Jan 12, 2009

Mr Keynes, Econometrics and the Economic Method

Does your statistical model have the correct specification? Are you sure your dependent variables do not suffer from collinearity? Do you think the linear specification is appropriate? How did you choose the number of lags in your equation? Did you check the structural stability of your estimated equation? How do you know your variables are not measured with errors? How do you account for non-measurable variables which may be crucial to your specification?

These set of questions is today in the mindset of any empirical economists (and it's also an easy set of questions to pose if you ever have to attend a conference in which an empirical paper is presented..). Anyway, these are the problems Keynes identified in a famous comment named "Professor Tinbergen' s method" appeared on "Economic Journal" in 1939. This comment provides an unusually harsh critique (unusual for academic standard) to Professor Tinbergen 's 1939 report to the League of Nations, "Statistical testing of the Business Cycle Theory", and in particular, to the chapter named "A method and its application to Investment Activity".

Econometrics has advanced a lot since then, and it is quite striking to see how Keynes' critiques have represented a sort of research agenda for future generation of econometricians. Keynes in his Comment, was also more fundamentally skeptical about statistical testing of economic theory, especially when it comes to the business cycle. How can one make sure that the model is all-encompassing? What type of induction can we make if different econometric specification can be derived from theory? How can you avoid the "slippery problem" of passing from an estimated model to casual inference?

Subsequent work has reduced the problem of casual inference by expanding the role of proabilistic models, in either classical or Bayesian setting, as Haavelmo pointed out in a 1943 reply, leading to the final take off of the discipline. What made Keynes so harsh toward his fellows? As explained by Garrone and Marchionatti in this paper, Keynes was harsh because he was skeptic about the path Economics was taking in the 1930s: it was seen as a tool to formalize mathematically and quantify statistically a complex sytem. This skepticism is still part of the discussion about the correct method of economic research. In his view, the economist's method should entail the use of statistics as a tool to provide a preliminary evidence to our logical thinking, to make less "impressionist" judgements. He did not discard Econometrics, but he thought of it as a way to test a certain theory derived which was to be derived from a small set of assumptions. He thought economic thinking should not be too formal, if this entails assumptions which are non plausible. So better to be ad-hoc and right than to be formally wrong. In conclusion, while econometricians took Keynes' critique at full value, economists have been less benign with him. It should not come as a surprise then, that still today, Le Figaro voted Keynes man of the year.

No comments: