مدیریت موجودی رقابتی در بازارهای خزانه داری

We decompose US Treasury bid-ask spreads into inventory, adverse selection and order processing costs by using the fact that inventory trades have different effects on spreads than do proprietary trades. We exploit this asymmetry and develop a technique to identify the three components of the spread in order to test three hypotheses: dealers make larger changes to inventory (1) following macroeconomic announcements (2) at the start and toward the end of the New York trading hours, and (3) when transaction sizes are relatively large. We test these predictions using GovPX data for on-the-run 2-year and 10-year Treasury Notes. All three predictions are supported. We also assess how primary dealers react to the Federal Reserve’s open market operations (OMOs). Our findings reveal interesting intraday patterns in the inventory component for both securities.

The market microstructure literature provides us with a general framework for decomposing the dealer’s bid-ask spread into adverse selection, inventory, and order processing costs.1 It is typically assumed that rational market makers incorporate all three costs in spreads.2 For the Treasury market, a strong case may be made ex ante for a large inventory component. This around-the-clock market is characterized by high-value trades that tend to cluster around the beginning and end of New York trading hours. Moreover, at any given time, many dealers in this market may be driven as much by self-preservation as by profiting from the spread. For instance, the market-making profits for primary dealers of US Treasuries are relatively minor compared to their profits or losses from taking on proprietary positions (Sundaresan, 1997).3 This might imply strong motives to reverse trades that are nonproprietary in nature. In fact, the motive to make the market may be mostly incidental; these dealers are afforded preferential treatment by the Federal Reserve Board in primary issues of Treasuries, but in turn are expected to serve as providers of liquidity in the secondary market (also see GAO, 1987). Since dealers making the secondary market for Treasuries may be uniquely motivated to control their inventories closely, these markets provide an excellent venue to examine the nature of the inventory component of the spread. We decompose US Treasury bid-ask spreads into inventory, adverse selection and order processing costs by using the fact that inventory trades have different effects on spreads than do proprietary trades. We exploit this asymmetry and develop a technique to identify the three components of the spread in order to test three hypotheses: the largest adjustments in the inventory component are made (1) following macroeconomic news announcements; (2) towards the start and end of New York trading hours; and (3) when transaction sizes are large. We test our three hypotheses using on-the-run 2-year and 10-year notes, employing intraday transaction data pertaining to inter-dealer trading. All three hypotheses are supported. Further, our framework permits us to assess how primary dealers react to the Federal Reserve’s open market operations (OMOs). Our findings reveal interesting intraday patterns in the inventory component for both securities. We allow for an adverse selection component in our decomposition because Treasury dealers may face information asymmetries, albeit in the less conventional sense. First, the bid or ask quote that a dealer submits to inter-dealer brokers is held firm for two minutes unless it is transacted upon, or a new and more favorable post occurs. The firmness of the quotes leaves the dealer vulnerable for a period of time during which market conditions may change.4 Additionally, dealers establish relatively large proprietary positions in Treasuries. Dealers commonly leverage their long positions via the repo market, and establish significant short positions to hedge against their positions in mortgage-backed securities (Sundaresan, 1997). Knowledge of large forthcoming transactions may allow these dealers to better predict the short-term evolution of prices, affording them a degree of market power (Cornell, 1993 and Huang and Stoll, 1997). Fleming (1997) provides the first detailed intraday analysis of the Treasury markets. He finds that trading volume and price volatility are highly concentrated during New York trading hours and less so during London and Tokyo trading hours. He concludes that Treasury markets, although in operation around-the-clock, tend to behave like US equity markets with limited trading hours. Our study adds to the growing literature that examines the behavior of spreads in Treasury markets. Pasquariello and Vega (2007) focus on the determinants of liquidity and price differentials between on-the-run and off-the-run US Treasury bonds. They note that mean spread differentials between off-the-run and on-the-run bonds are economically and statistically significant, even after controlling for other bond characteristics. Brandt and Kavajecz (2004) show that changes in the yield curve in Treasury markets may be related to the aggregation of heterogeneous private information. Green (2004) finds a significant increase in the informational role of trading following economic announcements. The focus of our study is on the inventory component of the spread.5

We set forth a framework for a three-way decomposition of the bid-ask spread in the multi-dealer market for US Treasuries. Our decomposition of the spread employs contemporaneous spread dynamics based on inventory regimes and allows us to test several hypotheses in these markets without having to track future price reversals. Specifically, we test whether dealers in this market make larger adjustments in the inventory component (1) following major macroeconomic news events, (2) towards the start and end of New York trading hours, (3) when transaction sizes are relatively large. Using on-the-run 2-year and 10-year Treasury notes we find that our predictions are supported by our data from July 1996 through June 1999. The evidence suggests that dealers offset increased risk during these times largely through inventory management. Further, our framework permits us to assess how primary dealers react to the Federal Reserve’s open market operations (OMOs). Our findings reveal interesting intraday patterns following OMOs in the inventory component for both securities. Our results are consistent with the findings of Manaster and Mann (1996) who examine futures, and are less consistent with the results found by Madhavan and Smidt, 1993 and Hasbrouck and Sofianos, 1993. The latter two studies examine equity markets. Manaster and Mann find that traders in futures markets aggressively manage their inventories and that individual market-maker inventories are mean-reverting. For example, they find that the median S&P 500 index trader reduces inventory imbalances by almost 50% in the next trade. Evidence that trade direction is negatively correlated with inventory is also found. This leads them to conclude that traders in futures markets adjust inventory levels much faster than traders in equity markets. In contrast, in equity markets specialists are reported to take several days (weeks) to adjust their inventories.