Lecture 1

THE 2X2 SMALL OPEN ECONOMY MODEL IN GAMS

Erinc Yeldan
http://www.bilkent.edu.tr/~yeldane/Ec562syl.html

1. An Open 2x2 Ricardo-Viner General Equilibrium Model with Sector Specific Capital.

2.The GAMS Code of the Model.

3.The Walras' Law, Normalization and the Exchange Rate.

4.The Production Sector and the Neo-classical Properties of the General Equilibrium System with
Sluggish Capital Adjustment over Time.

1.1 A 2x2 SmaIl Open Economy Model

In this section I first present an analytical model of an abstract model of an abstract two-sector,
two-factor (2x2) economy. The economy is assumed to be small in the world markets with both goods
being traded. There are two consumption goods, agricultural (A), and urban (N). The consumption
goods are assumed to be produced along a CES technology obeying neo-classical hypotheses. Each
sector uses capital, K, and labor, L, as primary inputs. It is assumed that the first sector is more labor
intensive.

Labor is treated perfectly mobile across sectors. Capital, however, is sector-specific, and once
installed, each sector is assumed to utilize its capital endowment up to the amount available. This
specification gives rise to differences of the "profit rates" across sectors. In the later stage of the
model, we hypothesize an environment where capital will move sluggishly out of the low-profit sector
to the high profit-sector, until when profit rates are equalized.

There are two households, corresponding to factor ownership: workers and capitalists. As an
initial condition, we assume that the worker household is endowed with 10 units of labor; and the
capitalist household is endowed with 15 units of physical capital. The wage rate of 0.289 clears the
labor market, and the capital market is in equilibrium with profit rates being at equal values initially at
0.336. Workers supply all their labor endowments to production activities, hence they do not have
leisure considerations. Both households have Cobb-Douglas type preferences over the two
commodities.

Initially the A-sector is assumed to be in excess demand (imported), and the N-sector has
excess supply (exported). World terms of trade is set at unity. The complete list of equations are
presented in Table 1.

---------------------------------------------------------------------------------------------------------

Table 2.1: Algebraic Equations of the 2x2 Small Open Economy Model

Prices: Pi = ER.PWi

FACTOR MARKETS

Production: Qis = Ai [α i K i− P + (1 − α i ) L−i P ]−1/ Pi

Labor demand: LDi = [( Pi / W )(1 − α i ) Ai− P ]σ QiS ,

1
With σi =
(1 + Pi )

Labor market equilibrium: ∑ LDi = L

S

Capital is sector specific: K i = K i

Wage income: Y W = ∑ W .LDi

Capitalist income: YK = ∑ ( Pi QiS − W .LDi )

COMMODITY MARKET S

Y 
Consumer demand: QhD,i = β h ,i  h 
 Pi 

Import (excess demand): Q M = ∑ Qh , A − QAS

Export (excess supply): Q E = QNS − ∑ QhD, N

Trade balance (Walras’ Law): PAW Q M − PNW Q E = 0
-----------------------------------------------------------------------------------------------------------

1.2. The 2x2 Model in the GAMS Syntax

In this section we introduce the GAMS code for solving our simple general equilibrium model
in a PC. The acronym GAMS, stands for the General Algebraic Modeling System; and is described in
detail in Brooke, Kendrick and Meeraus (1988). In what follows, we will introduce the main principles
and components of the program, referring the reader to the main documentation for further expertise.

Table 2.2 identifies the general structure of the GAMS program to implement the simple
model. While the sequencing of the commands may vary from model to model, the structure displayed
in Table 2.2 is quite straightforward, and we will be following this routine in our foregoing discussion.

First we introduce the SETS distinguished in our simple model. We identify three indexes: HH
for households: workers and capitalists; I for commodities: A for agricultural, and N for the industrial;
and TP for time periods, to be utilized in the "dynamic" stage of our modeling effort. We distinguish a
subset, T of TP, as "the current" time period, which will assume a particular realization over the range
of time periods, 1 through 50. The GAMS commands of the foregoing run as follows: :

26 *
27 ***SET DEFINITIONS
28 SET
29 HH INDIVIDUALS /WHH WORKER HOUSEHOLD /
30 KHH CAPITALIST HOUSEHOLD /
31
32 I GOODS / A AGRICULTURAL GOOD
33 N INDUSTRIAL GOOD /
34
35 TP TIME PERIODS /0*50/
36 T(TP) CURRENT TIME PERIOD
37
38
39

The elements of each set are introduced by use of slashes ("I"), and the SET identification is closed
with the use of the ubiquitous ";". The specific element that the index T will realize is to be introduced
below.
Table 2: General Structure of the GAMS Program

Component Description

SETS Declares the indexes of the model

PARAMETERS Declares the constants, parameters and the dummy variables to be used in the
program. The elements under this component are described by a brief
descriptive sentence following the declaration statement, Multi-dimensional
parameter arrays are declared with the TABLE command, with the "."
representing any item that can be freely red into the TABLE.

