compute global GDP using the SEM.
* Separate GDP into products, possible ex post from the QAM.
* Apply the estimated employment system.
* Aggregate employment and feed it into the SEM to run it again.
This process profits from the fact that employment is not an accounting variable. And the set can be run in a
single EViews, program.
o Finally, one can use the QAM to disaggregate the results of the SEM, without any feedback.
511
12 CHAPTER 11: STOCK-FLOW CONSISTENT MODELS
While the author has a long experience of building, using and maintaining structural models, and even Quasi-
Accounting ones, his interest in Stock-Flow Consistent models initiated just two years ago, in the course of a
project for Vietnam. Actually, this is the reason why our example will consider this country.
So this chapter will be even less affirmative than usual, and give even more the priority to technical issues,
compared to economic theory. One should not be surprised if we do not describe in detail the foundations of
this field, or provide a survey of existing models. This is done much better by specialists, in the references
(concentrating on actual surveys) provided at the end of the book.
What we intend here is to provide the reader with the information on how to build his own version of an SFC
model, leaving to him the choice regarding the formulations, the causalities, and even the identified elements,
as the decision to consider some variables as negligible is not always straightforward. The associated EViews
programs are available on demand and can be extended easily to a new case.
In the previous chapters, we have described the causalities and the balances in the “real” model, and
addressed the optimization behavior of agents, for instance in the choice of factors by firms, and the bargaining
in the determination of the wage rate. What we intend to do here is to extend this approach to the financial
field. By upgrading its description, our goal is to obtain a fully consistent and homogenous model, without
favoring one of the aspects.
This means that our approach will not change very much. We will establish the accounting framework for an
SFC model (or the SFC part of a traditional model), and provide the reader with the practical tools required to
build his own version. We will leave to him the development of the theoretical aspects, either on his own or
using the principles developed in the referenced papers.
In the following paragraphs, with a growing level of affirmation, we will provide a general description of the
field, describe the technical framework including a full table presentation, propose a set of accounting
causalities between concepts using exogenous keys, and consider possible endogenizations of these keys.
Then we will proceed a little farther, by showing how the real and financial sides can be merged, and their links
taken into account. We will reproduce some of our usual shocks, and observe how even a simplistic SFC
approach modifies and completes the diagnosis.
This chapter should be considered as a simple (maybe sometimes naïve) initial proposal. The comments we will
hopefully receive will help to update the present version. An actual cooperation with more advanced specialists
would even be a better solution and would be quite welcome.
12.1 THE GENERAL APPROCH
The Stock-Flow Consistent field of models presents four interesting characteristics compared to traditional
macroeconomic versions.
• They describe full economic tables, in which columns (agents) and lines (operations) are completely
balanced.
This means that for each transfer (such as wages or interests on loans) the model must describe the origin
(firms or households) and the destination (households or banks).
512
For goods and non-financial services, they must be part of vertical balances, and supply must be equal to
demand. However, lines cannot be balanced, as the transfer between buyer and supplier is not identified. For
instance, we do not know if an imported good will be bought by a firm or a household (the problem can be
partially solved by introducing an artificial intermediary agent).
Even for a single country model, this calls for a detailed financial account, describing cash (“high powered
money”), loans, deposits (both in local and foreign currency), bills and bonds, and equities.
This calls also for describing the impact of these elements on the real side itself.
Obviously, this feature guarantees model consistency, and the absence of “black holes”. Actually, it should be
required from any model, and has been present in most truly operational models, included the MESANGE
model managed by the French Ministry of Finance.
• They consider stocks. Each transfer must reduce a stock for an agent and increase a stock for another.
Financial stocks are not impacted by time, but not goods and non-financial services. Consumption goods
disappear as they are bought, and capital is subject to depreciation.
These stocks generate revenue, and influence behaviors, in particular ratios stabilizing mechanisms. The
stability of the model is improved and can be controlled.
• These stocks are consistent across agents. For instance, the amount of bonds issued by Government must
be equal to amount detained. But this is obviously guaranteed by the two previous conditions.
• Each transfer actually appears four times (a feature called “quadruple book-keeping”). A transaction by an
agent impacts its account, and (twice) the account of its partner in the transaction. For instance, if a
household runs a negative cash account it can ask for an equivalent bank loan. This transaction will appear
in the “cash” and “loans” lines in the Households column, and twice in the same lines in the Banks column.
In addition, the financial account introduces several behaviors of agents, maximizing their utility. In particular:
• Households decide how to spend their savings: housing investment, deposits, equities, bonds.
• Non-Financial Firms decide to borrow or issue equities.
• Government decides to borrow or issue bonds.
• The Rest of the World trades its currency, makes foreign direct investment (building and participations,
financial or non-financial).
This feature improves the model, including the non-financial blocks, as the revenue from financial assets
follows different rules.
But a survey of the present field of SFC models shows that they keep rather simple concerning the “real” part,
and in consequence use little or no econometrics (which is not generally present in financial blocks, anyway).
We propose to show how to improve this shortcoming. We shall give an example for this, admittedly limited as
to SFC features.
Even if our elements are not always pertinent, we hope to lay out a framework which can be used to produce a
fully consistent version, combining the real and financial aspects.
513
12.2 THE ACCOUNTING FRAMEWORK
Although the number of variables and equations can vary, the first condition is clear and mandatory: the model
must produce a table in which the lines (the operations) and columns (the agents) are both systematically
balanced.
This condition should actually apply to all models, and has actually been met by some of them, for instance the
MESANGE model of the French economy.
12.2.1 THE AGENTS
Let us start with the agents. As usual, the classification must adapt to the behaviors. A difference in behavior
(including the management of economic elements) must lead to the creation of a different category.
Taking into account basic operations calls at least for the following five:
• Households who receive wages and consume.
• Non-Financial firms which produce goods and non-financial services.
• Financial firms which manage loans and deposits.
• Government which provide non-market services and collect taxes.
• The rest of the world which trades with the country.
Now, this crude framework is not precise enough.
The Central Bank has a specific institutional role, and must be separated from other financial agents.
A large share of non-financial production comes from households (especially in Vietnam). But the associated
behavior is difficult to separate from that of actual firms. Identifying this production will lead to accounting
problems, starting with the transactions between them. For instance, the model will have to define the
intermediary consumption of the “household” good by the “firms” branch (like the rice bought by shops from
farmers). The practical solution (followed by most models) is to separate the productive role of households
from its other operations and merge it with firms. This introduces minor problems which we will present (and
solve) later.
However, there is a domain which should remain attached to households: the housing services, which include
the “fictitious” rents paid by real estate owners to themselves.
Finally, this classification presents an obvious dissymmetry: the rest of the world is described as a whole,
grouping all of its agents. Some SFC models will make the categories symmetric by identifying the same
elements for the “rest of the world”. Unfortunately, this is difficult to apply to an operational model, for an
obvious lack of information.
12.2.2 THE OPERATIONS
Concerning the operations, the degree of detail is not as clear. In practice, it must be made at three levels:
• The nature of the operation itself: for example social security benefits can be separated into
unemployment, sickness…
514
• The currency: one can consider local and foreign, for which a single currency will be used (the US
Dollar?).
• The horizon: short-term and long-term loans, or bills and bonds.
Our proposal will be associated to a medium case.
12.3 A PROPOSAL
We will now propose a framework taking into account the above considerations.
Compared to the single product model we have presented earlier, we have to:
• Adapt the real sector to a separation into three products.
• Create from scratch a financial block.
• And finally modify the “real” blocks to account for the new financial elements.
12.3.1 THE FRAMEWORK FOR THE NEW PRODUCTION BLOCK
The supply – demand equilibrium at constant prices is obtained for each product by:
GDP + intermediate consumption by the branch + imports
= Final local demand + intermediate consumption of the product + exports.
or using mnemonics.
(1) 𝐺𝐷𝑃𝑖 + 𝐼𝐶.𝑖 + 𝑀𝑖 = 𝐹𝐷𝑖 + 𝐼𝐶𝑖. + 𝑋𝑖
The index “i” stands for each of the three products (and branches) : non-financial firms, Firms and Households.
The dot (“.”) stands for the sum across the particular dimension (first product used, second branch using the
product).
12.3.1.1 At constant prices
• Final demand FD is the sum of its elements:
(2) 𝐹𝐷𝑖 = 𝐶𝑂𝐻.𝑖 + 𝐶𝑂𝐺𝑖 + 𝐼𝑃𝑖 + ⋯ ..
• Exports and imports come from the rest of the model.
• GDP balances equation (1)
• IC is given by applying a “technical coefficient” to value added:
(3) 𝐼𝐶𝑗,𝑖 = 𝑡𝑐𝑗,𝑖 . 𝑄𝑖
• Production is given simply by addition:
(4) 𝑃𝑅𝑖 = 𝐺𝐷𝑃𝑖 + 𝐼𝐶.𝑖
515
But we must consider that GDP is the sum of Value Added, VAT and tariffs (both at constant prices):
VAT at constant prices is obtained by applying a “constant” rate to final demand.
(5) 𝑉𝐴𝑇0𝑖 = 𝑟_𝑣𝑎𝑡0𝑖 ∗ 𝐹𝐷𝑖
For tariffs, we apply a rate to imports:
(6) 𝑇𝐴𝑅0𝑖 = 𝑟_𝑡𝑎𝑟0𝑖 ∗ 𝑀𝑖
Now value added can be obtained from GDP:
(7) 𝑄𝑖 = 𝐺𝐷𝑃𝑖 − 𝑉𝐴𝑇0𝑖 − 𝑇𝐴𝑅0𝑖
The “constant” rates can be approximated by the rate in the base year. However, it should also include the
change in the weightings across time. For instance, the VAT rate grows naturally as the weight of basic products
in household consumption decreases. The rate is actually called “using the regulations of the base year”.
As Government (non-market) GDP measures essentially the compensations of civil servants, its link to
intermediate consumption of Government is not so clear as other types. It is generally considered exogenous at
constant prices.166
Finally, the above makes it clear that the data for VAT must be separated from other indirect taxes if one wants
to identify GDP.
12.3.1.2 At current prices
The situation is different here, as we know the deflators for GDP (a decision of the local producing agent), and
the deflators for imports and exports (a decision of the local or foreign exporting agent). As we know all the
quantities, we can apply deflators to these three elements, and we get the value of demand for each product.
(8) 𝑋𝑉𝑖 = 𝑃𝑋𝑖 ∗ 𝑋𝑖
(9) 𝑀𝑉𝑖 = 𝑃𝑀𝑖 ∗ 𝑀𝑖
(10) 𝐺𝐷𝑃𝑉𝑖 = 𝑃𝐺𝐷𝑃𝑖 ∗ 𝐺𝐷𝑃𝑖
(11) 𝐼𝐶𝑉𝑗,𝑖 = 𝑡𝑐𝑗,𝑖 . 𝑄𝑉𝑖
166
(definition taken literally from the OECD website) General government final consumption can be broken
down into two distinct groups. The first reflects expenditures for collective consumption (defense, justice, etc.)
which benefit society as a whole, or large parts of society, and are often known as public goods and services.
The second relates to expenditures for individual consumption (health care, housing, education, etc.), that
reflect expenditures incurred by government on behalf of an individual household. This category of
expenditure is equal to social transfers in kind from government to households and so includes expenditure by
government on market goods and services provided to households. As goods and services produced by
government usually do not have a market price, the relevant products are valued at the sum of costs needed to
produce these goods and services. These costs mainly consist of compensation of employees, intermediate
consumption and depreciation. Final consumption of government can then be estimated as the difference
between on the one hand government output, and on the other hand payments made for goods and services
produced by government and the relevant output that is used for fixed capital formation.
516
This means that the equilibrium comes from FDV and its deflator:
(12) 𝐹𝐷𝑉𝑖 + 𝐼𝐶𝑉𝑖. + 𝑋𝑉𝑖 = 𝐺𝐷𝑃𝑉𝑖 + 𝐼𝐶𝑉.𝑖 + 𝑀𝑉𝑖
(13) 𝑃𝐹𝐷𝑖 = 𝐹𝐷𝑉𝑖 / 𝐹𝐷𝑖
(14) 𝑉𝐴𝑇𝑖 = 𝑟_𝑣𝑎𝑡𝑖 ∗ 𝐹𝐷𝑉𝑖
(15) 𝑇𝐴𝑅𝑖 = 𝑟_𝑡𝑎𝑟𝑖 ∗ 𝑀𝑉𝑖
(16) 𝑄𝑉𝑖 = 𝐺𝐷𝑃𝑉𝑖 − 𝑉𝐴𝑇𝑖 + 𝑇𝐴𝑅𝑖
12.3.2 THE FINANCIAL BLOCK
12.3.2.1 The financial operations
We shall follow a traditional separation of the way each agent distributes its (positive of negative) financial
capacity. However, it includes some simplifications, and the user is welcome to introduce or suppress minor
elements (which can have some impact on causalities).
• High Powered Money or cash issued by the Central Bank, and used by the agents (households, firms)
for their transactions.
• Deposits made by households
o In local currency in local financial firms
o In foreign currency in foreign banks (or local subsidiaries of foreign banks).
• Loans
o in local currency obtained by households, firms or Government from local financial firms.
o in foreign currency obtained by firms or Government from the rest of the world.
• Changes in bank reserves placed at the Central Bank
• Bonds and bills issued by Government, and bought by households, financial firms, the Central Bank
and the rest of the world.
• Equities issued by firms and bought by households, financial firms and the rest of the world.
• Foreign direct investment by the rest of the world in local firms of both types.
This leads to the following framework.
Financial account
Change in reserves in foreign currency DFOR_X DFOR_C
Change in High Powered Money DHPM_C DHPM_H DHPM_P DHPM_F
Change in deposits, local currency DDPL_F DDPL_H
Change in deposits, foreign currency DDPX_X DDPX_H
Change in loans, local currency DLOL_H DLOL_P DLOL_G DLOL_F
Change in loans, foreign currency DLOX_P DLOX_F DLOX_G DLOX_X
Change in bank reserves DRES_C DRES_F
Change in bills and bonds DBBO_G DBBO_H DBBO_F DBBO_C DBBO_X
Change in equities DEQU_P DEQU_H DEQU_F DEQU_X
Foreign direct investment FDIV_P FDIV_F FDIV_X
We hope the simplifications are not harmful.
12.3.3 THE ADDITIONAL CHANGES LINKED TO THE FINANCIAL ACCOUNT.
Taking into account these changes calls for two additions.
517
• The associated interests and dividends:
Primary allocation of income
Interests on deposits, local currency IDPL_F IDPL_H
Interests on deposits, foreign currency IDPX_X IDPX_H
Interests on loans, local currency ILOL_H ILOL_P ILOL_G ILOL_F
Interests on loans, foreign currency ILOX_P ILOX_F ILOX_G ILOX_X
Interests on bills and bonds IBBO_G IBBO_H IBBO_F IBBO_X
Interests NI_H NI_P NI_F NI_G NI_X
Dividends DIV_P DIV_H DIV_F DIV_X
• The stocks completing the “S” in SFC.
Balance sheet
Debt DEB_H DEB_P DEB_F DEB_G DEB_X
Productive capital KV_H KV_P KV_G
Stock of reserves in foreign currency FOR_C FOR_X
Stock of High Powered Money HPM_C HPM_H HPM_P HPM_F
Stock of deposits, local currency DPL_F DPL_H
Stock of deposits, foreign currency DPX_X DPX_H
Stock of loans, local currency LOL_H LOL_P LOL_G LOL_F
Stock of loans, foreign currency LOX_X LOX_P LOX_F LOX_G
Stock of bank reserves RES_C RES_F
Stock of bills and bonds BBO_G BBO_H BBO_F BBO_X BBO_C
Stock of equities EQU_P EQU_H EQU_F EQU_X
Foreign direct capital KFDIV_P KFDIV_F KFDIV_X
12.4 EVOLVING TOWARDS A FULL MACROECONOMIC MODEL
• In the “real” part of our model, we will need the elements presented above.
• An optimization of productive factors taking into account the relative costs of labor and capital.
• An impact of the output gap on investment and the production deflator, introducing again an
optimization process (long- and short-term respectively).
• An implicit choice of the optimal trade deflators by the exporting agent.
• Traditional Keynesian and price-wage loops fully formalized using the most logical framework
available.
• A full table for the six agents, balanced in lines and columns, and producing the financial capacities as
the last line.
As an example, we will build a macro-econometric model based using Vietnamese data, but in the very general
framework we have presented earlier. To comply with the SFC conditions, this model will consider three
productive sectors (and products): Financial, Non-Financial and Households (considering only the housing
sector).
We will start with a version in which the splitting by agents (such as the partition of households’ savings) uses
exogenous distribution keys.
We will consider incentives for deciding on these keys. The main one could be the rate of return of each
category, which would have to be made explicit and actualized. But we leave for now to specialists the choice
and formalizations.
Even without this feature, one of the key properties of this model will be the combination of stabilizing
processes coming from the real side (correcting the output gap, or maximizing profits by considering the
tradeoff between output and unitary margins) and the financial side (for instance considering the debt ratios,
or the tradeoff between the rates of return of instruments and their demand level).
The properties of this model will be observed through a series of shocks on:
• First on the real exogenous assumptions (such as Government demand).
• Some financial assumptions (such as the Central bank’s base rate).
518
And in future versions, we will shock:
• More financial assumptions.
• The basic keys (such as the share of equities bought in households’ savings).
12.4.1 A FULL PROPOSAL
In summary, our model will propose the following framework.
