Iconoclast.

Alright, first off what the fuck is a "DSGE" model?

The "D" stands for dynamic, meaning that it moves across multiple time periods. The "S" stands for stochastic, which means that some components are random, or unpredictable. The "G" stands for general (as opposed to partial), which means that it covers the entirety of an economy. The "E" stands for equilibrium, or essentially solutions to the model (A rock rolls down a hill and then rests at the bottom. This is an "equilibrium" point for the rock.)

Together that makes "Dynamic Stochastic General Equilibrium" models. They're about what you'd expect. Models of the economy that take into account everything I just mentioned. Different times periods, random shocks, and so on.

Certainly, as I've stated before, models are incredibly helpful when they're used to understand complex relationships in an economy. Not only that, but you can use them to make predictions about things that haven't happened yet. As well as other benefits, such as having a framework to understand policy with, etc.

Well, what's so bad about that? The problem isn't with DSGE models, per se, but theoretical modelling of economies in general. For example, models are exactly what you put into them. I mean, all you need are basic assumptions that make the internal organs of your model logically consistent, and then you're done. You certainly have to conform it to basic empirical data, but for the most part even if it doesn't no one will notice, since most models leave out many complicating factors (they're usually abstract representations that only measure very strict scenarios), and most empirical evidence isn't strong enough to completely falsify a model. So the more complex models with very subtle predictions will cling tenaciously to life.

Part of this is a problem due to empirical methods not being reliable enough to falsify hypotheses, part is inherent problems with assumptions and simplifications, part is because this shit is hard.

The point, however, is that there's a problem. As Sims pointed out in the last post, though, there are redeeming qualities, even if there are problems. Some models do have the ability to predict. We just need to get the right assumptions and complicate the model a bit to make it more realistic. The point is that as of right now we just need to get the theory right before we can use it for policy, and before we get the theory right we have to validate it with good econometric work.

Now, that's all I wanted to say here, but this is Harald Uhlig at the same conference defending (a bit more ingenuously) DSGE models:

"You know, physicists dream of a theory of everything...it's pretty pathetic...certainly doesn't help us understand the financial crisis."

I also really enjoyed his comparison of abstract art with abstract theory, although I don't think that aesthetically pleasing art and aesthetically pleasing theory are the same. One has the purpose of being beautiful, while the other has the purpose of predicting and understanding. Unless the point of modelling is so that we can make cool models. I'm down, but it seems a little inefficient. So
I hope "beautiful", with regards to economic models, means something else.