INITIAL DATA Data is introduced using the TABLE and SCALAR commands.

CALIBRATION Data is used to generate the structural parameters and the shift variables of the
algebraic model equations. The purpose here is to impose the algebraic system
of equations onto the data, so that the first "solution" of the model will reflect
exactly the "base-year" SAM data.

VARIABLES All variables, endogenous, or exogenous are declared here with brief descriptive
identifiers.

VARIABLE INITIALIZATION The Variables are initialized with the base-year data values. This
enables the algorithm to start its search for equilibrium from a "sufficiently
close" data point.

EQUATIONS The algebraic equations of the model are declared with descriptive statements.

EQUATION SPECIFICATIONS The declared EQUATIONS are specified in the GAMS language.

CLOSURE AND RESTRICTIONS Exogenous variables are "fixed" using the .FX appends. Bounds
are introduced using .LO and .UP. The overall macro closure
recognized in the model is implemented here.

SOLVE AND DISPLAY The MODEL statement declares the model by specifying which
EQUATIONS form the MODEL. SOLVE statement commands
GAMS to interface with the algorithm to search for a numerical
solution. The solution values are inputted to the display tables
and are documented using the DISPLAY command.

--------------------------------------------------------------------------------------------------------------
The parameters and the pseudo-variables to carry the initial values of the endogenous variables of the
model are declared under PARAMETERS. Constants are declared using the SCALAR command, with
data being entered between slashes such as /#.###/). Multi-dimensional parameters are declared by the
TABLE command, with the elements of the TABLE being declared according to the indexes ordered
in parentheses. If no SET index is described, the star ("*") can be used to declare "free" elements to be
used in that TABLE. The GAMS code of this section is as follows:

39
40 ***PARAMETER DEFINITIONS
41 PARAMETER
42 AD(I) SHIFT PARAMETER IN PROD FUNC
43 ALFA(I) CES FUNCTION SHARE PARAMETER
44 CLES(I,HH) SECTORAL CONSUMPTION DEMAND BY HOUSEHOLDS
45 PW(I) WORLD PRICES
46 RHO(I) PRODUCTION EXPONENTIAL PARAMETER
47 SIGP(I) PRODUCTION FUNCTION ELASTICITY OF SUBSTlTUTION
48
49 *DUMMIES TO HOLD INITIAL DATA
50 IGO(HH) HOUSEHOLD WEIGHTS IN THE SOCIAL WELFARE FUNCTION
51 KO(I) CAPITAL DEMAND IN SECTOR I
52 KSTAR(HH) HOUSEHOLD CAPITAL ENDOWMENT
53 LDO(I) SECTORAL LABOR DEMANDS
54 LSTAR(HH) HOUSEHOLD LABOR ENDOWMENT 5
5 PO(I) DOMESTIC PRICES
56 QSO(I) COMMODITY SUPPLIES
57 RPO(I) SECTORAL PROFITS
58 RKO(I) SECTORAL PROFIT RATE
59 VO(HH) HOUSEHOLD INDIRECT UTILITIES
60 WLO(I) SECTORAL WAGES
61 YO(HH) HOUSEHOLD INCOME
62 ;
63
64 SCALAR
65 WO NOMINAL WAGE RATE
66 QXO EXPORTS
67 QMO IMPORTS
68 ERO EXCHANGE RATE /1.0/
69
70
71 *TABLES TO DISPLAY RESULTS
72 PARAMETER
73 HOUSE(*,HH,TP) HOUSEHOLD RESULTS
74 PRODUCT(*,I,TP) PRODUCT MARKET RESULTS
75 SCALRES(*,TP) SCALAR RESULTS
76 RK(I) PROFIT RATE AT PERIOD T
77 QDT(I,HH,TP) HOUSEHOLD DEMAND TABLE
78 ;
79

Material Balances for the Simple General Equilibrium Model

Activities Commodities Factors Households

1.Agr 2.Ind 3.Rural 4.Urban 5.Labor 6.Capital 7.Worker 8.Capitalist 9.ROW SUM
Activities 1. Agriculture 2.206 0.000 2.206
2. Industry 0.000 5.721 5.721
Commodities 3. Rural 2.167 2.519 0.000 4.686
4. Urban 0.722 2.519 2.480 5.721
Factors 5. Labor 0.986 1.903 2.889
6. Capital 1.220 3.818 5.038
Households 7. Workers 2.889 0.000 2.889
8. Capitalist 0.000 5.038 5.038
ROW 9. Row 2.480 0.000 2.480
TOTALS 2.206 5.721 4.686 5.721 2.889 5.038 2.889 5.038 2.480
The data is introduced, mostly using TABLEs. The data usually come from an underlying Social
Accounting Matrix (SAM), and should, desirably, be in initial equilibrium. The initial data points of
our simple model are represented in the GAMS syntax as:

