اعتبار و تجارت بین فروشنده ها: تجزیه و تحلیل ساختار بازار خزانه داری اوراق قرضه

Trading generates not only information about the payoff of the assets traded, but also information about the traders themselves. Over time this information creates reputation. By using a unique dataset on the Treasury bond market, we derive a measure of reputation. This is then used to group dealers on the basis of their reputation and to analyze how they react to the reputation of other dealers. We show that the same type of trade, on the same asset, in the same market can generate different volume and volatility patterns depending on the type of dealers originating it. We also identify the “salient traders”. These traders, even if they do not originate the biggest volume of trade, have the highest impact on the market. These results have strong implications in terms of forecastability of future returns, volatility and overall trading volume because they show that most of the explanatory power of trades is due to salient traders.

Trading generates information. Traders learn about future asset payoffs and demand shocks, but also about each other. A dealer receiving an order not only acquires information about the traded asset, but also updates his beliefs about his counterpart. Over time this process generates reputation and reputation affects trading behavior. Therefore, traders’ reputation may help to explain volume and volatility in terms of market impact of trades originated by otherwise identical traders. In the present paper we empirically address this issue by directly inspecting the role played by dealers’ reputation on the mechanism of price formation in a dealership market. French and Roll (1986) argue that “the process of trading may induce volatility”. Since that paper dealers’ behavior and the interaction with market trading rules have been widely analyzed. However, the lack of data at a disaggregated level has made it difficult to properly test the role played by the existence of different types of dealers. For example, it is well known that large price movements in the Treasury Bond and FX markets are strongly affected by a release of public information (Andersen and Bollerslev, 1998; Fleming and Remolona, 1999), however it remains unclear how dealers’ interaction affects the dynamics of these adjustments. Madhavan et al. (1997) recognize that the pricing specification, and therefore volatility and volume, should contain a component that accounts for the way the trade has been intermediated. But they then generically attribute it to imperfections and market frictions without investigating it further.1 Bessembinder and Seguin (1993) suggested for the first time a connection between the volatility–volume relationship and the type of trader. More recently, Daigler and Wiley (1999), by observing the futures markets, identify two types of traders: the “in-pit” and “out-of-pit” traders. The former are the floor traders and clearing members who have an informational advantage due to the observation of the order flow. The latter are generically defined as “general public”. Trading by the informed dealers results in lower volatility, whereas trading by the general public results in increased volatility. In the FX market, Evans and Lyons (1999) and Lyons 1995 and Lyons 1997 identify a set of “microstructure-based” variables that help explain the exchange rate dynamics much better than the standard macroeconomic ones. In all these cases the classification of dealers is based on institutional characteristics (floor traders, clearing members and so on). The goal is limited to incorporating the institutional details of the market microstructure into the asset pricing literature, as opposed to directly classifying dealers in terms of their reputation or reaction to other dealers’ reputation. Indeed, while reputation has been studied from a theoretical perspective by Sobel (1985) and Benabou and Laroque (1992), no direct empirical investigation of it or estimation of its impact on the market has been carried out. Benveniste et al. (1992) argue that the information generated by the process of trading, by endowing dealers with private information about the other market participants, implicitly establishes reputation for the dealers. Madhavan and Cheng (1997), analyzing reputation in block trading, show how reputation affects the process of price formation. More recently, Battalio et al. (2001) argues that the profitability of the order flow depends on the identity of the broker who initiate the trade. 2 The empirical literature which is closer to our analysis in terms of the focus of the analysis and of the usage of data broken down at individual dealer level is the literature on interdealer trading (Gould and Kleidon, 1994; Reiss and Werner, 1995; Lyons, 1995; Hansch et al., 1998; Reiss and Werner, 1998). In particular, both Hansch et al. (1998) and Reiss and Werner (1998) use disaggregated data on dealers’ behavior in the London Stock Exchange to test dealers’ behavior. They mostly focus on testing the implications of the standard inventory model (Ho and Stoll, 1983) and analyzing dealers’ inventory management policies. We complement this literature by focusing on the informational dimension. We consider how the reaction of a dealer to an incoming order should depend not only on his inventory position and on the size of the order, but also on the identity of the dealer who has placed such an order. Repeated inter-dealer interaction generates information and creates reputation. We want to see whether reputation may affect the mechanism of price formation and, by itself, provide an alternative way of classifying the dealers. In order to investigate this issue, we focus on interdealer trading and use a unique dataset on the Italian Treasury Bond market that contains the individual transactions of all the dealers active on the market, disaggregated by dealer and type of bond. Unlike the US market, where the transactions are mostly over-the-counter, and therefore any analysis would have to be limited to a subsection of the market, the Italian market is a centralized and regulated one where all the transactions are concentrated. This allows us to observe the behavior of the market as a whole. In particular, we have available for each dealer in the market the complete set of his individual transactions in the secondary market, reported tick-by-tick and detailed by type of bond traded. Moreover, we have also information on the side that is originating the trade. This allows us to identify the informed dealers and to test the behavior of this selected small group of agents versus that of the other dealer. We are therefore able not only to endogenously define dealers’ reputation, but also to explicitly analyze the type of strategic interaction among them and to quantify the way it affects bond volatility and trading volume. We consider two dimensions of the impact of reputation on dealers’ behavior. First we focus on adverse selection and classify dealers in terms of their perception of the ability of the other dealers surrounding them (“confident” and “scared”). Alternatively, we group them in terms of the perception that the overall market has about their ability (“smart” and “dumb”). Then, we focus on how the reputation of the dealer placing an order affects the way the dealer who receives it assesses its informational content and strategically reacts to it. We therefore group dealers according to whether they try to hide the information they receive by placing orders with dealers less informed than those who have placed orders with them (“sneakies”) or they try to assess the quality of their information by placing orders with dealers more informed than those who have placed orders with them (“skeptics”). Among these classes, we identify the “salient” ones (i.e., scared, smart and skeptics), that is the dealers whose trades have the strongest market impact (in the Kyle definition). We show that in the very short run, the market is less deep when the trade has been originated by an salient dealer. We then analyze the cross-sectional and time-series differences in the impact of the trades originated by different types of dealers on overall trading volume and volatility in the periods following the transactions. We show that trade originated by salient dealers has a stronger impact over time. In particular, a smart dealer impacts on volume and volatility more than a dumb one. A scared dealer impacts on volume and volatility more than a confident dealer and a skeptic has a stronger impact than a sneaky one. Moreover, we show that the differential impact is still statistically significant at the end of the day and that the daily trade of the salient dealers impacts the market more than the one originated by the other classes of traders. While scared, smart and sneaky dealers tend to increase volatility, confident, dumb and skeptic dealers tend to reduce it. We show that at the daily level, the process of experimentation, by disseminating information earlier, reduces volatility. On the contrary, adverse selection (i.e., trading by scared and smart investors) drives volatility up. We argue that the fact that salient dealers (scared, smart and sneaky dealers) consistently have a stronger impact both in the short term and in the long term makes their trade an ideal leading indicator of future market conditions. We show that their trades have a higher out-of-sample power to forecast future returns, volume and volatility than the trade of the other classes of dealer as well as the overall aggregated trade, at every time horizon. Finally, we examine whether differences in impact correspond to differences in profitability. We show that dealers belonging to different classes display statistically different profits for all the trade-based classifications. In particular, we show that experimentation is costly as the skeptics display lower profits than the sneakies who immediately reap the benefit of their informational advantage. Adverse selection induces the scared dealers to enact trades which provide them with consistently lower profits than the ones of the confident investors. The smart investors make more profits. The paper is structured in the following way. In the next section we present our approach and define the role played by traders’ beliefs and reputation. In Section 3 we describe the market. In Section 4 we define and estimate information and reputation and classify traders on the basis of it. In Section 5 we formally lay out the econometric restrictions and report the results of the estimations. In Section 6 we calculate the profits of the different trading strategies. A brief conclusion will follow.

