What are DSGE models good for?

1/10/2014 07:35:00 PM
One potential problem with the vast DSGE literature could be the Euler equation itself. Several papers have shown that even advanced non-linear approximation techniques can still obscure important, uncertainty-driven phenomena.
Noah Smith has a post lambasting DSGE models. His evidence is perhaps less than airtight, but still fairly compelling: there are many ways in which private investors could make bunches of money by having accurate policy-conditional forecasts, yet after searching far and wide Smith did not find a single investor using a DSGE model. In fact, even the CBO--which specializes in policy-conditional forecasting--does not use DSGE models. Since the whole point of DSGE models is to produce better conditional forecasts of the effects of policies, this is hard to figure. The market is telling us that DSGE modelling provides forecasts that are at best no better than other purely data-driven methods. The value added of a microfounded macroeconomic theory appears to be quite low.

Of course, maybe all these private investors and the CBO just haven't adopted DSGE modelling yet. This is probably a long shot, but would be interesting to see whether any DSGE model could have consistently let investors make better bets about the effects of government policies.

Matt Yglesias goes further to say that DSGE is nothing more than a pointless intellectual exercise that persuades nobody ever. This has me wondering: are there any times when I have been persuaded by a DSGE model? Here's a couple of instances where I have been persuaded to revise my beliefs/understandings as a result of DSGE models.

  • Capital taxation

    Perhaps the most influential result of DSGE macro has been the Ramsey-Chamley result that the optimal tax rate on capital gains goes eventually to zero. The actual Ramsey-Chamley papers aren't quite DSGE--they're missing the S--but Chari, Christiano, and Kehoe (1994) show this is approximately true in the in a stochastic scenario. Obviously, that's not the end of the story--this post and it's comments will essentially give you a long list of ways we can make this result disappear--but the Chamley-Judd results did effectively disprove what many had previously thought about capital taxation policy, even if there are important caveats to consider.

  • Tax cuts in a liquidity trap

    Matt Yglesias actually mentioned this one as an example of saltwater trolling, but for me, at least, this is a case where a DSGE model upset my priors. If you had asked me what I think about stimulating the economy through an income tax cut, my answer would have been totally different before I read Eggertson's paradox of toil paper versus afterwords. To be sure, I'm not saying that I think the model in Eggertson's paper would yeild accurate policy forecasts--my guess is that cutting income taxes would still be expansionary because whereas Eggertson assumes Riccardian Equivalence, in reality RE probably does not hold. Nevertheless, the paper does tell us useful information about policy: because of the paradox of toil effect, even without RE spending programs are probably much more expansionary than tax rate cuts. If I were a financier, I would have made two bets: 1) the payroll tax cut of 2009 and 2010 would not be highly expansionary and 2) the termination of that tax cut in 2011 would not be highly contractionary. On a cursory glance, I think the data has broadly agreed with those two totally DSGE-based predictions.

  • Price flexibility is largely counter-productive in a liquidity trap

    Krugman's DSGE model says this best, though a weaker version of this result is inherent in all New Keynesian models. Basically, while more flexible prices might theoretically take you back to the long-run equilibrium sooner (though not in Krugman's model, as I understand it), it actually makes the trough even deeper. This completely overturns what I had previously expected.

There are many others, but those are what came to mind at the moment.