79
80 ***INITIAL DATA FROM SAM TABLE
81
82 TABLE MATBAL(*,I) MATERIAL BALANCES
83
84 A N
85 PW 1 1
86 WLO 0.986 1.903
87 QSO 2.206 5.721
88 LDO 3.413 6.587
89 KO 3.636 11.364
90 SIGP 0.8 0.4
91 ;
92
93 TABLE HHINC(*,HH) HOUSEHOLD PARAMETERS
94
95 WHH KHH
96 YO 2.889 5.038
97 LSTAR 10.0
98 KSTAR 15.0
99 IGO 1. O 1. O
100 ;
101
102 TABLE QDO(I,HH) HOUSEHOLD CONSUMPTION
103
104 WHH KHH
105 A 2.167 2.519
106 N 0.722 2.519
107 ;
108
109

We calibrate the model by using the algebraic equation system and the initial data points of the
model. Thus, using the base-year data as a point estimate, the outcome of this exercise reveals the
values of all the structural shift and share parameters of the model equations. The idea is to "force" the
program to choose those parameter values so that the initial solution of the program is an exact replica
of the base-year benchmark equilibrium data set compiled in the SAM accounts. These ideas are
formulated in the GAMS code as follows:

109 110 111 *##MODEL CALIBRATION 112

113 PW(I) = MATBAL{"PW" ,I);
114 PO(I) = ERO*PW{I);
115
116 LDO(I) = MATBAL{"LDO" ,I) ;
117 KO{I) = MATBAL("KO",I) ;
118 QSO(I) = MATBAL{"QSO" ,I) ;
119 WLO(I) = MATBAL(""WLO", I) ;
120 LS TAR (HH) = HHINC("LSTAR",HH) ;
121 KSTAR(HH) = HHINC("KSTAR",HH) ;
122
123
124 *FIRST CALCULATE THE WAGE RATE
125 WO = SUM{I, WLO(I) ) / SUM(HH, LSTAR(HH)) ;
126
127 *WE CALCULATE SECTORAL PROFITS BY DEDUCTING THE WAGE PAYMENTS
128 *FROM SECTORAL VALUE ADDED USING EVLER' S LAW (OBSERVE THAT THE
129 *PRODUCTION FUNCTION IS CRTS).
130
131 RPO ( I ) = PO ( I ) *QSO ( I) - WLO ( I) ;
132
133 *THEN WE CALCULATE THE PROFIT RATE
134 RKO(I) = RPO(I)/KO(I) ;
135
136 *##CALIBRATE THE PRODUCTION FUNCTION: CES(K,L)
137 SIGP(I) = MATBAL("SIGP",I);
138 RHO(I) = (l/SIGP(I))-l;
139
140 ALFA{I) = (KO(I)/LDO(I))**{-l-RHO(I))*{WO/RKO(I)) ;
141 ALFA{I) = 1/(1+ALFA(I)) ;
142
143 AD(I) = QSO(I) / ((ALFA(I)*KO(I)**(-RHO(I)) + (l-ALFA(I))*LDO(I)
144 **(-RHO{I)))**(-l/RHO{I))) ;
145
146 DISPLAY AD, ALFA, RHO;
147
148 *##CALCULATE HOUSEHOLD CONSUMPTION PARAMETERS
149 YO(HH) = HHINC{"YO",HH) ;
150
151 CLES(I,HH) = PO{I)*QDO{I,HH) / YO(HH) ;
152
153 DISPLAY CLES ;
154
155 *##(INDIRECT) UTILITY LEVELS
156 IGO(HH) = HHINC{"IGO",HH);
157
158 VO(HH) = PROD{I, QDO{I,HH)**CLES(I,HH) ) ;
159
160
161 *##EXPORTS AND IMPORTS
162 QXO = QSO{"N") - SUM(HH, QDO("N",HH)) ;
163 QMO = SUM(HH, QDO{"A",HH)) - QSO("A") ;

We first read in available data from the TABLE's and load the dummy variables with their
corresponding data points. (Lines 113 through 121). Observe that by setting the base-year prices at
unity, we choose the initial price index at unity, and treat all base-year nominal quantities as real
quantities, expressed in the base-year prices. This accounting procedure is especially helpful in later
simulations of the model, to distinguish between real versus nominal changes across the variables.

