Noncooperative Equilibrium and Market Signalling
研究了当买方对产品质量的信息少于卖方时，高质量产品卖方如何通过信号活动（如广告、教育、保险共保率）来区分自己，并分析了这种信号均衡的稳定性问题。
Among many recent developments in the economics of information, none has generated more controversy than the concept of market introduced by A. Michael Spence. If buyers are less well-informed about product quality than sellers, and no additional information is available, market-clearing prices must reflect some weighted average of product quality. Then if potential sellers of the highest quality products have the greatest opportunity costs and these costs exceed price, they will not enter the market. This adverse selection phenomenon is, however, offset if sellers of higher quality products can adopt activities that operate as a to potential buyers. Intuitively, an activity is a potential signal if entering into it has a lower marginal cost for sellers of higher quality products. For example, the fly-bynight operator faces a higher advertising cost per unit of sales than the new entrant who plans to build and then maintain his reputation. Or, in the labor market, productivity on the job is positively correlated with performance in school. Therefore, the higher productivity worker has on average a lower personal cost of obtaining a given set of educational credentials. Similarly, in purchasing insurance the marginal cost of accepting a higher coinsurance rate for a specific loss is lower for those with a lower probability of loss. But on closer inspection, it turns out that equilibria-that is, equilibria in which signalling is needed to distinguish product quality-do not have the stability properties of classical Walrasian equilibria. In papers by Michael Rothschild and Joseph Stiglitz, and the author (1975, 1977) it has been shown that unless the difference between quality levels is sufficiently large there is, for every set of signal-price pairs S, some alternative s' 4 S which, if offered by a single seller, would improve his lot. That is, there is no Cournot-Nash signalling equilibrium. One possible inference that might be drawn from this result is that signalling could not be an important phenomenon in a competitive economy. After all, if any attempt at signalling were to result in interference, in the form of competitive responses, there would be no opportunity for individuals to identify the correlation between the level of the signal and the underlying quality of the product. However this is too simplistic since the interference cannot take place until after signalling has been established. It is therefore more appropriate to suppose that an economy is initially in a state of informational equilibrium with signals reflecting product quality, and to ask whether the potential instability is likely to lead to a general collapse of the equilibrium. Following the approach adopted by Charles Wilson there have been several related attempts to introduce quasi-dynamic considerations in which every agent takes into account the reaction by other agents when contemplating a defection from the initial set of signal-price pairs S. It is then argued that the set S forms a stable equilibrium if, for every alternative s' that would make a single defector better off, this same signalprice pair would make the defector worse off after the reaction by other agents. Where these attempts differ is in the *University of California-Los Angeles. This research was supported by National Science Foundation Grant SOC-76-13443 to the Rand Corporation.