Grasping Reality on TypePad, by Brad DeLong

Buchanan: I do think that much of modern macroeconomics -- and I mean theory, not empirical work -- is a pretty spectacular failure. Research in this area doesn't look like science to me. It treats an economy as if each household and each firm make optimal decisions, thereby excluding most of the richness and heterogeneity of a real economy.

Indeed, there's something strange going on in this field. In 20-plus years writing about science, I've studied research in physics, biology, chemistry, psychology, anthropology and always found, after looking closely enough, that the models people use in these fields are usually well-motivated, make basic logical sense and get rejected if they don't fit the facts very well. Macroeconomics has been the one exception. 

I do like some macroeconomics research, just not the stuff in the mainstream. But Noah, I know you've criticized macroeconomics on many occasions. What do you think is right and wrong about the field? 

Smith: I certainly think there are lots of problems with macro. But criticizing it based on specific elements -- for example, assuming everyone’s perfectly rational -- won’t change much.

See, I think the biggest problem with macro is that models never get thrown in the trash. Researchers pump out theory paper after theory paper, most of which contradict each other. Sometimes we see a big, spectacular event like the 2008 crisis, which just couldn't happen in any of the popular, dominant, or Nobel-winning models. But then macroeconomists just pull some obscure 20-year-old paper off the shelf and say "Of course macro can explain this." Macro theory is sort of like a big collective effort to cover all the bases, not to find which models really work and which don't. 

So if people criticize macro by saying to add this or that feature, they’ll do what you ask. But the papers will just sit on the shelves, and when it comes time to make policy, people pick and choose the theories that suit their preconceptions.

Sure, that's unscientific. But what's the alternative? A macroeconomy is a big, complex system and no model will explain all of it at once. So how do we know when to toss out a theory? Suppose you write down a model that does a good job describing how job vacancies get created and filled, but does a poor job explaining trends in productivity. Do you toss it, or do you keep it? This is a question macroeconomists have never really made themselves answer. But I don't think anyone, including physicists or biologists, or macro critics for that matter, has a definitive answer to this question.

Buchanan: I like what you say: "Macro theory is sort of like a big collective effort to cover all the bases, not to find which models really work and which don't." That's more or less what some economists argued a couple of years ago, and they seemed to think it is OK. I think it's certainly a way to fill libraries with models, but that it's too safe, unambitious and not the way to make useful macroeconomics. 

Of course, models shouldn't aim to include everything, which is in any case impossible. A useful model is a simplified analogue of reality that neatly captures some important aspects of how something works, while leaving lots of less important detail aside. Economists should have lots of models attuned to different situations and problems.

But in trying to separate useful from non-useful models, macroeconomics could benefit by doing some things that are common in the rest of science. One big thing: empirical tests of macro theories could be much more severe, not only comparing model predictions to aggregate economic data, but also checking on the empirical plausibility of how the model works internally. Do the individuals and firms, for example, look and behave much as they do in the real world? It's the absurdity of so many macro models at this level that really gets macro critics going. Today's climate models, while far from perfect, are much more accurate than they were several decades ago,  because they've been tested not only at the level of their aggregate predictions, but also at the level of thousands of other details -- do they get atmospheric flows right, do they get heat transfer between ocean and atmosphere right, etc. In this way, less accurate elements have been eliminated and replaced with improved ones, gradually and systematically over time.

Surely something similar should be possible in macroeconomics.

Smith: You're right. This is what I call "getting the pieces right." Traditionally, going back to Milton Friedman, there has been this idea in macro that you don't have to worry about making the pieces of a model realistic -- you just take a bunch of assumptions, throw them in a bag, shake them up, and if the model that comes out kinda-sorta looks like it matches some parts of the overall economy, you get published, you get respected. 

But there are signs this is beginning to change. Recently, more top macroeconomists are taking a hard look at the data to figure out how consumers and companies really behave -- for example, how businesses make decisions about setting prices and hiring or firing workers. More people are trying to explain inequality. There's a recognition that consumers don't look infinitely far into the future. A lot of models have added finance. Partly this is a response to the availability of a lot more data, and part of it is the development of more sophisticated math techniques that can handle more features at once. And I think partly it’s in response to the criticism of macro in the wake of the crisis.

Note that this is the opposite of what a lot of macro critics are suggesting. Some, like Paul Krugman, advocate the use of very simple, old models, which are easy to use and interpret (especially in times of crisis). Others suggest tossing out the idea of individual decision-making as the foundation of macro models. Still others want to get rid of mathematical modeling. Compared to what these people are recommending, the approach of "getting the pieces right" seems pretty conservative. 

Buchanan: Recognizing that consumers don't look infinitely far into the future? That is a major advance! No, seriously, that's all good to hear. Perhaps my criticism is more conservative than I had thought. I'm certainly not in favor of getting rid of mathematics or of ignoring individual behavior. 

Another dreadful aspect of macro is economists' unreasonable fixation on general equilibrium models, which they rarely subject to the getting-the-pieces-right principle that you mentioned. The real macroeconomy comes about as people and firms and governments take actions of a zillion kinds, leading to aggregate outcomes, which then feed back to alter everyone's behavior. The general equilibrium approach builds a theory of this by turning away from this feedback, and instead looking to find sets of actions and expectations that are consistent with the overall aggregate patterns. In these nice equilibria, no one has further incentives to change their behavior or strategy.

Given the difficulty of the problem, this is OK as an initial crude stab at learning something. But it certainly shouldn't be the end of the road for macroeconomics. The obvious alternative is to explore the richer world of models in which the actions of individuals, firms and governments lead to aggregate outcomes, which may then demand new action and adjustment, creating further new outcomes and so on. This is not equilibrium, and it's a lot messier mathematically, but also looks a lot more like the real world, which make this kind of more open modelling and exploration possible. I'd like to see more economists start doing this kind of work, but many seem very hostile to the approach.

Smith: The biggest challenge right now for macro is to move toward more realistic micro, using data as a guide. And I also think there needs to be a culture change -- macro people need to be a lot more willing to just throw models out. That would lead to a shift in research from theory to empirics, at least for a while. And I think these two changes would go a very long way toward making macro look more like a typical science. I think instead of telling macro people which things to put in their models, we should focus more on pushing them to make this cultural change.