Noms Hhlds Firms Fin F Ctl Bk Govmt RoW Hhlds Firms Fin F Ctl Bk Govmt RoW
Production account
Exports of goods and services DLOL_H XV FCAP_G
Imports of goods and services MV
Intermediate consumption ICBV_H ICBV_P ICBV_F COGV ICPV_H ICPV_P ICPV_F
Gross value added / GDP GDPV_H GDPV_P GDPV_F GDPV_G
Ext. trade in goods and services XV-MV XV-MV
Operating account
Gross wages and salaries WAGE_H WAGE_P WAGE_F WAGE_G WAGE
Employers social contributions SC_H SC_P SC_F SC_G SCR_G
Workers social contributions SC_W SC_W
Tariffs TAR TAR
Taxes on production and imports OIT_H OIT_P OIT_F OIT
Subsidies SUBS SUBS_H SUBS_P SUBS_F
Gross operating surplus EBE_H EBE_P EBE_F EBE_G EBE_X
Gross mixed income REVQ REVQ
Primary allocation of income
Interests on deposits, local currency IDPL_F IDPL_H
Interests on deposits, foreign currency IDPX_X IDPX_H
Interests on loans, local currency ILOL_H ILOL_P ILOL_G ILOL_F
Interests on loans, foreign currency ILOX_P ILOX_F ILOX_G ILOX_X
Interests on bills and bonds IBBO_G IBBO_H IBBO_F IBBO_X
Interests NI_H NI_P NI_F NI_G NI_X
Dividends DIV_P DIV_H DIV_F DIV_X
Secondary allocation of income
Value Added Tax VAT_H VAT_P VAT_F VAT
Taxes on income and wealth TP_H+... TP_P TP_F TP+ICT
Social benefits SOCB SOCB
Current international cooperation COOP COOP
Gross disposable income RDB_H RDB_P RDB_F RDB_G RDB_X
Allocation of income
Final consumption expenditure COHV COGV
Gross savings GSV_H GSV_P GSV_F GSV_G
Capital account
Gross fixed capital formation IV_H IV_P IV_F IV_G
Change in inventories CIV
Financing capacity (+) FCAP_H FCAP_P FCAP_F FCAP_X
Financial account
Change in reserves in foreign currency DFOR_X DFOR_C
Change in High Powered Money DHPM_C DHPM_H DHPM_P DHPM_F
Change in deposits, local currency DDPL_F DDPL_H
Change in deposits, foreign currency DDPX_X DDPX_H
Change in loans, local currency DLOL_P DLOL_G DLOL_F
Change in loans, foreign currency DLOX_P DLOX_F DLOX_G DLOX_X
Change in bank reserves DRES_C DRES_F
Change in bills and bonds DBBO_G DBBO_H DBBO_F DBBO_C DBBO_X
Change in equities DEQU_P DEQU_H DEQU_F DEQU_X
Foreign direct investment FDIV_P FDIV_F FDIV_X
Balance sheet
Debt DEB_H DEB_P DEB_F DEB_G DEB_X
Productive capital KV_H KV_P KV_G
Stock of reserves in foreign currency FOR_X FOR_C
Stock of High Powered Money HPM_C HPM_H HPM_P HPM_F
Stock of deposits, local currency DPL_F DPL_H
Stock of deposits, foreign currency DPX_X DPX_H
Stock of loans, local currency LOL_H LOL_P LOL_G LOL_F
Stock of loans, foreign currency LOX_P LOX_F LOX_G LOX_X
Stock of bank reserves RES_C RES_F
Stock of bills and bonds BBO_G BBO_H BBO_F BBO_C BBO_X
Stock of equities EQU_P EQU_H EQU_F EQU_X
Foreign direct capital KFDIV_P KFDIV_F KFDIV_X
Of course, the model will also describe:
• The deflators.
519
• The transactions on goods and services at constant prices.
• Capital and labor.
12.4.2 THE CAUSALITIES
The above elements do not solve anything. They just map the domain on which we will work.
Two tasks remain.
• Defining the causalities between the financial elements (remember we are dealing with elements
at current prices, which must not remain exogenous). Exogenous (unidimensional) keys can be
used here.
• Endogenizing these keys.
From now we feel like an explorer travelling in uncharted territory, using the efficient instruments we can
possess. Help from local scouts would be great.
We would nevertheless make proposals, if only to show that addressing the first point does not lead to
completely unreasonable results.
The reader can use the following elements as he desires. The EViews programs are available for using and also
editing. We will welcome comments and maybe cooperation.
12.4.3 THE CAUSALITIES ( KEYS)
Our next step is defining the inner causalities and the closures. This will be done using what we think to be the
most logical sequence, but this can be questioned of course.
12.4.3.1 High Powered Money.
We will suppose that households and firms are able to obtain the cash they need from the Central Bank,
depending on their economic situation. This gives DHPM_H, DHPM_P and DHPM_F and the sum DHPM, closing
the line.
12.4.3.2 Households
We will suppose that households ask for (and obtain) loans, depending on their financial capacity, consumption
and housing investment, and they separate the remaining balance into buying bonds and equities, and making
deposits, both in local and foreign currency.
This closes the household’s column.
12.4.3.3 Deposits in foreign currency
We will suppose that these deposits are obtained from the rest of the world, with local banks as a simple
intermediary.
This closes the foreign deposits line.
12.4.3.4 Deposits in local currency
We will suppose that these deposits are made by households and held by local banks.
520
This closes the local currency deposits line.
12.4.3.5 Foreign reserves
We will suppose that in addition to providing HPM, the Central Bank buys foreign currency (or sells the local
one) to finance the foreign deficit, but also as a policy instrument.
The line is offset by the Rest of the World.
This closes the foreign reserves line.
12.4.3.6 Foreign Direct Investment
We will suppose that it is made by the Row in both financial and non-financial firms.
This closes the FDI line.
12.4.3.7 Government
We suppose that Government finances its deficit by obtaining loans and issuing bonds/bills.
This closes the Government column.
12.4.3.8 Foreign loans
We will suppose that they are made by the RoW to firms (both Financial and Non-Financial) and Government.
This closes the line.
12.4.3.9 Non-financial firms
We will suppose that they close their account by borrowing and issuing equities.
12.4.3.10 Bank reserves
The banks transfer reserves to the Central Bank in proportion to their loans.
This closes the reserves line.
12.4.3.11 The Central Bank
The Central Bank closes its account by buying bonds.
12.4.3.12 The rest
The remaining elements correspond to the Bills/Bonds and equities bought by financial firms and the Rest of
the World. The sum by agent is now fixed, and determining one element decides on the remaining three.
12.4.4 EXPLICITING THE BEHAVIORS
We shall now try to improve the above by proposing explanations to the elements.
As usual, one can consider two categories: behavioral or identity.
521
The identities normally represent closures, except maybe for the reserves held by the Central Bank, transferred
from Financial firms.
The behavioral equations can be separated again, into:
• Decisions depending only on non-financial variables. They are exogenous to the block.
• Decisions implying other financial elements. A causal sequence must be defined.
• Splitting financial decisions with a given sum. The share(s) are defined first, then the elements.
12.4.4.1 “Exogenous” decisions
One can consider 7 elements:
• Loans to households, depending on their consumption, their housing investment and (negatively) their
savings.
We will suppose that this decision is made before the other uses of the resources available.
• Loans by the RoW to firms (both types) and Government.
The decision should be made at constant (foreign) prices, depending possibly on the perceived present
health and prospects of the local economy.
Non-Financial Firms: they can depend on trade and financial capacity
Financial Firms: same
Government:: they can depend on financial capacity, but a share can come from international
cooperation, both international institutions or individual countries.
• Foreign Direct Investment.
This element is already present in the initial model. It depends on relative profitability and demand (local and
foreign) addressed to the country. It can be shared between the types of firms according to both criteria.
It influences total factor productivity and export potential.
• The change in reserves of the Central Bank
To the automatic impact of financial capacity can be added a behavioral element aiming at controlling
the exchange rate.
• The dividends paid to equities holders by the firms
As we consider only equities issued by non-Financial firms, dividends will come only from them. They
will represent a share on previous gross disposable income, and will be distributed to financial firms,
households and rest of the world according to their share of total equities.
12.4.4.2 First sharing
One can now apply three sharing processes.
12.4.4.2.1 Households
Households will share their savings (plus money from loans) between:
• Investment in housing
522
• Deposits in local currency
• Deposits in foreign currency
• Buying bills and bonds
• Buying equities
Of course, all these changes can actually be negative.
With such a number of opportunities, the sharing is rather complex.
The explanation could come from:
• The relative interest rates (and the expected returns from equities). The interest on housing
investment plays an opposite role.
• The expected evolution of equities.
• The expected evolution of the exchange rate.
The household account is now closed.
12.4.4.2.2 Foreign direct investment
Sharing between financial and non-financial firms should depend on relative profitability, and demand
potential for the non-financial sector.
12.4.4.3 First closures
As stated before, we can now determine total High-Powered Money (provided by the Central Bank), deposits in
local currency (in banks) and deposits in foreign currency (in banks but the money comes from the RoW).
We know enough now to determine the amount in the financial firms account which is subject to reserves, and
the change in reserves in the Central Bank. However, the ratio applied to stocks, so we must also consider the
change in the value of the previous stock
12.4.4.4 Second sharing
12.4.4.4.1 Government
The government deficit (minus the foreign loans) can be financed either by borrowing of issuing bills and
bonds.
The choice depends mostly on the interest rates.
Bills can be separated from bonds with different rates (depending on the horizon).
The Government account is closed.
12.4.4.4.2 Non-Financial Firms
They can finance the remaining sum by issuing equities or borrowing.
The choice is difficult, but one can consider:
• That the consequences of borrowing are clear.
• That the more you borrow the higher the interest rate.
523
• That issuing equities you will lose assets ownership, future profits and dividends (all unclear)
• That if your firms regain health, it is easier to pay back debt than buy back equities (at a higher price).
12.4.4.4.3 New closures
Now we know all elements in the Central Bank account except bills and bonds, which balance it accordingly.
Also, the loans by Financial Firms balance the line.
12.4.4.4.4 Final sharing
As stated earlier, the remaining bills/bonds and equities must be shared between local Financial Firms and the
RoW (mostly foreign financial firms).
This can be done in several ways, but once an agent has decided on the sharing, the behavior of the other is
fixed.
This balances the two remaining lines and columns
12.4.4.5 The list of equations in the financial block
The identities are completely defined, the logic of the behaviors state the potential explanations as a sum,
enclosed in parentheses, and following a “f”, as we have done for the real model.
Foreign Reserves
Equation [1] Foreign reserves: Central bank
dforr_c = fcap_x + dforra_c
Equation [2] Foreign reserves : Rest of the World
dforr_x = - dforr_c
High Power Money
Equation [3] High Power Money : Households
dhpm_h = f * (hdi_h)
Equation [4] High Power Money : Non Financial Firms
dhpm_p = f * (gdpmv_p)
Equation [5] High Power Money : Financial Firms
dhpm_f = f * (gdpmv_f)
Equation [6] High Power Money : Central Bank
dhpm_c = dhpm_h + dhpm_p + dhpm_f
Foreign Direct Investment
524
Equation [7] Foreign Direct Investment : Non Financial Firms
fdiv_p = f * (ur_p + rprob_p)
Equation [8] Foreign Direct Investment : Financial Firms
fdiv_f = f * (rprob_f)
Equation [9] Foreign Direct Investment : Rest of the World
fdiv_x = fdiv_p + fdiv_f
Deposits in local currency
Equation [10] Deposits in local currency : Households
ddpl_h = r_ddpl_h * (fcap_h + dlol_h)
Equation [11] Deposits in local currency : Financial Firms
ddpl_f = - ddpl_h
Deposits in foreign currency
Equation [12] Deposits in foreign currency : ]Households
ddpx_h = r_ddpx_h * (fcap_h + dlol_h)
Equation [13] Deposits in foreign currency : Rest of the World
ddpx_x = - ddpx_h
Loans in local currency
Equation [14] Loans in local currency: ]Households
dlol_h = f * iv_h
Equation [15] Loans in local currency : Non Financial Firms
dlol_p = r_dlolp * (dhpm_p + fdiv_p + dlox_p)
Equation [16] Loans in local currency : Government
dlol_g = (1 - r_dbbo_g) * (fcap_g + dlol_g)
Equation [17] Loans in local currency : Financial Firms
dlol_f = dlol_h + dlol_p + dlol_g
Loans in Foreign currency
525
Equation [18] Loans in foreign currency : Non Financial Firms
dlox_p = f * (@pch(gdpv_p) + @pch(ur_p) + fcap_x / gdpv_p)
Equation [19] Loans in foreign currency : Financial Firms
dlox_f = f * (@pch(gdpv_f) + fcap_f / gdpv_f)
Equation [20] Loans in foreign currency : Government
dlox_g = f * (fcap_g / gdpv)
Equation [21] Loans in foreign currency : Rest of the World
dlox_x = dlox_p + dlox_f + dlox_g
Bills and Bonds
Equation [22] Bills and Bonds : Households
dbbo_h = r_dbbo_h * (fcap_h + dlol_h)
Equation [23] Bills and Bonds : Government
dbbo_g = r_dbbog * (fcap_g + dlol_g)
Equation [26] Bills and Bonds : Rest of the World
dbbo_x = dbbo_g - (dbbo_h + dbbo_f - dbbo_c)
Equation [27] Bills and Bonds : Financial Firms
dbbo_f = r_dbbo_f * (fcap_f + fdiv_f + ddpl_f + dres_f - (dhpm_f - dlol_f - dlox_f))
Equation [32] Bills and Bonds : Central Bank
dbbo_c = dforr_c + dbbo_c - dhpm_c
Equities
Equation [24] Equities : ]Households
dequ_h = (1 - r_ddpl_h - r_ddpx_h - r_dbbo_h) * (fcap_h + dlol_h)
Equation [25] Equities : Non Financial Firms
dequ_p = (1 - r_dlolp) * (dhpm_p + fdiv_p + dlox_p)
Equation [28] Equities : Rest of the World
dequ_x = dequ_p - dequ_h - dequ_f
526
Equation [29] Equities : Financial Firms
dequ_f = (1 - r_dbbo_f) * (fcap_f + fdiv_f + ddpl_f - (dhpm_f - dlol_f - dlox_f))
Reserves
Equation [30] Reserves : Financial Firms
dres_f = r_dres_f * dlol_f
Equation [31] Reserves : Central Bank
dres_c = dres_f
12.5 THE SIMULATIONS
We shall present the properties of the present version, by applying shocks to assumptions. If our comments
will be somewhat detailed and technically sophisticated, one must not conclude that we are proposing this
version as final.
In practice we shall make eight shocks, the first two traditional shocks concentrating on demand and supply
features of the model.
The shocks will be represented and commented for:
1. An increase in government investment, by 1 point of market GDP.
2. A decrease in the VAT rate, by one point of the rate.
3. A devaluation of the Vietnamese Dong by 1%
We have also produced, but we will not show the results, for:
4. An increase in Foreign Direct Investment by one point of the capital level.
5. A decrease in the rate of tariffs applied to foreign products, by one point of the rate.
6. A decrease on the rate of tariffs applied by the rest of the world to Vietnamese products, by one point
of the rate.
7. An increase in quotas applied to foreign products.
8. An increase in quotas applied by the rest of the world to Vietnamese products.
None of these shocks apply to elements in the financial part of the model, although some influences transit
through it, like the interests from various financial objects purchased by the agents.
All these shocks start in 2021, and are sustained over the whole period.
For each shock we shall produce 6 graphs:
1 - The supply demand-equilibrium: GDP, Final demand, exports and imports at constant prices, plus the GDP
deflator.
2 - The production process: Investment, capacity, firms employment, rate of use of capacities.
3 – Some important ratios: rate of use, unemployment rate, real wage cost, capital-labor ratio.
4 - External trade at current prices (1): Exports, imports, export – import ratio
5 - External trade summary (2): Export - import ratio at constant and current prices, terms of trade.
527
6- Prices : of demand, of GDP, of imports, of exports, wage rate
THE SHOCKS
1 A demand shock: An increase in government consumption
Starting in the year 2021 we increase Government demand by 1% of the GDP forecasted for the same period.
This will make interpretation easier than for a constant shock; in particular concerning the “Keynesian
multiplier”.
The definition of this element, in practice the most frequently used to characterize a model, is quite simple. In
the presence of a given shock on demand, coming for instance from Government consumption, it measures the
ratio between the change in GDP produced by the shock and its initial size. If the ratio is higher than one, it
means that the economic effects of the shock on the economy are higher than the negative ones. Of course,
this figure has also to be interpreted.
In our case, the “Keynesian multiplier” will be obtained naturally by considering the relative percent change of
GDP itself. If GDP increases by more than 1%, it means that the response is higher than the shock.
We observe in Graph 1 that the multiplier presents a limited cycle (with a period of around 40 years)
converging to around 0.35 in the long run. This is consistent with other models associated with a very open
economy.
1.4
1.2
Exports to GDP
1.0 Imports to Final demand
0.8
0.6
0.4 The trade ratios at constant prices
0.2
96 98 00 02 04 06 08 10 12 14 16
Of course, using Production and Total demand as a denominator would bring the value down (to around 0.55
on 2017).
Final demand presents the same evolution, around a higher level, with an average value of about 1%, which
means that the positive effect of endogenous demand is offset by the loss on external trade.
In the short run, the change in imports follows the same dynamics, at a lower level. Exports presents a very
limited decrease.
528
These elements can be easily explained, considering the model as a whole.
As local demand and GDP are of the same order of magnitude (their difference is net exports, relatively small
compared to GDP), the initial (or ex-ante) shock increased demand by roughly 1%. This means that the impact
of model mechanisms reduces the ex-post value of GDP growth. Let us detail this aspect.
Graph 2 shows that for the non-financial product, both factors (capital and employment) grow significantly,
requiring a strong effort on investment in the short run. As explained earlier, firms expect production to
represent a “normal” share of capacities, defining a target on their rate of use. If production increases, a gap
will appear, which firms will try to close by increasing capacities through investment and labor. However, this
process is slowed by the cautiousness of firms, and the technical difficulty in implementing their decisions. This
inertia applies especially to capital, for which the consequences of wrong decisions are more costly. It takes a
long time (around 10 years) for capacities to reach. But then the inertia on decisions (particularly capital) leads
to overshooting, starting a cyclical process with a decreasing amplitude.
However, capital and employment evolutions diverge from capacity. In the beginning, employment grows
more, as adaptation to the target is faster (the change in capital comes only from investment, itself quite inert).
