System Dynamics versus General Equilibrium Models

[Justus Gallati]

I'm looking for papers which compare system dynamics and general equilibrium models (or other models commonly used in economics). The comparison should demonstrate where these approaches are applicable (fields of applications), what the underlying assumptions are (rationale), and what the specific potentials and limitations of the methods are. Looking forward to your comments and hints.
Thank you and best regards
Justus Gallati

[Nathan Forrester]

You might want to take a look at my 1982 dissertation: A Dynamic Synthesis of Basic Macroeconomic Theory: Implications for Stabilization Policy Analysis. The thesis integrates the IS-LM, Aggregate Supply/Aggregate Demand, multiplier-accelerator, and inventory-adjustment models often taught in basic macroeconomics courses. The four models are combined into a 10th-order dynamic model used to evaluate various approaches to stabilization policy.
http://dspace.mit.edu/handle/1721.1/15739
nathan@f3-analytics.com

[Thomas Fiddaman]

John Sterman's thesis also considers equilibrium vs. disequilibrium issues. You can search for it at the same MIT dspace link.

Static general equilibrium models may be the most common approach in economics, but there are increasingly many dynamic general equilibrium models (like MIT's EPPA) that combine short run equilibrium with long run dynamics in a few processes, like capital investment. The macro literature, particularly the endogenous growth thread, tends towards intertemporal optimization models, which are really just equilibrium plus perfect foresight, so it's worth looking at. The hot approach on the fiscal side of macro these days seems to be dynamic stochastic general equilibrium models.

Also take a look at Frank Ackerman's "Still Dead After All These Years," which is a critique of general equilibrium. As I recall, it's primarily technical (e.g., GE models don't have unique solutions). Personally, I think there are also major practical problems. My nutshell critique is that the typical GE (a bunch of nested CES production functions, plus an algorithmic search for a vector of market-clearing prices) is somewhat useful for understanding things like the incidence of tax changes on various sectors in the short run, or the relative magnitude of direct vs indirect effects of some policy. However, as soon as you consider any policy that's bigger than marginal tinkering, particularly for normative policy design, you're in trouble. First, it's likely that the elasticities that define your model are all wrong, because a variety of perception and action delays, and feedbacks like endogenous technology, are neglected. Thus seeming short-run rigidity of the economy that is in reality due to a mix of behavior and "physics" of the production landscape is attributed entirely to the "physics" portion, which makes it seem impossible to change anything without reducing welfare. Second, if unmodeled delays and feedbacks are in fact prevalent, preferences are endogenous, and institutions make rules that distort local incentives, the whole premise of equilibrium is silly. It's quite likely that the economy is far from equilibrium, and that while pressures move it in the right direction, social and technological forces evolve the landscape faster than the visible market economy can really approach equilibrium.

[Oleg Pavlov]

Dear Colleagues, considering the topic of this discussion, I thought you might be interested that this year the Economics Chapter is planning to host the Economics Roundtable during the summer conference in St. Gallen. We will discuss the following topic: “Will economists ever embrace system dynamics? In other words, is there anything about system dynamics that is inherently anti-economics that would prevent economists from embracing it? Would there ever be such a subfield of economics called "system dynamics economics" or “feedback economics” (just like "behavioral economics").”

[Justus Gallati]

Thanks for all these wonderful comments and considerations. With regard to the forum envisaged for the St. Gallen conference (and maybe in a more general sense) I would argue that we also should consider where the common ground between SD and economics (and in particular with modelling in economics) is. This might provide a good starting point to enter a discussion with economics. Say it like this: in the overlapping areas the two approaches should provide comparable results; further one should be able to argue why SD is capable to tackle problems outside this overlap in an appropriate way. As such the discussion could go in a direction to clarify the fields of applications and the rationales underlying the different methods.