Customer Reviews

Estimation and Inference in Econometrics

5 star:		(6)
4 star:		(4)
3 star:		(2)
2 star:		(2)
1 star:		(0)

 

 

This is one of the best books on econometrics published in the past few years. The authors use the theory of vector spaces (projection operators in a Euclidian space) to show how the intuition behind the General Linear Model extends in a natural way to more complex nonlinear models. The authors demonstrate that sophisticated maximum likelihood (or simulated maximum likelihood) estimation algorithms are essentially repeated applications of the linear projection operators seen in regression context. The result is a unified theory of econometrics which takes readers from a "cookbook" level of statistical sophistication to a more mature "model building" orientation.

In short, this is one of the most refreshing treatments of econometrics I've seen in many years. University instructors -- particulary those teaching doctoral level courses -- should seriously consider adopting this as a text.

We were recommended to use this book as a complement to Hayashi, which we had used as our initial primary text for the 2nd and 3rd quarter of a first-year graduate econometrics sequence.

I think I would have found the exposition here rather challenging had this been my initial text. A few comparisons between the two books:

H - GMM as organizing principle.
D&M - Least squares as organizing principle.
I think the latter was in many ways a more intuitive way of viewing these techniques (for me), but perhaps provides a less fully integrated view of the estimators.

H - Matrix algebra and first order conditions as justifying estimation techniques.
D&M - Geometric projection as justifying estimation techniques.
The geometry is a powerful tool for understanding these concepts, but I think serves me better as a complement rather than a primary motivator.

H - Treats homoskedasticity and lack of serial correlation as special cases.
D&M - Treats heteroskedasticity and serial correlation as extensions of iid models.

H - Treats nonlinear models as extensions.
D&M - Treats linear models as special cases.

H - Offers a large number of economic applications.
D&M - Basically entirely theoretical in its justification of theorems and techniques.
This would be among the most frustrating things about using D&M as a primary text.

Just a few thoughts that might be useful to someone considering this book. The organization around least squares is very useful, I think, and a geometric intuition for econometrics must be a powerful tool as one progresses in the field.

I do not know better book on nonlinear estimation and inference in econometrics.

Overall the book is very well written and relatively easy to understand, considering its subject. However, if you have not been introduced to linear econometrics, the book can become very hard, mainly if the reader is not acquainted with matrix algebra.

The first chapter on the geometrics of regression is simply marvelous, although a better picture is in Ruud's.

The style is someway formal, but different from the traditional lemma-theorem-proof-corollary way. This makes the book easier to read.

Future improvements include:

a. More examples (please);
b. Make the early 2 chapters on asymptotics clearer;
c. Extend the GMM approach interconnecting it with other chapters (it's more general);
d. Put exercises, with solutions, with selected solutions, whatever, but exercises, including computational ones;
e. Some economics - this does not mean applications per se, but it means to explain where and why such techniques are necessary in the real world.

In the first year of my PhD, teachers recommended this book. I must admit that I don't appreciate too much the geometrical approach of several topics. Instead, I liked a lot the introduction given to unit root analysis and to simulation procedures. I think it's a more readable book than Greene's, and much more fun to work with. I conclude that this book is very useful for people beginning a PhD, but not for undergraduate people.

Green's textbook was the assigned text when I took my econometrics sequence. Like many others, I found it not well written and the explanations are pretty bad. Also, Green's is priced sky-high (around $100 for a brand new copy).

Davidson and MacKinnon is different. Both expositions and explanations are clear and easy to follow. I was so delighted after picking up a copy from the libarary. This is the one econometrics students should have. The price is also hard to beat. The reason I think it is not widely adopted is because of the geometric analysis of regression (Chapter 2). But if you don't like geometrics, you can simply skip it.

An improved version of this book is just published under the new title "Econometric Theory and Methods". This new version contains a chapter on unit-root and cointegration, as well as some new numerical methods. I urge interested buyers to take a look at the new version.

As several other readers, I am not crazy about the geometric approach used in this book, although it is certainly very original. I love the the way the non-linear estimation is discussed. True, it is much better than Greene...

It is a very readable presentation of econometrics, although maybe lacking some mathematical rigor. The geometric view is an excellent instrument for showing "how things work" when talking about least squares. As a consequence, it opens the reader's mind to generalized least squares and its geometry parallel, bringing us a very clear and pleasant reading. A good companion book would be Econometric Methods, by Prof. Johnston.

This book is good for students of economics and statistics , is advanced , the mathematics necesary are mediun level, has great deal dedicated for optimization and is intituive.

It is a very good theory book for 1st and 2nd year graduate students. I suggest to read this book after Green or Davidson & MacKinnon to learn the intuition behind basic techniques in econometrics.

It's a nice piece of work.
There is no one like this.
The only problem is the way the contents are presented. There is no a logical order that help us in a course. I agree that there is not a clear structured inside the chapters or in the entire work. But this is the book that reach the deepest point being readable. Another books are better structured or more intutive but too superficial or old-fashioned.
With the modern computers and software the old classical books based on small sample theory are unsuitable. Davidson and MacKinnon point us to the econometry of the future.
It would be a good idea to combine this book with Berndt's one on applied econometrics, plus a good software like Stata 8 or matrix-based programming software like MATLAB.
That's the best way to access the econometry.