Could be, but I doubt it.

Photographer: Scott Eells/Bloomberg

Krugman Gets the Big Picture Right

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion.
Read More.
a | A

In about 1980, there was a big change in the way academic economists did macroeconomics. The old method was to write down a system of equations representing macroeconomic quantities (gross domestic product, investment, etc.).  The new way was to write down an optimization problem, representing the decision-making of one or more agents in the economy. This approach is called DSGE (short for Dynamic Stochastic General Equilibrium). The old approach is sometimes called IS-LM which refers to the name of two curves in a graph in a classic model; they are also referred to as Old Keynesian models.

 Since 2009, there has been a big debate in the economics blogosphere about whether academic econ took a wrong turn when it switched tactics. The main advocate of this view is Paul Krugman, who is a force of nature unto himself. According to Krugman, IS-LM models do a good job of explaining the essential qualitative features of what we’ve seen since the 2009 financial crisis. Krugman’s most prominent opponent in this debate is University of Chicago professor John Cochrane. In a recent post, Cochrane calls for advocates of the IS-LM approach to back up their claims with quantitative data and publish it in reputable peer-reviewed journals:

The right way to address this is with models -- written down, objective models, not pundit prognostications -- and data. What accounts, quantitatively, for our experience?…Models confront data in the pages of the [American Economic Review], the [Journal of Political Economy], the [Quarterly Journal of Economics], and Econometrica. If old-time Keynesianism really does account for the data, write it down and let's see.

It’s a legitimate challenge. But in fact, the challenge has already been met! In 1992, macroeconomist Jordi Gali, then of Columbia University, published a paper called “How Well Does the IS-LM Model Fit Postwar U.S. Data?” in the Quarterly Journal of Economics (QJE).

In that paper, Gali compares an IS-LM type model with a more modern type of statistical model called a structural vector autoregression, or VAR (this is the type of model that Christopher Sims won the 2011 Nobel Prize for inventing). Gali uses the VAR to pick out “shocks,” or surprise events, that happened to the U.S. economy in the past. He then analyzes how the old-school IS-LM model predicts that the economy will respond to those shocks, and finds that the predictions are pretty accurate.

In other words, even as far back as 1992, it was clear that IS-LM type Old Keynesian models fit the data pretty well.

So why haven’t academics kept using IS-LM models? Why haven’t they even followed up on the 1992 Gali paper? One reason is because of the Lucas critique.

The Lucas critique -- named after University of Chicago economist Robert Lucas -- says that even if a model fits the data, you can’t use it to make policy recommendations. Suppose you look in the past and see that every time the government increases spending, growth goes up. Doesn’t that mean that fiscal stimulus works? Not necessarily -- maybe it only works when people aren’t expecting it! Maybe if you raise spending on purpose in order to try to boost GDP, the correlation will suddenly break down.

I’m not saying that this is actually true of fiscal stimulus. But in general, it could be true of any policy action. This implies that if you really, really want to be sure that your policy will have the intended effect, you need to use a model that is totally structural -- a model that won’t suddenly change when you try to use it, because it’s based on things that don’t change when policy changes, such as consumer tastes or technology. Modern DSGE models are generally assumed to be structural (though some have their doubts). Old-style IS-LM models are obviously not structural, and that’s why they were abandoned, not because they failed to fit the data.

So Krugman is probably right that IS-LM models explain the past as well as anything else we’ve got. Cochrane is right to demand that IS-LM be tested against the data, but so far those tests have turned out favorably. On the other hand, if you think the Lucas critique is a big deal -- as most academic macroeconomists do, rightly or wrongly -- then just fitting the facts isn’t enough. You need to explain why those facts will hold up even if the government changes its policies. IS-LM, in its traditional form, doesn’t come with any such explanation. And unless and until it does, academia is unlikely to embrace it.

As a final note, the Federal Reserve Board still does use something quite a bit like an IS-LM model. For the Fed, at least, the Lucas critique isn't the only game in town.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the author on this story:
Noah Smith at nsmith150@bloomberg.net

To contact the editor on this story:
James Greiff at jgreiff@bloomberg.net