In the medium run, the order is reversed by the increase in the relative cost of labor, coming from the reduced
unemployment and the partial indexation on the value-added deflator, the element with the highest increase
among prices (see Graph 6). In the end the increase in employment is quite small, and capital increases by 0.3%
(representing in the long run an increase in investment of the same amount).
Graph 3 shows that the changes in the labor-capital ratio and their relative cost converge, a result consistent
with our Cobb-Douglas assumption.
Consumption grows even slower (around 0.25%). There are several reasons for this inertia.
• Employment adapts gradually to the production increase.
• The purchasing power of wages is affected initially by non-unitary indexation on prices, the gap
building up with additional inflation. This loss will disappear as the price increase stabilizes.
• The short-term sensitivity of consumption to revenue is lower than one.
• Inflation reduces the purchasing power of current savings.
• A large share of household revenue is not indexed on activity.
The real wage rate profits from the decrease in unemployment, but this has an adverse moderating effect on
employment itself, as we have seen.
The initial decrease in unemployment is the main explanation for the evolution of prices (Graph 6). In the short
and medium terms, the higher inflation comes from tensions on the rate of use (with higher sales perspectives,
firms feel that they can apply higher prices), and on the labor market through the unemployment rate.
One moderating element is the increase in labor productivity, due to the inertia facing an increase in
production. But we have seen that the adaptation of labor to production was rather fast. 167 This element
appears essentially in the first periods, creating very temporary deflation.
167
As to capital, the effect is more complex: the investment effort, destined to generate capacities for the
future, is higher in relative terms than the present increase in GDP. But in the cost we only consider the share
of the present period, through a correcting coefficient.
529
The loss in competitiveness will reduce quantities sold. Let us now consider the trade balance (Graphs 4 and 5).
We know that the trade prices have a different (calibrated) sensitivity to local inflation (0.2 for imports, 0.7 for
exports). And the short- and long-term competitiveness elasticities are lower than one: 0.53 for exports, 0.46
for imports. But the change in local prices is not negligible, especially in the long-term.
The evolution of exports is the simplest to explain they are essentially sensitive to the loss in price
competitiveness, with the highest elasticity. They also follow the rate of use of capacities, but the effect is
limited by the (calibrated) coefficient and the evolution of the rate itself, which returns to the base value at the
end of the sample. In addition, exports are favored by the higher share of FDI capital
Imports are more affected: they follow local demand, with the addition of losses in market shares, coming first
from the limits on available capacity (decreasing with time) then from the gradual loss in price competitiveness,
which compares the local production price to rather stable import prices. Both effects roughly offset each other
during the period, and the profile of imports follows demand. The level is intermediate between final and
intermediate demands (the latter roughly indexed on GDP).
Concerning the balances, the terms of trade are little affected, and the loss is similar on both nominal and real
balances.
Now, as to the Government budget, the additional expenditures lead to some revenue, coming mostly from the
taxes on production. But as borrowing finances this measure, the interests paid will also increase.
We can see that the revenue effect wins in the short run, but the accumulation of debt reverses the order
quickly, to a level which could be considered unsustainable (limiting the reliability of the shocks’ message).
In the above explanation, we have evoked FDI. It actually plays an important (and original) role.
First, it accounts for the low decrease in exports, as it creates capacities oriented in that direction.
It also explains the initial deflationary effect, as the higher productivity of both factors reduces costs.
In the medium-term, the effect decreases, but remains significantly favorable.
12.5.1.1.1 The financial elements
Now, let us observe the consequences for the financial account, the main originality of our model. One must be
aware that although the financial account is completely consistent and balanced, and follows an explicit causal
sequence, the agents take decisions by applying exogenous shares. Applying truly endogenous behaviors will
be introduced in a later version.
For this, we will use the global economic table, concentrating on the financial account.
One must also realize that elements appear at current prices. For shocks prioritizing deflators (such as a
decrease in the VAT rate, or a devaluation) this diverges from the usual comments on real values.
In the following we have neglected the smallest elements, but one can check that the table is balanced both in
lines and columns.
We shall first consider the results for the first period of the shock, 2021. We shall then move to the medium-
term (2025) and the long-term (2050).
The agent directly concerned is Government.
530
Its investment is the most changed element in the table (1.023). As it must be financed without compensation,
interests grow already (0.123). Some revenue increases: social contributions (0.005), tariffs (0.046), direct taxes
(0.007). The resulting deficit (0.973) has to be compensated by issuing bonds (0.438), local borrowing (0.438)
and foreign (0.097), the latter allowed by the trade deficit.
For firms, the situation improves with GDP (0.315), reduced by wages (0.063), taxes (0.053): The gross
operating surplus increases by 0.201, 0.058 of which goes to individual workers. Taking into account VAT
(0.116), and the increased investment and inventories (0.076), financial capacity decreases by 0.044. But the
funds from foreign loans (0.036) and FDI (0.050) allow equities and local loans to decrease (0.027).
The rest of the world uses the trade surplus (0.863) to buy bills and bonds (0.830), equities (0.234), and to loan
(0.133). The difference comes from the increase in reserves (0.198), and FDI (0.048).
Households use the additional revenue (0.130), interests (0.003), to consume (0.088), to invest in lodging
(0.008), to buy money (0.012), equities (0.005), bills (0.002), and to increase deposits in both currencies
(0.020).
Financial agents (including the Central Bank) have a mostly passive behavior.
The above elements are consistent with our framework and causalities. They can be criticized by specialists of
the field for:
• The causalities themselves.
• The values used for shares.
• Their exogeneity.
However, all these choices can be modified, provided the consistency of the accounting system remains.
Let us move now to the two later tables. As we have seen earlier, for a demand shock, most of the
consequences are obtained at the first period.
The main dynamic effects come from:
• The increase in investment, with an initial inertia and a later decrease as capacities adapt to the
target.
• The increase in household consumption as employment adapts gradually and the wage rate gains
purchasing power at every period.
• The accumulation of the Government debt, as the investment effort is made every year, with the
addition of growing interests.
All these elements are present in the tables, including an increase in the debt by 19 points of GDP which looks
quite unsustainable.
The situation should change when we endogenize the shares, as they should consider the new relative
profitability.
2 A supply shock: decreasing the VAT rate by one point
We suppose that the Government decreases the VAT rate by 1 point.
The results are quite similar to the previous shock on the French model.
531
Graph 6 shows that all prices decrease significantly. Of course, the highest impact is on the demand price (fully
depending on VAT). The wage rate (partially depending on it, and gaining in purchasing power through the
decrease in unemployment).
But all other prices decrease too. The value-added deflator is indexed on the cost of factors, not only wages but
also capital. The production price follows, and the prices of trade, depending on the latter.
We will use again the following graph:
One can observe the loop between the wage rate and the value-added deflator, moderated by trade deflators
(the import price is less sensitive to the GDP deflator). Compared to a complementary factors framework, the
margins target includes also the capital cost, moderating the role of wages but introducing an additional loop.
The gain in competitiveness favors exports (Graph 1), once capacities have started to adapt, creating growth
through demand in investment and consumption goods. The subsequent increase in imports (coming also from
intermediate goods and energy requested by exports) is accentuated at first by the initial bottlenecks. Then
competitiveness increases, capacity adapts (and even over adapts) reducing also the demand for equipment
goods. But the variation of imports remains significantly positive, higher in fact than exports. But the loss on
trade in real terms is much lower than for the previous shock, for a comparable gain on GDP.
In the medium run, the deflationary effect keeps increasing moderately (Graph 6), as well as the gains in
competitiveness.
Concerning the factors (Graph 2) capital and investment are again favored by the increase in the real wage
rate, as the previous case, but the main impact comes from the immediate decrease in the demand price. As
employment adapts faster than capital (which closes slowly with its target), capacity takes some time in
increasing, and the rate of use remains high.
The impact on the real trade balance (Graph 5) is negative, as the increase in GDP is followed by demand, both
final and intermediate. With the global loss on the terms of trade, the impact on the current balance is even
more negative.
532
As to the Government budget, its evolution is similar to the previous case, except for the interests, with a
growing importance of interests paid, which decrease at first with the rate itself (remember it is indexed on the
evolution of CPI, which decreases for the first two years). However, the cost is lower, and the gain on GDP is
higher: this instrument seems more efficient on that aspect.
As to FDI, its effect is more favorable here, as profitability increases: the numerator (profits) is not sensitive to
VAT (it is even positively affected as wages are), and the denominator (the value of capital at cost of renewal)
decreases. The profitability ratio increases.
Its contribution to the improvement of exports both through factor productivity and competitiveness explains
the persistence of the GDP increase.
12.5.1.1.2 The financial elements
Now, considering the financial account, we face perhaps the most important difficulty in the SFC approach: the
main channel of this shock is through prices, and appreciating the efficiency of the decision takes essentially
into account the evolution of quantities.
The table is presented at current prices, and each of these prices changes a lot following the decision, with
different intensities depending on the role of unchanged foreign prices. This is particularly visible in the ending
paragraph giving the decomposition of GDP: we see that if GDP decreases a little (0.016) this comes from a
balancing between an increase in value added (0.797), tariffs (0.107) and a decrease in VAT itself (-0.919).
Of course, a presentation at constant prices could be clearer, but this will not be easy as the change in the
value of deflators is not homogenous at all.
3 A financial shock: devaluating the Vietnamese Dong by 1%.
Let us first consider Graph 6.
We can see that all prices increase gradually, converging monotonously to a 1% value (we have checked that
this will be achieved a few years later). Only wages grow faster, due to the improved activity and the
subsequent gains in purchasing power
The hierarchy is quite logical: the foreign price increases by 1%. The import price follows as the most
influenced, then comes the export price, the consumption price (an average of local production and imports),
and the value-added price (the most “local”).
Graph 6 shows that the gains in competitiveness are important, if we compare the export price with the 1% line
(representing the increase in the foreign price in Dongs), and the import price with the production price. This
brings as usual exports, production, demand and imports.
These gains bring growth (Graph 1) coming mostly from local demand, and a small gain in net exports.
Concerning the trade balance, the change at constant prices is small with a changing sign, compared to the loss
on the terms of trade (Graph 5).
However, all these effects will disappear with time at all prices converge to the target.
12.5.2 CONCLUSION ON MODEL PROPERTIES
We can draw the following conclusions:
533
The model converges over as long a period as we want, with stabilizing growth rates. The consequences of
changes in assumptions stabilize too.
The properties look reasonable in the short and long run, including the role of instruments (sign and size), and
especially:
• The stabilization of the rate of use of capacities (convergence of value added and capacities).
• The variation of prices, relative to each other.
• The link between factors and their costs.
• The contribution of factors to capacity growth.
• The presence of cycles, converging after a reasonable period.
• The differentiated role of policy instruments.
The introduction of a detailed financial account (both in variations and stocks) brings interesting features and
provides a better stability, even with exogenous shares.
The choices made in the values of shares and the causalities will need to be checked, and the shares remain to
be endogenized, following an optimizing framework.
534
Name Role Definition Agent
DBBO_C Eq32 Bills and Bonds Central Bank
DBBO_F Eq27 Bills and Bonds Financial Firms
DBBO_G Eq23 Bills and Bonds Government
DBBO_H Eq22 Bills and Bonds Households
DBBO_X Eq26 Bills and Bonds Rest of the World
DDPL_F Eq11 Deposits in local currency Financial Firms
DDPL_H Eq10 Deposits in local currency Households
DDPX_H Eq12 Deposits in foreign currency Households
DDPX_X Eq13 Deposits in foreign currency Rest of the World
DEQU_F Eq29 Equities Financial Firms
DEQU_H Eq24 Equities Households
DEQU_P Eq25 Equities Non-Financial Firms
DEQU_X Eq28 Equities Rest of the World
DFORR_C Eq1 Reserves in foreign currency Central Bank
DFORR_X Eq2 Reserves in foreign currency Rest of the World
DHPM_C Eq6 High Power Money Central Bank
DHPM_F Eq5 High Power Money Financial Firms
DHPM_H Eq3 High Power Money Households
DHPM_P Eq4 High Power Money Non-Financial Firms
DLOL_F Eq17 Loans in local currency Financial Firms
DLOL_G Eq16 Loans in local currency Government
DLOL_H Eq14 Loans in local currency Households
DLOL_P Eq15 Loans in local currency Non-Financial Firms
DLOX_F Eq19 Loans in foreign currency Financial Firms
DLOX_G Eq20 Loans in foreign currency Government
DLOX_P Eq18 Loans in foreign currency Non-Financial Firms
DLOX_X Eq21 Loans in foreign currency Households
DRES_C Eq31 Reserves at the Central Bank Rest of the World
DRES_F Eq30 Reserves at the Central Bank Financial Firms
FDIV_F Eq8 Foreign Direct Investment Financial Firms
FDIV_P Eq7 Foreign Direct Investment Non-Financial Firms
FDIV_X Eq9 Foreign Direct Investment Rest of the World
DFORRA_C Exog Policy correction to Foreign Reserves Central Bank
FCAP_F Exog Financial Capacity Financial Firms
FCAP_G Exog Financial Capacity Government
FCAP_H Exog Financial Capacity Households
FCAP_X Exog Financial Capacity Rest of the World
GDPMV_F Exog Market GDP Financial Firms
GDPMV_P Exog Market GDP Non-Financial Firms
GDPV Exog GDP Total
GDPV_F Exog GDP Financial Firms
GDPV_P Exog GDP Financial Firms
HDI_H Exog Household Disposable Income Households
IV_H Exog Investment Households
R_DBBO_F Exog Share of Bills and bonds in the residual Financial Firms
R_DBBO_G Exog Share of Bills/Bonds versus Local loans Government
R_DBBO_H Exog Share of Bills and bonds in the residual Households
R_DDPL_H Exog Share of Deposits in Dongs in the residual Households
R_DDPX_H Exog Share of Deposits in Foreign Currency in the residual
Households
R_DLOL_P Exog Share of Loans in Dongs in the residual Non-Financial Firms
R_DRES_F Exog Ratio of Reserves in Loans Financial Firms
RPROB_F Exog Profits rate Financial Firms
RPROB_P Exog Profits rate Non-Financial Firms
. UR_P Exog Rate of use of capacities 535 Non-Financial Firms
13 SOLVING A MODEL FOR TARGET VALUES
In the previous version of the book, we have presented a method allowing to switch the exogenous /
exogenous status of model variables without editing the actual code, using a quasi-Newton algorithm.
After a detailed description of the methodology, we provided some practical examples and a subprogram
allowing to apply it in a simple way.
Although this remains an option, starting with EViews 12, one can perform a similar task using simple and
official menus and commands.
We shall start with a detailed presentation of our method, then give the syntax of both new EViews options,
which are not yet documented from a theoretical point. We shall also provide examples.
Unfortunately, the User’s Guide provides few elements on the methodology. So I have chosen the following
approach. I will:
• Develop my methodology, assuming it is the same as the one used by EViews 12.
• Illustrate it on practical examples.
• Show how to use EViews commands to solve the same problem.
This is also the easiest approach for me of course. It will also allow users of other packages to replicate
the method.
13.1 THE METHOD USING A PROGRAM
13.1.1 INTRODUCTION
We start by presenting a general method allowing making a model reach given values for a set of endogenous
variables, by freeing the values of an equivalent set of exogenous elements.
This method can treat single or multi period simulations, and rational expectations can be considered at a small
price in complexity (but a higher one in computation time, especially in the last case)
At first, we shall consider solving a backward-looking model for a single period.
13.1.2 THE PROBLEM
Let us consider the model:
𝑦 = 𝑓(𝑦, 𝑥)
with:
536
• y vector of endogenous with dimension m
• x vector of endogenous with dimension n
We have suppressed the time index, as past (and future) values do not change in the solving process.
Let us suppose we want to obtain:
𝑦𝑠 = 𝑦𝑠∗
For a subset of y of dimension p
through the endogenization of a subset of x with the same dimension, called x s.
13.1.3 THE LINEAR CASE
If f was a linear function:
𝑦 =𝐴⋅𝑦+𝐵⋅𝑥+𝑏
we would have
𝑦 = (𝐼 − 𝐴)−1 (𝐵 ⋅ 𝑥 + 𝑏)
Let us call C the sub-matrix (𝐼 − 𝐴)−1 ⋅ 𝐵 (dimension (p x n) limited to the lines of 𝑦 𝑠 et 𝐶 𝑠 the sub-matrix (p x
p) of 𝐶 associated to the elements of𝑥 𝑠 .
We have:
𝑦 𝑠 ∗ = 𝐶 𝑠 ⋅ (𝑥 𝑠 ∗ − 𝑥 𝑠 0 ) + 𝐶 ⋅ 𝑥 0 + 𝑐
The values allowing reaching 𝑦 𝑠 ∗ are obtained immediately through:
𝑥 𝑠 ∗ = 𝑥 𝑠 0 + (𝐶 𝑠 )−1 (𝑦 𝑠 ∗ − 𝐶 ⋅ 𝑥 0 − 𝑐))
537
0
𝑥 𝑠 ∗ = 𝑥 𝑠 0 + (𝐶 𝑠 )−1 (𝑦 𝑠 ∗ − 𝑦 𝑠 )
The C matrixes can be obtained easily by a matrix transformation, once A, B and b are known.
13.1.4 THE NON-LINEAR CASE
Let us now consider the non-linear case. We still have:
𝑦 = 𝑓(𝑦, 𝑥)
and we still want to get:
𝑦𝑠 = 𝑦𝑠∗
We shall start by supposing that the model is quasi-linear and apply the above method. The C matrix becomes:
𝜕𝑔(𝑦)/𝜕𝑔(𝑥)
Where g represents the function 𝑦 = 𝑔(𝑥) where the y vector has been eliminated from the right term.
It is unfortunately clear that identifying the g function is difficult if not impossible in practice. On the other
hand, the values of the Jacobian are easy to obtain, or its approximation through finite differences.