Next we calculate the wage rate by dividing the total wage payments by aggregate employment
to get wage earnings per unit of labor (line 125). Assuming constant returns to scale production
technologies, factor payments exhaust he total sectoral value added. We get aggregate profits as a
residual value added, by subtracting wage costs.

The CES production technology is calibrated to the production and employment data in lines
through 137 to 144. In this process the Elasticity of substitution is given from "outside", desirably from
econometric estimation. In lines 140 and 141, α İ of the CES equation is calibrated using the first order

conditions for labor employment to maximize producer profits; while the shift parameter, Ai , is
calibrated in lines 143 to 144. The values of these parameters are DISPLAYed in line 146.
 Household consumption shares by sectors are found in line 151, under the assumption that the
consumer preferences are Cobb-Douglas (line 158). Finally excess demand equations are implemented
for calibrating the initial data points for exports (162) and imports (163).

Variable are declared under line 167 using the POSITlVE VARIABLE command, and under
line185, using the simple VARIABLE command. The first usage ensures ha the corresponding values
will always be bounded from below during the solution search procedure, as economists do not like to
work with negative quantities.

164
165
166 ***VARIABLE DEFINITIONS
167 POSITIVE VARIABLE
168 PINDEX PRICE INDEX
169 ER NOMINAL EXCHANGE RATE (CONVERSION FACTOR)
170 P(I) DOMESTIC PRICES
171 RP(I) SECTORAL PROFITS
172 QS(I) REAL OUTPOT SUPPLIES
173 LD(I) SECTORAL LABOR DEMANDS
174 K(I) SECTORAL CAPITAL DEMANDS
175 W NOMINAL WAGE RATE
176
177 Y(HH) HOUSEHOLD INCOME
178 V(HH) INDlRECT HOUSEHOLD UTILITY
179 IG(HH) HOUSEHOLD WEIGHTS ON THE SOCIAL WELFARE FUNCTION
180
181 QD(I,HH) REAL HCUSEHOLD DEMAND
182
183 ;
184
185 VARIABLE
186 OMEGA SOCIAL WELFARE FUNCTION VALUE
187 QX REAL EXPORTS (EXCESS SUPPLY)
188 QM REAL IMPORTS (EXCESS DEMAND)
189 ;
190
The variable in line 186 (OMEGA) is a cooked-up variable to initiate the non-linear solver
algorithm. This variable is optimized under the SOLVE statement (line 327 below). However, since
the Walrasian system is a square system of simultaneous equations, with the number of independent
equations equal to that of the independent variables, the algorithm searches for an "optimum" value for
OMEGA with no avail. It ought to declare "equilibrium solution" once all the equation restrictions are
satisfied, and it cannot improve any further. In this respect, for models of this class, specification of
OMEGA is entirely neutral on the solution values. However, for models _not in the Walrasian square
system tradition, the way OMEGA is defined will be important, as the algorithm will try to optimize it
treating the problem as one of "constrained optimization". Here we defined OMEGA as the weighted
sum of the indirect utility levels of the private households. Note that the weights, IGhh ,are entirely
exogenous, and there is no mechanism in the model for private households to affect them. Such a
mechanism would turn the model into a non-square system, bringing it closer to the models discussed
in the endogenous political economy literature. [Cite our work, and our friends here].

We declare EQUATIONs by appending specific names to each class of algebraic equations. If

preferred the equation names can be numbers, corresponding to the order of presentation in an
accompanying text:

242
243
244 ***EQUATION DEFINITIONS
245 EQUATION
246 INDEX PRICE NORMALIZATION RULE
247 PRICE DOMESTIC PRICES
248 PRODUCTION SUPPLY OF GOODS
249 LABOR DEMAND FOR LABOR
250 LMEQL LABOR MARKET EQUILIBRIUM
251
252 WINC WORKER HOUSEHOLD INCOME
253 KINC CAPITALIST HOUSEHOLD INCOME
254 UTILW WORKER-HOUSEHOLD INDlRECT UTILITY FUNCTION
255 UTILK CAPITALIST-HOUSEHOLD INDIRECT UTILITY FUNCTION
256
257 DEMANDA HOUSEHOLD DEMAND FOR A GOOD
258DEMANDN HOUSEHOLD DEMAND FOR N GOOD
259 EXPORT EXCESS SUPPLY
260 IMPORT EXCESS DEMAND
261 * TRDBAL TRADE BALANCE (omitted!!!, by way of Walras'Law)
262
263 OBJ OBJECTIVE FUNCTION (TOTAL SOCIAL WELFARE)
264 ;
265

Observe the correspondence between the "names" of the EQUATION's and the model
description in Table 1 above. Further observe that we have deleted the trade balance equation from the
system by using the comment "*" in the first column in line 261. The equation is written only for
convenience to the reader and will be ignored by the solver. Furthermore, we know from theory that
the value of excess demands vanish in such a system (Walras' law), and the TRDBAL equation is
indeed redundant. We will comment more on this issue further below when we introduce the full scale
CGE model for Turkey.

