In this episode of “Where is the General Theory of the 21st Century?”, we talk to Professor Dani Rodrik, the Ford Foundation Professor of International Political Economy at Harvard’s John F. Kennedy School of Government, and leading economist on research covers globalization, economic growth and development, and political economy.
–On Macroeconomics and the “One true Model”–
Q: You have mentioned in your book that even Keynes’s General Theory is not a grand theory according to the above definition. May I ask if you think the case is that economics, given its nature as a social science, is not suitable in aiming to generate grand theories?
R: I do think that trying to come up with a “general theory,” say, of business cycles or employment, is not just a fool’s errand in the sense that it is not feasible; I think it also gets us astray. Because what happens is, instead of entertaining a variety of models that might be applicable at different times, we tend to fixate on a particular model that does well for a while; then, when the nature of the underlying realities change, and we need to shift the focus to another model, we are caught unprepared.
This is in some ways the history of macroeconomics. We discarded the Keynesian Model in the 1970s when we have a series of shocks on the supply side, and the Keynesian models did not seem to do a very good job explaining the simultaneous rise in inflation and unemployment. We built new models that were much more careful about expectations and its formation, which were in advance over the Keynesian Model in certain dimension; but it was also a regress in other dimensions because it downgraded the role of animal spirit, investor psychology, multiple equilibria and the demand side.
Then, when we find ourselves in the 2008 financial crisis, where we had demand-led recession and unemployment. All our models were sort of the neo-classical varieties, and they are not particularly helpful.
I still see this tendency among macroeconomists to develop the “one true model.” I think this is an unhelpful way of thinking about macroeconomics.
We need models with all the intertemporal bells and whistles, and in some kinds of policy analyses, those would be extremely important. In some other types of applications, ISLM would work just as well. Instead of trying to come up with ‘the model” – the grand unifying model – we need to carry in our mind a collection of models. Depending on the context, the nature of shocks, and which causal mechanism seem to be much more important, we try to do a better job of figuring out how to switch from one model to the next.
Q: You have mentioned the diagnostic approach in your book. Do you agree that in fields other macroeconomics, economists are more acceptive to the “diagnostic approach” which incorporates a set of models, instead of pursuing the “one true model”?
R: I think in some parts of the profession, the tendency to search for “the model” is relatively less strong.
In general, the more explicitly empirical and explicitly microeconomic the field is, it has less urges for the universal models. In labor economics and development economics, in particular, where the institutional details of the application highly matter; and the empirical tools of randomization or well-identified models have become very common.
In these areas, labor and development, I think there are much greater willingness to entertain multiplicity of models, and not be looking for the single model. I think it is a question of degree other than kind. There is a general tendency to think that the most recent model has been the most relevant ones. However, it is still true that macro probably is where I see the desire for “what is the right model” most prominently displayed.
Maybe that is because in some sense macroeconomics has failed us. The latest generation of macroeconomics thinking had very little to say about the possibility of, say, the global financial crisis and how to react to it. I think there is a greater sense of crisis in macroeconomics that macro has failed after a period of self-congratulation.
Just before the great financial crisis, macroeconomics has been in a very self-congratulatory mood. Macroeconomists think that the business cycle has been conquered; they think they understood the basics of it, and have a good idea of how fiscal and monetary policy ought to be used, and so forth. So perhaps, the disappointment is the greatest in macro.
I think the reaction has been the wrong one. The problem was not that the neo-classical model and all the dynamic optimization frameworks were completely worthless. It was just that we were applying the wrong model. We should have gotten a little bit away from Lucas and Surgent, and gone back toward Keynes or the New Keynesian models, or some other models with financial frictions.
I think what distinguishes a good economist from a pure ideologue is that a good economist can navigate between different mental frameworks and different models and knows that he has to be flexible. The framework he chooses is driven by the circumstances, not from prior notions of what a good model is.
Q: Some macroeconomists I have talked to would counter your arguments, saying that it would create many ad hoc theories.
For example, in development economics, the use of experiments possibly generates many ad hoc results that can’t be generalized. What do you think there could be a problem that if there are too many ad hoc theories, we can be able to choose among them?
R: No, I think what we are lacking is the way to systemically thinking about how to use the different models that we have. How do we distinguish between the models that have a chance of being useful, and those are not going to be useful, so we can navigate among the models that we have.
Obviously, if we have thousands and thousands of models, that is not a practical solution. However, I think the kind of craft, or the art, of economics, is the ways that we use the models that are the ones that proved handy; therefore we use them more, without necessarily discarding those that are “useless.”
The notion of ad hoc is very misleading in this kind of context because every model is ad hoc, in the sense that every model is a mix of a massive number of simplifying assumptions.
What is more ad hoc? To assume a representative agent that lives infinitely? Or to assume that a consumption function where individuals consume a fixed fraction of their income? I can imagine circumstances that the first assumption is more out of place than the second one. When you are thinking about issues of dynamic about future policy changes, we may want to use a model with forward-looking individuals; assuming representative individual with a very long horizon may not be a bad approximation under certain circumstances. It might be an appropriate assumption.
However, in other cases, when you have different groups of consumers that access very different types of financial markets and consumption possibilities, where distribution matters, heterogeneity matters, I think the representative agent assumption can be misleading.
So, I would never want to say that one is the right specification and the other one is ad hoc. I think that is just misleading terminology.