دقت اطلاعات، هزینه های معامله، و حجم معاملات

Most theoretical models of trade (Pfleiderer, 1984; Grundy and McNichols, 1989; Holthausen and Verrecchia, 1990; Kim and Verrecchia, 1991; Blume et al., 1994) imply that the trading volume prompted by a public announcement is positively related to the announcement's precision. Relying upon this notion, empirical researchers interpret high trading volume as an indication that an announcement is highly informative. We argue that such interpretations are not, in general, correct. In a world with transaction costs, the relation between information precision and trading volume is ambiguous and can be negative. This explains why, in empirical tests using data from actual markets, the relation between announcement precision and trading volume is not monotonically positive, even though in laboratory experiments it is. Our results imply that trading volume reactions to public announcements are most sensitive to announcement precision among low-transaction cost securities and in low-cost trading regimes.

Can we tell when a public announcement conveys meaningful information to investors? And just what can be inferred from a trading volume reaction to new information? For nearly 25 years, researchers have had answers to these questions. Beaver (1968), Morse (1981), Bamber (1986), and Bamber and Cheon (1995), for example, all interpret earnings announcements that are accompanied by high trading volume as conveying more information to investors than announcements that generate low trading volume. Supporting such interpretations, several rational expectation models demonstrate that trading volume increases with the precision of investors' information.1 The notion that high volume accompanies informative announcements has taken roots in both the theoretical and empirical literature. Blume et al. (1994), for example, use it to advance a rationale for using trading volume in technical analysis. Hence, the conventional wisdom holds that meaningful news generates trading volume, and that trading volume is a useful measure of a public announcement's information content. Unfortunately, however, this conventional wisdom is incorrect. In this paper we demonstrate that one should not expect, as a matter of course, a monotonically positive relation between announcement precision and the corresponding trading volume. As a result, the interpretation of trading volume reactions to public announcements, and our understanding of which types of announcements convey the most information to investors, are more complicated than recognized in most empirical research. The key to our argument is the cost of trade. Consider as an example Kim and Verrecchia's (1991) model. In this model precise public announcements cause investors to agree more about the asset's value, homogenizing investors' private valuations of the risky asset. Precise public announcements also increase each investor's confidence about his or her private valuation. Bolstered by such confidence, investors become more willing to take speculative positions. In a world of costless trading the net result of these effects is to increase trade. At the extreme, a very precise public announcement could leave investors' valuations differing by only a penny, but each investor would be virtually certain that he or she was correct. Betting on their precise beliefs, investors would take extremely large speculative positions, thereby generating large trading volume. If transactions are costly, however, investors will not dicker over the last penny. Informative announcements will homogenize valuations, but many potential gains from trade will be outweighed by the transaction costs. At the extreme, a very precise and informative announcement will generate no trade, because investors' valuations will converge to such an extent that they lie within the bid–ask spread. In general, when transactions are costly, the most precise and informative announcements will trigger the fewest trades. The effect of transaction costs can explain the available empirical evidence regarding trading volume and information precision. When transaction costs are zero, our model predicts that highly informative public announcements cause investors' beliefs to converge and simultaneously generate large trading volume – the same prediction as in previous models. This is exactly what Gillette et al. (1999) find in an experimental market in which transaction costs are zero. In actual markets with positive transaction costs, however, our model predicts that trading volume is not monotonically and positively related to information precision and the convergence of investors' beliefs. This result, which other models do not generate, also is consistent with available evidence. Ziebart (1990) and Bamber et al. (1997), for example, find that trading volume is negatively related to the convergence in analysts' forecasts around earnings releases. Barron (1995) reports that trading volume is negatively related to convergence in analysts' forecasts in general. Barron et al. (2003) find no significant relation between abnormal trading and a measure of announcement precision that is derived from Barron et al. (1998). And Wasley (1996) finds that trading volume is first increasing, and then decreasing in the precision of management earnings forecasts, with the most precise forecasts generating the lowest trading volume reactions. All of these findings are inconsistent with the conventional notion that trading volume increases with information precision and the convergence of beliefs. They are consistent, however, with our model of trading volume reactions in the presence of transaction costs. In the next section, we develop our argument in a pure exchange model with transaction costs. The model assumes the same information environment used by Kim and Verrecchia (1991), in which investors have heterogeneous prior private information and receive a common signal from a public announcement. Our general proposition regarding the importance of transaction costs, however, is not limited to this one type of informational setting. To demonstrate that the proposition applies broadly, in Section 3 we examine an informational environment based on the Holthausen and Verrecchia (1990) model. In this framework, investors have homogeneous priors but interpret the public announcement differently. Regardless of one's specific assumptions about investors' information and beliefs, the presence of transaction costs stands conventional wisdom on its head, as trading volume no longer is monotonically related to information precision. In Section 4 we contrast our results to several theoretical papers that also yield non-monotonic relationships between volume and precision, and argue that the transactions cost argument is most consistent with the available evidence. Section 5 concludes the paper with a discussion of our argument's implications for empirical research.

