Mar 27, 2008

Does Advanced Econometrics Have any Sex Appeal ?

Hello Everyone,

now that I have everyone's attention, I would like to bring up a topic that concerns anyone pursuing empirical research in their Ph.D. At one point or another we have all set down and wondered where the credibility of any econometric research is, given that once one goes beyond OLS it is possible to create almost any result.

While this topic is not as recent as the sub prime crises I think that it is worth discussing it a bit.
There is a very nice paper by Larry Summers : The Scientific Illusion in Empirical Macroeconomics. While the paper is from 1991, he brings up a very important point. Namely that most of the research applying very advanced econometrics does not really advance our thinking professionally. One really good example he gives is about the short run real effects of monetary policy. In a pioneering study Friedman and Schwartz show the real effects of monetary policy with a very simple narative study. In fact this really changed thinking about economics as a science, but do VARs have the same contribution?

I think it would have been good for me to have read this at the beginning of my Ph.D. rather than towards the end..... I do tend to agree with Summers that sometimes in empirical research, the method starts to dominate. I mean there are very, very many VAR studies for instance, but how many of them did actually contribute to the way we think about economics?

These days I tend to think that the best empirical papers will change the readers perception of the world. Look for instance at the original sin literature. Fantastic contribution with just OLS regressions.

An even older piece of research about econometrics and whether it is a science or not is also David Hendry's classic article: Econometrics: Alchemy or Science? from 1978 in Economica.

Enjoy and share your opinion about modern econometrics in applied research.


Pierre-Louis said...

I couldn't agree more...Levitt has some brilliant papers that use very simple Econometrics... Econometrics can be used with brillance, and most of the time, this is simple maths but super intuitive arguments...for example, when some guys wanted to study the link between trust and growth, instead on focusing on crazy econometircs to control for endogeneity, they used the right data: trust witihin second generation migrants...heres the Vox column

Sebastian said...

Amazing, finally a point we can all agree upon. I completely subscribe to this point as well. In any case data is often "constructed": just think about quarterly government deficits. Once, you heard how this or other data is created, you stop wondering why some people find certain patterns using such data...I guess many know what I talk about.
In either case simple descriptive statistics and OLS can go a far way. Nevertheless, there are cases where there is reason enough to depart from this technique, but the end of the exercise should remain on a research question which goes above applying fancy techniques.

Sebastian said...
This comment has been removed by the author.
Cam said...

Fascinating articles by Hendry and Summers on the relevance of econometrics.

I agree completely .......while doing the literature review for my Masters Thesis, I found so many instances where people would use bright shiny new crazy econometric techniques seemingly without any reason, apart from that they were new and might be fun to implement....sometimes the use of crazy empirical techniques seem so unnecessary and completely divorced from reality..