There’s a good debate going on about the usefulness of macro models, and in particular whether the so-called New Keynesian models let us down or even helped bring on the financial crisis and the Great Recession.

This weekend Noah Smith argues that the new Keynesian models were built for the “last war” and thus missed macro-financial connections or the zero interest rate bound needed to understand the Great Recession and slow recovery. Simon Wren-Lewis writes that the models omitted key factors by focusing too much on rational expectations and other micro-foundations which “crippled the ability of New Keynesians to understand subsequent real world events.” And back last summer Paul Krugman in Macroeconomists at War wondered if these models were being used much at all, saying “It would be interesting to know how many graduate departments were in fact teaching New Keynesian macro in 2008. My guess is that a fair number weren’t.” Underling the debate is an implicit view that New Keynesian models deserve much of the blame for the poor economic performance in recent years, and should be “taken to the woodshed” as Chris House put it.

For what it’s worth, here are my thoughts on these issues based on years of building, using, and teaching such models.

A defining characteristic of New Keynesian models is the combination of “wage-price rigidities” and “rational expectations.” Early examples are the stylized models by Stan Fischer and Ned Phelps and me (published back-to-back in the JPE) and the empirically estimated econometric models based on staggered price setting. Old Keynesian models prior to this did not have rational expectations, and new classical models did not have sticky prices. The Mankiw and Romer book New Keynesian Economics was a collection of such models. An up-to-date compilation of models since then—with special focus on empirical models used in practice for policy in many different countries—is contained in Volker Wieland’s model data base, and the new paper by English, Lopez-Salido, and Tetlow explains the Fed’s current New Keynesian model.  Of course some New Keynesian models are better than others; some are still quite stylized, but others are fully estimated and fit the data well.

What was the motivation for the new models? They were originally designed to be an improvement over existing policy evaluation models (old Keynesian and new classical) which could then be used to find better policies than the failed policies of the 1970s. In these models, monetary policy had important effects and rational expectations techniques made the models more useful for many policy issues. The new models were not designed to explain the Volcker recessions of the early 1980s. As defined here, the models came before those recessions.

What about the rules versus discretion debate? As a technical matter, the very nature of the rational expectations assumption built into the models meant that the policy evaluation was originally about alternative policy rules rather than discretion—interest rate rules, money growth rules, exchange rate rules, etc.

What about the zero bound? It was not ignored.  From the start of research on interest rate rules, the zero bound on the interest rate had to be taken into account in the policy rule simulations. In my 1993 book (Chapter 6) the interest rate was set at a small positive value (say .01) whenever the formula gave a rate below that value.

What about graduate teaching? Some graduate programs began teaching such models early on. Jim Tobin asked me to team teach a graduate course with him in 1980 to get the ball rolling at Yale. Fischer and Olivier Blanchard taught the models at MIT. Ben McCallum taught them at Carnegie-Mellon. David Romer, Carl Walsh and Mike Woodford wrote textbooks which featured such models. Several Stanford grad students (John Williams, Volker Wieland, and Andrew Levin) learned about the models and went to the Fed where they helped introduce them there. I continued to teach a course based on these models after I returned from Washington in 2005 and still do. Lars Svensson, Jordi Gali and others were teaching the models in Europe. Of course the ideas took a while to catch on in some schools, and in a few cases they never caught on.

What about the macro-financial connection?  Research by Tobin and Brainard, Gurley and Shaw, Brunner and Meltzer was available when the new Keynesian models arrived, and Ben Bernanke, Jordi Gali and Mark Gertler had already made enough progress in the 1990s that Mike Woodford and I included it in the Handbook of Macroeconomics (Volume 1C, Chapter 21), published in 1999 long before the Great Recession. (Mark Gertler and Nobu Kiyotaki will be doing a “follow-up” chapter in the forthcoming new Handbook of Macroeconomics edited by Harald Uhlig and me.)   But my experience was that when it came to fitting models to data the lack of good flow of funds data led researchers to focus on financial market prices rather than quantities, with the financial frictions appearing as interest rate spreads. Recent theoretical research and better data are improving this situation. Of course, economists should try to improve their models to make then more useful for policymakers, and it is certainly a worthy task to improve our understanding of the credit channel.

One can criticize the rational expectations/sticky price models by saying that they do not admit enough rigidities, or have only one interest rate, or do not have money or credit in them, or simplify too much with rational expectations. But simplified versions of models, which frequently boil down to only three equations, should not be confused with more detailed models used for policy. Many of the models listed in Volker Wieland’s model data base are more complex, have time varying risk premia in the term structure of interest rates, an exchange rate channel, and more than one country.

I do not see the evidence that these models led policy makers astray or were a cause of the financial crisis. To the contrary I have argued that the general policy recommendations of these models—which generally took the form of particular monetary policy rules for the interest rate instrument—were not followed by policy makers in the years leading up to the crisis though they followed them during the Great Moderation. Ignoring the recommendations was the problem rather than the recommendations themselves. These models did not fail in their recommendations. Rather the policymakers failed to follow the recommendations.