One just has to solve the model for starting values of x, apply in sequence small shocks to each of the
exogenous, and compute the associated solution. We get:
𝜕𝑔(𝑦𝑖 )/𝜕𝑔(𝑥𝑗 ) ≈ (𝑔𝑖 (𝑥 + 𝛥𝑥𝑗 ) − 𝑔𝑖 (𝑥))/𝛿𝑥𝑗
where 𝛥𝑥𝑗 is a vector of dimension n with null values except for 𝑥𝑗 , changed by 𝛿𝑥𝑗 .
To get the equivalent of matrix C, we just have to select in the Jacobian the columns and lines associated with
𝑥 𝑠 and 𝑦 𝑠 .
538
The solution uses the same formula as the linear case:
𝑥 𝑠 ∗ = 𝑥 𝑠 0 + (𝐶 𝑠 )−1 (𝑦 𝑠 ∗ − 𝑦 𝑠 0 )
One shall notice that solving the model and computing the changes are only required for the exogenous
concerned, and the inversion applies to a much smaller (p x p) matrix.
The only difference with the linear case is that the solution is not exact, due to the linear approximation. But
one just has to extend the process, starting from the new values obtained for x (actually limited to 𝑥 𝑠 ). The
process remains exactly the same and can be repeated as many times as necessary. And as the linear
approximation becomes more and more accurate (in term of properties) when we get closer to the solution
(especially if the model is locally convex), the share of error eliminated each time becomes higher and higher,
and convergence should be reached after a few iterations.
We identify here a Newton type method, which is characterized by a supralinear convergence property. An
interesting feature is that we know the actual precision of the solution, contrary to other algorithms which
consider only the speed of convergence.
13.1.5 A FEW QUESTIONINGS
Now, several questions appear:
The method we suggest looks quite simple for a package which does not provide an automatic process (such as
Troll), and represents an efficient alternative to more complex (and less efficient) tools. One can question why
it is not proposed right away by packages and applied more often by modelers (unless we are not up to date as
to the present fashions).
Several answers can be considered:
• It does not work in practice, or only in very simple (non-operational) cases. We shall prove it works, at
least for simple but operational models. There is no reason (but it remains to be proved) that it does
not work also for the majority of reasonable cases.
• It does not work always. But then we can at least try, and maybe improve it in “difficult” cases. We can
expect two main cases for failure:
o The model is highly nonlinear. This is one of the problems with the Newton method, which can lead
the solution to absurd values if the Jacobian computed is very far from its “exact” value. It is essential
to start from values rather close to the solution, which should be the case in operational cases, if the
forecasting quality of the model is high enough to provide spontaneously reasonable, if not
acceptable, solutions.
o The Jacobian is highly collinear, and cannot be inverted, or suggests once inverted very high changes in
the exogenous (at least two of them). This case is less problematic, as another method (« by hand »
included) will face the same problems, and here at least the problems will be evidenced immediately,
and maybe interpreted through the observation of the matrix. The time gained can be used to look in
other directions.
• It does not apply to the problem. Indeed, the method is limited to reaching of a given set of n
endogenous values by freeing n exogenous elements. A professional modeler might want to spread
the charge of the process on more exogenous elements. But if the weight of the contributions is
539
known, the method can still be applied: several exogenous will be affected, but the algorithm looks
only for one value. For instance, if the rates of social contributions by employers and workers are used
simultaneously with the same variation, this variation becomes the unique endogenous element.
More problematic is the case of limits on changes, set by the user 168: a given exogenous is used until it reaches
a limit, in which case the computation switches to a new element. This case can also be treated automatically,
but the program become more complex, and the solution is not guaranteed within the acceptable range (but
neither “by hand” anyway).
• Finally, if the method supposes backward looking processes, it can also be applied to forward looking
models. Two techniques can be considered, associated to the current methods for the simple case:
o Fair Taylor: we iterate on the period-by-period solution, initializing future values by thus of the last
iteration (at the beginning, the historical, theoretical, or previously computed values).
o Lafarge: we solve the global models where equations are repeated along the time dimension.
But now we have to consider two nested loops, and the process can become unacceptably slow. Specific
improvements could be found.
13.1.6 A FEW IDEAS ON CHOSING THE INSTRUMENTS
In choosing both targets and instruments, a few principles can be established, based on the general idea that
the Jacobian matrix should be as regular as possible.
• The targets should be as independent as possible. For instance, using two correlated deflators, or
demand + imports, will be dangerous.
• Even more important, the instruments should have different impacts. In the same way, using two
instruments increasing household revenue (like social security payments and the income tax rate) can
lead to problems.
• The difference between the base solution and the targets must not be contradictory. For instance, it
will be difficult to improve the budget balance and increase GDP.
• The instruments should have an important impact. Reaching a realistic target must not need
unrealistic variations.
• And obviously the target itself should be realistic….
13.1.7 A TEST ON A VERY SMALL MODEL.
We shall present here the conditions and results of a test conducted on a very small model, called pic_b, with
18 equations (5 estimated) used by Jean Louis Brillet for his course in modeling techniques.
We have supposed a change in GDP (called Q) and imports at constant prices (M) of respectively +5% and -5%
compared for the historical values for the same period. The change in values goes much beyond usual cases,
and their direction is rather opposed: an increase in GDP generates imports, even if all things being equal an
increase in imports reduces GDP.
168
Mostly for sociological reasons : it not reasonable to expect an increase in the income tax by 20% to
accepted by households.
540
We shall use as targeting variables the real wage rate WR and world demand WD. These elements are rather
independent, first from each other (this should increase their efficiency and accuracy, as a team with different
skills), and also from the targeted elements.
We will give:
• The EViews program, clearly short in statements (but not in comments). It can apply to any model, the
only information requested (in addition to the access to the mode itself) being the period concerned,
the lists of targeted variables (with their values) those of targeting elements.
• The results from iterations. In this list, the “_cur” suffix is associated with current values (of
endogenous and exogenous) and the “_star” suffix to the endogenous targets.
The « delta » values represent the difference between target and present solution, for each targeted
endogenous. We observe that we need in practice 4 or rather 3 iterations (the first one consisting in simulating
the model with the initial values). The relative higher gain is generally obtained at the second.
As an indication, running the program takes 0,003 seconds on a portable computer of average power.
We shall the replicate this test on a larger model.
541
13.1.7.1 The EViews program
'============================================
' This program computes for a set of exogenous variables
' the values
' which allow a set of endogenous to reach given solutions
' The two sets must have the same dimension
' The pic_b model is used as an example
' but the program is quite general
' The present program applies to a single period
' but it can be generalized quite easily
'=============================================
' We set the directory
cd "c:\program files\eviews5\pic\"
' Results will be dumped as text to a file called jacob.txt
' overriding any previous elements
output(t,o) jacob
'
' We create a specific workfile called pic_star
' from the original pic_b
' We make sure no file of this name is open at present
'
'
close pic_b.wf1
close pic_star
wfopen pic_b
wfsave pic_star
' We create a new model
' Basically, the purpose is to present the model used as an example
delete(noerr) _pic_b
model _pic_b
_pic_b.append cap=pk*k(-1)
_pic_b.append ur=q/cap
_pic_b.append q+m=fd+x
_pic_b.merge eq_i
_pic_b.append log(prle_t)=c_prle(1)+c_prle(2)*(t-2002)+c_prle(3)*(t-t1)*(t<t1)+c_prle(4)*(t-t2)*(t<t2)
_pic_b.append led=q/prle_t
_pic_b.merge eq_le
_pic_b.append lt=le+lg
_pic_b.append rhi=wr*lt+xhr*q
_pic_b.append co=rhi*(1-sr)
542
_pic_b.merge eq_ic
_pic_b.append fd=co+i+gd+ic+rfdx*q
_pic_b.append res_m=log(m/(fd+0.5*x)) -1.322108*log(ur)+0.419518*log(compm)-
0.012582*(@trend(60:1)*(t<=2002)+@elem(@trend(60:1),"2002S2")*(t>2002))
_pic_b.merge eq_m
_pic_b.append res_x=log(x/wd)+0.686672*log(ur)-4.87E-
05*(@trend(60:1)*(t<=2002)+@elem(@trend(60:1),"2002S2")*(t>2002))
_pic_b.merge eq_x
_pic_b.append k=k(-1)*(1-dr)+i
' Now we start the actual program
' We shall use the period 2001S1
smpl 2001S1 2001S1
' We define the lists of elements which change status
' exogenous to endogenous
' endogenous to exogenous
' The numbers must be the same
' If not, the program stops with a message
group g_vexog1 wd wr
group g_vendo1 q m
!t1=g_vexog1.@count
!t2=g_vendo1.@count
if !t1<>!t2 then
statusline !t1 exogenous !t2 endogenous: stop
stop
endif
' nstar is the common number
scalar nstar=g_vexog1.@count
' We delete Scenario 1 (it exists)
_pic_b.scenario(d) "Scenario 1"
' We build Scenario 0 as the base
' It will contain the current solution
' The associated suffix will be "_0"
' We declare all the overriden exogenous
' (with the currently best solution)
' (note: we start with a blank list and fill it one by one)
_pic_b.scenario(n) "Scenario 0"
_pic_b.append assign @all _0
_pic_b.override
for !i=1 to g_vexog1.@count
%1=g_vexog1.@seriesname(!i)
_pic_b.override(m) {%1}
next
543
' Now we create the alternate scenarios
' for computation of the Jacobian
' They are numbered 1 to nstar (here 2)
' The suffix is "_n"
' The exogenous are the same as Scenario 0
' Note: all exogenous have to be overriden
' but only one will change from the _0 value
' for a given scenario
for !i=1 to nstar
_pic_b.scenario(n,a={!i},i="Scenario 0") "Scenario {!i}"
_pic_b.append assign @all _{!i}
next
' Now we give the values for the targets
' We used arbitrary numbers
smpl 2000S1 2001S1
genr q_star=q*(1+0.05*(t=2001))
genr m_star=m*(1-0.05*(t=2001))
' Now we define the Jacobian
matrix(2,2) jacob
' Now we initialize the exogenous instruments
' used to reach the target
for !i=1 to g_vexog1.@count
%1=g_vexog1.@seriesname(!i)
genr {%1}_cur={%1}
genr {%1}_0={%1}
for !j=1 to g_vexog1.@count
genr {%1}_{!j}={%1}
next
next
' And the endogenous target variables
for !i=1 to g_vendo1.@count
%1=g_vendo1.@seriesname(!i)
genr {%1}_0={%1}
genr {%1}_cur={%1}_0
next
' We set the maximum number of iterations
' and the convergence criterion
scalar nitmax=20
scalar sconv=1e-6
' We initialize as control variables
' the number of iterations
544
' And the variable controlling the convergence
!niter=1
!iconv=0
' We double the elements so they can be displayed
scalar niter=0
scalar iconv=0
' We display several elements for control
print g_vexog1 g_vendo1
print *_cur *_star
'---------------------------------------------------------
' Now we start the loop
'---------------------------------------------------------
smpl 2001S1 2001S1
' It will run while convergence is not achieved
' and the maximum numbre of loops is not reached
while !niter<=nitmax and !iconv=0
' We test for convergence
' We suppose it is achieved (!iconv=1)
' Then we look at the relative difference
' between the present solution and the target
' for each targetted endogenous
' If the relative difference is higher than the criterion
' for any target
' convergence is not yet reached (!iconv=0)
' else we stop
!iconv=1
scalar iconv=1
for !j=1 to g_vendo1.@count
%1=g_vendo1.@seriesname(!j)
genr delta_{!j}=({%1}_star-{%1}_cur)/{%1}_cur
scalar delta=@elem(delta_{!j},"2001S1")
if @abs(delta)>sconv then
!iconv=0
scalar iconv=0
endif
next
if !iconv=1 then
stop
endif
' We increment the iteration counter
545
!niter=!niter+1
scalar niter=niter+1
' We display the differences
print delta_*
' We compute the base solution
' defined as Scenario 0
_pic_b.scenario "Scenario 0"
solve _pic_b
' The solution values are set as current
for !i=1 to g_vendo1.@count
%1=g_vendo1.@seriesname(!i)
genr {%1}_cur={%1}_0
next
' We display the current solution (exogenous + endogenous) and the targets
print niter
print *_cur *_star
' Now we compute the Jacobian
' We start a loop on the shocked instruments
for !i=1 to g_vexog1.@count
' For each exogenous, we compute the whole set of overriding instruments
' shocking only the current one by 0.1 %
' while the others keep their base value
' This is necessary to implement the current changes
for !j=1 to g_vexog1.@count
%2=g_vexog1.@seriesname(!j)
genr {%2}_{!i}={%2}_cur*(1+.001*({!i}={!j}))
next
' We solve the model under Scenario !i
_pic_b.scenario "Scenario {!i}"
_pic_b.solve
' Now we compute the relative change in the target endogenous
' This will give a column of the Jacobian matrix
for !j=1 to g_vendo1.@count
%2=g_vendo1.@seriesname(!j)
_z=(({%2}_{!i}-{%2}_cur)/{%2}_cur)/0.001
jacob(!j,!i)=@elem(_z,"2001S1")
next
next
' This ends the computation of the Jacobian
' We compute its inverse
matrix jacob_inv=@inverse(jacob)
546
' We apply to each exogenous a change equal to
' the inverse of the jacobian
' times the remaining error.
' This gives a new solution for the exogenous
' We start the process again until convergence (hopefully)
for !i=1 to g_vexog1.@count
%1=g_vexog1.@seriesname(!i)
genr {%1}_old={%1}_cur
for !j=1 to g_vendo1.@count
%2=g_vendo1.@seriesname(!j)
genr {%1}_cur={%1}_cur+{%1}_old*jacob_inv(!i,!j)*({%2}_star-{%2}_cur)/{%2}_cur
next
genr {%1}_{!i}={%1}_cur
genr {%1}_0={%1}_cur
next
wend
13.1.7.2 The results
Iteration 0: Initial values
=================================================
obs 2000S1 2000S2 2001S1
=================================================
WD 411615.4 432806.5 439347.6
WR 0.028897 0.029101 0.029511
Q 1344432. 1365342. 1385977.
M 356344.0 383838.0 389639.0
=================================================
=================================================
obs 2000S1 2000S2 2001S1
=================================================
M_CUR 356344.0 383838.0 389639.0
Q_CUR 1344432. 1365342. 1385977.
WD_CUR 411615.4 432806.5 439347.6
WR_CUR 0.028897 0.029101 0.029511
M_STAR 356344.0 383838.0 370157.0
Q_STAR 1344432. 1365342. 1455275.
=================================================
=========================
547
obs 2001S1
=========================
DELTA_1 0.050000
DELTA_2 -0.050000
=========================
Iteration 1: solution with the initial exogenous
=========================
obs 2001S1
=========================
M_CUR 409107.1
Q_CUR 1383326.
WD_CUR 439347.6
WR_CUR 0.029511
M_STAR 370157.0
Q_STAR 1455275.
DELTA_1 0.052012 The model solution gives larger errors.
DELTA_2 -0.095208 We have not started targetting yet.
=========================
Iteration 2
=========================
obs 2001S1
=========================
M_CUR 365029.1
Q_CUR 1441045.
WD_CUR 887589.2
WR_CUR 0.012164
M_STAR 370157.0
Q_STAR 1455275.
DELTA_1 0.009875
DELTA_2 0.014048
=========================
Iteration 3
=========================
obs 2001S1
=========================
M_CUR 370184.2
Q_CUR 1455269.
548
WD_CUR 893720.4
WR_CUR 0.012746
M_STAR 370157.0
Q_STAR 1455275.
DELTA_1 4.35E-06
DELTA_2 -7.34E-05
=========================
Iteration 4: Convergence!
=========================
obs 2001S1
=========================
M_CUR 370157.0
Q_CUR 1455275.
WD_CUR 893915.9
WR_CUR 0.012738
M_STAR 370157.0
Q_STAR 1455275.
DELTA_1 2.25E-07
DELTA_2 5.42E-08
13.1.8 A TEST ON A LARGER MODEL
We reproduced the test on cases closer to operational.
Note: applying the program to a new model is immediate, provided it is available in an EViews workfile along
with the required data. One has only to update the two lists of targets and instruments, set the values for the
targets, and call for the workfile. In all it should take less than ten minutes.
First case: the quarterly model of the French economy, which we have just used to illustrate modelling under
EViews. It contains 88 equations, 11 of which are behavioral. Its specifications are quite traditional, and all
behavioral estimations are validated by econometrics. It features a Keynesian and a price-wage loops.
We have also selected a more operational case. We ask the model to reach.
• An unemployment rate of 6%
• An inflation rate of 2%.
• A null trade balance (exports = imports at current prices).
• A budget deficit of 1% of GDP.
For this we shall use the following instruments:
• The rates of social contributions by both firms and workers.
• Government investment at constant prices.
549
• The number of civil servants.
The algorithm converges again, this time in 5 iterations. Computation time is 0.231 seconds.
Second case: a multi country yearly model for the Andean community (5 countries including Venezuela and a
simplified rest of the world). It includes 770 equations, 11 x 5 of which are behavioral. Its specifications are
again quite traditional, and all behavioral estimations are calibrated this time. It formalizes completely the
trade flows.
This time we set as targets the Government deficit of each country (at 3% of GDP) and as instruments in
sequence:
• Government investment
• The VAT rate
• The rate of social security contributions by workers
• The rate of social security contributions by firms.
• The exchange rate
• An increase in the long-term interest rate.
THz system converged in all cases, taking between 3.7 and 5.2 seconds (5 iterations most of the time).
and finally:
• targets: the Government deficit of each country (at 3% of GDP) and balanced trade (net exports=0)
• instruments: Government investment and subsidies to firms.
converges in 7 .2 seconds.
13.2 USING THE EVIEWS INTEGRATED COMMAND
Of course, this new feature is better integrated in the software (but limited to EViews). As usual, you can work
through menus and programs.
The syntax for the program version is:
Name-of-the model.control(options) control target trajectory
You can either:
• give the full name of the control, target and trajectory series. You do not need to define options.
