Friday, July 23, 2010

Paul Krugman and the limitations of historical analysis

I really like this post from Paul Krugman both for the reasoning and, more importantly, his awareness of the limitations.

So, about the first point: economic history is a great source of evidence about how the economy works – in fact, pretty much our only source. And the RR project, drawing on evidence from much further back and farther afield than usual among economists, was a great idea.

But there are usually major problems with historical analysis, no matter how much data you have, because it’s very difficult to isolate the things you’re interested in. There’s an old line to the effect that everything in the economy affects everything else, in at least two ways; this gives enormous room for spurious correlation. Econometrics is supposed to provide ways to disentangle the effects of multiple factors, but it’s difficult, and my sense is that few big arguments in economics have ever been settled by multiple regression analyses, let alone by all the sophisticated techniques developed these past three generations.

But Reinhart-Rogoff is relatively robust to these problems. Why? Because it focuses on extreme events. Financial crises are very big things, sharply concentrated in time. As a result, it’s reasonably certain that the economic developments in the aftermath of a financial crisis were driven by that crisis, not by other stuff that may have been going on.

Of course, not all economists are so careful when it comes to drawing their conclusions.

No comments:

Post a Comment