This is a very good introductory econometrics textbook for the mathematically well-prepared. No prior knowledge of econometrics or statistics is assumed, and the discussion of the necessary probability and statistics concepts is integrated into the main text rather than being relegated to appendices. All you need to read this book is a good knowledge of linear algebra and calculus. Once you finish it you will have a firm grasp of the basic methods and models used by econometricians and be prepared for going to more advanced sources like Wooldridge's Econometric Analysis of Cross Section and Panel Data or Hamilton's Time Series Analysis

Throughout the book Davidson and MacKinnon focus on developing intuition rather than on mechanical calculation. In particular, their geometric approach to ordinary least squares estimation is a must read. By focussing on the geometry and making clever use of the Frisch-Waugh-Lovell theorem, they make the properties of OLS very intuitive. Many of the standard results usually proved by opaque matrix algebra in other books, become clear and easy to prove in this framework.

The book also has the advantage of covering topics like GMM estimation, the bootstrap and numerical methods that cannot be found in older textbooks.

Yet, I have three quibbles with this book.

The first, minor one, is that its treatment of time series methods is too short, and unlike the rest of the book tries to trade off depth for breadth.

The second, bigger problem with this book is that it is entirely about econometric 'theory'. It teaches you how to find estimators and test statistics with good properties for particular models. But it does not train the student at all in the applied/methodological aspects of econometrics: given that I have a vague question about economic phenomena in mind, and given a bunch of data, how do I proceed? What questions can be meaningfully asked, how to choose between alternative models, how to present and interpret results, are questions that are given a short shrift in this book. Even data-based exercises are few and seem to have been reluctantly included.

The third problem with this book is that it completely ignores the Bayesian approach to econometrics. Though this is in line with the general frequentist dominance of the econometrics profession, I feel that without at least an introduction to the Bayesian approach, the training of an econometrician will remain one-sided.

The first two shortcomings of this book can be addressed by complementing it with Hayashi's Econometrics. Many interesting papers on methodology can be found in the book Modelling Economic Series edited by Granger.