• Specify suffixes for the trajectory and (computed) control series.
options = (create, tsuffix=suffix, csuffix=suffix)
In case of 2 or more sets, you can use:
550
Name-of-the model.control(options) control1 control2 target1 target2 trajectory1 trajectory2
Name-of-the model.control(options) control1 target1 trajectory1 control2 target2 trajectory2
Name-of-the model.control(options) control1 control2 target1 trajectory1 target2 trajectory2 …..
13.2.1 AN EXAMPLE BASED ON OUR SMALL MODEL
We will present the method using the small described in chapter 13.
We will use Q (GDP) as target, and EC_M (the residual of imports) as instrument.
First, we will solve the model, and set the trajectory target (Q) as 5% higher than the result.
pic_b.append assign @all _b
pic_b.solve
genr q_t=q_b*1.05
Then we apply the method, specifying “_c” as the suffix for the instrument series found.
pic_b.control(create,csuffix=_c,tsuffix=_t) q ec_m
Then we solve the model overriding the instrument with the value found.
pic_b.append assign @all _c
pic_b.override ec_m
pic_b.solve
The result is indeed equal to the trajectory.
The case is favorable as the change in the instrument is small (and the staring point was close to the target).
In our first experiments, the algorithm can stall if the gap is high.
551
13.2.2 USING A MENU
Using a menu, the logic is exactly the same.
First, you access the model page.
In the “Proc” item, you select “Solve Controls for Target”. You obtain a menu in which you specify the
elements:
552
We solve the model again, using the value found for the instrument.
Obviously, it allows to meet the target.
13.2.3 WITH TWO TARGETS
Let us now suppose we want to reach given values to investment M and investment I, through the
determination of residuals EC_I and EC_M, over the 2020 – 2100 period, we will access the model item, then
“Procs” and “Solve Controls for Targets"
553
Percentage change (%)
Control Minimum Q1 Q2 Q3 Maximum
EC_M -2698.312 -417.858 -415.299 -390.636 -263.469
EC_I 53.399 110.343 120.146 121.203 1211.488
Level change
Control Minimum Q1 Q2 Q3 Maximum
EC_M 0.0062 0.00919 0.00977 0.00983 0.0635
EC_I 4.67e-05 9.64e-05 0.000105 0.000106 0.00106
554
Control EC_M Values
.07
.06
.05
.04
.03
.02
.01
.00
-.01
2020 2030 2040 2050 2060 2070 2080 2090 2100
Original Modified (EC_M_VAR)
Control EC_I Values
.0012
.0010
.0008
.0006
.0004
.0002
.0000
2020 2030 2040 2050 2060 2070 2080 2090 2100
Original Modified (EC_I_VAR)
13.3 RESPECIFICATION OF ENDOGENOUS AND EXOGENOUS VARIABLE S
As we have not considered using this option in practice, we will only rely on the EViews integrated command.
This feature provides you with the edited version of the model.
The syntax is:
model_name.fliptype(dest-name=name) variables [@lock fixed-variables] [@lockafs] [@lockinnovs]
555
• The “dest-name” parameter allows to create a new model with the given name.
• The “@” modifiers forbids the method to modify
o A list of variables
o The equations with add-factors
o The equations with innovations. This is useful for eliminating estimated equations, which generally
imply a given causality.
In menu mode, one accesses the model menu, and uses the item ”Flip Endog/Exog Variables” under “Proc”.
13.3.1 THE FIRST EXAMPLE
It switches imports and its residual. This exogenizes M and computes the residual which between its value and
the result from the model formula.
One can also reach a given GDP by endogenizing Government demand. The supply-demand equilibrium gives
556
now Final demand FD, and GD as the residual substracting Household consumption, Investment, and the
Change in inventories. All these elements come from the model solution, making it economically consistent.
13.3.2 A MORE COMPLEX EXAMPLE
Considering two switches, one can endogenize capital productivity PK and the wage rate WR to reach values of
Imports M and Household consumption CO.
The two new chains are independent (one linked to supply, the other to demand). Imports give the rate of use
UR, which gives capacity CAP and capital productivity PK. CO gives Household revenue RHI which gives the
wage rate WR. This is rather logical.
13.3.3 A LESS LOGICAL ONE
Now, if we endogenize capital productivity PK and World demand WD to reach values of GDP Q and exports X,
the process becomes less logical, as it transits through imports:
557
But of course one cannot object if the goal is only technical: to reach a solution.
It remains to be seen how this method affects the change of convergence for both Newton and Gauss-Seidel.
This can be the goal of a further study.
558
13.4 THE CONSEQUENCES FOR THE SOLVING PROCESS
This is here that the choice of method has the highest impact.
• The “multivariate” method” uses a quasi-Newton algorithm, probably similar to the one detailed in
the above paragraphs. The process is not accessible to the user. It should be fast (it is on the following
example).
If it does not converge (a proven possibility), the remedies are limited. The main option consists in a change in
the instruments, or a reduction of the gap between the target and the base solution, by using a less ambitious
target or making a better guess.
One should remember that the problems should come from non-linearities, and the difference between the
Jacobians (or rather their inverse) at the present values and the solution.
As the user has no control over the process, it is quite difficult to interpret the problems and ask for additional
information, as in a normal application.
• The “respecification” method create a new model version. The solving process is the responsibility of
the user, who can choose between Gauss-Seidel, Broyden or Newton.
For information on the differences between methods and their problems, please refer to Chapter 7.1.
Basically, solving problems rely on the linearity of the model, which has a different impact according to the
algorithm. The Newton method solves a succession of linearized versions, and a quasi-linear model will
converge quickly. Whereas the Gauss-Seidel process can lead away from the solution of a linear model, in a
“cobweb” manner.
This means that for Gauss-Seidel, the problems can come from the eigenvalues of the Jacobian, which can be
quite different from the original ones. It the new model framework is not checked for logic, is quite possible
that the highest modulus is higher than one. If that happens, one should examine the whole set of new
causalities, and try a different respecification. As we shall see, blocking some equations from the process is
available as an option.
For Newton (and maybe Broyden) one should check that the process does not create equations with non-linear
properties. Here logical interpretation can help. This will happen for instance if employment depends on
unemployment, a volatile element.
14 A LIST OF USEFUL SERIES FOR PRODUCING A SINGLE COUNTRY, SINGLE PRODUCT
MODEL
We shall give now a list of series normally useful for producing a simple but operational model. Of course, this
is only a proposal, designed as a starting point to which the user is welcome to apply as many changes as he
wishes.
559
Just as we have done for model specifications, we think it is easier to start from a proposal, even rather far
from what the modeler has in mind, than from nothing. Convergence to an acceptable version can start
immediately, by a series of acceptations, rejections and changes based on an actual solution.
14.1 THE PERIODICITY
It should be quarterly, annual if quarterly data is not available. This is not so important for population, or for
stock variables: capital, housing, credits and debts.
14.2 GENERAL ELEMENTS
If several series are available for the same concept (as unemployment) the first priority is to get one series, the
second to get all available series with their definition. Of course, if several sets of series are available for the
same domain (like exports and imports in both local currency and dollars) the priority is to make the whole set
as consistent as possible. This means for instance that being part of the major set of data used can be
considered more important than perfect adaptation to the concept required.
We are not considering product decomposition. But oil could be separated in terms of production, exports,
and relation to the State budget.
The agent decomposition is:
Households
Firms (private and public)
Financial institutions (non-state)
State (including local administrations and Social Security)
Rest of the World.
14.3 THE SERIES
The most important series are underlined.
Some of the variables can be computed from others. They can be provided or not.
We need
14.3.1 THE SUPPLY - DEMAND VARIABLES AT CONSTANT PRICES (WHOLE ECONOMY)
Total local demand (including intermediary consumption)
Intermediary consumption by use and product
Total local demand
Household consumption at constant prices
Housing investment by households at constant prices
Housing investment by the State at constant prices
Productive investment by firms at constant prices
Government investment at constant prices
Government consumption at constant prices
Changes in inventories at constant prices.
560
Exports at constant prices (constant terms)
Imports at constant prices (constant terms)
GDP at constant prices
14.3.2 THE SAME ELEMENTS AT CURRENT PRICES
14.3.3 THE SAME ELEMENTS DECOMPOSED INTO PRODUCTS
All elements have an additional dimension (3) except for
intermediate consumption (3 x 3)
productive investment (3 x 3)
14.4 PRODUCTIVE CAPACITY
Productive capital at constant prices
If not, capital depreciation at constant prices
Productive capital (material) at constant prices
Productive capital (building) at constant prices
Rate of use of capacities (ratio)
14.5 EMPLOYMENT
Employment of firms (global)
Employment of households
Employment of financial institutions
Employment of the State (including local administrations)
Self employed
Family workers
Unemployment
Unfilled job vacancies
Weekly work duration
Population
Population in age of working (or 15-65 years)
14.6 PRICE DEFLATORS
The deflators associated to the supply - demand equilibrium (but we can compute them ourselves if we have
elements at both current and constant prices).
The consumer price index
The production price index
The yearly wages (global)
The yearly wages (firms)
The yearly wages (State)
561
The yearly wages (employed by households)
The yearly wages (by product)
14.7 HOUSEHOLDS ACCOUNT
All in current terms
Revenue before income tax
Revenue after income tax
Revenue from wages
Individual workers revenue
Pensions
Social security benefits (global)
Social security benefits (illness).
Social security benefits (family).
Revenue from housing
Interests received.
Income tax
Social security contributions
Interests paid.
Housing capital
14.8 FIRMS ACCOUNT (ALL TYPES OF FIRMS)
All elements in current terms
Value added
Intermediary consumption
Tax on production
Wages
Social security contributions paid by firms
Revenue of individual workers
Dividends paid
Dividends received
Interests paid
Interests received
Tax on profits
Productive investment
Change in inventories
Balance
14.9 REST OF THE WORLD
Trade balance
Financing capacity
Balance of services
562
Exchange rate to the dollar
Exchange rate to the partners of the country.
Balance of payments
Capital movements
Wage transfers of the country’s expatriate workers.
Demand in the Rest of the World (weighted by the share of countries in local exports).
14.10 GOVERNMENT ACCOUNT (ELEMENTS NOT DESCRIBED EARLIER)
All elements in current terms
VAT
Other indirect taxes
Tariffs
14.11 FINANCIAL DOMAIN
The following applies to a traditional model. For Stock-Flow Consistent versions, the requirements are much
more important. They are developed in the specific chapter.
Short-term interest rate
Long-term interest rate
Interest rate on consumption
Interest rate on housing investment
All elements in current terms
Stock of debt for firms (global)
Stock of debt for firms (private)
Stock of debt for firms (public)
Stock of debt for the State
Stock of debt for households
Capital movements
Foreign direct investment (in current terms)
Foreign direct investment (in constant terms)
563
15 A LIST OF USEFUL SERIES FOR A MODE L WITH SEVERAL SECTORS
15.1 THE PERIODICITY
Again, it should be quarterly. But this is not so important for population, or for stock variables: capital, housing,
credits and debts.
15.2 GENERAL ELEMENTS
If several series are available for the same concept (as unemployment) the first priority is to get one series, the
second to get all available series with their definition. Of course if several sets of series are available for the
same domain (like exports and imports in both local currency and dollars) the priority is to get one consistent
set.
The agent decomposition is:
Households
Firms (private and public)
Financial institutions (non-state)
State (including local administrations and Social Security)
Rest of the World
In addition to non-market services (separated in the first version) the product decomposition is:
Agriculture, fishing and forestry
Manufacturing
Market services
This decomposition applies to series noted *
In addition, separations should be considered for:
Agricultural production into firms and family
This applies to production, labor, investment, wages, and revenue
Manufacturing trade, into energy and non-energy
This applies to local and world demand, imports and exports at constant and current prices.
15.3 THE SERIES
The most important series are noted by *.
Some of the variables can be computed from others. They can be provided or not.
We need
15.3.1 THE SUPPLY - DEMAND VARIABLES AT CONSTANT PRICES (WHOLE ECONOMY)
* Total local demand (including intermediary consumption)
* Intermediary consumption by use and product
* Total final local demand
* Household consumption at constant prices
* Housing investment by households at constant prices
564
* Housing investment by the State at constant prices
* Productive investment by firms at constant prices
* Government investment at constant prices
* Government consumption at constant prices
* Changes in inventories at constant prices.
* Exports at constant prices (constant terms)
* Imports at constant prices (constant terms)
* GDP at constant prices
15.3.2 THE SAME ELEMENTS AT CURRENT PRICES
All elements have dimension 3 except for
Intermediate consumption (3 x 3)
Productive investment (3 x 3)
Change in inventories (3 x 3)
15.4 PRODUCTIVE CAPACITY
* Productive capital at constant prices
If not, capital depreciation at constant prices
* Productive capital (material) at constant prices
* Productive capital (building) at constant prices.
* Rate of use of capacities (ratio).
15.5 EMPLOYMENT
* Employment of firms (global).
Employment of households
Employment of financial institutions
Employment of the State (including local administrations)
* Wage earners in firms (global).
Employment of households
Employment of financial institutions
Employment of the State (including local administrations)
* Self employed
* Family workers
Unemployment
Unfilled job vacancies
Weekly work duration
Population
Population in age of working (or 15-65 years)
565
15.6 PRICE DEFLATORS
The deflators associated to the supply - demand equilibrium (but we can compute them ourselves if we have
elements at both current and constant prices).
The consumer price index
The production price index
* The yearly wages (global)
* The yearly wages (firms)
The yearly wages (State)
The yearly wages (employed by households)
The yearly wages (by product)
15.7 HOUSEHOLDS ACCOUNT
All in current terms
Revenue before income tax
Revenue after income tax
* Revenue from wages
* Individual workers revenue
Pensions
Social security benefits (global)
Social security benefits (illness).
Social security benefits (family).
Revenue from housing
Interests received.
Income tax
Social security contributions
Interests paid.
Housing capital
15.8 FIRMS ACCOUNT (ALL TYPES OF FIRMS)
All elements in current terms
* Value added
* Intermediary consumption
* Tax on production
* Wages
* Social security contributions paid by firms
* Revenue of individual workers
* Dividends paid
* Dividends received
* Interests paid
566
* Interests received
* Tax on profits
* Productive investment
* Change in inventories
* Balance.
15.9 REST OF THE WORLD
Trade balance
Financing capacity
Balance of services
*Exchange rate to the dollar
Exchange rate to the partners of the country
Balance of payments
Capital movements
*FDI
*Wage transfers of the country’s expatriate workers.
*Demand in the Rest of the World (weighted by the share of countries in local exports).
15.10 GOVERNMENT ACCOUNT (ELEMENTS NOT DESCRIBED EARLIER)
All elements in current terms.
* VAT
* Other indirect taxes
* Tariffs
15.11 FINANCIAL DOMAIN
Short-term interest rate
Long-term interest rate
Interest rate on consumption
Interest rate on housing investment
All elements in current terms
Stock of debt for firms (global)
* Stock of debt for firms (private)
* Stock of debt for firms (public)
Stock of debt for the State
Stock of debt for households
* Capital movements
* Foreign direct investment (in current terms)
* Foreign direct investment (in constant terms)
567
16 RECENT ADDITIONS TO SESSION MANAGEMENT
The following features are not specially used for modelling
16.1 SNAPSHOTS
Snapshots allow you to keep the contents of a workfile as a hidden element, and compare them to the current
items.
Clicking on a given item will give the actual changes (the numerical difference for a series).
Snapshots can be created automatically (with a given frequency) or on request.
Accessing a snapshot, you can have it replace the current workfile.
This feature is described in detail in the manual.
16.2 CLOUD MANAGEMENT
16.2.1 GOOGLE DRIVE
You can easily access a Cloud directory.
For instance for a Google Drive you can use
Open>Foreign Data as Workfile>New Location>
Then authorize and get access to the whole directory.
568
17 USING EVIEWS MENUS TO BUILD AND MANAGE AN ECONOMETRIC MODEL
In this chapter, we shall review the functions allowing to produce and use a model under EViews. To make the
process clearer and shorter, we will use a very small model, with only two equations, one behavioral and one
identity.
We will concentrate here on using menus, not the most efficient way as we have explained, and should be
obvious already. Programming makes the task easier and (more importantly) allows more options. In some
cases, it is the only way to work in a clear, fast and secure way.
But going through the following example, or perhaps his own similar case, one can rain in the different
functions. Then comparing the following example with the same tasks addressed by the program, and keeping
it as a reminder, should make the process easier.
Let us remember also that EViews provides, in the “Capture” window, a sequence of the commands performed
during the session, including the program translation of the menu commands. This means that by saving this
text one can reproduce all the tasks as a program, obviously after some editing.
17.1 GENERAL MENU FEATURES
17.1.1 COMMAND CAPTURE
This new function represents an important improvement in the management of sessions.
As we will see, modelling under EViews relies mostly on programming (except for the research for the best
behavioral formula). However, this process calls very often for menu functions, in particular when one wants to
get information in case of problems or questionings. For instance
• If the numerical profile of a computed series looks strange, one can rely on a graphical display.
• One can consider adding an explanation to an estimated equation, or changing its sample period.
This means that the user is quite frequently switching between running programs and calling for menu
elements.
Before version 9, only the typed commands were retained by EViews (in the Log window). One had to note (by
hand!) the accesses to menu functions.
Now the log contains also the transcription of the calls to menus, translated as commands. This means that the
sequence of statements contains the entire set of function calls.
This set can be saved as a program, which describes all the tasks between two points of a particular session. In
most cases, this program will give the same results as the ones obtained in the session. Of course, this is not
always true, for instance as an element used in the beginning might have been modified later.
An additional good point is that one gets a transcription of menu commands as program elements. The syntax
can be used later in programs, independently. This is very useful as the syntax of some commands (for instance
in graph management or model solving) can be quite complex.
17.1.2 WINDOW MANAGEMENT AND DOCKING
569
EViews 9 provides a large number of features for managing windows. This allows to display one the same
screen several elements, like a spreadsheet, a graph and a program. One particular feature is that the
command window can be displayed anywhere, and can be separated from the Capture window, allowing to
control both the sequence of commands and the sequence of tasks.
