PREFACE

1. A discussion of aims and methodology of econometrics in Chapter 1.

2. A critique of conventional significance levels in Chapter 2.

3. A discussion of direct versus reverse regression and inverse prediction in Chapter 3.

4. A thorough discussion of several practically useful results in multiple regression (Chapter 4), some of which are not to be found even in specialized books on regression.

5. Discussion of several new tests for heteroskedasticity, choice of linear versus log-linear models, and use of deflators (in Chapter 5).

6. A thorough discussion of the several limitations of the Durbin-Watson test, showing that it is almost useless in practice, and a discussion of Sargans test, ARCH test, LM test, etc. (Chapter 6).

7. A critical discussion of the use and limitations of condition numbers, variance inflation factors (VlFs), ridge regression, and principal component regression in the analysis of multicollinearity (Chapter 7). Many of these techniques are often used uncritically, because computer programs are readily available.

8. The use of dummy variables to obtain predictions and standard errors of predictions, the relationship between discriminant analysis and the linear probability model, the logit, probit, and tobit models (Chapter 8).

9. Inference in under-identified simultaneous equation models, criteria for normalization, and tests for exogeneity and causality (Chapter 9).

10. The discussion of partial adjustment models, error correction models, rational expectations models, and tests for rationality (Chapter 10).

11. Reverse regression, proxy variables (Chapter 11).

12. Different types of residuals and their use in diagnostic checking, model selection, choice of F-ratios in the selection of regressors, Hausmans specification error test, tests of nonnested hypotheses (Chapter 12).

These are not new topics for advanced texts but these topics (most of which are never dealt with in introductory texts) are important in econometric practice. The book explains them all with simple models so that students (and researchers) who have not had exposure to advanced texts and advanced courses in econometrics can still learn them and use them.

I have avoided any special discussion of computer programs. Nowadays, there are several computer packages available that one can choose from. Also, they change so rapidly that the lists have to be expanded and updated very often. The instructors manual will provide some guidance on this. 1 feel that it is more important to know the answers to the questions of what to compute, why to compute, how to interpret the results, what is wrong with the model, and how we can improve the model. Learning how to compute is rather easy. In the 1960s this last question received more attention because computer technology had not progressed enough and not many efficient computer programs were readily available. With the advances in computer technology and the large number of computer programs readily available, everyone can easily learn "how to compute." That is why 1 have tried to minimize the discussion of

PREFACE Xvii

computer programs or the presentation of computer output. Moreover, there is no single program that will do all the things discussed in the book. But by simple adaptation many of the computer programs available will do the job.

Instructors using the book might find it difficult to cover all the material. However, it is always better to have some topics in the book that instructors can choose from depending on their interests and needs. Chapters 1 through 9 cover the basic material. The last three chapters, on models of expectations, errors in variables, and model selection, are topics that instructors can pick and choose from. A one-semester course would cover Chapters 1 to 6 (or 7). A two-semester course would cover Chapters 1 to 9 and parts of Chapters 10-12. In either case Chapter 2 need not be covered but can be used as a reference.

Second Edition

1. Addition of matrix algebra. Many who have used the first edition found the omission of matrix algebra a handicap. Since the purpose of the book was to convey the basic ideas without using the "crutches" of matrix notation. I have decided to retain the original format and add matrix algebra in the appendixes to Chapters 2, 4, 5, 7, 9, and 12. The appendix to Chapter 2 gives all the necessary background in matrix algebra and there are some exercises at the end of the appendix.

2. Exercises on statistical inference. Chapter 2 presents a review of statistical inference. Those who used the book found the review too short, but expanding it would have made the book unnecessarily long. However, students are expected to have had a beginning statistics course. To make the review more useful, a list of questions has been added to Chapter 2. Students will get a sufficient review by attempting to solve these questions.

3. Addition of chapters on time series, unit roots, and cointegration. One major drawback of the first edition has been the complete omission of time-series analysis. This has been corrected by adding Chapter 13. In addition, the important recent developments of unit roots and cointegration have been covered in Chapter 14. There is currently no book that covers these topics at an introductory level.

Some instructors might want to cover Chapters 6, 10, 13, and 14 together because they are all on time-series topics. There is no need to cover the topics in the order in which they appear in the book.

At some places I have a reference to my earlier book: Econometrics (New York, McGraw-Hill, 1977) for details. I saw no point in reproducing some proofs or derivations where they were not absolutely necessary for understanding the points being made. At several others, there is a footnote saying that the result or proof can be found in many books in econometrics, and I give a reference to my earlier book with page numbers. However, the same result can be often found in other standard references, such as:

XVUl

PREFACE

J. Johnston, Econometric Methods (New York, McGraw-lHill), 3rd ed., 1984.

J. Kmenta, Elements of Econometrics (New York, The Macmillan Co.), 2nd ed., 1986.

H. Theil, Principles of Econometrics (New York, Wiley), 1971.

G. G. Judge, C. R. Hill, W. E. Griffiths, H. Lutkepohl, and T. C. Lee, Theory and Practice of Economics (New York, Wiley), 2nd ed., 1985.

E. Malinvaud, "Statistical Methods of Econometrics" (Amsterdam, North Holland), 3rd ed., 1976.

Since I did not find it practicable to give detailed page numbers for every book, I have just referred to my earlier book.

I would like to thank Jack Repcheck for first encouraging me to write this book and Ken MacLeod and John Travis for bringing the first edition to completion. 1 would also like to thank Caroline Carney for initiating this second edition and Jill Lectka and David Boelio for their encouragement in bringing it to completion.

1 would like to thank Richard Butler, Melanie Courchene, Jinook Jeong, Fred Joutz, Sang-Heung Kang, In-Moo Kim, Jongmin Lee, Wanki Moon, Marc Ner-love, Mark Rush, W. Douglas Shaw, Robert Trost, Ken White, Jisoo Yoo, and many others for their comments and suggestions.

1 would like to thank Kajal Lahiri, Scott Shonkwiler, and Robert R Trost for their help on the first edition.

1 would like to thank In-Moo Kim for his invaluable help in preparing this second edition. Without his assistance, this edition would have taken much longer to complete. 1 would also like to thank Ann Crisafulli for typing several versions of the additions and corrections cheerfully and efficiently. Finally, I would hke to thank all those who used the book and also passed on helpful comments.

G. S. Maddala