تهاجم و کمیت سفارش : چگونه آنها در یک بازار سفارش محدود تعیین می شود؟

Dealers trading in a limit order market must choose both the order aggressiveness and the quantity for their orders. Since little research has considered how dealers make this trade-off, we empirically investigate how dealers jointly make these decisions in the foreign exchange market using a unique simultaneous equations model. Our model uses an ordered probit model to account for the discrete nature of order aggressiveness and a censored regression model to capture the quantity decision recognizing the clustering of orders at the smallest available quantity, $1 million. Using two currency pairs with very different trading characteristics, we find evidence of a trade-off between order aggressiveness and quantity. We also find a significant role being played by factors related to the levels of information asymmetry and liquidity in the dealers’ choices of both the order aggressiveness and quantity.

In a limit order market, traders submitting an order must choose both the order aggressiveness (price) and the quantity. Order aggressiveness is defined according to the execution priority, with market orders being the most aggressive because they are executed immediately at the best prices currently standing in the market. Limit orders are less aggressive because they are submitted at a set price, so their execution is not guaranteed and they follow strict price and time priorities in execution. Despite orders being defined by both price and quantity, much of the literature focuses on the relationship between various factors and order aggressiveness (i.e., Griffiths et al., 2000 and Ranaldo, 2004, and more recently Ellul et al., 2007 and Cao et al., 2008). In the majority of existing studies, the quantity dimension is either ignored or treated as one of the explanatory variables rather than as a choice variable. Although the existing studies recognize the costs associated with the risk of being picked-off or not having trades executed, they frequently assume trades of a standard size and therefore ignore a key component of the overall cost to submitting orders. Studies investigating order aggressiveness have found relationships between market volatility, the bid-ask spread, and the depth of the limit order book on the order aggressiveness decision (e.g., Al-Suhaibani and Kryzanowski, 2000, Ahn et al., 2001, Hall and Hautsch, 2006, Hall and Hautsch, 2007 and Foucault et al., 2007). Although some studies have empirically examined traders’ quantity decisions (e.g., Biais et al., 1995, Al-Suhaibani and Kryzanowski, 2000 and Moulton, 2005), their models do not consider the simultaneity of the price-quantity decision and the constraints in placing orders of different sizes. The theoretical literature addressing these issues includes studies such as Easley and O’Hara (1992) and Glosten (1994) and more recent papers such as Dridi and Germain (2004), Goettler et al. (2005) and Foucault et al. (2007). All discuss a role for quantity and price but do not explicitly examine the dynamic trade-offs that are made between price and quantity.1 The ability to change the size of the order at different prices plays an important role in determining the price schedule (e.g., the market clearing price) over time and, as a result, in determining the overall cost to submitting orders at different levels of aggressiveness (the overall cost is the feature of most relevance to traders when submitting their orders). Our study therefore seeks to extend the existing literature by empirically examining some of the findings from these studies and how they relate to the price-quantity decision. It is important for researchers and practitioners to gain an understanding of the factors that influence the price-quantity decision in order to obtain some useful information about the ways liquidity is provided to the market and what factors lead dealers to demand or supply liquidity. We accomplish this by simultaneously estimating traders’ choice of order aggressiveness and quantity to more completely model traders’ order submission strategies. We consider both factors simultaneously because traders choose both order aggressiveness and quantity when submitting orders in a limit order market. Although order aggressiveness has been modeled before, the inclusion of the joint role of quantity is new. In our model, order aggressiveness is estimated using an ordered probit model (as in Griffiths et al. (2000), Cao et al. (2008), and Ranaldo (2004)), thereby accommodating the discrete nature of price aggressiveness. 2 Quantity is modeled using a censored regression framework, which allows our model to accommodate a wide variety of order sizes as well as the clustering of orders at $1 million. Our model therefore captures both the simultaneous nature of the price-quantity decision as well as many of the empirical features of the data—price clustering, quantity censorship, etc. Beyond being the first study to present a detailed empirical model that, when compared to the existing literature, more completely models the order submission process, our study also contributes to the market microstructure literature by using foreign exchange data on two currency pairs – the Deutschemark–U.S. dollar and the Canadian dollar–U.S. dollar – that possess very different trading characteristics. We use data on firm orders from one of the world's largest foreign exchange electronic broker systems. This system allows dealers to submit and cancel market orders and limit orders of different sizes at different price levels. In this way, we extend the existing microstructure literature that currently focuses on equities markets, especially less liquid markets.3 To ensure the robustness of our results and investigate the possible role for differences in liquidity on our results and those existing in the literature, we consider data from both the very actively traded Deutschemark–U.S. dollar market (referred to below as the DM) and the less actively traded Canadian dollar–U.S. dollar market (referred to below as the Canadian dollar). Using foreign exchange data allows us to benefit from some of the differences in the structures of the equities and foreign exchange markets.4 Investigating the order submission process for two currency pairs that have very different trading characteristics also allows us to also examine the impact of differences in asset liquidity on the order submission decision. The foreign exchange market is a useful market to consider because dealers are able to submit both market and limit orders 24 h a day. Hence, our data is not impacted by formal opening and closing procedures or by other institutional mechanisms that may influence order submission and execution. Consequently, our analysis is able to more carefully examine the influence of changes in the state of the order book, market conditions and liquidity on the dealers’ order submission decisions. We can also examine the role within the order submission process of different types of private versus public information. Every order submitted over the trading day is contained in our dataset,5 but dealers can observe only the best price and corresponding quantity posted on both sides of the market in addition to information on the last five transactions. To investigate the potential differences in the roles of public and private information, we estimate a model using only the publicly available information on orders at the best quoted prices and past transactions; we also provide an extended model in which we include information that is unobservable to traders on the orders behind the best quoted prices. This process allows us to determine whether the information from the orders not directly observable by dealers will, in turn, affect the dealers’ price and quantity decisions. If this extra information turns out to be irrelevant in our analysis, then such a finding would support the fact that dealers often use other sources, such as their private customer base, to obtain relevant information on the state of the market, i.e., information they use in making their decisions at each point in time. The inclusion of this data in our analysis therefore allows us to examine the level of information available to dealers, as well as how it influences their decision-making and, consequently, the degree of market efficiency. We find that dealers’ willingness to submit aggressive orders varies by currency. For the Canadian dollar, dealers are more willing to submit aggressive orders than they are for the DM. For both currencies, the less aggressive orders are more concentrated in the middle of the day when the quantity is at its peak. These results suggest that dealers are more interested in execution speed around market open and close, especially for the less liquid asset, i.e., the Canadian dollar. Next, we find that the influence of different characteristics of the order book on dealers’ order submission decisions varies across the DM and Canadian dollar. For the Canadian dollar, the order submission decisions appear to be designed to profit from the supplying of liquidity. The decisions in the DM market, however, are more consistent with standard arguments related to traders wishing to minimize transaction and execution risks, especially in the presence of asymmetric information. Consequently, our results for the DM are relatively consistent with those from existing studies but our results are less similar for the Canadian dollar, where the value of providing liquidity is likely much higher for dealers and thus plays a larger role in their order submission decisions. When examining the role of private information on the order submission decisions, we find that it played only a limited role. The results are significant only for the DM and confirm our earlier findings that were obtained using only the publicly available price and quantity information. Our paper develops as follows. In Section 1 we discuss the data we used. Section 2 presents our empirical model and our hypotheses. Section 3 discusses the results from the basic model. Section 4 considers how the performance of the market affects the submission and cancellation decisions. Conclusions are offered in the final section.