The window containing the messages from a program (if any) is also available directly as a clickable item.
17.1.3 SNAPSHOTS
Snapshots allow you to keep the contents of a workfile as a hidden element, and compare them to the current
items.
Clicking on a given item will give the actual changes (the numerical difference for a series).
Snapshots can be created automatically (with a given frequency) or on request.
Accessing a snapshot, you can have it replace the current workfile.
This feature is described in detail in the manual.
17.1.4 LOGGING
This feature is quite interesting for controlling execution of programs.
17.2 A VERY SMALL MODEL
We will base the document on a very small model, admittedly artificial, but nevertheless based on actual data.
We will consider a self-sufficient economy, with no external trade. We will suppose that final demand is divided
into public (called G) and private (called CO169), without separating the latter. This means in particular that we
will not identify investment, capital and therefore supply.
These elements are measured at constant prices.
G will be exogenous, and CO endogenous, defined by a behavioral equation. Total final demand (equal to GDP)
will be called Q.
This leads to the model:
[1] CO = f(Q)
[2] Q = CO + g170
169
Logic would name it C, but this name is reserved by EViews for generic coefficients.
170
When this is possible, we will restrict capital letter to endogenous.
570
17.3 THE DATA
Obviously, these variables cannot correspond to actual series, as any country has an external trade.
We will use World Bank data for France, for GDP and private consumption, which we will identify as demand.
We will suppose that Government demand balances equation 2.
As we will only present basic features, we have not changed the presentation from the previous version of the
book. Obviously, we are not producing an operational model, or even a model using actual concepts.
17.4 CREATING THE WORKFILE
Before taking any step, we need an EViews workfile (just as you need a desk to start working). This file will be
the location where you keep the elements in your project, starting with the data itself.
You will access to the workfile creation window by calling for
File > New> Workfile.
This opens the following menu:
Our data will have a regular quarterly frequency, with observations from 1975 to 2010.
571
We will call our workfile “nano” with a page called “base”
This will create a workfile, which appears as:
We can see that:
• The current page is called “base”
• The (quarterly) data can start from 1975 to 2010.
• Any following statement will apply to the full sample.
• The file contains already two elements:
o c which will contain the generic coefficient vector.
o resid which will contain the residual from the last estimation process.
As we have made no estimation yet, c is filled with zeroes and resid by NAs.
These names cannot be given to another item.
572
17.4.1 PROGRAMMING
The equivalent statement in a program can be:
wfcreate(wf=nano_q,page=base) q 1975q1 2010q4
The only advantage of programming is that the stored statement can be repeated edited.
17.5 READING THE DATA
There are several ways to read the data into EViews, even in some cases from the same data set.
We will import the data from the file “nano_q.xls” by calling for:
Import>
Calling “Open” (or “Ouvrir”) gets you the following:
573
You can check that the data complies with the sample decided at workfile creation for the sheet “base”.
You get two series in columns: FRA_CPV and FRA_GDPV. These series are for private consumption and GDP
respectively (V stands for constant prices). If series were in rows, the “Read series by row” should have been
ticked.
You can finish (“Terminer”).
But you get the message:
If you click “Yes” (we will do it) the series in the “data” page will be linked to their counterpart in the Excel file.
This means that if the Excel file is modified, the change (if any) will apply to the workfile too.
The two series will appear in the page.
574
If we click on “Details +/- we get additional information:
But of course the page will contain less series (not important here).
By double-clicking on a series (fra_cpv for instance) we get the values:
575
Or a graph:
576
17.5.1 PROGRAMMING
The equivalent statement will be :
smpl 1975q1 2010q4
read(t=xls,b2) nano_q.xls
Again, the advantage is that the statement can be repeated, as an element of a program producing several
tasks (perhaps creating a full model).
17.6 CREATING THE PAGE
We will not create the model (and the associated items) in the page “base”. There are two main reasons for
this.
• We do not want to risk modifying the original data.
• We want to limit the number of items to the ones actually used.
The new page will be called “model”. EViews will propose the same characteristics as the original page “data”.
577
Of course it contains only “c” and “resid”.
We must give the page access to the original data.
17.6.1 PROGRAMMING
The programming equivalent will be :
pagecreate(page=base) q 1975q1 2010q4
17.7 LINKING THE DATA
Of course, the simplest option would be to copy the original series in the new page. There is a cleaner and
more efficient solution.
We start by specifying in the Command window:
578
We get:
We can see that the two variables appear with a pink question mark.
Double clicking on CO we get
This means that the values for CO should be obtained in another source, which has not been yet defined.
The second phase will define the source.
And displaying the series will give (along with the values).
17.7.1 PROGRAMMING
The programming equivalent uses four statements
579
link CO
link Q
CO.linkto base::FRA_CPV
Q.linkto base::FRA_GDPV
The advantage of programming begins to appear.
17.8 DISPLAYING THE DATA
17.8.1.1 Object preview
This feature allows to display the characteristics of any element in a workfile, giving much more information
than a spreadsheet or a graph presentation.
This feature allows to display the characteristics of any element in a workfile, giving much more information
than a spreadsheet or a graph presentation.
It is called by selecting an element and right clicking on the second item (below “open”).
For the GDP of our model, we can get:
580
This feature can be applied to a group.
581
17.9 CREATING THE MODEL
We can now proceed to the model. For this we will start with a blank model.
For this, we will use:
Object>New object>Model
And specify “nano” as the name.
582
The resulting model “nano” has no equations.
Now we will enter our equations.
But the equation for CO is not estimated yet.
Still, we want to show our intention through a tentative formula.
This could be represented by:
CO = f(Q).
This introduces two problems.
• First, it will not be accepted by EViews as we call for a non-existent function f.
A simple solution is to modify slightly the formula into:
CO = f*(Q)
583
• But in this case f will be counted as a (missing) exogenous.
To avoid this, we will define f as a scalar in the command window.
If the Command window is not present, you can call it through
Window>Display Command Capture Window
And clicking on the screen image on the left.
Now f and nano are present in the workfile.
Let us go back to the model by clicking on “nano”.
To access the model text, we need to click on “text” and enter the text:
584
Now we can view the model using “View”.
We get the following message:
Of course, we answer “Yes”
We can use the following views.
Equations.
The model has indeed two equations.
Variables.
The model has three variables, two endogenous, one exogenous.
Source text
The original text.
585
Block structure.
The model has a single block, which is simultaneous. This means that with any ordering, the block presents at
least one loop, and all variables in the block participate in this loop. The model solution cannot be obtained in
one pass.
This subject is developed extensively in the book.
17.9.1 PROGRAMMING
Creating the model will use:
if @isobject("nano") then
delete nano
endif
model nano
nano.append Q=CO+g
scalar f=0
nano.append CO=f*(Q)
This option allows repetition, edition and inclusion in a wider program.
Also, can check (without testing himself) if the program exists, and delete it in this case171.
17.10 CHECKING THE MODEL
Of course, without a behavioral formula the model cannot be simulated. But we can still check that:
• The data is present
• The syntax is acceptable.
• The identity holds true.
171
Adding the two equations to the same existing model would create a model with four equations and two
endogenous variables, which would be refused by EViews.
586
Actually the data is not yet fully defined. We still need values for G. For this we can rewrite equation (2) as:
G = Q – CO
The three above conditions are obviously met (a syntax error would produce a message at compiling).
17.11 ESTIMATING CONSUMPTION
17.11.1 A VERY SIMPLE FORMULA
We can start by supposing that consumption shows a constant elasticity to GDP. The growth of household
consumption can follow that of household revenue, itself following the growth of GDP. We will try this, using
the command window.
Of course, we could also use:
Quick> Estimate Equation
Our argument here is that the Command window is easier to edit, in case of successive estimations of the same
concept. Its justification was much stronger until EViews 7, as it was the only way to have access to the full (and
reproductible) sequence of the session statements.
Let us try, limiting the period to the historical values (admittedly our data set is rather old).
587
26.2
26.0
25.8
.02 25.6
.01
25.4
.00
25.2
-.01
-.02
-.03
-.04
76 78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
The obvious problem comes from the autocorrelation of residuals. It can be treated using an autoregressive
process.
smpl 1975q1 2004Q4
ls log(co) log(q) c ar(1)
Please remember that our goal is not to present sophisticated methods but rather a sequence of simple steps.
588
26.2
26.0
25.8
25.6
.015 25.4
.010 25.2
.005
.000
-.005
-.010
-.015
76 78 80 82 84 86 88 90 92 94 96 98 00 02 04
Residual Actual Fitted
The results improve a lot, both in terms of correlation and global quality.
However, a problem remains: if we use this formula in the long run, consumption will grow steadily slower than
GDP, leading to a 100% savings rate…. And introducing an additional trend does not change things.
17.11.2 IMPROVING THE FORMULA
Without resorting to sophisticated elements, we can try to improve the formula both in theory and model
properties, which request consumption to grow at the same speed as revenue (and GDP) in the long-term.
At the same time, theory implies that in the first periods the short-term elasticity is lower than one, as
households take time in adapting to fluctuations of their revenue.
Obviously the only option is to consider two elasticities: one unitary (the long run), one supposedly lower (the
short run).
589
Clearly this calls for an error correction model, the simplest form being:
Dlog(CO) = a . Dlog(Q) + b Log(CO(-1)/Q(-1)) + c
But to be allowed to follow this process, the ratio between CO and Q must be stationary.
This is checked simply by
Displaying the log of the ratio in the Command window:
show log(co/q)
and use “Unit root test” calling for a Dickey Fuller test allowing for a trend (we will limit the
number of lags to 4).
590
The test allows to non-stationarity a very low probability (1.6%).
We can now estimate the error correction equation.
First, to make the trend clearer (and the handling of time elements simpler) we will define a
variable which takes the value of the year in the first quarter, and grows by 0.25 for each
following period.
591
Then we estimate the equation172
The short-term elasticity is 0.77, and the gap is closed at each subsequent period by 20%.
172
Note : We could also have computed the residual from the test, and use it as a lagged element in the
equation.
592
We can now name the equation (as eq_co) and specify a label
To introduce the equation, we must edit the model text, replacing the equation by the name of the
equation, preceded by “:” as:
593
The characteristics do not really change.
• co appears (as lagged) on the right hand side of its equation.
• t is a new exogenous
17.11.2.1 Programming
The above techniques look quite fine as long as one is evaluating potential formulas. But
when a choice has been made, it introduces several problems.
• The formula will be used a large number of times. Obviously, it has to be stored
somewhere.
• But updating the initial version must be possible.
o The data changes through data set extension, new methodology or new field.
o New ideas can show up.
o The equation is used for a new case (a different country.).
• If several equations are estimated, the coefficient vector must be characteristic of
the equation, as well as its name and its residual.
In addition:
• Knowing the name of the dependent variable, one must have access to the equation
elements.
• Looking at the elements, one must be able to associate them to an equation.
This is why we propose the following framework, conditional on the “programming” option.
594
• Name the equation, the coefficient vector and the residual according to the
dependent variable.
• Introduce these elements on the equation.
We will use for the CO equation:
genr ec_co=0
coef(10) c_co
equation eq_co.ls(p) dlog(co)=c_co(1)*dlog(q)+c_co(2)*(log(co(-1)/q(-1))
-c_co(3)*t-c _co(4))+ec_co
genr ec_co=resid
nano.drop co
nano.merge eq_co
One will note that before the estimated equation is merged into the model, one has to
delete it in its text form. This is done simply by dropping the dependent variable from the
model.
However, to make the process clearer and limit mistakes, we suggest creating the model
again from scratch, by deleting it and specifying again the whole set of equations. In
programming mode, the cost is negligible.
17.12 SOLVING THE MODEL
To solve the model, you need to open it (by double clicking) and click on the “solve” item:
595
Let us solve the model over the largest period possible (restricted to 1975Q2 to 2010Q4 as
there is one lag).
We shall perform a solution both deterministic (no random residuals) and dynamic (solving
one period uses the simulated values for the lagged elements).
By clicking on “Edit Scenario Options” we get:
596
We can now access the following menus:
Overrides : we override no variable. This would allow to replace the values of a variable by alternative ones,
stored in a series with the same name but an additional suffix (more on this later).
Excludes : we exclude no variable. This would exclude the endogenous from the solving process, and use its
workfile value (which must be present).
597
Aliasing : This sets the names of the objects created. For now, the important element is the suffix of the
“Deterministic Solved Endogenous” : _0.
A few explanations:
• The method for solving is “Broyden”.
• It works (no error messages).
• It takes less than 5000 iterations to obtain convergence.
• At convergence, no residual is higher than 10-6 (in relative absiolute velue).
598
The number of iterations can be obtained by using « Diagnosis » and clicking on “Display……”
We get:
599
The method needs 4 to 5 iterations to converge.
Looking at the workfile; we see that two new variables have been created: co_0 and q_0.
They contain the results of the simulation (remember that the suffix associated with
“Baseline” is _0).
We can produce a graph for CO (or Q) with actual and simulated values.
We open both series as a group and call for “Graph”
600
The results are rather good until 2004 (the estimation period), but a gap appears afterwards.
17.12.1 PROGRAMMING
With programming, the sequence is simply ;
smpl 1975q2 2010q4
solve(d=d) nano
17.13 A SHOCK
Now let us suppose we want to see what happens if an assumption changes, for instance if
we increase Government demand by 1% of GDP, starting in 1981.
We need to change:
601
• The Scenario: Baseline does not allow to override variables. We must change it.
• The suffix associated with overridden elements. It will also be used for solution
results.
We will need of course also to create the variable.
In the window “Active” we chose “Scenario 1”
In “Edit Scenario Options” we chose “Aliasing”. We can see that the suffix associated with “Scenario
1” is “_1”.
602
In “Edit Scenario Options” we chose “Override” and, in the window, we specify “g”:
In the “Command” window we create a variable called g_1 increased from g by 1% of GDP
starting in the first quarter of 1981. This is easier controlled by the availability of “t”.
smpl @all
genr g_1=g+0.01*q*(t>=1981)
g_1 appears now in the workfile.
We solve the model again.
603
Model: NANO
Date: 06/01/17 Time: 18:40
Sample: 1975Q2 2010Q4
Solve Options:
Dynamic-Deterministic Simulation
Solver: Broyden
Max iterations = 5000, Convergence = 1e-08
Scenario: Scenario 1
Solve begin 18:40:32
Solve complete 18:40:32
Now we can produce a graph for the difference.
604
We will leave the interpretation for later, when we solve the model over the future, its natural field.
17.13.1 PROGRAMMING
Programming a shock is easier. We shall use:
smpl 1975Q2 2010Q4
nano.append assign @all _b
solve(d=d) nano
genr g_v=g+q*.01*(t>=1980)
605
nano.scenario "scenario 1"
nano.append assign @all _v
nano.override g
solve(d=d) nano
To produce a graph, we will create explicit variables.
For this, we will use a loop. The statements between “for” and “next” will be executed for each value after the
“for”, replacing in turn the “%1” by the value, and dropping the brackets.
for %1 CO Q
series dv_{%1}={%1}_v-{%1}_b
series pv_{%1}=100*dv_{%1}/({%1}_b+({%1}_b=0))
next
graph pv_co pv_q
17.14 FORECASTING
We shall now use the model over the period for which it has been created: the future
17.14.1 CREATING A PAGE
We need a new page with the future periods. We will call it “forecast” and extend the period
to 2100.
We will copy the « model » page first , as « forecast ».
We shall use :
Proc>Copy/Extract to current page>By value to New Page or Workfile
606
Then we change the periods in the page using:
Proc>Structure/Resize Current Page
607
17.14.1.1 Programming
We shall use
pagecopy(page=forecast)
pagestruct(end=2100Q4) *
17.14.2 THE ASSUMPTIONS
The model will compute the endogenous on the future. But for this it needs values for the
exogenous variables.
Our model contains two exogenous : Government Demand g and time t.
We will suppose that g grows by 0.5% at each quarter (roughly 2% per year) after 2010Q4.
608
17.14.2.1 Programming
In a program, we will simply use:
smpl 2010Q1 2100Q4
genr t=t(-1)+0.25
genr g=g(-1)*1.005
Note that we must also forecast the residual for the consumption equation. As it applies to a
Log, it works as a multiplicative factor. We will keep it constant over time.
genr ec_co=ec_co(-1)
17.14.3 BASIC SIMULATION
Simulating the model follows the same statements as before.
609
The model solves without problem.
Model: NANO
Date: 06/02/17 Time: 13:07
Sample: 2005Q1 2100Q4
Solve Options:
Dynamic-Deterministic Simulation
Solver: Broyden
Max iterations = 5000, Convergence = 1e-08
Scenario: Scenario 1
Solve begin 13:07:22
2005Q1 Convergence after 4 iterations
2005Q2 Convergence after 4 iterations
2005Q3 Convergence after 4 iterations
2005Q4 Convergence after 4 iterations
2006Q1 Convergence after 4 iterations
2006Q2 Convergence after 5 iterations
2006Q3 Convergence after 5 iterations
2006Q4 Convergence after 5 iterations
--------------------------------------------------------------
2099Q1 Convergence after 4 iterations
2099Q2 Convergence after 4 iterations
2099Q3 Convergence after 4 iterations
610
2099Q4 Convergence after 4 iterations
2100Q1 Convergence after 4 iterations
2100Q2 Convergence after 4 iterations
2100Q3 Convergence after 4 iterations
2100Q4 Convergence after 4 iterations
Solve complete 13:07:24
17.14.3.1 Programming
smpl 2005Q1 2100Q4
nano.scenario "Baseline"
nano.override
solve(d=d) nano
17.14.4 A SHOCK
We will make the same shock as before.
We shock again Government demand by 1% of the baseline GDP, starting in 2011 this time.
611
612
613
We can add the relative variation of g.
We edit the graph somewhat.
