The primary objective of the third edition of Essentials of Econometrics is the same as that of the previous two editions, namely, to provide a user friendly introduction to econometric theory and techniques. The intended audience is undergraduate economics majors, undergraduate business administration majors, M.B.A. students, and others in social and behavioral sciences where econometrics techniques, especially the techniques of linear regression analysis, are used. The book is designed to help students understand econometric techniques through extensive examples, careful explanations, and a wide variety of problem material.

Following the suggestions of many students and teachers, I have reorganized the topics discussed in this text and provided numerous new examples. Where necessary, computer printouts from various statistical packages are also included.

The salient features of the various chapters are as follows:

Part I now consists of four chapters rather than three, which are as follows: Chapter 2 reviews fundamental concepts of probability, probability distributions, and random variables. New features of this chapter are the introduction of the Venn diagram, Bayes' theorem, and the distinction between probability mass function (PMF) and probability density function (PDF). Several new examples are discussed in the text and in the exercises.

Chapter 3 discusses the major characteristics of probability distributions, such as the expected value, variance, covariance, correlation, conditional expectation, conditional variance, and skewness and kurtosis. The chapter then discusses how these characteristics are measured in practice in a given sample. This leads to a discussion of the sample mean, sample variance, sample covariance, sample correlation, sample conditional mean and conditional variance, and sample skewness and kurtosis.All these concepts are well illustrated.

Chapter 4 discusses four important probability distributions that are heavily used in econometrics, namely, the normal distribution, the t distribution, the chi-square distribution, and the F distribution. Major features of these distributions are discussed and illustrated in this chapter.

Chapter 5 discusses the twin topics of estimation and hypothesis testing. The major ideas introduced in this chapter are point and interval estimation, test statistic, the sampling distribution of test statistic, confidence interval, Type I error, Type II error, properties of estimators, one-tail and two-tail tests, null and alternative hypotheses, level of significance, and the p value. All these concepts are carefully discussed and illustrated. The ideas introduced in this chapter are critical for the study of econometrics in the rest of the book.

Part II, consisting of chapters 6 through 10, deals with the linear regression model, the "bread-and-butter" tool of econometrics.

Chapter 6 discusses the basic ideas of linear regression in terms of the simplest linear model, namely, the two-variable model. In such a model we study the behavior of one variable, called the dependent variable, in relation to another variable, called the independent or explanatory variable. A crucial distinction is made between the population regression model and the sample regression model, because in practice we have a sample from some population. In this chapter we show how to estimate the population regression model on the basis of the sample data.

Chapter 7 considers the topic of hypothesis testing in the context of what is generally known as the classical linear regression model (CLRM). This model makes several simplifying assumptions. We examine carefully the nature of these assumptions and their relevance.

Chapter 8 extends the two-variable linear regression model to the multiple regression models; that is, models that have more than one explanatory variable. We consider both estimation and hypothesis testing of multiple regression models, still working within the framework of the CLRM.

The linear regression model, whether two-variable or multivariable, only requires that the parameters of the model be linear; the variables entering the model themselves need not be linear. Chapter 9 considers a variety of models that are linear in the parameters (or that can be made so) but are not necessarily linear in the variables. In this chapter we show how and where such models can be used.

Explanatory variables included in a regression model are often qualitative in nature, such as gender, color, and religion. Chapter 10 shows how such variables can be measured and what role they play in regression modeling.

Part III, consisting of Chapters 11 through 14, considers several practical aspects of the linear regression model. The CLRM is based on several simplifying assumptions that may not hold in any concrete application. In this chapter we try to find out what happens if one or more of the CLRM assumptions are relaxed or not fulfilled.

Since an assumption of the CLRM is that the model chosen in practice is correctly specified, the fundamental question is how we find the correct model. In Chapter 11 we discuss this topic in some detail by first considering the attributes of a "good" model and then finding out the consequences of fitting the wrong model. We discuss various types of model misspecification errors.

One of the assumptions of the CLRM is that the explanatory variables entering a model are not linearly related. If there is such correlation among the explanatory variables, we face the problem of multicollinearity. In Chapter 12 we examine the consequences of multicollinearity, which is a commonly encountered problem in regression analysis.

Another assumption of the CLRM is that the error term in the linear regression model has constant variance (technically, this is the assumption of homoscedasticity). In cross-sectional data this assumption may not be always tenable. If the error variance is not constant, we have the situation of heteroscedasticity. In Chapter 13 we examine in depth the consequences of heteroscedasticity.

Another assumption of the CLRM is that the error term in the linear regression model is not correlated with its past value(s). If there is such a correlation, we have the case of autocorrelation. In Chapter 14 we show that this assumption may not hold in data involving time series. In this chapter we examine in depth the consequences of autocorrelation.

In Chapters 11 through 14 we adopt a common format: First we discuss the nature of the problem, next we discuss its consequences, then we discuss the diagnostic tools used to detect the presence of the problem, and finally we suggest remedies to correct the problem.

