Most Helpful Customer Reviews
|
|
103 of 104 people found the following review helpful:
5.0 out of 5 stars
Best "Intuition" for and Explaination of Econometrics, May 10, 2000
By A Customer
Unfortunately, I found this book at the END of too many PhD courses where I was swamped by assumptions the instructor and various authors were making. If I had only read this book FIRST - and then read Green, Greene, Johnston, Goldberger, etc., I would have gotten much more out of the courses with less stress. Peter Kennedy writes the type of book that students dream of finding (and I dream of writing someday). Each chapter is in 3 parts: 1) Overview of what and why, 2) Some more detail and 3) The nitty gritty that you'll worry about in Amemiya's book and others. This is the perfect book for PhD students interested in learning econometrics as a tool instead of an area of research (i.e. developing new models). Once you learn the basics, you can go to Greene's Econometrics Analysis for the details regarding implementation. My recommendation for "reading" this book is to whip through Part 1 of each chapter for the best overview of econometrics ever. (Peter Kennedy is an excellent writer, so this is actually an enjoyable and interesting experience.) Then revisit the relevant chapters before tackling the assigned readings for your course.
|
|
|
40 of 42 people found the following review helpful:
5.0 out of 5 stars
well written classic on econometrics, July 1, 2001
This is the fourth edition of a very popular text for an introductory graduate level course in econometrics. Although designed for econometricians and economics majors, the book has a lot to offer the statistician (time series analyst). There is good coverage of both the classical econometric models and the classical ARIMA time series models. The difference, as Kennedy points out, is that most univariate statistical time series models use only the past history to model and forecast the future while the econometric models emphasize the inclusion of economic predictor variables and not the past history. However, in recent years, and partly because in fair-fight forecasting competitions the Box-Jenkins time series methods have done better than the econometric models, the econometricians are beginning to incorporate the Box-Jenkins approach in their models. As Kennedy points out,the new theory of multivariate ARIMA models is providing the econometricians with a methodology that is similar to their simultaneous equation models. One nice feature of the book is that it treats classical linear regression theory early, highlights the key assumptions and then provides specific chapters that cover how to deal with the violations of the assumptions taken one by one. The book is clear, up-to-date and has an excellent bibliography. It introduces the structural econometric time series approach along with multivariate Box-Jenkins methodolgy. Advanced topics such as dealing with roots on the unit circle in Box-Jenkins models and cointegration are covered. Also robust estimation procedures are discussed. It even introduces bootstrap methodology and the Bayesian approach to inference. There is some coverage and some warnings about neural networks. Models for count data, duration, linear structural equations and instrumental variables are all presented in an introductory way. Emphasis is placed early on the concept of sampling distributions for estimators. A clear understanding of sampling distributions is essential to understanding classical frequentist statistical approaches. Much confusion can arise when these concepts are glossed over.
|
|
|
33 of 35 people found the following review helpful:
4.0 out of 5 stars
A good overview of the subject, April 14, 2006
Econometrics is now a respectable topic, both in the financial industry, where it is used extensively, and in academia. Like most efforts to model phenomena in the real world, especially those that attempt to model the behavior of human agents, econometrics has had its share of critics. These critics pointed out some of the failures of the econometric models, and some of their criticism was justified. However, there have been successes as well, if one realizes that the success of a model should be determined by what a model is actually developed for.
The author of this book is fully aware of what modeling is all about, and gives a very interesting overview of the major mathematical techniques used in econometrics. He characterizes econometrics as a study of how to obtain a good estimator in a situation or problem at hand that must be estimated. He recognizes that any criteria for what is "good" is somewhat subjective, but a "good" estimator it is generally believed must be computationally cost effective, unbiased, efficient, and robust. The author gives detailed discussions of these criteria in the book, and throughout most of the book more detailed mathematical derivations take place in the notes at the end of each chapter. The discussions can be a bit wordy at times in places outside of the notes for this reason. The book includes of course discussions on least squares, nonlinear regression, and Bayesian estimation of parameters. These are all topics that are fairly standard in the literature, but the author also includes discussions on topics such as neural networks and kernel estimation. An extensive list of exercises is included at the end of the book. For practitioners, the author includes a list of "ten commandments" that should be respected when doing applied econometric analysis.
No guide on econometric techniques would be complete without a discussion on how to analyze time series, and in this one that author points out the differences between how econometricians analyze time series and how traditional time series analysts do. The arrival of studies indicating problems with the approaches of the econometricians resulted in an explosion of research activity, some of which is reviewed by the author. This includes discussions of the Box-Jenkins method, ARIMA (autoregressive integrated moving average) models, VAR (vector autoregression), and error-correction (ECM) models. Interestingly, and close to the truth in practice, the author views model selection as being an art form, the correct choice of which is highly dependent on the experience of the modeler. Also interesting is his discussion on the `structural economic time series approach' (SEMSTA), which arose when econometricians realized their methods were being outperformed by Box-Jenkins methods, and which can be described as a synthesis of the two. When SEMSTA is simplified by omitting the moving average component, one obtains the VAR model. The author discusses in some detail the controversies behind the use of VAR, due to its assumption that all variables be endogenous. Both the ARIMA and VAR models are viewed as being successful in econometrics due to their ability to deal with the dynamics of the economy, even though they ignore the role of long-run equilibria. When terms are included in these models to represent the extent to which the long-run equilibrium is not met, one obtains the error-correction models. The author discusses an explicit example of how to obtain an ECM representation when there is linear relation occurring in the long run. Embedded in all the discussions on time series is the problem on how to deal with nonstationary data, the latter of which econometricians ignored historically, due to their belief that econometric analysis was not affected by nonstationary variables, and due to the unavailability of studies that indicated that most macroeconomic data obeys a `random walk' and is therefore nonstationary.
The author also gives a brief outline of forecasting techniques in econometrics and how to assess their accuracy. He emphasizes that the choice of how to evaluate the accuracy of the forecasting model depends on the actual purpose of the forecast. If a large degree of error can be tolerated, this may motivate the choice of one criterion for accuracy over another. Unfortunately forecasting is viewed by many as an activity that should guarantee high or even infinite accuracy. Since no forecasting model can guarantee this, and since a perusal of the historical record on forecasting shows that most of them have "missed the target", forecasting is viewed with ever-increasing skepticism (this is especially true for the current controversy over climate forecasting and global warming). There needs to be an objective study that compares the accuracy of the forecasting models and which also compares their utility in prediction over and above what is typically called "intuition" or some other equally subjective ability. Other than a brief discussion on neural networks, the use of machine intelligence to do forecasting is not discussed in the book. It is becoming more popular to use artificial intelligence in forecasting, but it remains to be seen whether using it is more advantageous than simulation or Monte Carlo techniques, both of the latter being dependent essentially on randomization and requiring minimal intelligence.
|
|
|
Most Recent Customer Reviews
|
by Dominick Salvatore
by Jeffrey Wooldridge
by Jeffrey M. Wooldridge
by Jeffrey Wooldridge
by Fumio Hayashi
See all Editorial Reviews