614
615
17.14.4.1 Programming
Our statements will be similar to the previous case.
smpl 2005Q2 2100Q4
616
nano.scenario "scenario 1"
genr g_v=g+q_p*.01*(t>=2011)
nano.append assign @all _v
nano.override g
solve(d=d) nano
smpl 2005Q2 2100Q4
for %1 CO Q
series dv_{%1}={%1}_v-{%1}_0
series pv_{%1}=100*dv_{%1}/({%1}_0 +({%1}_0=0))
next
17.15 A FULL PROGRAM
The sequence of programming statements we have used can be collected into a full
program, which can contain comments explaining each step in the process.
We can also use elements which are not so easy to specify by menu (especially if typing
them must be repeated). For this reason, the program we propose is a little more
complicated. The differences are:
17.15.1 A RESIDUAL IN THE ESTIMATED EQUATION
In the consumption equation, we shall introduce a residual call ec_co. It will be set to zero
before estimation, and to the observed residual after it is performed.
Note that we have blocked the time trend after the estimation period, to allow the solution
to follow a steady state path (Q and CO will grow at the same rate in the long run).
' The name of the residual connects it with the variable
genr ec_co=0
' We create a vector of coefficients
' The name of the vector also connects it with the variable
coef(10) c_co
' We specify the equation using coefficient and residual
' The name of the equation again connects it with the variable
' The equation is estimated immediately using least (not necessarily ordinary) squares
equation eq_co.ls(p) dlog(co)=c_co(1)*dlog(q)+c_co(2)*(log(co(-1)/q(-1))-c_co(3)*(t-2004)*(t<2004)-
c_co(4))+ec_co
' The residual is given the value of the left side - right side
genr ec_co=resid
617
17.15.2 A RESIDUAL CHECK
We will check that the residual takes a zero value for all equations.
nano.scenario "scenario 1"
nano.append assign @all _s
' We compute a zero residual with the suffix "_s"
genr ec_co_s=0
' We declare that ec_co must use an alternate value
' The suffixed variable must override the base one
nano.override ec_co
solve(d=d) nano
' We compute the difference between simulation and history
for %1 CO Q
series ds_{%1}={%1}-{%1}_s
series ps_{%1}=100*ds_{%1}/({%1}+({%1}=0))
next
17.15.3 COMPUTING BASE FUTURE VALUES FOR THE ENDOGENOUS
We also initialize endogenous variables using the same assumption. This will give us an
alternate starting point and allow to check if the solution deviates from the theoretical
value.
As q and co are linked to the values in the model page (which are in turn linked to FRA_GDPV
and FRA_CPV) we must “unlink” them to allow changes.
unlink q
unlink co
genr q=q(-1)*1.005
genr co=co(-1)*1.005
17.15.4 A BASELINE SOLUTION USING A SCENARIO
smpl 2005Q1 2100Q4
nano.scenario "scenario 1"
nano.append assign @all _p
nano.override
solve(d=d) nano
17.15.5 PRODUCING A SHOCK FROM SCENARIO
618
smpl 2005Q2 2100Q4
genr g_v=g+q_p*.01*(t>=2011)
nano.append assign @all _v
nano.override g
solve (d=d) nano
17.15.6 COMPUTING DIFFERENCES USING A LOOP
smpl 2005Q2 2100Q4
for %1 CO Q
series dv_{%1}={%1}_v-{%1}_p
series pv_{%1}=100*dv_{%1}/({%1}_p+({%1}_p=0))
next
17.15.7 THE PROGRAM
'
' This program describes the functions associated to the building of a macroeconometric model under
EViews
' Although the model is extremely simple, all the main functions are treated
' ----------- The model
'
' (1) Private consumption CO is a function of GDP : CO=f(Q). This function will be estimated
' (2) GDP is the sum of private and public consumption g : Q=CO+g. This equation is supposed to be
exact (or at least consistent with the data)
'------------- The program
' First we set the directory
' This is useful if we manage several projects at the same time
' This guarantees that programs will access the right files
' files will be stored in the right place
' no file outside the directory will be wrongly affected
'
cd "c:\users\jean louis brillet\google drive\book_8\nano"
' We direct the output to nano.rtf
' This means that all "printed" output (mostly the results of tests for which the "p" option has been
specified)
' will go to the specified file
' r calls for the RTF format (the richer, read by Word and the like). "t" would call for character text.
' o specifies that any previous results stored under the same name will be overriden
619
output(r,o) nano
'
' We create the work file nano_q (quarterly from 1975q1 to 2010q4) with a page called "origin"
' This page will contain the original series
' If the workfile exists, it will be open
' But if it exists and is open, a new version will be open in a new window
' This is obviously dangerous, to avoid it we ask any open version to be closed
' ***** Creating the framework and the data
close nano_q.wf1
wfcreate(wf=nano_q,page=base) q 1975q1 2010q4
' We also destroy any existing element (to start from a blank file)
delete *
'
' We read 2 series of the French economy (OECD data) from nano_q.xls
' The data is organized as a table with series in colums, and its upper left cell is B2
' (second row, second colum)
'
smpl 1975q1 2010q4
read(t=xls,b2) nano_q.xls 2
' Now we create a page for the model elements
pagecreate(page=model) q 1975q1 2010q4
' We start creating the data
'
' We create a time trend : the yearly date for the first quarter, then +0.25 per quarter
smpl 1975q1 1975q1
genr T=1975
smpl 1975q2 2010q4
genr T=t(-1)+0.25
'
' We create the model data from two orginal series :
' CO from FRA_CPV : private consumption at 1995 prices
' Q from FRA_GDPV : gross domestic product at 1995 prices
' g as the difference between Q and CO
smpl 1975q1 2010q4
' We will create CO and Q by reference to original elements
' First we define them as "linked" variables
link CO
link Q
' Then we create the link giving the reference to page and the series
CO.linkto base::FRA_CPV
Q.linkto base::FRA_GDPV
' g is not available as such. We have to use the equation reversed
620
genr g=Q-CO
' We create a first version of the model
' just to check its structure
' and the fact that it can be estilated
'
' We ask EViews to destroy any existing version
' to start from scratch
' We do it only if the model exists
' (this avoids an error message)
'
if @isobject("nano") then
delete nano
endif
' We create a new model called "nano"
model nano
' We add the first equation (identity)
nano.append Q=CO+g
' For the second, we just declare our intention
' We would like to write CO=f(Q)
' but EViews will call for the function "f"
' So we have to trick EViews into accepting the syntax
' This is the closest we have found
' f is defined as a scalar to avoid its identification as a (spurious) exogenous
scalar f=0
nano.append CO=f*(Q)
'
' We produce a residual check
' The model is broken into single equation ones
' and each model is solved based on actual values
'
' We use the largest period for which CO, Q and g are all available
' equivalent to the availability of their product
smpl 1975q1 2004q4 if co*q*g<>na
' We ask the results to be given the suffix "_c"
nano.append assign @all _c
' We solve the model using the option "f"
solve(d=f) nano
' We check that the identity holds true
' by computing the difference between actual and computed
' We also know that the behavioral equation can be estimated
genr dc_q=q_c-q
genr pc_q=dc_q*100/q
621
' ****** Estimating the equation
' We estimate the equation
' First, a very simple case (linear in logs)
smpl 1975Q1 2004q4
ls(p) log(co) log(q) c
' Equivalent to
ls(p) log(co)=c(1)*log(q)+c(2)
' The first two positions of the "c" vector will be filled with estimated values
' as well as the "Resid" series with the residual (actual left hand side - estimated right hand side)
' second, with an autoregressive term
ls(p) log(co) log(q) c ar(1)
' Now, a more sophisticated form using error correction
' We suppose that the long-term elasticity between CO and Q is unitary
' but we leave it free in the short run
' First, we test the stationarity of Log(CO/Q)
uroot(trend,p ) log(co/q)
uroot(trend,p,pp) log(co/q)
' It works (the negative T statistic is high enough)
' Now two techniques
' ***** The fast one (good for the exploratory phase)
ls(p) dlog(co) dlog(q) log(co(-1)/q(-1)) t c
' problem : estimates will be erased by the next estimation
' and where is the equation?
' a little better : the equation is stored
equation eq1.ls(p) dlog(co) dlog(q) log(co(-1)/q(-1)) c t
' ****** the slow one (good once the choice is made)
' We create a residual, with 0 as the initial value
' The name of the residual connects it with the variable
genr ec_co=0
' We create a vector of coefficients
' The name of the vector also connects it with the variable
coef(10) c_co
' We specify the equation using coefficient and residual
' The name of the equation again connects it with the variable
' The equation is estimated immediately using least (not necessarily ordinary) squares
' To delay estimation, one must drop the ".ls"
equation eq_co.ls(p) dlog(co)=c_co(1)*dlog(q)+c_co(2)*(log(co(-1)/q(-1))-c_co(3)*(t-2004)*(t<2004)-
c_co(4))+ec_co
' The residual is given the value of the left side - right side
genr ec_co=resid
' We also compute the difference between the estimated and observed variables
622
genr co_h=exp(c_co(1)*dlog(q)+c_co(2)*log(co(-1)/q(-1))+c_co(3)*t+c_co(4))
genr dh_co=co-co_h
genr ph_co=dh_co*100/co
' **** creating the model
' Now we have our model
' We delete the old version
delete nano
' We create a new blank one
model nano
' We introduce the same identity as before
nano.append Q=CO+g
' And we merge the estimated equation
' "merging" also includes statistics
nano.merge eq_co
' Now we have to check model consistency again
' We shall use the suffix "_d"
nano.append assign @all _d
solve(d=f) nano
' We will use a loop (first form)
' It starts with "for" followed by a parameter and a list, and ends with "next"
' The statements in between will be repeated
' as many times as there are elements in the "for" list
' each time the parameter is replaced by the current element
' The brackets are just for locating the parameter
' they are formally dropped after the replacement
' A trick: if a variable is naturally zero (like a contemplated tax with a present zero rate) the error will
be zero too
' Computing the relative error in EViews will divide 0 by 0 generating an error, with a message
' For us, 0/0
for %1 CO Q
genr dd_{%1}={%1}-{%1}_d
genr pd_{%1}=100*dd_{%1}/({%1}+({%1}=0))
next
' Solving the model
' Assessing the reliability over the past
' Gives an idea of the reliability of forecasts
' In the future, we will not know the residuals
' So we set the residual at its most probable value : zero
' ****** The "old" way
' We store the residual value under an alternate name
genr ec_co_h=ec_co
623
' We set the base value to zero
genr ec_co=0
' We solve the model with the suffix "_s"
nano.append assign @all _s
solve(d=d) nano
' We recover the estimation residual
genr ec_co=ec_co_h
' This method is tedious, and runs the risk of losing the residual
' The "scenario" way
' We decide on a scenario, with the suffix "_s"
nano.scenario "scenario 1"
nano.append assign @all _s
' We compute a zero residual with the suffix "_s"
genr ec_co_s=0
' We declare that ec_co must use an alternate value
' The suffixed variable must override the base one
nano.override ec_co
solve(d=d) nano
' We compute the difference between simulation and history
for %1 CO Q
series ds_{%1}={%1}-{%1}_s
series ps_{%1}=100*ds_{%1}/({%1}+({%1}=0))
next
' Now we test the economic properties of the model (!)
' by measuring the response of the solution to a shock on assumptions
smpl 1964Q1 2004Q4
' We still use "Scenario 1"
' We must drop any current overriding (in our case ec_co)
' This is done by declaring a blank list
nano.override
' The base solution (should take the historical values)
smpl 1975Q2 2004Q4
nano.append assign @all _b
solve(d=d) nano
' We create an alternate g, increased by one point of GDP
' starting in 1980, to check that with a zero shock we get the base solution
genr g_v=g+q*.01*(t>=1981)
nano.scenario "scenario 1"
' Now the suffix is "_v"
nano.append assign @all _v
' We override g
624
nano.override g
solve(d=d) nano
' We compute the difference between "_v" and "_b"
smpl 1979q1 2004q4
for %1 CO Q
series dv_{%1}={%1}_v-{%1}_b
series pv_{%1}=100*dv_{%1}/({%1}_b+({%1}_b=0))
next
' Now we work on the future
' We copy the page under a new name
pagecopy(page=forecast)
' We extend the period to 2100
pagestruct(end=2100Q4) *
' We extend the assumptions
' First, the time trend!
smpl 2005Q1 2100Q4
genr t=t(-1)+0.25
' The residual is kept at its last value
smpl 2005Q1 2100Q4
genr ec_co=ec_co(-1)
' g will grow at 0.5% per quarter
genr g=g(-1)*1.005
' We also initialize endogenous variables using the same assumption
' This will give us an alternate starting point
' and allow to check if the solution deviates from the theoretical value
unlink q
unlink co
genr q=q(-1)*1.005
genr co=co(-1)*1.005
' We solve the model with the suffix "_p"
' making sure that no overriden value is called
smpl 2005Q2 2100Q4
nano.scenario "scenario 1"
nano.append assign @all _p
nano.override
solve(d=d) nano
' Again, we produce a shock, starting in 2006
genr g_v=g+q_p*.01*(t>=2006)
nano.append assign @all _v
nano.override g
solve(d=d) nano
' and we compute the deviations from the base
625
smpl 2005q1 2100Q4
for %1 CO Q
series dv_{%1}={%1}_v-{%1}_p
series pv_{%1}=100*dv_{%1}/({%1}_p+({%1}_p=0))
next
series pv_g=q_p/g*(t>=2006)
pageselect forecast
save nano_q
626
627
628
18 INDEX
access (to data), 70, 78, 79, 86, 343 convexity of models, 40
aggregation, 57, 79, 80, 81 cumulated shock, 280
algorithm, 40, 47, 53, 81, 215, 216, 219, 222, 224, 228, cyclic dynamics, 318
231, 232, 233, 240, 241, 242, 266, 335, 337
damping factors, 224, 235, 236, 238, 239, 254, 255, 261,
analytic (shocks), 271, 280, 336, 337 424
autocorrelation, 149, 150, 171, 184, 185, 199, 200, 203, data correction, 44, 270
269, 588
data definition, 83, 344
average (solution), 312, 313
data format, 70, 77, 78, 86
average absolute error, 267
data gathering, 53, 69
behavioral equations, 21, 24, 25, 51, 134, 270, 311, 336
Deleau and Malgrange, 266, 321
bias, 267, 312
deterministic (solution), 312
block-recursive, 228
differentiability of models, 36
blocks (model), 228
dimension, 47
bootstrap, 312
direct access to data, 32, 78
Broyden (algorithm), 219, 220, 228, 243, 246, 259, 306
disaggregation, 79, 80
CGE, 15, 16, 57, 286
discretization, 28
complex shocks, 52, 337
divergent dynamics, 235, 264, 266, 318, 319, 321, 324
Computable General Equilibrium, 14, 15
documentation of data, 83
constants with dimension, 47
documentation of models, 62
continuous time models, 32
dummy variables, 28, 29, 159, 161, 181, 270, 390, 405,
convergence (study of), 219, 221, 223, 224, 228, 231, 406, 407
232, 235, 239, 242, 266, 319, 321
Durbin - Watson statistic, 150
convergence criterion, 221, 223, 268
dynamic models, 29
convergence of the solution, 49, 221, 223, 226, 231, 238,
240, 242, 266, 319, 321 dynamic simulations, 58, 267, 268, 271, 320
convergence problems, 49, 221, 226, 228, 233, 265, 266 economic policy, 21, 52, 268, 334, 337, 338, 343
convergent, 234, 242, 318 economic policy (shock), 52, 337, 338, 343
629
efficiency, 313 Gauss-Seidel, 47, 215, 216, 221, 223, 224, 230, 231, 233,
235, 242, 243, 247, 250, 255, 257, 259, 260, 261, 262,
eigenvalues of the Jacobian, 232, 240, 242, 266, 313, 264, 293, 306, 437, 445, 504
317, 318, 320, 321, 328
Gauss-Seidel (algorithm), 228, 231, 232, 236, 240, 241,
eigenvectors of the Jacobian, 240, 318 242, 243, 252, 258, 261, 264
elimination, 32, 266 Gilli-Rossier method, 228
endogenous (variables), 40, 224 global compatibility (specifications), 40
error (estimation), 26, 267, 270, 271, 272, 312, 336 graphs, 83, 267, 343, 344, 345
error (simulation), 26, 133, 134, 135, 210, 223, 249, 264, histograms, 245, 344
265, 267, 268, 269, 270, 271, 272, 311, 312, 313, 335,
336 homogeneity, 46
error correction, 61, 150, 152, 156, 163, 169, 170, 171, horizon, 58, 59, 267
214, 265, 272, 279, 288, 326, 328, 330, 356, 403, 417,
423, 426, 427, 436, 472, 490, 511, 590, 591, 622 horizontal (aggregation), 57
error correction models, 328, 330 Households, 18, 83, 84, 104, 110, 116, 234, 319, 352,
361, 363, 396, 476, 480, 481, 482, 513, 514, 515, 518,
estimated coefficients, 25, 135, 210, 211, 233, 265, 267, 520, 522, 524, 525, 526, 531, 560, 562, 564, 566
270, 311, 312
identifiability, 49
estimation, 24, 25, 26, 32, 51, 62, 135, 146, 267, 269,
271, 272 identified (formulations), 219
estimation methods, 32, 54, 271 identities, 22, 51, 272
exogeneity, 264 ILO, 87, 88, 95
exogenous (variables), 36, 44, 45, 46, 51, 57, 67, 108, incidence matrix, 224, 225, 227
116, 287, 291, 292, 334, 338, 443, 444, 502, 503, 520,
527
information, 83, 343, 344
explained (variable), 25, 147
instruments, 21, 313, 334, 335, 338, 342, 343
explanatory (variable), 23, 227, 270, 271
internal (methods), 55
ex-post (simulations), 267, 272
International Monetary Fund, 72, 87, 88, 333, 334
external methods, 55
interpretation, 56, 345
external shocks, 338, 344
iterations (simulations), 215, 216, 218, 219, 221, 222,
224, 226, 227, 231, 232, 234, 236, 237, 240, 241, 242,
field (models), 56, 57, 62 262, 264, 541, 599
finite differences, 218, 240, 241, 242 Jacobian matrix, 218, 219, 228, 232, 233, 237, 240, 241,
242, 312, 323, 337, 539
forecasts, 339, 424, 500
lagged elements, 15, 30, 31, 32, 62, 160, 196, 228, 267,