Part IV discusses two advanced topics. Chapter 15 considers the topic of simultaneous equation models and Chapter 16 discusses a potpourri of selected single equation regression models. The material in these chapters may be somewhat advanced for the beginner, and the instructor may assign these chapters on an optional basis.

Chapter 15 on simultaneous equation regression models considers situations in which there is interaction between variables so that the distinction between a dependent variable and an independent variable(s) is often blurred. In such cases we need to consider estimation methods beyond the commonly used method of ordinary least squares.

Chapter 16 discusses several topics on the increasingly important field of time series econometrics. In data involving time series, the critical concept is that of a stationary time series. In this chapter we discuss the meaning of this term and show its practical importance. In this chapter we also discuss the logit model in which the dependent variable is a dummy variable, taking a value of 1 or 0. In Chapter 10 we have already seen how to handle explanatory variables if they are dummy variables. However, unlike in Chapter 10, when the dependent variable is a dummy, there are some tricky estimation problems. In this chapter we examine these problems and show how the logit model can handle them. The logit model is used quite frequently in applied research.

MATHEMATICAL REQUIREMENTS

In presenting the various topics, I have used very little matrix algebra or calculus.I firmly believe that econometrics can be taught to the beginner in an intuitive manner, without a heavy dose of matrix algebra or calculus. Also, I have not given anyproofs unless they are easily understood. I do not feel that the nonspecialist needs to be burdened with detailed proofs. Of course, the instructor can supply the necessary proofs as the situation demands. Some of these proofs are available in my Basic Econometrics (McGraw-Hill, 4th ed., 2003).

SUPPLEMENTS AID THE PROBLEM SOLVING APPROACH

Supplements include a Student Solutions Manual with detailed solutions to the300+ end-of-chapter questions and problems in the text. Students can purchase it, and it is available free to adopters on a CD or at the Instructors Center on the website. In addition to providing the data to students and solutions to instructors, The Website contains the datasets from the text in a variety of formats and valuable weblinks for the student.For the instructor there are solutions to the questions and problems in the textand Powerpoints of the figures from the text. In addition to the data CD that is provided with the text, there is an Eviews CD containing a Student Version of Eviews and the data from the problem in Eviews format.

COMPUTERS AND ECONOMETRICS

It cannot be overemphasized that what has made econometrics accessible to the beginner is the availability of several user-friendly computer statistical packages. The illustrative problems in this book are solved using statistical software packages, such as Eviews, EXCEL, MINITAB, and STATA. Student versions of some of these packages are readily available. The data posted on the website and on the CD, however, can be read by many standard statistical packages, such as LIMDEP, RATS, SAS, and SPSS.In Appendix B we show the outputs of Eviews, EXCEL, MINITAB and STATA, using a common data set. Each of these software packages has some unique features, although some of the statistical routines are quite similar.

IN CLOSING

To sum up, in writing Essentials of Econometrics my primary objective has been to introduce the wonderful world of econometrics to the beginner in a relaxed but informative style. I hope the knowledge gained from this book will prove to be of lasting value in the reader's future academic or professional career and that the reader's knowledge learned in this book can be further widened by reading some advanced and specialized books in econometrics. Some of these books can be found in the selected bibliography given at the end of the book.

ACKNOWLEDGMENTS

My foremost thanks are to the following reviewers who made very valuable suggestions to improve the quality of the book.

Jin Man Lee, University of Illinois, Chicago
Houston Stokes, University of Illinois, Chicago
Yin-Wong Cheung, University of California at Santa Cruz
Harumi Ito, Brown University
Donald Waldman, Colorado University at Boulder
Reuben Kyle, Middle Tennessee State University
Harvey D. Palmer, University of Mississippi
Robert J. Gitter, Ohio Wesleyan University
Stephen Stageberg, Mary Washington College
Shady Khody, California State University at Pomona
Necati Aydin, Florida A&M University
Norman Swanson, Rutgers University
Richard Startz, University of Washington
Rachel Friedberg, Brown University

For their constant support and encouragement, I am thankful to the following colleagues in the Department of Social Sciences at West Point:

Brigadier General Daniel Kaufman
Colonel Russ Howard
Colonel Mike Meese
Major Paul Kucik
Major John Hansen
Dudley Dean
Dennis Smallwood
Don Snider

I am very grateful to Andrew Paizis for his help in the preparation of the Solutions Manual and to Jo May for all her help on the book. I am grateful to Lucille Sutton, my editor at McGraw-Hill, for helping me through all the editions of this book. I am also grateful to Jackie Grabel, editorial assistant at McGraw-Hill, for keeping me on track and helping me in many ways. I would be remiss if I did not acknowledge my debt to Jan Nickels, my copy editor, for doing such a superb job of editing a rather demanding manuscript, what with all those subscripts, superscripts, summation symbols, the acronyms, and, not to mention, my colloquial English.

Finally, I want to thank my wife Pushpa, and my daughters Joan and Diane, my sister in law Sushila Bushci-Gidwani, my brother-in-law Joseph Buschi, and my son-in-law Charles Chesnut for all their behind-the-scenes help and encouragement in the preparation of this edition. It is hard to overemphasize the family support.