In this paper we investigate two of the most important components of a dealer’s order submission decision: the aggressiveness of the price and the quantity at which dealers are willing to transact. Despite the importance of the quantity decision for every order submitted, this aspect of the order submission decision has received relatively little attention in the microstructure literature. Using a simultaneous equation model, we investigate the factors that influence the dealers’ order submission decisions. Our information comes from firm quotes submitted to one of the largest electronic brokerage systems for the Deutschemark–U.S. dollar and the Canadian dollar–U.S. dollar. Since this data set encompasses information on every order submitted over the trading day (which includes information not visible to all market participants), we can investigate the role of public and private information on the order submission decision for a very heavily traded – and thus very liquid – currency and from a less heavily traded or relatively illiquid currency. We find a negative relationship between order aggressiveness and quantity – order size tends to be smaller when the order is more aggressive. By submitting smaller orders, traders submitting market orders avoid the potentially higher cost of execution that results from having to walk up/down the order book for the order to be executed. The result here is consistent with the theoretical model of Goettler et al. (2005) . We do not find that order aggressiveness itself plays a significant role in theorder size decision. Consequently, it is important to study both the order aggressiveness and quantity parts of the order submission decision. Consistent with other studies, we also find a significant role for changes in market conditions on the order submission decision; however, we find that several of the relationships differ between the DM and the Canadian dollar. Most of the differences appear to be related to the differences in liquidity of the two currencies. For the more liquid DM, the market reacts to changes in many of the factors in a manner consistent with many theoretical studies (e.g., Parlour, 1998; Foucault, 1999; Foucault et al., 2005 and empirical studies such as Bae et al., 2003; Ranaldo, 2004; Ellul et al., 2007 ). However, the reactions are not the same for the less liquid Canadian dollar, where the changes in market conditions and the state of the limit order book can be interpreted differently. For the Canadian dollar, dealers can profit more by supplying liquidity, so they react differently to changes in market conditions than they would, for example, in the already very liquid DM market. For the private information on the orders behind the best price, we found that the impact is smaller than for the information at the best price, but it still impacts both the order aggressiveness and the order size decision, even though the information at the best price is unobservable to market partic- ipants. When the behind best depth increases on the same side of the market, order aggressiveness and order size decrease significantly. The intuition is that the depth behind the best price could reveal information on the underlying efficient price. Recent studies such as Bloomfield et al. (2005) , Kaniel and Liu (2006) and Goettler et al. (2005, 2009) show that informed traders optimally submit limit orders, so the depth behind the best price represents bets backed up by real money on the future direction of price change. As a result, traders submit smaller and less aggressive orders when the behind best depth increases. Our analysis provides an important step in improving our understanding of how dealers make their price-quantity decisions when submitting limit orders. By using a simultaneous equations model, we can clearly see the components of this decision-making process. Our use of the Deutschemark and the Canadian dollar further helps us by highlighting the significant role played by the supply and demand for liquidity in order aggressiveness and order size decisions, as well as in their interactions.