Sunday, October 4, 2009

Asymmetric information and the REE (amv)

Hohenheim hosted a workshop on the history of macroeconomics. All guests were excellent and the papers presented nicely combined the past of our science with our current research frontiers. It's just great to be an economist and to grow into the profession: there are plenty of nice and interesting people, all much smarter than myself, so that there is a lot of learning experience whenever they cluster. From my personal research background, I was expecially happy to see Pascal Bridel in Hohenheim, a true expert in general equilibrium analysis (inter alia) and a really, really subtle and challenging guy.

However, another of our outstandingly warm and competent guests proved to be the more challenging to me. After my little talk on Wicksell, Hayek, and Robbins and their respective contributions to optimal and endogenous growth Muriel Dal-Pont and I discussed the relation between asymmetric information and Rational Expectation Equilibrium (EER).

Muriel's point - if I get it right - is that following Stiglitz/Grossman there is no need to associate the REE with Arrow-Debreu equilibria. My first reaction was to think of the kind of contrained Pareto optimality highlighted by Stigler/Becker and other Chicagoans. But as Muriel correctly argued, Stiglitz/Grossman goes beyond such transaction cost reasoning. In a Stiglitz/Grossman world, asymmetric information leads to strategic behaviour, that is, people trying to game the system (I may add: this goes hand in hand with the assumption of imperfect competition. If competition is perfect, that is, if every agent of the system faces a perfect outside option, strategic behaviour vanshies despite asymmetric information. Further, in large economies strategic behaviour earns little if any payoff, since all coalitions are dominated by some other coalitions. The weakly dominant strategy is to play price-taker, even if you are not ... common knowledge is however important). If rational people game the system, so Muriel, REE can describe allocations not related to the constrained Pareto efficiency developed by Stigler and Becker.

My argument against Muriel is this: she is completely right in so far as we can develop partial equilibrium models which combine rational expectations with Stiglitz/Grossman-like market failures. But as I understand their research right, they themselves argue that on a general level a rational expectations equilibrium does not exist, given their assumptions on asymmetric information. Strategic behaviour simply cannot be handled by pure Arrow-Debreu schemes. I would further argue that the existence of REE in the case of asymmetric information may still be proven if we could make more use of Hurwicz's mechanism design and check for institutional settings that ensure incentive compatibility and minimize the payoff to strategic behaviour. So, since I argued strictly on a general equilibrium level, I keep my position that if REE exist, it must be the Arrow-Debreu-Stigler-Becker style of constrained optimum that it describes. In a global model, in constrast to partial analysis, either you assume asymmetric information and skip the general equilibrium notion, or you have to skip asymmetric information so as to keep REE.

See also my comments here.