P1: FCG

FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS

Foreword xvii

AL
Preface to the Fourth Edition xix

RI
Part I Introduction and the Linear Regression Model 1

TE
CHAPTER 1
What is Econometrics? 3
What is in this Chapter?
1.1 What is econometrics?
MA 3
3
1.2 Economic and econometric models 4
1.3 The aims and methodology of econometrics 6
ED

1.4 What constitutes a test of an economic theory? 8

Summary and an outline of the book 9
References 10
HT

CHAPTER 2
Statistical Background and Matrix Algebra 11
IG

What is in this Chapter? 11

2.1 Introduction 11
R

2.2 Probability 12
Addition rules of probability 13
PY

Conditional probability and the multiplication rule 14

Bayes’ theorem 14
CO

Summation and product operations 15

2.3 Random variables and probability distributions 17
Joint, marginal, and conditional distributions 17
Illustrative example 18
2.4 The normal probability distribution and related distributions 18
The normal distribution 19
Related distributions 19
2.5 Classical statistical inference 21
2.6 Properties of estimators 22
Unbiasedness 23
Efficiency 23
Consistency 23
Other asymptotic properties 24

v
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

[ vi ] CONTENTS

2.7 Sampling distributions for samples from a normal population 26

2.8 Interval estimation 26
2.9 Testing of hypotheses 28
2.10 Relationship between confidence interval procedures and tests of hypotheses 31
2.11 Combining independent tests 32
Summary 33
Exercises 33
Appendix: Matrix algebra 40
Exercises on matrix algebra 56
References 57

CHAPTER 3
Simple Regression 59
What is in this Chapter? 59
3.1 Introduction 59
Example 1: Simple regression 60
Example 2: Multiple regression 60
3.2 Specification of the relationships 61
3.3 The method of moments 65
Illustrative example 66
3.4 The method of least squares 68
Reverse regression 71
Illustrative example 72
3.5 Statistical inference in the linear regression model 76
Illustrative example 78
Confidence intervals for α, β, and σ 2 78
Testing of hypotheses 80
Example of comparing test scores from the GRE and GMAT tests 81
Regression with no constant term 83
3.6 Analysis of variance for the simple regression model 83
3.7 Prediction with the simple regression model 85
Prediction of expected values 87
Illustrative example 87
3.8 Outliers 88
Some illustrative examples 90
3.9 Alternative functional forms for regression equations 95
Illustrative example 97
*3.10 Inverse prediction in the least squares regression model1 99
*3.11 Stochastic regressors 102
*3.12 The regression fallacy 102
The bivariate normal distribution 102
Galton’s result and the regression fallacy 104
A note on the term “regression” 105
Summary 105

1 Here and below, the * indicates that this is an optional section.

 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS [ vii ]

Exercises 106
Appendix: Proofs 113
References 126

CHAPTER 4
Multiple Regression 127
What is in this Chapter? 127
4.1 Introduction 127
4.2 A model with two explanatory variables 129
The least squares method 129
Illustrative example 132
4.3 Statistical inference in the multiple regression model 134
Illustrative example 135
Formulas for the general case of k explanatory variables 139
Some illustrative examples 140
4.4 Interpretation of the regression coefficients 143
Illustrative example 146
4.5 Partial correlations and multiple correlation 146
4.6 Relationships among simple, partial, and multiple correlation coefficients 147
Two illustrative examples 148
4.7 Prediction in the multiple regression model 153
Illustrative example 154
4.8 Analysis of variance and tests of hypotheses 155
Nested and nonnested hypotheses 157
Tests for linear functions of parameters 158
Illustrative example 159
4.9 Omission of relevant variables and inclusion of irrelevant variables 160
Omission of relevant variables 161
Example 1: Demand for food in the United States 163
Example 2: Production functions and management bias 163
Inclusion of irrelevant variables 164
2
4.10 Degrees of freedom and R 165
4.11 Tests for stability 169
The analysis of variance test 169
Example 1: Stability of the demand for food function 170
Example 2: Stability of production functions 171
Predictive tests for stability 174
Illustrative example 174
4.12 The LR, W, and LM tests 176
Illustrative example 177
Summary 178
Exercises 180
Appendix 4.1: The multiple regression model in matrix notation 187
Appendix 4.2: Nonlinear regressions 193
Appendix 4.3: Large-sample theory 196
Data sets 202
References 207
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