The EQUATIONs are specified in the GAMS syntax in lines 269 through 299:

265
266
267 ***EQUATION SPECIFICATIONS
268
269 INDEX.. SUM(I, P(I)*0.5) =E= PINDEX;
270
271 PRICE(I).. P(I) =E= PW(I)*ER ;
272
273 PRODUCTION(I).. QS(I) =E= AD(I) * ( ALFA(I)*K(I)**(-RHO(I))
274 + (l-ALFA(I))*LD(I)**(-RHO(I)) )**(-l/RHO(I)) ;
275
276 LABOR(I).. LD(I) =E= ( (P(I)/W)*(l-ALFA(I))*AD(I)**C-RHO(I)) )
277 **SIGPCI)*QS(I) ; 278
279 LMEQL.. SUM(I, LD(I) ) =E= SUM(HH, LSTAR(HH)) ;
280
Z81 WINC.. Y("WHH") =E= SUMCI, W*LD(I));
282
283 KINC.. Y("KHH") =E= SUMCI, P(I)*QS(I)-W*LD(I) ) ;
284
285 UTILW.. VC"WHH") =E= PROD(I, QD(I,"WHH")**CLES(I,"WHH")) ;
286
287 UTILK.. V("KHH") =E= PROD(I., QD(I,"KHH")**CLES(I,"KHH")) ;
288
289 DEMANDA(HH).. P("A")*QDC"A",HH) =E= CLES("A",HH)*Y(HH) ;
290
291 DEMANDN(HH).. P("N")*QD("N",HH) =E= CLESC"N",HH)*Y(HH) ;
292
293 EXPORT.. QX =E= QS("N") - SUM(HH, QD("N",HH) ) ;
294
295 IMPORT.. QM =E= SUM(HH, QD("A",HH) ) - QS("A") ;
296
297 *TRDBAL.. PW("N")*QX =E= PW("A")*QM ;
298
299 OBJ.. OMEGA =E= SUM(HH,IG(HH)*V(HH) ) ;
300

These correspond to the algebraic specification documented in Table 2.1 above. Observe,
however, that we have two extra equations: INDEX (line 269) and OBJ (line 299). The latter one is
used by the solver in an attempt to optimize over OMEGA. The INDEX equation defines the
normalization rule recognized in the model, something that we have not mentioned for the system so
far. Since the excess demand functions, hence the overall real quantities, are homogeneous of degree
zero in prices, we have to specify a normalization rule for our system to be able to express the nominal
quantities in terms of a "unit of account". In equation INDEX we normalize the system over a price

index benchmark, PINDEX, using equal weights for each sector: ∑ Pwi i = PINDEX. In line 315,
below, we fix the value of PINDEX at unity to complete the normalization.

305
306 *##MODEL RESTRICTIONS
307 *CAPITAL IS SECTOR SPECIFIC
308 K. FX ( I ) = K. L ( I) ;
309
310 *HOUSEHOLDS HAVE NO CONTROL OVER SOCIAL WELFARE WEIGHTS
311 IG.FX(HH) = IG.L(HH) ;
312
313 *##PRICE NORMALIZATION
314 *CHOOSE ONE OF THE FOLLOWING AS NUMERAIRE
315 PINDEX.FX = PINDEX.L;
316 * ER.FX = ER.L ;
317 * P.FX("N")= P.L("N") ;
318

Observe that under lines 316 and 317 alternative normalization procedures are proposed, but
not implemented. Line 316 treats the exchange rate as the numeraire; whereas 317 fixes the price of N-
good, regarding it as the numeraire commodity. Under both cases we would comment on the line 314
using ("*"), thus the overall price level would turn to an "endogenous" variable.

In the GAMS section above, capital stocks are declared sector-specific by adding the .FX(I)
next to their variable names. Thus, line 308 signifies that the variables K i are fixed at their level
values, K.L(I).
 The model is created and declared in line 322 under the name GENEQM. Between slashes, we
note that all equations specified thus far are part of the current model:

3ı9
320 ***MODEL DEFINITION
321
322 MODEL GENEQM OPEN ECONOMY GENERAL EQUILIBRIUM MODEL IALLI ;
323

The model is solved using the SOLVE statement in line 327. GAMS interfaces with the non-
linear program (NLP) solver to maximize OMEGA. But as we have noted earlier, the system forming a
square block of equations, all given by equality . constraints, stops short of any "optimization" attempt
once the constraints are satisfied.