We studied the impact of reputation on dealers’ behavior. We identified different classes of dealers defined in terms of their perception of the ability of the other dealers surrounding them as well as in terms of the perception that the overall market has about their ability. We also grouped them in terms of the way they strategically react to the informational content of the incoming trade on the basis of the reputation of the dealers placing it. Among these dealers, we identified the “salient” ones, that is the dealers whose trades have the strongest market impact. We showed that in the very short run, the market is less deep when the trade has been originated by a salient dealer and that the trade originated by salient dealers has a stronger impact over time. Moreover, we showed that the differential impact is still statistically significant at the end of the day and that the daily trade of the salient dealers impacts the market more than the one originated by the other classes of traders. Differences in impact seem to correspond to differences in profitability, that is dealers belonging to different classes display statistically different profits. We used the fact that salient dealers consistently have a stronger impact both in the short term and in the long term, to construct leading indicators of future market conditions. In particular, we showed that the trades of the salient dealers have a higher out-of-sample power to forecast future returns, volume and volatility than the trade of the other classes of dealer as well as the overall aggregated trade, at every time horizon. These results provide many implications for future theoretical research. One possibility is the formal relaxation of the assumption of market anonymity, so as to make the determination of prices for the market maker and the reaction of the dealers dependent on the identity of the dealer placing the incoming order. Moreover, it would be interesting to directly analyze the implications of allowing the market maker to optimally choose between changing the bid–ask spread and directly placing orders with other dealers. Also, our analysis has been restricted to the trade in the secondary market. We can conjecture that dealers may coordinate their behavior in both the primary and the secondary market when there is an auction. This would involve an analysis of the joint trade in the two markets as well as an investigation of how the standard models of bidding behavior at the auction have to be changed once the reputation developed in the secondary market is accounted for. These results also may help to shed some light on the interaction between the type of market structure and its institutional features. Indeed, reputation and dealers strategic interaction depend on the amount of information existing in the market and therefore on the degree of transparency dealers deal with. It would be also interesting to consider how regulation should tackle the issue of endogenous development of reputation among market participants. In particular, it is possible that some types of regulations and disclosure rules may prevent the development of reputation or may make it very short-term. If reputation favors a quick impounding of information in prices, this enforced transparency may paradoxically hamper market efficiency.