[ viii ] CONTENTS

Part II Violation of the Assumptions of the Basic Regression Model 209

CHAPTER 5
Heteroskedasticity 211
What is in this Chapter? 211
5.1 Introduction 211
Illustrative example 212
5.2 Detection of heteroskedasticity 214
Illustrative example 214
Some other tests 215
Illustrative example 217
An intuitive justification for the Breusch–Pagan test 218
5.3 Consequences of heteroskedasticity 219
Estimation of the variance of the OLS estimator under heteroskedasticity 221
5.4 Solutions to the heteroskedasticity problem 221
Illustrative example 223
5.5 Heteroskedasticity and the use of deflators 224
Illustrative example: The density gradient model 226
5.6 Testing the linear versus log-linear functional form 228
The Box–Cox test 229
The BM test 230
The PE test 231
Summary 231
Exercises 232
Appendix: Generalized least squares 235
References 237

CHAPTER 6
Autocorrelation 239
What is in this Chapter? 239
6.1 Introduction 239
6.2 The Durbin–Watson test 240
Illustrative example 241
6.3 Estimation in levels versus first differences 242
Some illustrative examples 243
6.4 Estimation procedures with autocorrelated errors 246
Iterative procedures 248
Grid-search procedures 249
Illustrative example 250
6.5 Effect of AR(1) errors on OLS estimates 250
6.6 Some further comments on the DW test 254
The von Neumann ratio 255
The Berenblut–Webb test 255
6.7 Tests for serial correlation in models with lagged dependent variables 257
Durbin’s h-test 258
Durbin’s alternative test 258
Illustrative example 258
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS [ ix ]

6.8 A general test for higher-order serial correlation: The LM test 259
6.9 Strategies when the DW test statistic is significant 261
Autocorrelation caused by omitted variables 261
Serial correlation due to misspecified dynamics 263
The Wald test 264
Illustrative example 265
*6.10 Trends and random walks 266
Spurious trends 268
Differencing and long-run effects: The concept of cointegration 270
*6.11 ARCH models and serial correlation 271
6.12 Some comments on the DW test and Durbin’s h-test and t-test 272
Summary 273
Exercises 274
References 276

CHAPTER 7
Multicollinearity 279
What is in this Chapter? 279
7.1 Introduction 279
7.2 Some illustrative examples 280
7.3 Some measures of multicollinearity 283
7.4 Problems with measuring multicollinearity 286
7.5 Solutions to the multicollinearity problem: Ridge regression 290
7.6 Principal component regression 292
7.7 Dropping variables 297
7.8 Miscellaneous other solutions 300
Using ratios or first differences 300
Using extraneous estimates 300
Getting more data 301
Summary 302
Exercises 302
Appendix: Linearly dependent explanatory variables 304
References 311

CHAPTER 8
Dummy Variables and Truncated Variables 313
What is in this Chapter? 313
8.1 Introduction 313
8.2 Dummy variables for changes in the intercept term 314
Illustrative example 317
Two more illustrative examples 317
8.3 Dummy variables for changes in slope coefficients 319
8.4 Dummy variables for cross-equation constraints 322
8.5 Dummy variables for testing stability of regression coefficients 324
8.6 Dummy variables under heteroskedasticity and autocorrelation 327
8.7 Dummy dependent variables 329
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

[x] CONTENTS

8.8 The linear probability model and the linear discriminant function 329
The linear probability model 329
The linear discriminant function 332
8.9 The probit and logit models 333
Illustrative example 335
The problem of disproportionate sampling 336
Prediction of effects of changes in the explanatory variables 337
Measuring goodness of fit 338
Illustrative example 340
8.10 Truncated variables: The tobit model 343
Some examples 344
Method of estimation 345
Limitations of the tobit model 346
The truncated regression model 347
Summary 349
Exercises 350
References 352