326
327 SOLVE GENEQM MAXIMIZING OMEGA USING NLP;
328
329 *SOLUTION VALUES AT TIME 1
330
331 T(TP) = NO;
332 T("ı") = YES ;
333
334 RK( I) = (P.L(I)-QS.L(I)-W.L-LD.L(I)) i K.L(I) ;
335
336 SCALRES("WAGERATE",T) = W.L ;
337 SCALRES("EXPORTS",T) = QX.L;
338 SCALRES("IMPORTS",T) = QM.L;
339 SCALRES("SOCWELFARE",T) = OMEGA.L;
340 SCALRES("RELATlVEPR",T) = P.L("A")/P.L("N") ;
341 SCALRES("EXCRATE",T) = ER.L ;
342
343 PRODUCT("DOMPRICE",I,T) = P.L(I);
344 PRODUCT("WRLDPRICE",I,T)= PW(I) ;
345 PRODUCT("LABDEMAND",I,T) = LD.L(I);
346 PRODUCT("CAPITAL",I,T) = K.L(I);
347 PRODUCT("OUTPUT",I,T) = QS.L(I);
348 PRODUCT("KLRATIO",I,T) = K.L(I)/LD.L(I) ;
349 PRODUCT("PROFITS", I,T) = P.L(I)*QS.L(I)-W.L*LD.L(I) ;
350 PRODUCT( "PROFITRATE" , I, T) = RK( I);
351
352 HOUSE("INCOME",HH,T) = Y.L(HH);

353 HOUSE("UTILITY",HH,T) = V.L(HH);

354 HOUSE("WLFREWGHTS",HH,T) = IG.L(HH) ;

355 HOUSE("LSTAR",HH,T) = LSTAR(HH);

356 HOUSE("KSTAR",HH,T) = KSTAR(HH);
357
358 QDT(I,HH,T) = QD.L(I,HH);
359
360 DISPLAY SCALRES, PRODUCT, HOUSE, QDT ;

We specify the set T as the current period ("1") in lines 332 and 333.
The first one says that T does not assume any element of TP, while the next one amends it stating that
T takes the value of "1". We calculate the Sectoral profit rates in line 335 as part of preparation for the
display tables. We load the solution values to our previously prepared display tables through lines 336
and 359. We document them by using the DISPLAY command in line 360. We next utilize the model
in conducting a series of simulation experiments to highlight some further attributes of the Walrasian
system at hand.

1-3. Policy Exercises with the Simple Model: Normalization and the Exchange Rate

Given the discussion of the previous section, we first analyze the behavior of the exchange rate
in this economy. To do so we perturb the world terms of trade in favor of the A-good by 20%. We
trace out the adjustments in the economy under two distinct "normalization" rules: first, we choose the
N-good as the numeraire of our simple economy (Experiment E1A). Thus, with this specification we
set P("N")=PW("N")=1.0. As a second alternative, we continue to utilize the unit-simplex as our
overall normalization rule (Experiment E1B).

The overall GAMS program is kept intact as it was introduced above. However, under E1A, the
"Price normalization" section is changed to:
313
314 *##PRICE NORMALIZATION
315 *CHOOSE ONE OF THE FOLLOWING AS NUMERAIRE
316 * PINDEX.FX = PINDEX.L;
317 * ER.FX = ER.L ;
318 P.FX("N")= P.L("N") ;
319
320

Given that P.L("N") = 1.0 from line 114 of the GAMS program, we effectively choose N-good
as the numeraire in line 318.
We introduce the experiment as a continuation of the program so that the model "searches" for the new
equilibrium starting from the previous solution values:

383
384 *INTRODUCE EXPERIMENT HERE: INCREASE PW(A)BY 20%
385 PW("A")=1.2*PW("A") ;
386
387 SOLVE GENEQM MAXIMIZING OMEGA USING NLP;
388

The model results are documented in Table 1.3. In what follows we provide a brief depiction of the
equilibrium relationships ongoing in our simple economy.

With more favorable terms of trade, we get the Samuelson-Stolper results vis-a-vis to labor
incomes. Since the A-good is more labor intensive, labor numerations increase, as employment
increases in the production of A. The A-profit rate increases in nominal terms, due to improved terms
of trade, while that of N-sector decreases.

With N-price set at unity, we observe no change in the nominal exchange rate. The domestic terms of
trade are stilI given by the foreign terms of trade, PW("A")/PW("N") = 1.20. It is in fact clear that,
under such a specification the exchange rate can play no role in this economy. For, assuming away the
transportation costs, the domestic prices are given by:

Pi = PWi ER (1)

thus, the terms of trade become:

Pi / Pj = PWi / PW j (2)
The exchange rate has no influence in the relative price system of such an economy, and under
conditions of balanced trade, it only translates foreign currency units into domestic units.

