Working Through Kennedy's Econometrics Textbook
I run across people asking about this book all the time, usually grad students who have been handed it by their program coordinator and don't quite know how to use it. Jeffrey Kennedy's An Introduction to Econometrics takes a different approach than most econometrics textbooks. Rather than deriving proofs before applying them, Kennedy introduces the material through concrete examples and computational exercises, often with datasets that mirror real research situations. The structure is not linear in the way you might expect. A chapter on simple regression will reference a multiple regression diagnostic from Chapter 7, which will ask you to check assumptions using data from Chapter 3. You work through problems while the theoretical pieces fit around them. It is not the cleanest layout, but it mirrors how actual empirical work happens.
A Guide To Econometrics Kennedy
If you are trying to find the companion materials, Kennedy's site has worked datasets, code snippets, and solutions for most editions. The textbook goes through several editions, so make sure your code matches your version. The third edition has slightly different notation in the instrumental variables section compared to the second, which caused confusion when I was preparing a course back in 2019. Students would paste the example code, run it, and get errors that made no sense until someone realized we were mixing editions. The book's strength is in the applied exercises. Each chapter ends with a set of problems that push you to actually estimate models rather than just manipulate expressions on paper. The weak spot, which nobody likes to admit, is the treatment of time series. If your program requires serious time series work, you will need supplemental reading. Kennedy covers ARIMA models at a surface level, and the section on cointegration is thorough but assumes you have already seen it done elsewhere. One specific issue I ran into: the treatment of heteroscedasticity-robust standard errors in the later chapters assumes familiarity with matrix algebra that the earlier chapters deliberately avoid. Students who skipped the appendix end up stuck trying to understand why the covariance estimator changes form once you add weighting. The workaround is straightforward. Go back to the appendix, work through the derivation for OLS with heteroscedasticity, then return to the chapter. It takes about twenty minutes and saves you from spending three hours confused.
The diagnostic plots in Kennedy are useful but occasionally misleading if you read them the same way you would a standard statistics textbook. The residual versus fitted plot behaves differently under certain forms of misspecification than it does under pure heteroscedasticity, and the book acknowledges this without emphasizing it enough. I stopped relying on visual inspection alone after a project where I spent two days chasing a heteroscedasticity problem that turned out to be an omitted variable. The Breusch-Pagan test was significant, the plot looked bad, but the root cause was a variable I had left out because it seemed irrelevant at the time. Once I added it, the residuals behaved. For learning the material, the book works best when paired with actual software. Kennedy provides examples in multiple packages, and I would recommend picking one and following along rather than switching between them. The syntax details differ, but the conceptual steps remain the same. Stata users get cleaner output formatting. R users get more flexibility with visualization. Both are adequate. The point is to pick one and stop second-guessing it after two weeks. If you are looking to download supplementary files, search for the official Kennedy econometrics website linked from the publisher page. Third-party sites host PDFs, but many are outdated. Edition mismatches cause real problems with problem numbers and dataset versions. I once had a student spend an entire weekend on an assignment only to realize the dataset he was using did not match the problem set because he downloaded the wrong edition's supplement. A few minutes of verification at the start would have prevented that.
Get the Full Details
The book is not the most rigorous text available. Wooldridge covers more ground with deeper theoretical treatment. Gujarati gets you through the basics faster if you want a traditional approach. But Kennedy is useful when you need to move from theory to application quickly, which is exactly the situation most graduate programs put students in during their first year. Read the appendices first if you are struggling with the notation. Work through the early problems before moving on. Treat the diagnostic sections as reference material you return to, not as chapters to read cover to cover in order. And do not skip the data exercises, even if they seem tedious. The ability to actually produce estimates with your own hands is what separates people who understand econometrics from people who can repeat definitions on an exam.
