تجزیه و تحلیل بیزین از اطلاعات ارزشمند معامله گران دوگانه در بازارهای آتی

We take a closer look at the question of whether dual traders in futures markets are indeed informed traders. Underpinning this question is the intuition that a dual trader's decision to trade on his own account is not random, but is endogenously determined by his expectations of trading profits related to this decision. We employ a simultaneous equations model with two endogenous variables: (1) the binary decision of own account trading (or not), and (2) the trading profit resulting from his own account trading. Our test of whether dual traders are informed traders comprises of estimating the correlation between the error terms of the two equations in our model, where one error term proxies for a dual trader's unobserved private information and the other captures his abnormal profit. Upon estimating the model, using the Bayesian approach, we find no evidence of significant correlation between a dual trader's private information and his abnormal profit. Overall, dual traders appear to be uninformed traders with distinct trade-related characteristics.

In the wake of the seminal work by Kyle (1985), Glosten and Milgrom (1985), and Easley and O'Hara (1987), there has developed a large body of research examining how informed traders impact asset prices and market liquidity (see O'Hara, 1995 for a summary). Unfortunately, the implications from this body of work are difficult to test, due to the simple fact that informed trading is unobservable. Consequently, empirical researchers have had to resort to inferring informed trading through indirect means, such as through trading profits (Fishman and Longstaff, 1992), through trade sizes (Barclay and Warner, 1993), through trade sizes and trader types (Chakravarty, 2001), and through the timing of trades (Lee et al., 1993). In the current paper, we adopt a new empirical technique to investigate if dual traders are informed traders. Dual trading is an age-old custom in futures markets whereby some floor traders are allowed to trade both for themselves and for their customers. Our investigation is motivated by an ongoing Congressional debate on whether dual traders should enjoy such a privilege (see Chakravarty and Li, 2001) and the extant literature, both theoretical and empirical, that is silent on the specific question addressed here.2 Underpinning the Congressional debate is the issue of whether these traders can (and do) take advantage of their privileged position of observing their customers' orders to make personal trading profits. We employ a simple, yet powerful, test to investigate if dual traders are informed traders. In particular, we jointly examine a dual trader's own account trading decision and profitability by employing a simultaneous equations model with a binary endogenous variable—the decision of whether to trade on his own account—and a trading profit variable. In the own account trading equation, the error term captures the dual trader's (unobservable) private information. In the profit equation, the error term captures his abnormal profit. A significant correlation between the error term in the own account trading decision equation and the error term in the profit equation would indicate that the dual trader possesses (unobservable) private information that leads to abnormal profit. An important feature of our modeling framework is that we isolate the abnormal trading profit associated with a dual trader's personal trades from his overall trading profit and correlate the abnormal profit with the unobserved private information (if any) of the dual trader. In contrast, Chakravarty and Li (in press) examine if dual traders are informed traders by regressing their own account trades on variables capturing information (derived from observing their customers' orders), liquidity supply, and inventory control, while Fishman and Longstaff (1992) infer dual traders' information from the overall trading profit associated with their personal trades. Our data are time series of audit trails at the Chicago Mercantile Exchange (CME) during the first half of 1992 compiled by the Commodity Futures Trading Commission (CFTC). The data provide information on trade time, price, quantity, trade direction (buyer or seller) and the trader's identification. They are used internally by the CFTC for regulation and enforcement purposes. We estimate our system of equations on each of the 101 most active dual traders in the data using Bayesian techniques. We find weak evidence, at best, to suggest dual traders are informed traders. That is, there is no significant correlation between a dual trader's abnormal trading profit and his unobservable private information. We, however, identify an inventory control effect in dual traders' own account trading. That is, dual traders are more likely to buy for (sell from) their personal account if their inventory level is below (above) zero. We also find that not all of dual traders' own account trades are correlated with positive trading profits. Finally, there is strong evidence to suggest that dual traders are distinctly heterogeneous in their trade-related characteristics. Overall, our results imply that dual traders are uninformed—a finding that has important policy implications, in light of the ongoing Congressional debate on imposing personal trading restrictions on dual traders. Our research is most closely related to, and supports the conclusions of, Chakravarty and Li (in press) who examine, at the individual trader level, the timing and the determinants of dual traders' personal trades. Using correlation statistics and time series regressions, Chakravarty and Li find that there is an absence of any trade timing by dual traders in relation to the execution of their customers' orders, and the determinants of dual traders' personal trades appear to be liquidity supply and inventory rebalancing. In this paper, we focus specifically on the informativeness of dual traders, by estimating the statistical correlation between the private information of dual traders and their abnormal trading profits. The plan for the rest of the paper is as follows. Section 2 describes the empirical model used to examine dual trader informativeness and provides relevant details on Bayesian estimation. Section 3 introduces the data and defines the explanatory variables. Section 4 reports our findings. Section 5 concludes.

Using detailed audit trail transaction data compiled by the CFTC and a simultaneous equations modeling framework, we investigate if dual traders are informed traders. Our study goes significantly beyond existing research. In particular, we recognize and account for the potential endogeneity between the own account trading decision of a dual trader and his trading profit. Our methodology also allows us to isolate the abnormal trading profit associated with a dual trader’s personal trades and to compute its correlation with the unobserved private information (if any) of the dual trader. The estimation of our simultaneous equations model is performed on each of the 101 dual traders in our sample, using the MCMC method. Most significantly, we find that dual traders do not possess any significant private information.We also uncover strong evidence that dual traders are heterogeneous in terms of trade-related characteristics. Overall, our results have important policy implications in that they cast doubt on the notion that dual traders are informed traders. Rather, dual traders appear to be observationally distinct and uninformed.