ساختار بازار، کارآیی اطلاعاتی و نقدینگی: مقایسه تجربی بازارهای مزایده و فروش

We report the results of 18 market experiments that were conducted in order to compare the call market, the continuous auction and the dealer market. Transaction prices in the call and continuous auction markets are much more efficient than prices in the dealer markets. The call market shows a tendency towards underreaction to new information. Execution costs are lowest in the call market and highest in the dealer market. The trading volume and Roll's (Journal of Finance (1984) 1127–1139) serial covariance estimator are inappropriate measures of execution costs in the present context. The relation between private signals, trading decisions and trading profits is analyzed.

Growing competition forces stock exchanges to react to customer needs. The design of the trading mechanism is the most important determinant of market performance. The central issues are the computerization of the trading process and the choice between an order-driven and a quote-driven mechanism. This paper focusses on the second issue. Its purpose is to compare the principal alternatives, namely, the call auction, the continuous auction and the competitive dealer market. Recent changes that occured in major European stock markets indicate that the issue of which trading mechanism is best suited is far from being resolved. For example, the London Stock Exchange replaced its quote-driven trading system with the electronic order-driven system SETS in 1997. On NASDAQ, public limit orders now compete with dealer quotations. In France and Germany, on the other hand, dealers were introduced to provide additional liquidity to the electronic continuous auction markets NSC and XETRA, respectively. Continuous trading for less liquid stocks in the French CAC system (the predecessor of NSC) was replaced with two daily call auctions in 1992. On the other hand, the stocks listed on the French Nouveau Marché were transferred from a call market to the electronic continuous auction system in 1998. The London Stock Exchange considered to have less liquid stocks traded on SETS while, almost at the same time, Deutsche Börse AG announced that, for a number of less liquid stocks, continuous trading in XETRA would be replaced with call auctions. Call auctions are also frequently used to establish opening prices whereas closing call auctions and intradaily call auctions are less common. This evidence suggests that more empirical research into the relative advantages of the principal trading mechanisms is needed. The existing empirical literature has mainly focussed on comparing the liquidity of continuous auction and dealer markets. The issue of informational efficiency has rarely been addressed. This is, to a large part, due to data limitations. These limitations can be overcome in experimental research. In this paper we report the results of a series of 18 market experiments with a total of 216 participants. The experimental method allows to vary the trading mechanism under ceteris-paribus conditions. Different results can therefore be attributed to the design of the trading system. The asset value and the information each market participant holds is known to the experimenter. This allows to directly address the issue of informational efficiency. The experiments described in the present paper are designed in a way that enhances the comparability of the results to those obtained from field studies. The findings can be summarized as follows. Consistent with the practice of many exchanges to start trading with a call auction, opening prices in the call market are closer to the true value of the asset than opening prices in the continuous markets. The difference to the dealer market is significant whereas the difference to the continuous auction market is not. Generally, transaction prices in the call and continuous auction markets are closer to the asset value than are transaction prices in the dealer market. The results for the latter are largely attributable to high bid–ask spreads. Once the spread is eliminated from the data by averaging prices or by using midquotes instead of prices, the information provided by the time series of prices in the dealer market is very precise. Analysis of the determinants of price changes reveals that prices in the call market underreact to new information. No such pattern is found in the price series from the continuous trading mechanisms. Transaction costs as measured by the bid–ask spread are lowest in the call market and highest in the dealer market. This result is consistent with evidence obtained from real world stock markets. Both the trading volume and Roll's (1984) serial covariance estimator are shown to be inadequate measures of execution costs in the present context. Higher trading volume is associated with higher rather than lower spreads. The Roll measure systematically overstates the execution costs in the call market. The quality of the private information traders receive influences the composition of individual portfolios and the trading volume. A relation between signal quality and profits is only found in the dealer market. The paper is organized as follows. Section 2 provides a brief summary of the literature, Section 3 describes the experimental design and procedures, Section 4 presents the experimental results and Section 5 concludes.

This paper reports the results of a series of market experiments that were conducted in order to compare the call market, the continuous auction market and the dealer market. The design incorporates asymmetric information but guarantees that the ex-ante quality of the private signals of all traders is identical. Therefore, the aggregation of diverse information can be analyzed in the absence of insider trading. It is found that opening prices in the call market are closer to the true value of the asset than opening prices in the continuous auction and the dealer markets, although the difference to the continuous auction is not significant. The practice of many exchanges to start trading with a call auction can thus be justified. Single transaction prices in the call and continuous auction markets are generally much more efficient than prices in the dealer market. The latter is, however, very efficient when average prices are analyzed. Averaging the prices of a trading period largely eliminates the bid–ask spread. The conclusion, therefore, is that prices in a dealer market convey information of high quality, but at the expense of high transaction costs. The call market, although depicting small pricing errors, shows a systematic tendency towards underreaction to new information. This result is consistent with the predictions of theoretical models such as Kyle (1985) and the empirical results reported in Amihud et al. (1997). An analysis of market liquidity leads to the conclusion that execution costs are lowest in the call market and highest in the dealer market. This result does not depend on how execution costs are measured. It is consistent with empirical evidence obtained from field data. The trading volume turns out to be a poor proxy for the execution costs because higher trading volume is associated with higher rather than lower spreads. The experimental results confirm the prediction that Roll's (1984) serial covariance measure overstates the execution costs in call markets. The accuracy of the signals traders receive influences the composition of individual portfolios and the trading volume. A relation between signal quality and profits is only found in the dealer market. This result is consistent with efficient price discovery in the call and continuous auction markets. There are several promising ways to extend the research presented in this paper. First, similar comparisons of trading mechanisms can be performed using a design that incorporates insiders. Second, the design of the trading mechanisms can be varied. Possible variations are the introduction of indicative prices or an open order book in the call market or of varying numbers of market makers in the dealer market.