Now the foregoing results can be contrasted with those of the ElE, where the overall price
index was parametric to reflect normalization around the price simplex:

295
296 *##PRICE NORMALIZATION
297 *CHOOSE ONE OF THE FOLLOWING AS NUMERAIRE
298 PINDEX.FX = PINDEX.L;
299 * ER.FX = ER.L ;
300 * P.FX("N")= P.L("N") ;
301
302

So by restricting PINDEX to its level (1.0), we set an aggregate inflation benchmark to the domestic
prices. Clearly the domestic prices must now assume different levels than their foreign counterparts in
absolute values. This is because while the domestic prices satisfy the domestic price level
constraint, set at 1.0; the weighted sum of world prices add up to 1.1

( PWA wA + PWN wN = 1.2*0.5+1.0*0.5 = 1.1). The exchange rate must adjust

such that:

∑ P w = PINDEX (3)
i i

where by definition:

∑ PW .ER.w = PINDEX (4)

i i

Thus, ER satisfies:

ER = PINDEX/ ∑ Pw
i i (5)
 Observe from Table 3 that PA = 1.091, whereas PWA = 1.2; and PN =0.909, with PWN =1.0.

However, the relative domestic terms of trade, PA / PN is N A N still equal to 1.20 (=1.091/0.909),
hence across the two experiments there are no real effects. Levels of real production and consumption
are the same even though nominal incomes are different. But by the homogeneity of the system, the
two solutions are exactly the same given an appropriate choice of the normalization units. The
"appropriate" choice of normalization units can be found actually by applying the
"purchasing-power-parity" price level deflated (PPP-PLD) concept of the exchange rate. Accordingly,
by relation (5), the equilibrium value of the exchange rate is 0.909 (=1.0/1.1). That is, as the world
"inflation rate" accelerates to 10% in excess over the domestic inflation rate, the nominal exchange
rate must appreciate by 10% to 0.909, to keep the PPP-PLD exchange rat e constant. The latter is still
given by the relative terms of trade, which cannot diverge from the world price ratio by way of small
country hypothesis.

Hence in this model, the exchange rate is nothing but a macro index, translating world prices
into domestic prices, with no significance on the relative price system nor the real variables.

In the pure trade theory literature, the real exchange rate is defined as the "relative value of the
basket of traded good prices to the nontraded (home) good prices". If we extend our analytical model
to include a nontraded, home good, H, the price normalization rule would yield:

∑ PW .ER.w + P w
i i H H =PINDEX. (6)

where the summation is assumed to be over the traded goods, A and N, and

the home good is added with ∑w İ

+ wH = 1.0.

In this system the relationship between the PH and the ER becomes:

PH = PINDEX / wH − (∑ PWi wi / wH ) ER

Thus,

∂PH / ∂ER = − (∑ PWi wi / wH ) 〈 0

Consequently, under the PINDEX normalization rule, nominal changes in the exchange rate will lead
to a fall in the price of the home good. What is more, the relationship will be linear given by the term
in parenthesis in (7). (Dervis, de Melo & Robinson, 1982, chp 6, further note that, by setting ER to the
real exchange rate will itself amount to a normalization rule. This method sets the home good price to
the real exchange rate index).

The foregoing discussion stresses the importance of the choice of the normalization rule in the
model. It is clear that the normalization procedure is an artifact of the homogeneity of the general
equilibrium system, and has no effect on the real quantities. What is important to note is that the
nominal variables should always be analyzed in relation to the choice of the numeraire. In particular if
a nominal quantity is assumed fixed, that implies fixity only in terms of the numeraire and one has to
be careful in specifying equilibrium relation with the accompanying solution and the numeraire price
index.

1-4. The Properties of the General Equilibrium System under Dynamic Adjustment of the
Capital Stocks.

Most results of the pure theory of trade (viz. Samuelson, 1949; Stolper and Samuelson, 1941;
Rybczynki, 1955) have been derived under conditions of fully flexible factors of production. In this
section we illustrate the dynamics of our system with sector specific capital adjusting sluggishly over
time in response to the profit rate differentials across sectors. In so doing we will introduce more
examples on the GAMS syntax, and reinstate some of the classic propositions of the trade theory
which were introduced in the previous section.

We foresee the following adjustment rule for capital over "time":

 π i (t ) − π j (t ) 
∆K (t ) = φ   K j (t )
 π i (t ) 
K i (t + 1) = K i (t ) + ∆K (t )
K j (t + 1) = K j (t ) − ∆K (t )
where φ is a parameter showing the degree of responsiveness of the capital

stock to profit rate, Π i , differentials.