CHAPTER 9
Simultaneous Equation Models 355
What is in this Chapter? 355
9.1 Introduction 355
9.2 Endogenous and exogenous variables 357
9.3 The identification problem: Identification through reduced form 357
Illustrative example 360
9.4 Necessary and sufficient conditions for identification 362
Illustrative example 364
9.5 Methods of estimation: The instrumental variable method 365
Measuring R 2 368
Illustrative example 368
9.6 Methods of estimation: The two-stage least squares method 371
Computing standard errors 373
Illustrative example 375
9.7 The question of normalization 378
*9.8 The limited-information maximum likelihood method 379
Illustrative example 380
*9.9 On the use of OLS in the estimation of simultaneous equation models 380
Working’s concept of identification 382
Recursive systems 384
Estimation of Cobb–Douglas production functions 385
*9.10 Exogeneity and causality 386
Weak exogeneity 389
Superexogeneity 389
Strong exogeneity 389
Granger causality 390
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS [ xi ]

Granger causality and exogeneity 390

Tests for exogeneity 391
9.11 Some problems with instrumental variable methods 392
Summary 392
Exercises 394
Appendix 396
References 400

CHAPTER 10
Diagnostic Checking, Model Selection, and Specification Testing 401
What is in this Chapter? 401
10.1 Introduction 401
10.2 Diagnostic tests based on least squares residuals 402
Tests for omitted variables 402
Tests for ARCH effects 404
10.3 Problems with least squares residuals 404
10.4 Some other types of residual 405
Predicted residuals and studentized residuals 406
Dummy variable method for studentized residuals 407
BLUS residuals 407
Recursive residuals 408
Illustrative example 409
10.5 DFFITS and bounded influence estimation 411
Illustrative example 413
10.6 Model selection 414
Hypothesis-testing search 415
Interpretive search 416
Simplification search 416
Proxy variable search 416
Data selection search 417
Post-data model construction 417
Hendry’s approach to model selection 418
10.7 Selection of regressors 419
2
Theil’s R criterion 421
Criteria based on minimizing the mean-squared error of prediction 421
Akaike’s information criterion 422
10.8 Implied F-ratios for the various criteria 423
Bayes’ theorem and posterior odds for model selection 425
10.9 Cross-validation 427
10.10 Hausman’s specification error test 428
An application: Testing for errors in variables or exogeneity 430
Some illustrative examples 431
An omitted variable interpretation of the Hausman test 433
10.11 The Plosser–Schwert–White differencing test 435
10.12 Tests for nonnested hypotheses 436
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

[ xii ] CONTENTS

The Davidson and MacKinnon test 437

The encompassing test 439
A basic problem in testing nonnested hypotheses 440
Hypothesis testing versus model selection as a research strategy 440
10.13 Nonnormality of errors 440
Tests for normality 441
10.14 Data transformations 441
Summary 442
Exercises 444
Appendix 446
References 447

CHAPTER 11
Errors in Variables 451
What is in this Chapter? 451
11.1 Introduction 451
11.2 The classical solution for a single-equation model with one explanatory variable 452
11.3 The single-equation model with two explanatory variables 455
Two explanatory variables: One measured with error 455
Illustrative example 459
Two explanatory variables: Both measured with error 460
11.4 Reverse regression 463
11.5 Instrumental variable methods 465
11.6 Proxy variables 468
Coefficient for the proxy variable 470
11.7 Some other problems 471
The case of multiple equations 471
Correlated errors 472
Summary 473
Exercises 474
References 476

Part III Special Topics 479

CHAPTER 12
Introduction to Time-Series Analysis 481
What is in this Chapter? 481
12.1 Introduction 481
12.2 Two methods of time-series analysis: Frequency domain and time domain 482
12.3 Stationary and nonstationary time series 482
Strict stationarity 483
Weak stationarity 483
Properties of autocorrelation function 484
Nonstationarity 484
12.4 Some useful models for time series 485
Purely random process 485
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS [ xiii ]

