RDP 8707: Asymmetric Information and Bid-Ask Spreads in the Eurocurrency Markets 1. Introduction

It is sometimes suggested that when market participants have unequal access to information, asset prices may not be fully efficient in the sense of reflecting the combined information of all agents. A strong theoretical basis for this view is provided in an influential paper by Grossman and Stiglitz (1980). They argue that when information is costly, it will not in general be optimal for all participants in the market to acquire it; in equilibrium, some proportion of agents will invest in information, and the pricing system will be sufficiently noisy so as to allow those investors to earn an additional return which compensates them for information costs.

Whether or not asymmetric information effects are quantitatively important is of.course an empirical question. One way of approaching this question is to specify a joint hypothesis concerning informational efficiency and the determination of equilibrium asset prices, and to test such a hypothesis. This is the standard method used in the empirical literature on market efficiency. Whilst not denying the value of this approach for other purposes, it seems unlikely that these tests will tell us very much about the specific issue of asymmetric information. Rejections of the joint hypothesis, when they occur, are always ambiguous in their interpretation, and the possibility that asymmetric information effects are important is only one of the competing explanations for a rejection. The aim of this paper is to develop a method for detecting asymmetric information effects directly. The method will be to analyse the influence of the information structure on the determination of the equilibrium bid-ask spread of a security dealer. Data on bid-ask spreads in the foreign exchange markets will then be used to test for significant informational effects in these markets.

The optimal bid and ask prices of a security dealer have been studied by a number of writers. Several have considered this problem as one of optimal inventory management, for example Garman (1976), Amihud and Mendelson (1980) and Ho and Stoll (1981). In contrast to these earlier papers, Copeland and Galai (1983) and Glosten and Milgrom (1985) consider the bid-ask spread as a purely informational phenomenon, and it is on that approach that the present paper builds. In the Glosten-Milgrom model, the dealer is assumed to be risk neutral and competitive (i.e. he expects to make a zero profit on every transaction). The dealer's customers are either “informed” or “uninformed”; in other words they are motivated to trade either by liquidity considerations or by the possession of superior information to the dealer. Glosten and Milgrom are able to show that when an equilibrium exists, the dealer will in general need to set a non-zero spread between bid and ask prices in order to break even. Intuitively, the purpose of the spread is twofold. First, a large spread acts as a disincentive to the informationally motivated traders; secondly, it secures an expected profit from the liquidity traders, which can offset the expected losses to those informational traders who remain willing to transact.

The Glosten-Milgrom idea is important because it suggests a way in which the effects of asymmetric information in financial markets might be detected empirically. Their work suggests that the size of the bid-ask spread is related to the accuracy of the dealer's information, and to the number of liquidity traders relative to informational traders. If we suspect that informational asymmetries may have an influence on the mean and variance of asset prices, then the use of data on bid-ask spreads provides an obvious way of testing for such effects.

The theoretical part of this paper aims to develop in more detail the empirical implications of asymmetric information in a dealership market. To do this it takes a somewhat simpler theoretical framework than Glosten and Milgrom, by assuming that asset returns are normally distributed, and that asset demand functions are linear. This ensures that the dealer's expected profit function is continuous, and allows us to find necessary and sufficient conditions for equilibrium to exist, which Glosten and Milgrom were unable to do. The approach also does away with the somewhat artificial distinction between informed and uninformed traders, and leads to some fairly straightforward empirical implications.

The basic model and its equilibrium are described in sections 2 and 3. Section 4 focusses on how the empirical implications can be tested, and contains results using bid-ask spreads in the Eurocurrency markets. Conclusions are summarised in section 5.