Thursday, September 15, 2016

Whither Agent-Based Macroeconomic Modeling?

Last week at this time I participated in a conference called "Economics, Economic Policies, Sustainable Economics in View of the Crisis." which took place at the Universita della Polytechnic in Ancona, Italy.  The main host was Mauro Gallegati, a prominent agent-based modeler, econophysicist, and more general complexity economist (also a sometime coauthor of mine).  It went on for three days with parallel session and most of the participants from Italy, although also from across Europe, including Russia, and beyond to such places as India and Australia.  The more well-known plenary speakers were Duncan Foley, Bruce Greenwald, Alan Kirman, David Colander, and me.  While there were papers on many subjects, including a bunch on ecological economics and sustainability, a very main focus was about macro modeling and how agent-based modeling (ABM)  relates to other kinds such as DSGE, VAR, and old reduced form many equation ISLM models, these latter three reportedly what get looked at seriously at the Fed and most other central banks.  A big question was given the problems with those and the hopes for ABMs, why are they not getting adopted as a fourth model in those settings or even a replacement for one of the others?

A conference in Italy is an especially appropriate place to raise these questions as it has been perhaps the world's center of ABMs, especially for macroeconomics, as well also being a major center for econphysics and complexity economics more generally.  Gallegati runs a group and his students have spread all over the country.  There is a major group in Milan led by Domenico Delli Gatti that often works with Gallegati's group (when they coauthor it is the "Gattis," with Gatti meaning "cat").  There is another group in Pisa, led by Giovanni Dosi, and another in Genoa led by Silvano Cincotti, with reps from all these groups (and some others from other countries) all there.

The situation may be seen by the group in Genoa, where the group developed the EURACE model, which Cincotti spoke, on, perhaps the most widely studied and used macro ABM there is.  He used it to show how increasing the Basel Accord capital requirements on banks could increase financial fragility in the system, a good ABM kind of result not al that easy to get from other kinds of models and certainly very macro and central banky.  It involved agents moving into the shadow banking sector, with some unsurprising results.

The cynic in me says that they did that model to try to lobby against stricter EU requirements on the Italian banking system, which is very fragile and near collapse.  The borderline bank is Monte dei Paschi di Siena, the world's oldest bank, dating from the 1400s, which has a fabulous Renaissance art collection based on pieces given to them as collateral from debtors who failed. More generally the Italian economy seems to be stagnant, going along, without people looking miserable in the streets or whatever, although I noticed some places closed that used to be open in Ancona.  The most romantic is an old hotel, the Roma e Pace, which hired Joseph Stalin in around 1906 as a doorman, but then fired him for being "troppo timido," (too timid), what a hoot.  They had a newspaper article in their lobby on that, but cannot see it anymore.  Gallegati also said that Mussolini had an affair there at one point with a famous Italian actress.  Oh well.

Anyway, there are these pretty interesting ABMs out there that seem to be able to do interesting stuff, but somehow they are not being picked up by the central banks.  Furthermore, I heard rumors that funding from the EU and INET and some other places may be cut for this kind of research.  If this turns out to be the case, I think it will be too bad.  In the end, it may be that the rival that is holding it off is not DSGE, which everybody dumped on at the conference, but the much less visible but still used old ISLM ones. This might make Paul Krugman glad. Atheortical VARs just seem to be too useful for very short term forecasting to get kicked aside.

There were other interesting debates, between Foley and Kirman and me over general equilibrium theory (with an eye to DSGE applications) and between Colander and Kirman over spontaneous self-organization, this arising from the recent article in the JEL by Kirman about Colander's recent book on Complexity and Public Policy.  Kirman really slammed spontancous self-organization, much pushed by Hayekian Austrians, but Colander pushed back enough that Alan actually said, "Well, maybe I need to get reorganized."

Anyway, I think there is still potential for macro ABMs, so will be sad and frustrated if the flow of money to study these does get shut down.  It will look like a major triumph for the mainstream establishment after all the hullabaloo over the huge crash that happened eight years ago in just a few days.  We should have learned more.

Barkley Rosser

7 comments:

  1. Barkley,
    I'm very interested in ABM, but I have a question. Is it normal in actually existing ABMs to include things like innovations and Product Lifecycles in them. I mean to see products and processes as agents.

    ReplyDelete
  2. I wish you had written this post 10 years ago

    ReplyDelete
  3. reason,

    Not normally but some do, especially those coming out of the group based in Pisa and led by Giovanni Dosi. That group is very into such things.

    cook,

    Well, ten years ago I was pushing ABMs. This post would not have been written then because then I was more optimistic about their prospects. A sub-text here is that I am less optimistic and frustrated by apparent recent developments.

    ReplyDelete
  4. I want to thank pgl for wisecrack he made in the comments section over on Economistsview, where Mark Thoma linked to this post and there was some fairly extensive commentary.

    He said that it looked like the conference had a lot of the "gnomes of heterodoxy" in it. I like that one and thank him for coining it, :-).

    A Gnome of Heterodoxy

    ReplyDelete
  5. by 'prospects' you mean funding?

    ReplyDelete
  6. In the piece I expressed what I heard from some other that indeed future funding proospects for building and testing these models may be less available than in the recent past.

    ReplyDelete
  7. Jorge Vianna MonteiroSeptember 29, 2016 at 8:11 AM

    I am sorry but that seems to me a very complex/unreadable prose ... I certainly prefer Romer's way of presenting his critic.

    ReplyDelete

Spam and gaslight comments will be deleted.