The session I moderated at the INET Conference was called "What Kind of Theory to Guide Reform and Restructuring of the Financial and Non-Financial Sectors?" (the panel was Franklin Allen, Sheila Dow, Axel Leijonhuvfud, and Joe Stiglitz). During the introduction, I made (or meant to make) the following points:
... If you believe, as I do, that macroeconomics needs to change, there are three possible ways to proceed.
First, we could try to reform the DSGE model used widely today. How much would it help to do one or more of the folowing: (i) Within the DSGE model, develop better connections between the real and financial sectors. In particular, the model should allow for the endogenous collapse of financial intermediation. Recent models of financial frictions and endogenous leverage cycles give an indication of how to proceed. (ii) Replace rational expectations (and the efficient markets hypothesis) with a better approximation of how expectations are actually formed. One possibility along these lines is to added learning to the models. Another is behavioral economics. (iii) Replace the representative agent assumption with heterogeneous agents. It's hard to have realistic financial markets with one agent. However, adding heterogeneity is not as simple as it might seem. Generically, having heterogeneous agents in a model makes it difficult to aggregate across individuals – once the representative agent assumption is dropped you cannot, for example, guarantee that uniqueness or stability will appear at the aggregate level even if individual agents are well-behaved neoclassical agents. There are clever ways to allow for heterogeneity without sacrificing the ability to aggregate, but they aren’t fully satisfactory. If we are going to go this route, then more cleverness is needed. (iv) You may be surprised to learn that regulations such as capital requirements have very little theoretical backing -- they are largely ad hoc (see the talk by Franklin Allen). If the problem wasa failure of regulation and not a failure of the model more generally, then perhaps better models of regulation are all that is needed (or, if you believe the crisis was caused by the Fed's pursuit of low interest rates, perhaps all we need is a better model of monetary policy). However, this brings up the question of whether the DSGE structure is an adequate foundation for models of regulation, and I am not convinced that it is. (v) This wasn't part of my remarks, but a physicist spoke at the conference and one of his main points was that financial markets are dictated by power laws, not the Gaussian distributions that are commonly assumed in theoretical work (often for analytical convenience). Is addressing this problem all that is needed?
The second way we might proceed is to adopt new models. Possibilities along these lines are: (i) To develop network (complexity) models and their associated measures of network characteristics such as centrality and degree distribution that can be used to estimate the risk of network failure. (ii) There is George Soros' Reflexivity theory, and (iii) there is the theory developed by Frydman and Goldberg, Imperfect Knowledge Economics. (Both of these had been discussed in earlier sessions.) (iv) We could begin the modeling process at the aggregate level and give up the insistence that the models be microfounded. This would, among other things, avoid the problems associated with aggregating from realistic microfoundations discussed above.
Third, take a whole new approach to theory. (i) The call for pluralism that Sheila Dow will talk about falls into this category (see here for more on this). (ii) Economics could give up trying to model itself after physics as it existed a century or more ago, drop the "natural" language, and embrace the methods of the "softer" sciences. These disciplines have already successfully addressed many of the problems that economists face. ...
Since these were introductory remarks, I ended there in order to allow the panel as much time as possible. (I should also note that the distinction between fixes to the existing model and building a new model are not always clear, e.g. imperfect knowledge economics could have also been listed as a variation on the existing model.)
My view on all of this is that I am torn between the first two options -- fixing the existing model and building a new one -- but fortunately work on the two options is not mutually exclusive. There's no reason why one group of researchers can't try to fix the model we are using now while another group works on a much different theoretical structure. Presumably, in the end, the better model will win out.
Many people are working on fixing the existing structure -- macroeconomists are already tooled up for this, so it's a natural progression -- what is needed is more acceptance within the profession (at journals in particular) for alternative theoretical models and different approaches to economic modeling (one thing that come up at the conference is if there any role for articles without math in top journals).
Outside of the work on fixing the existing DSGE model (where I think developing better connections between real sector and financial intermediation, and endogenous leverage cycles are the first things to do), my first choice of where to go next would be to investigate network models. There is already progress in this direction, and I think these models have a lot of potential for characterizing the risk within financial networks in a way that would be helpful for regulators. But while these models look promising in some areas, it's hard for me to see how network models could constitute a brand new macroeconomics more generally (they seem more of a complement than a substitute for existing models). So there's a lot more work than that to be done.