Modern Macro Fails the Market Test

Some fresh, fresh Minnesota water.

Photo by David Cannon/Getty Images

I wrote in December about how “freshwater” macroeconomic models have failed the market test and are never used in private industry even though a model that actually did what these models purport to do would be extremely valuable. Noah Smith extends the point in two ways, one by rebutting academics’ most common excuse for why the private sector shuns these models and the other by observing that it’s actually all dynamic stochastic general equilibrium (DSGE) modeling that the private sector shuns, not just the freshwater variety.


It seems to me, though, that this is really just a problem for freshwater macro. Saltwater DSGE macro, as best as I can tell, is just a kind of highbrow trolling.

Freshwater macro is trying to derive a substantive conclusion (policy intervention to smooth the business cycle is counterproductive) from a methodological point. Saltwater DSGE is trying to show that you can get salty results from freshwater methods, so the freshwater program fails. But this is because it turns out that a competent practitioner can produce a DSGE model that proves anything at all about the world. Brown University’s Gauti Eggertsson is a specialist at this. Want a model in which people becoming harder-working is bad for the economy? He’s got one. Or a model in which the way to end a depression is to deal crippling supply-shocks to the economy? He’s got one. I’m not saying these models are wrong, it’s just that a basic survey of the literature is going to teach you that smart modelers can model whatever they want. Your actual reasons for believing one model or the other would have to come from somewhere else. Sensible people see that as a matter of historical fact the New Deal gestalt succeeded in greatly increasing economic growth, so they’re interested in investigating exactly what aspects of the New Deal had that result.

And that’s, I think, what you see in the private sector’s rejection of DSGE modeling. Even in the academic work, the models themselves aren’t genuinely driving anyone’s thinking so you might as well skip to whatever it is that’s actually changing minds.