We implement this specification by using the LOOP and ABORT commands of the GAMS
syntax:

297
298 SET TT(TP) TIME PERIODS FOR DYNAMIC -SOLUTIONS
299
300 LOOP (TT,
301
302 SOLVE GENEQM MAXIMIZING OMEGA USING NLP;
303
304 TCTP) = NO;
305 TCTT) = YES ;
306
307 RK(I) = CP.L(I)-QS.L(I)-W.L-LD.L(I)) i K.L(I) ;
308 DK = ED*(1- RK("N")/RK("A")) - K.L("N") ;
309
310 SCALRES("CAPADlUST",TT) = DK;
311 SCALRES("WAGERATE" ,TT) = W.L.;
312 SCALRES("EXPORTS",TT) = QX.L;
313 SCALRES("IMPORTS",TT) = QM.L;
314 SCALRES("SOCWELFARE",TT) = OMEGA.L;
315 SCALRES("RELATIVEPR",TT) = P.L("A")/P.L("N") ;
316 SCALRES("EXCRATE",TT) = ER.L ;
317
318 HOUSE("INCOME",HH,TT) = Y.L(HH);
319 HOUSE("UTILITY",HH,TT) = V.L(HH);
320 HOUSE("WlFREWGHTS",HH,TT) = IG.L(HH) ;
321 HOUSE("LSTAR",HH,TT) = LSTAR(HH);
322 HOUSE("KSTAR",HH,TT) = KSTAR(HH); "
323 QDT(I,HH,TT) = QD.L(I,HH);
324
325
326 PRODUCT("PRICES",I,TT) = P.L(I); :
327 PRODUCT("LABDEMAND",I,TT) = LD.L(I);
328 PRODUCT("CAPDEMAND",I,TT) = K.L(I);
329 PRODUCT("OUTPUT",I,TT) = QS.L(I);
330 PRODUCT("KLRATIO",I,TT) = K.L(I)/LD.L(I) ;
331 PRODUCT("PROFIIS",I,TT) = P.L(I)-QS.L(I)-W.L-LD.L(I) ;
332 PRODUCT("PROFITRATE",I,TT) = RK(I);
333
334 DISPLAY $(ABS(RK("N")-RK8"A")) LT 0.0005) AD, ALFA, SCALRES, PRODUCT, HOUSE,
335 QDT;
336 ABORT $(ABS(RK("N")-RK("A")) LT 0.0005)
337 "PROFIT RATES ARE EQUALIZED WITHIN TOLERANCE" ;
338
339 *ADJUST CAPITAL STOCKS IN RESPONSE TO PROFIT RATE DIFFERENTIALS
340
341 K.FX("A") = K.L("A") + DK;
342 K.FX("N") = KoL("N") - DK;
343
344
345 *#END OF LOOP
346 );
347

Thus we first create a new set TT over the time periods set, TP. Calculate the profit rates;
RK(I), in line 307. Capital adjustment is set at line 308, through DK. ED stands for φ above, and is set
at 1.30:

SCALAR DK CAPITAL ADJUSTMENT FROM K("N") TO K( "N" );

ED ELASTICITY OF CAPITAL ADJUSTMENT /1.30/ ;

the LOOP starts at line 300. Un less ABORTED, it continues over again. The ABORT statement at
336 checks whether the absolute levels of the profit rate differentials are less than some 0.0005 units.
If so, the algorithm stops, with the message, "PROFIT RATES ARE EQUALIZED WITHIN
TOLERANCE". If not, capital stocks are adjusted in lines 341 and 342, and the LOOP continues once
again with a new SOLVE statement in the next "period". The DISPLAY statement precedes the
ABORT statement in line 334 because, with ABORT, GAMS stops without any further action. This
would leave the equilibrium results unprinted.

Using this apparatus, we illustrate the main propositions of the trade theory in turn.

-----------------------------------------------------------------------------------------------------------
Table. 1.3 Short Run Adjustments to the 20% Rise in the World Price of A-Good

N-Good is
Numeraire Fix PINDEX
Base Value (Exp E1A) (Exp E1B)
Domestic Prices
Sector A 1.000 1.200 1.091
Sector N 1.000 1.000 0.909
Terms of trade (PA/PN) 1.000 1.200 1.200
Aggregate price level
Domestic 1.000 1.100 1.000
Foreign 1.000 1.100 1.100
Exchange Rate 1.000 1.000 0.909
Labor Employed
Sector A 3.413 3.818 3.818
Sector N 6.587 6.182 6.182
Real production
Sector A 2.206 2.319 2.319
Sector N 5.721 5.598 5.598
Wage rate 0.289 0.321 0.291
Profit rate
Sector A 0.336 0.428 0.390
Sector N 0.336 0.318 0.289
Exports of N 2.480 2.210 2.210
Imports of A 2.480 1.841 1.841