FRB Richmond — Interview with Dani Rodrik

Curiosity about what makes nations grow and develop is as old as the economics profession itself, having motivated Adam Smith's 1776 The Wealth of Nations. Development economics as a distinct field, however, is relatively young.

Its progress has come in fits and starts. Widespread support for state-directed investment after World War II was met by backlash with growing awareness of communism's weaknesses and with the Latin American debt crisis of the 1980s. The Asian miracle then convinced the world that free trade and open markets would solve poor nations' ills, and aid to such countries was conditioned on their efforts to liberalize. Liberalization without attention to context, too, is now seen as a failure.

The work of economist Dani Rodrik suggests that policymakers may have been asking the wrong questions. He argues that development is not a "one size fits all" proposition: The individual circumstances of countries determine the success or failure of aid, liberalization, and other efforts to prod development. Moreover, his work has contributed to the modern idea that institutions matter as much as any single policy, natural endowment, or economic structure.

That is not the only area in which Rodrik, a native of Turkey, has questioned convention. Few topics unite economists like the virtues of free trade, so it was notable when Rodrik made a case for the limits of globalization in the 1997 book Has Globalization Gone too Far? In it, he argued that globalization in the extreme can harm social stability by threatening the institutions that underpin it.

Rodrik has also taken a critical look at the economics profession, and in particular, how economists can best provide practicable guidance to policymakers. As his views on development suggest, one theme is that economists should be wary of hubris — and should be clear to policymakers about the limits of the profession's knowledge.
Rodrik is currently at the Institute for Advanced Study, but he will be returning to Harvard University in mid-2015.....

FRB Richmond — Econ Focus
Interview with Dani Rodrik
ht/ Greg Mankiw

Here's an excerpt on economic modeling if you don't want to read the whole article, which is a bit longish but worth the read, and important for economics as a discipline as well as approaching development economics:

The root of it is the problem that the profession has more or less the wrong idea about how economics as a science works. If you ask most economists, "What kind of a science is economics?," they will give a response that approximates natural sciences like physics, which is that we develop hypotheses and then we test them, we throw away those that are rejected, we keep those that cannot be rejected, and then we refine our hypotheses and move in their direction.

This is not how economics works — with newer and better models succeeding models that are older and worse in the sense of being empirically less relevant. The way we actually increase our understanding of the world is by expanding our collection of models. We don't throw out models, we add to them; the library of models expands. Social reality is very different from natural reality in that it is not fixed; it varies across time and place. The way that an economy works in the Congo is very different from the way that it works in the United States. So the best that we can do as economists is try to understand social reality one model at a time. Each model identifies one particular salient causal mechanism, and that salient effect might be very strong in the Congo but it may be very weak at any point in time in the United States, where we may need to apply a different model.

If you look at the progress of economics all the way from perfect competition to imperfect competition, from incomplete information to behavioral economics, at every step we have said, "Here are some additional realities for which we need newer models." Behavioral economics doesn't mean that we want to ignore models in which people are rational. There are plenty of settings where presuming people are behaving rationally is still the right way to go.

When you look at economics in that way, as a collection of models, then what does it mean to say that economics knows something about the world? Economists know how to think about various causal mechanisms that operate as part of social reality, but what they're very bad at in practice is navigating among the models describing them. How exactly do I pick the right model for a given setting? This is a craft because the evidence never settles it in real time. We have these periods of fads where we say the New Keynesian or the Neoclassical model explains everything. We lose sight of the fact that models are highly context-specific and we need to be syncretic, simultaneously carrying many models in our mind.

Certainly, there are things in economics that can be said to be fairly universal; many of those are actually quite innocuous, such as the assertion that "incentives matter." I think we can often also agree after the fact: We can say that the Soviet system was economically inefficient, or that the 2009 stimulus package of President Obama reduced unemployment. These are things on which there is a fair amount of consensus, and rightly so, because the evidence is more or less in. But the vast range of propositions over which economists agree in public don't have that kind of support, and that's where I think we often get into trouble.

EF: So when you start thinking about, for instance, the Congo, how do you know that you have enough place-specific evidence of what's going on to select the right model?

Rodrik: The answer is: not very well and only imperfectly.

What you do is apply a number of strategies. One is to look at whether the critical assumptions of a model seem to track the context well. The emphasis here is on critical assumptions, because every model, and therefore every explanation, is necessarily a simplified version of reality, so there will always be assumptions that don't make sense. A critical assumption is one that if you were to relax it in the direction of greater realism, it would change the result significantly. Should I apply a model with imperfect competition or perfect competition — the question there is going to be, will it change significantly the result of the question that I am asking?

Another strategy is seeing whether the comparative static properties of the model are borne out beyond the specifics that you're interested in. So in the case of the Congo, you might say, "The problem is that the private sector is investing so little. Is it because they don't have access to finance, or is it because they are in a poor investment climate?" Those are two different models of growth, if you will. One would be based on low savings, the other would be based on, let's say, poor institutions or poor property rights. One of the comparative static implications of these two models would be that when foreign aid comes in, or there are remittances, does it go into consumption or does it go into investment? If it's a saving constraint model, it would go into investment. If it's a return constraint model, it's going to consumption. The way these economies actually behave tells you something about what the underlying model might be.

These are diagnostic techniques that can help us navigate among alternative models, but above all it requires judgment.....

That's one of the ways in which economics works best; models are like case studies that give you a sense of the possibilities. If something that ought to have happened isn't happening, then you have to dig deeper and figure out which assumption of the model is being violated.

A mistake that I think the very empirical end of the profession makes is thinking that we can simply run the right experiment to determine causality, and then we can do this model-free. In fact, nothing is model-free. Anytime you are thinking about a causal relationship, you have some model in the back of your mind, and that model has a bunch of explicit and implicit assumptions that are hidden into it. And you better recognize what those are.

Put differently, most economists think that they work in an exclusively deductive method, meaning they form a hypothesis and then test it. In fact, much of the best economics involves moving back and forth between the deductive and the inductive. They look at the world and find an irregularity — the inductive stage — and then formulate a model or a theory to explain it, the deductive stage. Of course, you cannot test that model on the observation you have just made because the model has been created for the purpose of explaining that thing. You use the model to generate additional implications you have not thought of before, and then you see whether those additional implications are borne out or instead how you need to modify the model to make those additional implications empirically verified. But very few papers you read in research journals will actually tell you that's the mold in which they were generated....