THE EFFECTS OF COMMODITY PRICE STABILIZATION on an individual consumer's welfare has been controversial ever since the issue was first analyzed by Waugh [15]. The early approach of Waugh, which was based explicitly on expected consumer's surplus and ignored the production side of the economy, has since been generalized in several respects. For instance, Turnovsky, Shalit, and Schmitz (hereafter T-S-S) [14] have suggested an approach that utilizes the indirect utility function of a single individual. By focusing on a single consumer and comparing risk-no risk situations where prices are random variables, they derive interesting and useful results that express conditions for the desirability of price stabilization in terms of the familiar Arrow-Pratt measure of relative risk aversion, price and income elasticities, and budget shares. Although the approach of T-S-S provides useful information about an individual's preference for or against price stability, it does not provide information about a group of heterogeneous consumers' preference for price stability. The approach of T-S-S also assumes that perfect price stabilization is possible; indeed, in many cases government policy may serve to partially stabilize prices, but not perfectly stabilize prices.2 Thus the analysis of T-S-S does not address comparisons of risk-risk situations (the comparisons of unstable versus partially stabilized prices) nor comparisons when there are heterogeneous individuals. Recently, Newbery and Stiglitz [9, 10] have used stochastic dominance rules to analyze mean preserving partial price stabilization schemes. The purpose of the present note is to extend the T-S-S and Newbery-Stiglitz analysis by providing comparisons of partial price stabilization policies that affect multiple prices in non-mean preserving ways. Since an excellent survey of the stabilization literature appears in Newbery and Stiglitz [10], we proceed with the results of our analysis.

Let S denote a set of consumers who have identical, nondecreasing, ordinal, quasiconcave utility functions u: XS(t) -* u[X (t)], where XX(t) is the vector of n goods consumed by individual s at time t. Consumers shop at different stores and hence may pay different prices for commodities.4 Let PS(t) and yS(t) denote exogenous vectors of commodity prices and income, respectively, faced by individual s at time t. This does not preclude the existence of a subset, 5, of consumers who all face the same prices, i.e., P5(t) = pS (t) for all s, s'c S. At each instant in time consumers attempt to

[This paper finds in closed form a noisy rational expectations equilibrium for a class of economies with many risky assets and analyzes the properties of such equilibria. Because of the various interactions between the assets, phenomena appear that do not arise in models with a single risky asset. For example, an asset's equilibrium price might be decreasing in its own payoff and/or increasing in its own supply; an asset might be a Giffen good; a higher price for an asset (holding other prices fixed) might be "bad news" for the asset's payoff; and even for assets in fixed supply, uncertainty about other assets' supplies may prevent their prices from being fully revealing.]

In a series of eleven markets, sellers possessed products that were exogenously designated as either grade "regular" or grade "super." Supers were valued more by buyers but grade could not be observed by buyers prior to purchase. Sellers could add costly units of quality to their products that were observable and valued by buyers. The data are analyzed with perfect information models, signaling equilibrium models, and pooling models. A variety of behaviors are observed across the eleven markets. Signaling is observed in most markets with some markets approaching the most efficient signaling equilibrium. Pooling or partial pooling occurs in a few markets. The performance seems to be sensitive to the relative cost of signaling and the market institutional setting.

On montre a partir de l'evidence de Monte Carlo que les differences dans la forme fonctionnelle des restrictions non lineaires sont importantes dans les petites tailles d'echantillons