Random walk 486

Moving average process 486
Autoregressive process 488
Autoregressive moving average process 490
Autoregressive integrated moving average process 491
12.5 Estimation of AR, MA, and ARMA models 492
Estimation of MA models 492
Estimation of ARMA models 492
Residuals from the ARMA models 494
Testing goodness of fit 494
12.6 The Box–Jenkins approach 496
Forecasting from Box–Jenkins models 497
Illustrative example 499
Trend elimination: The traditional method 500
A summary assessment 500
Seasonality in the Box–Jenkins modeling 502
12.7 R2 measures in time-series models 503
Summary 506
Exercises 506
Data sets 507
References 508

CHAPTER 13
Models of Expectations and Distributed Lags 509
What is in this Chapter? 509
13.1 Models of expectations 509
13.2 Naive models of expectations 510
13.3 The adaptive expectations model 512
13.4 Estimation with the adaptive expectations model 514
Estimation in the autoregressive form 514
Estimation in the distributed lag form 515
13.5 Two illustrative examples 516
13.6 Expectational variables and adjustment lags 520
13.7 Partial adjustment with adaptive expectations 524
13.8 Alternative distributed lag models: Polynomial lags 526
Finite lags: The polynomial lag 527
Illustrative example 530
Choosing the degree of the polynomial 532
13.9 Rational lags 533
13.10 Rational expectations 534
13.11 Tests for rationality 536
13.12 Estimation of a demand and supply model under rational expectations 538
Case 1 538
Case 2 539
Illustrative example 542
13.13 The serial correlation problem in rational expectations models 544
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

[ xiv ] CONTENTS

Summary 545
Exercises 547
References 548

CHAPTER 14
Vector Autoregressions, Unit Roots, and Cointegration 551
What is in this Chapter? 551
14.1 Introduction 551
14.2 Vector autoregressions 551
14.3 Problems with VAR models in practice 553
14.4 Unit roots 554
14.5 Unit root tests 555
The Dickey–Fuller tests 556
The serial correlation problem 556
The low power of unit root tests 557
The DF-GLS test 557
What are the null and alternative hypotheses in unit root tests? 558
Tests with stationarity as null 559
Confirmatory analysis 560
Panel data unit root tests 561
Structural change and unit roots 562
14.6 Cointegration 563
14.7 The cointegrating regression 564
14.8 Vector autoregressions and cointegration 567
14.9 Cointegration and error correction models 571
14.10 Tests for cointegration 571
14.11 Cointegration and testing of the REH and MEH 572
14.12 A summary assessment of cointegration 574
Summary 575
Exercises 576
References 579

CHAPTER 15
Panel Data Analysis 583
What is in this Chapter? 583
15.1 Introduction 583
15.2 The LSDV or fixed effects model 584
Illustrative example: Fixed effect estimation 585
15.3 The random effects model 586
15.4 Fixed effects versus random effects 589
Hausman test 589
Breusch and Pagan test 590
Tests for serial correlation 590
15.5 Dynamic panel data models 591
15.6 Panel data models with correlated effects and simultaneity 593
15.7 Errors in variables in panel data 595
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

CONTENTS [ xv ]

15.8 The SUR model 597

15.9 The random coefficient model 597
Summary 599
References 599

CHAPTER 16
Small-Sample Inference: Resampling Methods 601
What is in this Chapter? 601
16.1 Introduction 601
16.2 Monte Carlo methods 602
More efficient Monte Carlo methods 603
Response surfaces 603
16.3 Resampling methods: Jackknife and bootstrap 603
Some illustrative examples 604
Other issues relating to the bootstrap 605
16.4 Bootstrap confidence intervals 605
16.5 Hypothesis testing with the bootstrap 606
16.6 Bootstrapping residuals versus bootstrapping the data 607
16.7 Non-IID errors and nonstationary models 607
Heteroskedasticity and autocorrelation 607
Unit root tests based on the bootstrap 608
Cointegration tests 608
Miscellaneous other applications 608
Summary 609
References 609

Appendix 611

Index 621
 P1: FCG
FM JWBK384-Maddala September 14, 2009 15:48 Printer: Yet to come

xvi