Tuesday, July 28, 2009

DSGE frustrations

This is a follow-up on an earlier post on DSGE models. The Economist while less shrill than Willem Buiter they portray an somewhat accurate picture of the state of many DSGE models:

In many macroeconomic models, therefore, insolvencies cannot occur. Financial intermediaries, like banks, often don’t exist. And whether firms finance themselves with equity or debt is a matter of indifference. The Bank of England’s DSGE model, for example, does not even try to incorporate financial middlemen, such as banks. “The model is not, therefore, directly useful for issues where financial intermediation is of first-order importance,” its designers admit. The present crisis is, unfortunately, one of those issues.

The bank’s modellers go on to say that they prefer to study finance with specialised models designed for that purpose. One of the most prominent was, in fact, pioneered by Mr Bernanke, with Mark Gertler of New York University. Unfortunately, models that include such financial-market complications “can be very difficult to handle,” according to Markus Brunnermeier of Princeton, who has handled more of these difficulties than most. Convenience, not conviction, often dictates the choices economists make.

Convenience, however, is addictive. Economists can become seduced by their models, fooling themselves that what the model leaves out does not matter. It is, for example, often convenient to assume that markets are “complete”—that a price exists today, for every good, at every date, in every contingency. In this world, you can always borrow as much as you want at the going rate, and you can always sell as much as you want at the going rate.

Some points however:

1. DSGE macro models with financial intermediation are not trivial to build. If they were we would see more of them. Bernanke-Gertler type models while useful in looking at the financial accelerator channel have not been featured at all (?) in the current crisis.
2. In general, DSGE models are hard to extend hence the statement: "they prefer to study finance with specialised models designed for that purpose" reflects this situation.
3. Complete markets and all that are assumed yes, but even 10 years ago when I was an graduate student, economists were beginning to build incomplete markets models with heterogeneity (e.g. Huggett, Aiyagari, Krusell, etc.). Unfortunately, these economists prefer to use these models to "explain stylized facts" (mainly income distribution) rather than to build a full blown macro model.

The article further goes on to quote David Colander who thinks the future is in agent based models. I am partial to this but I think in the medium term the return might be into medium scale econometrics models that fell out of vogue due to the Lucas critique which while devastating remains merely a theoretical argument without real empirical foundations especially in short and medium run models.

No comments: