It's taken me a few weeks to work through Joshua Angrist and Jorn-Steffen Pischke's Mostly Harmless Econometrics: An Empiricist's Companion. Although econometrics was one of my specialisms in graduate school, a million or so years ago, no econometrics text was going to be a light read. But I must say this one came quite close. I'd recommend it to the entire range of empirical economists, from those still in training to those who, like me, have only a hazy memory of statistical theory and stick to our tried and tested methods of estimation. For although it's by no means a comprehensive treatment of econometrics – for example, you'd certainly want to supplement it with Hendry-style time series methods – it is an excellent guide to how to do basic regression/IV/panel data estimation really well. In particular, it demonstrates through many examples how to bring about a happy marriage between one's underlying model and the data which might or might not confirm the researcher's hypotheses. There is a strong emphasis on clarity of thought about how to approach testing a model, including the fact that tests of causal relationships imply a counterfactual world, and about what estimation strategies will allow that hypothesis to be tested rigorously. Both statistical significance and economic significance are taken into account. All the regression results presented are discussed not only in terms of their associated statistics but also – and how rarely one sees this – in terms of what the size of the coefficients indicates and whether the results are meaningful. I think the authors perhaps over-claim for the range of estimation problems that can be addressed in the way they set out, but this is a minor quiblle – especially as they include some jokes. Funny jokes, not just anoraks' jokes!

For people like me who defend economics against its legions of critics in the public debate, it is a constant worry and embarrassment that our subject is marred by so much very poor quality empirical work. Here is a very accessible book (relatively speaking, anyway) which can perhaps make some inroads into the harmful econometrics practised pretty widely by the profession. I also worry that many of us as we age fail to keep up with the econometric work done by younger colleagues; but just as senior bankers really should have made sure they understood CDOs and SIVs, we really should understand the technical basis for the empirical results we read about. Finally, it's a very reasonably priced trade paperback. No excuses!

The book was also usefully and favourably reviewed by Andrew Gelman, who writes an excellent statistics blog.