Bottom-up versus Top-down Macro

|Peter Boettke|

Greg Ransom linked to his discussion at Taking Hayek Seriously of this new paper in our comments, but the idea behind the paper might be too important to be left to the comments.

What do you think of the claim that rational expectations macroeconomics was the intellectual descendants of the market socialist models of the 1930s?  Not in terms of policy design, but in terms of way of conceiving of the problem and of the way that economics should be done.

Comments

[bill woolsey]

Pete:

This ties to your quote from Buchanan in a previous post.

The problem is the notion that economics is all about doing constrained maximization problems.

After that, the difference between a socialist manager ordered to produce where marginal cost equals price, and the typical businessman or worker setting prices and production based upon an optimal response to macro policy is distant.

Well, other than it being doubtful that it works out "optimally."

[Jerry O'Driscoll]

I agree with the thesis outlined in the summary of Paul De Grauwe's paper. (DeGrauwe is a solid macroeconomist.) The problem is not the math or even the expectations theory, but the general equilibrium framework. The Socialist planners fulfilled the logic of the Walrasian system.

On the intellectual history, the core idea of rational expectations can be traced back to Mises, Wicksell and Hayek. In Mises, it is Lincoln's Law. In Wicksell, it is a parable of the man who sets his watch fast so he won't miss his train. In Hayek, it is embedded in the proposition that market forces re-asert themselves in the crisis.

Lucas, et al. chose to model expectations in a GE framework and thereby made it top down. Recall what the alternative was at the time: crude adaptive expectations in Keynesian models. An implication was that you consistently could fool the market by inflation and in effect create real wealth. We must admit that RE was an improvement on that.

Again,to reiterate, rational expectations were imported into GE modles and that eliminated the decentralized logic in Mises-Wicksell-Hayek.

[amv]

I like the de grauwe paper. but he still remains top down, just starting from a lower top:

"A methodological issue arises here. The forecasting rules (heuristics) introduced here are not derived at the micro level and then aggregated. Instead, they are imposed ex post, on the demand and supply equations. This has also been the approach in the learning literature pioneered by Evans and Honkapohja 2001). Ideally one would like to derive the heuristics from the micro-level in an environment in which agents experience cognitive problems. Our knowledge about how to model this behaviour at the micro level and how to aggregate it is too sketchy, however, and I have not tried to do so." (p.7-8)

[amv]

BTW: from an Austrian perspective, the more illuminating paper of the munich conference is Alan Kirman's: http://www.cesifo-group.de/portal/page/portal/CFP_CONF/CFP_CONF_2009/Conf-es09-Illing/Papers/es09_Kirman.pdf

[Barkley Rosser]

I agree that Kirman's is the more important of the two papers.

[Sinclair Davidson]

I think the de Grauwe abstract is exactly right and uncontroversial. What is interesting, to my mind, is the extent to whick Arnold Kling's recalculation model (conjecture?) is a top-down model too.

[Pietro M.]

Although I have no problems with the pars destruens of the article, as it is wholly Austrian in its focus on compex market coordination with limited rationality and information, I fear that the pars construens may be even worse than walrasian macro.

In fact, in place of "angel-like" agents capable of understanding everything except the realization of stochastic shocks, we have "amoeba-like" agents which are only capable of reacting mechanically to past prices or reacting mechanically to the policy-makers' stated objectives. Where is entrepreneurship? Where is the market process? They have been apparently lost in translation during the process of translating Austrian market process theory into formal models of mechanical and uncreative bounded rationality.

The end result is the resurrection of Keynes's animal spirits, i.e., psychological explanations of economic phenomena.

Theorically, general equilibrium models at least show the market process of adjustment in its perfection. Albeit unrealistic, understanding the logic of general equilibrium is often a good starting point in understanding markets. Ricardian equivalence may not hold, but its theory reminds us of intertemporal effects that may be lost in less sophisticated views of the market. The same is true for rational expectations and other neoclassical interesting wishful thinking.

Where's the economic beef in assuming that markets MUST fail because agents are fool?

[amv]

I totally agree with Pietro. I may add two things: first, even if agents are perfectly rational, the global system produces so much nasty but exciting dynamics (out of equilibrium) that it is impossible to even think of convergence in the aggregate. thus, outside partial analysis, there is simply no need to impose irrationality or the like on agents to make the system more troublesome. second, the statistical learning approach by evans and honkapohia rests on the assumption of structural stability. that is, convergence is guaranteed by the fact that reality is by-and-large invariant while agents understand more of more of it simply as time passes by. if agents are 'rational' (in the rational-expectation-sense where they now all contigent states yet to come and the equilibrium prices associated with them), convergence to equilibrium is the fastest. If agents are stupid, backward-looking, etc. the adjustment is slower. but global stability is a priori assumed. instead, think of franklin fisher and his emphasis on the speed of convergence for global stability. further, endowments change during transition, thus structural stability is a bad idea. statistical learning is as good as it gets in dsge models. but it has little to do with reality. see the kirman paper I linked above.