Previous theoretical models (e.g., Kim and Verrecchia, 1991) predict that trading volume increases with the precision, or information content, of a public announce- ment. These models reflect the intuition that increased precision makes all traders more confident about their private valuations, encouraging greater speculative posi- tions and higher trading volumes. Empirical researchers rely on this presumed rela- tion between precision and trading volume when they use trading volume data to make inferences about the information content of certain public announcements.In this paper we show that the relation between information precision and trading volume is sensitive to transaction costs. Increased precision bolsters each investor s confidence in his or her private valuation, but it also homogenizes investors beliefs and causes their demand-prices to converge. In a world of costless trading, highly precise announcements encourage investors to take extreme positions even though their valuations of the asset differ by small amounts. If trades are costly, however, any potential gains from trade can be swamped by transaction costs. At the extreme, highly precise announcements can cause very little trading because they cause inves- tors valuations to converge. In their investigation of price changes and trading volume, Holthausen and Ver- recchia (1990) propose that trading volume: (i) increases with an announcement s precision and (ii) decreases with investors consensus about the information. Verrec- chia (1981), demonstrates that, because of investors heterogeneous risk tolerances, condition (ii) is neither necessary nor sufficient to generate trading volume. One way to think about our analysis is that, because of transaction costs, increases in in- vestor precision (condition (i)) also are not sufficient to generate increased trading volume. Thus, care must be taken in deriving inferences from trading volume data about the precision of public announcements. As pointed out in the introduction, our argument helps reconcile several otherwise puzzling empirical results involving the trading volume reactions to public an- nouncements. Wasley (1996), for example, finds that the trading volume reactions to earnings forecast announcements from managers are not monotonic in the an- nouncement precision. In particular, trading volume first increases, then decreases with the announcement precision. Wasley notes that this is inconsistent with the Kim and Verrecchia (1991) model. Our model, however, demonstrates that Wasley s results are consistent with an information environment that is similar to that of Kim and Verrecchia (1991), but when transaction costs are allowed to be positive. Transaction costs also can explain the apparent discrepancy between empirical and experimental findings regarding the trading volume reactions to public announcements. Using data from an experimental market, Gillette et al. (1999) find that trading volume is positively correlated with the convergence of investors forecasts following an an- nouncement. Noting that convergence in investors beliefs is a measure of an announce- ment s precision, Gillette et al. conclude that this result is consistent with the Kim and Verrecchia (1991) model. Using data from actual security markets, however, Ziebart (1990), Bamber et al. (1997), and Barron (1995), find that trading volume is negatively related to convergence in analysts forecasts. Barron et al. (2003) find a negative relation even after controlling for the potentially confounding effects of changes in consensus. The apparent discrepancy is attributable to differences in transaction costs. Gillette et al. s experimental market has zero transaction costs. It yields results that are consistent with the Holthausen and Verrecchia (1990) and Kim and Verrecchia (1991) models because it mimics these models zero-transaction cost environments. Empirical data from actual markets, however, yield a non-monotonic volume–preci- sion relation because trades are costly in the real world. Our analysis yields several implications for empirical tests. The fact that transac- tion costs differ across stocks implies that the sensitivity of trading volume reactions to the precision of public announcements also will differ cross-sectionally. For exam- ple, Stoll and Whaley (1983) and others report that transaction costs are inversely related to firm size. This implies that, holding constant the characteristics of the pre- disclosure information environment, trading volume will be less sensitive to the pre- cision of public announcements regarding small firms (where transaction costs are large) than for public announcements regarding large firms. A second empirical implication rests on the observation that the per-share cost of trading options typically is lower than that of trading in the underlying securities. When the cost of trading the underlying security is relatively high, a highly precise announcement will tend to generate little trading in the security, but may generate trading in the options market, where transaction costs are smaller. This implies that the ratio of the trading volume reaction in the options market to the trading volume reaction in the underlying asset market should be higher for precise public announce- ments than for less precise announcements. Not only do transaction costs vary across firms, but they also vary across inves- tors and have been declining rapidly over time, especially for small investors. An- other empirical implication of our analysis is that the sensitivity of trading volume reactions to the precision of public announcements should be increasing over time, as transaction costs have decreased.