formal transformations, 81 271, 303, 306
630
lags, 320, 321 non - linearity (equations), 34
lengthening (of data), 82 normalization (equations), 47, 49
linearity (equations), 46 normalized (equations), 210, 225, 268
linearity (estimation), 147 normative (scenarios), 334, 338
linearity (models), 33, 35, 37, 54, 232, 237, 240, 242 OECD, 87
linearization (Newton), 217, 241 one-time shocks, 280
linearized (models), 317 optimal control, 342
Long-term, 58 organization of the modelling process, 61, 62, 85
loop (variables), 227, 228, 231, 233, 236, 240, 241, 242, override, 275, 276, 277, 278, 283, 296, 297, 437, 440,
266 441, 543, 597, 602, 606, 611, 617, 618, 624, 625
maintained (shock), 280, 340 override statement, 275
means, 44, 216 parameters, 21, 24, 36, 147, 334, 342
medium-term, 15, 16, 57, 272, 273, 285, 286, 287, 291, periodicity, 27, 28, 33, 58, 79, 81, 82
294, 317, 337, 338, 340, 447
physical units, 69
Medium-term, 58
pies, 344
methods of analysis, 311, 317, 321, 328
preliminary processing of data, 79
minimal (set of loop variables), 228
presentation of results, 267, 343, 344
minimum (set of loop variables), 228
processing (of data), 79
model analysis, 33, 55, 57, 58, 280, 311, 313, 317, 321,
328 processing of problems, 53, 85, 248
model continuity, 36, 37 Quasi-Accounting, 14, 15, 17, 57, 243, 501, 504, 505, 512
model dynamics (analysis of), 31, 318, 321, 328, 330 R square (test), 147
Monte Carlo, 311 random term, 25, 26, 311
national accounts, 53, 56, 59, 78, 359, 497, 501, 504 rational expectations, 32, 281, 305, 306, 536
National Accounts, 82 recursivity, 226
Nepomiaschy-Ravelli, 228 residual (estimation), 26, 135, 312, 336
Newton (algorithm), 216, 219, 221, 223, 228, 236, 237, residual check, 132, 134, 135, 249
240, 241, 242, 243, 247, 262, 264, 293, 306, 504, 539
resolution, 33, 215, 228, 335
nomenclature, 62, 81, 343
Ritz-Jordan (algorithm), 216, 264
non - linear (model), 312
631
scenarios, 52, 58, 333, 334, 338 suppression (of data), 82
scientific applications of models, 55, 311, 342 tables, 83, 343, 344
seasonality, 28, 29, 81 temporal (aspect), 70
shocks, 46, 52, 54, 271, 280, 318, 320, 324, 329, 333, term (short, medium, long), 20
336, 337, 338, 343, 344
tests (estimation), 28, 148, 270
Short-term, 58
theoretical models, 32, 54, 59, 342
simulation error due to coefficients, 273, 311, 312
time dimension, 27
simulation error due to residuals, 134, 311, 336
trend (estimates), 148, 269
simulations, 58, 266, 267, 268, 271, 313, 334, 336, 344
trend forecasts, 333
size of models, 31, 44, 53, 54, 56, 62, 228, 241, 242, 266,
281, 342 types of error, 26, 134, 270, 271, 311, 312, 313
smoothing (of data), 81 types of model, 52
software aspects, 77, 78, 83, 85, 228, 344 uncertainty, 311, 312, 313
specification (of the model), 51, 135, 146 uniqueness of the solution, 39
specification constraints, 51 United Nations, 99
stability (dynamics), 241, 318, 324 units (models), 268
standard error (estimation), 26, 313 update (of data), 83, 85
standard error (simulation), 267, 313 use (models), 52, 58, 241
static (models), 29 validation of the model, 52, 266, 342
static (simulations), 267 variables, 19, 216, 268, 271
steady state, 214, 286, 296, 324, 329, 424, 617 vertical (aggregation), 57
steady state paths, 318, 329 World bank, 87
Stock-Flow Consistent, 14, 15, 16, 17, 350, 368, 466, 467, World Bank, 71, 73, 86, 87
503, 512, 563
X13 ARIMA, 28
632
You might also like
- Management Science The Art of Spreadsheet ModelingNo ratings yetManagement Science The Art of Spreadsheet Modeling5 pages
- The Gamification of Learning and Instruction Kapp en 19782100% (1)The Gamification of Learning and Instruction Kapp en 197827 pages
- Ghysels, Eric - Marcellino, Massimiliano - Applied Economic Forecasting Using Time Series Methods-Oxford University Press (2018)No ratings yetGhysels, Eric - Marcellino, Massimiliano - Applied Economic Forecasting Using Time Series Methods-Oxford University Press (2018)617 pages
- ProdEconRIntroduction To Econometric Production Analysis With R (Draft Version)100% (1)ProdEconRIntroduction To Econometric Production Analysis With R (Draft Version)277 pages
- Solutions To Selected Problems in Machine Learning: An Algorithmic Perspective0% (1)Solutions To Selected Problems in Machine Learning: An Algorithmic Perspective4 pages
- Sustainable Preparation of Metal NanoparticlesNo ratings yetSustainable Preparation of Metal Nanoparticles243 pages
- Structural Econometric Modelling Methodology (Eviews)100% (1)Structural Econometric Modelling Methodology (Eviews)497 pages
- The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees instant download100% (1)The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees instant download70 pages
- Applied economic forecasting using time series methods Ghysels All Chapters Instant Download100% (2)Applied economic forecasting using time series methods Ghysels All Chapters Instant Download66 pages
- Applied economic forecasting using time series methods Ghysels - The latest updated ebook is now available for download100% (1)Applied economic forecasting using time series methods Ghysels - The latest updated ebook is now available for download55 pages
- The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees download100% (2)The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees download54 pages
- 2015 Bookmatter MathematicalMethodsAndModelsInNo ratings yet2015 Bookmatter MathematicalMethodsAndModelsIn5 pages
- Fundamentals of Drilling Engineering: MCQs and Workout Examples for Beginners and EngineersFrom EverandFundamentals of Drilling Engineering: MCQs and Workout Examples for Beginners and Engineers5/5 (2)
- Full download Introductory Econometrics for Finance (R Guide) Wichmann pdf docx100% (17)Full download Introductory Econometrics for Finance (R Guide) Wichmann pdf docx55 pages
- ResearchMethods - 2015-Research Methods For Economics and Related Studies PDFNo ratings yetResearchMethods - 2015-Research Methods For Economics and Related Studies PDF186 pages
- Econometric Model Selection A New Approach (Antonio Aznar Grasa (Auth.) ) (Z-Library)No ratings yetEconometric Model Selection A New Approach (Antonio Aznar Grasa (Auth.) ) (Z-Library)264 pages
- Introductory Econometrics for Finance (R Guide) Wichmann instant download100% (1)Introductory Econometrics for Finance (R Guide) Wichmann instant download56 pages
- Introductory Econometrics for Finance (R Guide) Wichmann - Quickly download the ebook to never miss any content100% (2)Introductory Econometrics for Finance (R Guide) Wichmann - Quickly download the ebook to never miss any content68 pages
- Instant Access to (Ebook) The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods by Michael Rees ebook Full Chapters100% (4)Instant Access to (Ebook) The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods by Michael Rees ebook Full Chapters81 pages
- financial-modeling-and-forecasting-notesNo ratings yetfinancial-modeling-and-forecasting-notes90 pages
- Dynamic Macroeconomic Modeling With MatlabNo ratings yetDynamic Macroeconomic Modeling With Matlab38 pages
- Smoothness Priors Analysis of Economic and Financial Time SeriesNo ratings yetSmoothness Priors Analysis of Economic and Financial Time Series193 pages
- (eBook PDF) Principles of Econometrics, 5th Editioninstant download100% (4)(eBook PDF) Principles of Econometrics, 5th Editioninstant download49 pages
- Instant Download The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees PDF All Chapters100% (5)Instant Download The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees PDF All Chapters66 pages
- Econometrics - Theory and Applications With EViews - 2005 (Vogelvang) PDF100% (17)Econometrics - Theory and Applications With EViews - 2005 (Vogelvang) PDF379 pages
- Engineering Optimization: An Introduction with Metaheuristic ApplicationsFrom EverandEngineering Optimization: An Introduction with Metaheuristic ApplicationsNo ratings yet
- Adaptive Approximation Based Control: Unifying Neural, Fuzzy and Traditional Adaptive Approximation ApproachesFrom EverandAdaptive Approximation Based Control: Unifying Neural, Fuzzy and Traditional Adaptive Approximation Approaches5/5 (1)
- Translation Quality Assessment: An Argumentation-Centred ApproachFrom EverandTranslation Quality Assessment: An Argumentation-Centred ApproachNo ratings yet
- Quantitative Finance: Its Development, Mathematical Foundations, and Current ScopeFrom EverandQuantitative Finance: Its Development, Mathematical Foundations, and Current ScopeNo ratings yet
- Guidelines for Investigating Chemical Process IncidentsFrom EverandGuidelines for Investigating Chemical Process IncidentsNo ratings yet
- The Mathematics of Derivatives: Tools for Designing Numerical AlgorithmsFrom EverandThe Mathematics of Derivatives: Tools for Designing Numerical Algorithms3/5 (1)
- Guidelines for Chemical Process Quantitative Risk AnalysisFrom EverandGuidelines for Chemical Process Quantitative Risk Analysis5/5 (1)
- iPhone SDK 3 Programming: Advanced Mobile Development for Apple iPhone and iPod touchFrom EverandiPhone SDK 3 Programming: Advanced Mobile Development for Apple iPhone and iPod touchNo ratings yet
- Mary's Scientific Knowledge : Luca MalatestiNo ratings yetMary's Scientific Knowledge : Luca Malatesti23 pages
- Good Day, Grade 9!: "Believe That You Can and You'Re Halfway There." - Theodore RooseveltNo ratings yetGood Day, Grade 9!: "Believe That You Can and You'Re Halfway There." - Theodore Roosevelt25 pages
- Nour M. Khattar Civil Engineer Master of Science in Civil EngineeringNo ratings yetNour M. Khattar Civil Engineer Master of Science in Civil Engineering2 pages
- Hco Document Naming Procedures: (2 Character Max.) (3 Character Max.) (2 Character Max.)No ratings yetHco Document Naming Procedures: (2 Character Max.) (3 Character Max.) (2 Character Max.)1 page
- Thompson, Travis Zeiler, Michael D MacCorquodale, Kenneth - Analysis and Integration of Behavioral UnitsNo ratings yetThompson, Travis Zeiler, Michael D MacCorquodale, Kenneth - Analysis and Integration of Behavioral Units408 pages
- (Part 1) Introduction To Management Challenges For EngineersNo ratings yet(Part 1) Introduction To Management Challenges For Engineers23 pages
- Implementation and Applications of xDSL Technology 1st Edition Philip Golden instant download100% (1)Implementation and Applications of xDSL Technology 1st Edition Philip Golden instant download42 pages
- CIBANO 500 Circuit Breaker Testing PDF A Guide For Performing The Anti 2021 ENUNo ratings yetCIBANO 500 Circuit Breaker Testing PDF A Guide For Performing The Anti 2021 ENU12 pages
- Bài Tập Rút Gọn Mệnh Đề Quan HệNo ratings yetBài Tập Rút Gọn Mệnh Đề Quan Hệ2 pages
- 1st Puc Physics Model Paper 2022-23 by Bvvs Puc VidyagiriNo ratings yet1st Puc Physics Model Paper 2022-23 by Bvvs Puc Vidyagiri17 pages
- Management Science The Art of Spreadsheet ModelingManagement Science The Art of Spreadsheet Modeling
- The Gamification of Learning and Instruction Kapp en 19782The Gamification of Learning and Instruction Kapp en 19782
- Ghysels, Eric - Marcellino, Massimiliano - Applied Economic Forecasting Using Time Series Methods-Oxford University Press (2018)Ghysels, Eric - Marcellino, Massimiliano - Applied Economic Forecasting Using Time Series Methods-Oxford University Press (2018)
- ProdEconRIntroduction To Econometric Production Analysis With R (Draft Version)ProdEconRIntroduction To Econometric Production Analysis With R (Draft Version)
- Solutions To Selected Problems in Machine Learning: An Algorithmic PerspectiveSolutions To Selected Problems in Machine Learning: An Algorithmic Perspective
- Stochastic Claims Reserving Methods in InsuranceFrom EverandStochastic Claims Reserving Methods in Insurance
- Structural Econometric Modelling Methodology (Eviews)Structural Econometric Modelling Methodology (Eviews)
- The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees instant downloadThe Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees instant download
- Applied economic forecasting using time series methods Ghysels All Chapters Instant DownloadApplied economic forecasting using time series methods Ghysels All Chapters Instant Download
- Applied economic forecasting using time series methods Ghysels - The latest updated ebook is now available for downloadApplied economic forecasting using time series methods Ghysels - The latest updated ebook is now available for download
- The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees downloadThe Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees download
- Fundamentals of Drilling Engineering: MCQs and Workout Examples for Beginners and EngineersFrom EverandFundamentals of Drilling Engineering: MCQs and Workout Examples for Beginners and Engineers
- Full download Introductory Econometrics for Finance (R Guide) Wichmann pdf docxFull download Introductory Econometrics for Finance (R Guide) Wichmann pdf docx
- ResearchMethods - 2015-Research Methods For Economics and Related Studies PDFResearchMethods - 2015-Research Methods For Economics and Related Studies PDF
- Econometric Model Selection A New Approach (Antonio Aznar Grasa (Auth.) ) (Z-Library)Econometric Model Selection A New Approach (Antonio Aznar Grasa (Auth.) ) (Z-Library)
- Introductory Econometrics for Finance (R Guide) Wichmann instant downloadIntroductory Econometrics for Finance (R Guide) Wichmann instant download
- Introductory Econometrics for Finance (R Guide) Wichmann - Quickly download the ebook to never miss any contentIntroductory Econometrics for Finance (R Guide) Wichmann - Quickly download the ebook to never miss any content
- Instant Access to (Ebook) The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods by Michael Rees ebook Full ChaptersInstant Access to (Ebook) The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods by Michael Rees ebook Full Chapters
- Smoothness Priors Analysis of Economic and Financial Time SeriesSmoothness Priors Analysis of Economic and Financial Time Series
- (eBook PDF) Principles of Econometrics, 5th Editioninstant download(eBook PDF) Principles of Econometrics, 5th Editioninstant download
- Instant Download The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees PDF All ChaptersInstant Download The Essentials of Financial Modeling in Excel: A Concise Guide to Concepts and Methods Michael Rees PDF All Chapters
- Econometrics - Theory and Applications With EViews - 2005 (Vogelvang) PDFEconometrics - Theory and Applications With EViews - 2005 (Vogelvang) PDF
- Engineering Optimization: An Introduction with Metaheuristic ApplicationsFrom EverandEngineering Optimization: An Introduction with Metaheuristic Applications
- Adaptive Approximation Based Control: Unifying Neural, Fuzzy and Traditional Adaptive Approximation ApproachesFrom EverandAdaptive Approximation Based Control: Unifying Neural, Fuzzy and Traditional Adaptive Approximation Approaches
- Translation Quality Assessment: An Argumentation-Centred ApproachFrom EverandTranslation Quality Assessment: An Argumentation-Centred Approach
- The Science and Engineering of Thermal Spray CoatingsFrom EverandThe Science and Engineering of Thermal Spray Coatings
- Quantitative Finance: Its Development, Mathematical Foundations, and Current ScopeFrom EverandQuantitative Finance: Its Development, Mathematical Foundations, and Current Scope
- Guidelines for Investigating Chemical Process IncidentsFrom EverandGuidelines for Investigating Chemical Process Incidents
- The Mathematics of Derivatives: Tools for Designing Numerical AlgorithmsFrom EverandThe Mathematics of Derivatives: Tools for Designing Numerical Algorithms
- Guidelines for Chemical Process Quantitative Risk AnalysisFrom EverandGuidelines for Chemical Process Quantitative Risk Analysis
- iPhone SDK 3 Programming: Advanced Mobile Development for Apple iPhone and iPod touchFrom EverandiPhone SDK 3 Programming: Advanced Mobile Development for Apple iPhone and iPod touch
- Good Day, Grade 9!: "Believe That You Can and You'Re Halfway There." - Theodore RooseveltGood Day, Grade 9!: "Believe That You Can and You'Re Halfway There." - Theodore Roosevelt
- Nour M. Khattar Civil Engineer Master of Science in Civil EngineeringNour M. Khattar Civil Engineer Master of Science in Civil Engineering
- Hco Document Naming Procedures: (2 Character Max.) (3 Character Max.) (2 Character Max.)Hco Document Naming Procedures: (2 Character Max.) (3 Character Max.) (2 Character Max.)
- Thompson, Travis Zeiler, Michael D MacCorquodale, Kenneth - Analysis and Integration of Behavioral UnitsThompson, Travis Zeiler, Michael D MacCorquodale, Kenneth - Analysis and Integration of Behavioral Units
- (Part 1) Introduction To Management Challenges For Engineers(Part 1) Introduction To Management Challenges For Engineers
- Implementation and Applications of xDSL Technology 1st Edition Philip Golden instant downloadImplementation and Applications of xDSL Technology 1st Edition Philip Golden instant download
- CIBANO 500 Circuit Breaker Testing PDF A Guide For Performing The Anti 2021 ENUCIBANO 500 Circuit Breaker Testing PDF A Guide For Performing The Anti 2021 ENU
- 1st Puc Physics Model Paper 2022-23 by Bvvs Puc Vidyagiri1st Puc Physics Model Paper 2022-23 by Bvvs Puc Vidyagiri