I really hope that the conversation in the comments to the post What's Wrong with Modern Macroeconomics: Conference papers will continue:
Barkley Rosser said... I should have guessed that de Grauwe might have a good paper.
So, Mark, what have you to say to the assembled masses there that you are willing to report back to us about the spot-on paper by Alan Kirman?
reason said in reply to Barkley Rosser... Barkley,
thanks for the tip. The Kirman paper is my reading on the train now.
Mark Thoma said... One thing I learned from it is that I need to read the old papers by Sonnenschein (1972), Mantel (1974), and Debreu (1974) since these papers appear to undermine representative agent models. According to this work, you cannot learn anything about the uniqueness of an equilibrium, whether an equilibrium is stable, or how agents arrive at equilibrium by looking at individual behavior (more precisely, there is no simple relationship between individual behavior and the properties of aggregated variables - someone added the the axiom of revealed preference doesn't even survive aggregating two heterogeneous agents).
I need to learn the full extent to which this work undermines the whole microfoundations approach (hence Kirman's call to study the properties of networks so as to generate endogenous cycles from phase transitions rather than trying to model individual agents from first principles - that's not his sole reason for wanting to turn to this approach, but it's part of it).
I didn't understand that extent to which representative agent models are an analytical convenience to work around this problem (the DSGE theorists who understood this kept quiet about it).
You can get interactions among agents while maintaining identical agents, i.e. network effects do not necessarily require heterogeneity, but most interesting cases, it seems to me, do involve both heterogeneity and agents whose decisions are interdependent.
(Robert Solow said all you get is that excess demands have to sum to zero, i.e. Walras Law and a couple of other properties, but it's not much).
So that was the most important thing I learned, and it's something I should have known already. Once I learn a bit more about the results in these papers, I hope to post something about it. (A small part of my remarks wondered if learning models might not help to overcome this problem since they might give you a path to the equilibrium, and the number of equilibrium paths might be determinate and equal to one giving uniqueness, but it was pure speculation).
Barkley Rosser said... Mark,
The missing man in the critique of the representative agent model is Michael Jerison of SUNY-Albany. In his much-cited JEP paper on "Whom does the representative agent represent?", Kirman used (and cited) an example from a still unpublished paper by the sadly neglected Jerison.
Kirman is an interesting figure in all this as he was originally a general equilibrium theorist (well, originally a game theorist, and then a GE theorist). While he has differences with his old GE comrades, they all respect his critique because it does start with the SMD theorem, which is well known to all of them and very fundamentally disruptive. The DSGE modelers conveniently cover that one up, along with some other major problems.
Of course, in a world of multiple equilibria, learning may lead you anywhere.
Sebastion said... A good reference for anyone with access to the Mas-Colell et al micro textbook is their chapter 4 on aggregate demand and in particular chapter 4D on the existence of a representative consumer. Unfortunately, chapter 4 is usually NOT being taught in standard micro classes
Roberto Cruccolini said... And it might be good after having read chapter 4 to continue with chapter 17E in MasColell, called "Anything goes: The Sonnenschein-Mantel-Debreu Theroem". There you are, even in the advanced-micro bible you can read about those results, which are also at least extremely interesting for modern "microfounded" macro.
I think, this phenomenon of forgetting and/or neglecting former knowledge, e.g. the whole discussion and aspects of aggregation, which is really central to the methodology of modern macro, as the notion of microfoudations via optimizing agents was one of the core-arguments of New Classical Makro Revolution, is deeply unsettling. You could also add the oblivion of coordination & interaction problems, of discontinuities & emergence as probably central aspects of makroeconomics, which seem to be the reasons, why macro was once thought to be necessarily a different approach than micro.
There seem to be two ways to deal with this finding. One is to complain about the way modern macro has developed, and to suggest other/better solutions; this is, what we see most of the time right now.
That is of course worthwhile and understandable, but there remains a strange aspect: nearly all of nowadays criticisms were already mentioned 20 or 30 years before (recall Solow 1978 at the same conference as Lucas & Sargent, or Summers 1986 in response to Prescott, or Blinder 1987, or, which is sort of funny, Kirman 1989 & 1992 and - again - 2009,...)
And this leads to the interesting question, why modern macro/new classical methodology & thinking was so successful in conquering the field:
why did economists think, that Lucas 1976 said something new, given the reflections of Marshall how to theorize given ever changing structures, given Haavelmos ideas on the autonomy of economic relations, given the debates between Keynes and Tinbergen of econometrics and structural instability, and so on?
And why did they follow him in applying a Walrasian program using the representative-agent methodology given all these challenging aggregation results (see for the makro-production-function Fisher 1969 or Fisher&Felipe 2003 & 2006 and the bunch of literature to the Cambridge Capital Controversies and of course the literature interpreting the Sonnenschein-Mantel-Debreu results, f.e. Rizvi 1994 or 2006)
And in what sense does it make sense to describe modern macro models as "microfounded", if at the same time, you need some Friedman-1953-as-if argumentation to justify & make plausible your way of modeling, which is often referred to, that the inner functioning of your model is a black box and unknown, only built to generate predictions, not less. but certainly not more, in the sense that we think the way of modeling is reasonably realistic and corresponding to some mechanisms in the real world. Why should we, or why is this commonly called "microfoundations"??
johnchx said... MT wrote: "One thing I learned from it is that I need to read the old papers by Sonnenschein (1972), Mantel (1974), and Debreu (1974) since these papers appear to undermine representative agent models."
Yes; I have the impression that it's been clear for a very long time that representative agent models lack microfoundations -- that is, that classical micro assumptions are insufficient to support the existence of a representative agent, that the necessary conditions are unknown, and that the known sufficient conditions are extremely restrictive (e.g. perfectly homogeneous agents). I've tended to view this as evidence that many of those advancing the banner of microfoundations simply weren't serious.
If one really cared about understanding macro fluctuations in terms of the behavior of individual households and firms, one would study the behavior of individual households and firms; a representative agent framework would be entirely unsatisfactory.
I think it's also telling that most of the famous names associated with microfoundations in macro have been theoreticians rather than microeconometricians. I suspect that we may be able to learn more about real microfoundations of macro from, say, James Heckman than from Lucas or Barro.
Am I wildly off-base?
Herman said... Mark Thoma,
May I suggest elevating Roberto Cruccolini's very interesting comment above as a separate new post.
reason said in reply to Herman... Or even on Steven Keen's book "Debunking Economics" which covers most of the same territory in a very readable fashion.
Update: More from Alan Kirman: Economic theory and the crisis.