Anticipated and Repeated Shocks in Liquid Markets* Dong Lou London School of Economics [email protected] Hongjun Yan Yale School of Management [email protected] Jinfan Zhang Yale School of Management [email protected] This Draft: May 2013 * We are grateful to Geert Bekaert (the Editor) and three anonymous referees for their constructive comments. We also thank Viral Acharya, Nick Barberis, Mikhail Chernov, Pierre Collin-Dufresne, Greg Duffee, Darrell Duffie, Michael Fleming, David Goldreich, Jeremy Graveline, Robin Greenwood, David Hirshleifer, Owen Lamont, Peter Kondor, Arvind Krishnamurthy, Andrew Metrick, Lasse Pedersen, Christopher Polk, Tyrone Smith, Jeremy Stein, Dimitri Vayanos, Lei Xie, Tong Yu, and seminar participants at Beijing University, London School of Economics, Temple University, University of California Irvine, University of Lugano, University of Massachusetts Amherst, Yale University, Econometric Society Annual Meetings 2011, Roundtable on U.S. Treasury Markets and Debt Management, Western Finance Association Annual Meetings 2011, China International Finance Conference 2011, Society for Economic Dynamics Annual Meetings 2011, and SAC Capital for helpful comments. We are also grateful to Pengjie Gao, Andrea Vedolin, and Tong Yu for their generous help with some of the data used in this paper. Lou acknowledges financial support from the Paul Woolley Center at the London School of Economics. Yan received support from a Whitebox Advisors Research Grant and from the National Natural Science Foundation of China (grant 70932003). Anticipated and Repeated Shocks in Liquid Markets Abstract This paper examines how anticipated and frequently repeated shocks are absorbed in liquid financial markets. We show that Treasury security prices in the secondary market decrease significantly in the few days leading up to Treasury auctions and recover shortly thereafter, even though the time and amount of each auction are announced in advance. The issuance cost to the Treasury Department is estimated to be between 9 and 18 basis points of the auction size, or over half a billion dollars for note issuance alone in 2007, most of which can be attributed to the price pressure effect around auction days. These results are linked to dealers’ limited risk-bearing capacity and the imperfect capital mobility of end-investors, highlighting the important role of market frictions even in very liquid financial markets. JEL Classification Numbers: G12. Keywords: Liquidity, Capital mobility, Supply shocks, Treasury auctions. I. Introduction In this paper, we empirically examine how anticipated and frequently repeated supply shocks are absorbed in liquid financial markets. In particular, we examine the temporary price impact of Treasury security auctions on the secondary Treasury and repo markets. This may appear as a surprising agenda: Treasury auctions are conducted every month, of which the exact time and amounts are announced in advance, so these events are largely anticipated. Given the size and liquidity of the U.S. Treasury and repo markets, one might expect no appreciable price impact from these anticipated events. In sharp contrast to this conventional view, our evidence reveals significant temporary price effects in both markets around Treasury auctions. Specifically, we find that Treasury security prices in the secondary market tend to dip in the few days leading up to Treasury auctions and recover shortly thereafter. An intuitive way to describe this price dip and recovery pattern is to compare Treasury returns around auctions: For example, the 5-day cumulative return of an on-the-run 2- year Treasury note before a 2-year note auction is, on average, 8.89 (t = 2.93) basis points lower than the 5-day post-auction return of the same security. 1 This pattern holds true for off-the-run securities, as well as other maturities. These findings have potentially important implications. First of all, they suggest that the cost of security issuance borne by the Treasury Department is an order of magnitude larger than what is implied by prior research on Treasury auction underpricing. This strand of research compares the auction price with a benchmark price on the auction day, and finds the underpricing to be less than 1 basis point of the 1 An on-the-run Treasury security is the most recently issued security of a given maturity. The first off- the-run and the second off-the-run securities are the second and third most recently issued securities of a given maturity, respectively. 1 auction size (Goldreich, 2007). Our findings suggest that, due to the price impact resulting from Treasury auctions, these benchmark prices on auction days are already depressed. Therefore, while the traditional approach is appropriate for measuring underpricing in auctions, it does not reflect the total issuance cost borne by the Treasury Department. To incorporate this price impact into our cost measure, for each n-year note auction (n = 2, 5, 10), we take the average price of the n-year note in the secondary market on the tth day before and tth day after the auction as the benchmark price, with t ranging from 1 to 10. That is, our measure reflects the amount of money the Treasury Department could have saved were it able to issue Treasury securities at the average secondary market price during the several days before and after each auction. For t = 5, for example, our estimates of Treasury issuance costs for 2-, 5-, and 10-year notes are 9.07, 16.81, and 18.43 basis points of the auction size, respectively. According to these estimates, the total cost of issuing Treasury notes alone in 2007 is $649 million. While it is by no means clear whether part of the estimated cost can be saved through better designs of the Treasury selling mechanism, recognizing such a cost is a necessary first step to understand and improve the efficiency of Treasury auctions. Second, our findings suggest that the frictions behind the market’s response to demand/supply shocks are important even in very liquid financial markets. In particular, we compute the return of a long-short strategy that exploits these reoccurring swings in Treasury prices around auction days. By going short the on-the-run 2-year note and long a duration-matched portfolio comprising the on-the-run 6-month Treasury bill and 10-year note during the 10 days before each 2-year note auction, and holding the reverse positions during the 10 days after, one can achieve an annualized Sharpe ratio of 0.84, 2 after accounting for bid-ask spreads and repo funding costs. This is comparable to many well-known asset pricing anomalies, such as currency carry trade and momentum in the U.S. stock market. Our evidence that even the Treasury market is unable to quickly absorb these anticipated and repeated shocks suggests that market frictions play a perhaps more important role in the financial system than previously thought. 2 Our evidence is consistent with the interpretation that Treasury auctions cause temporary price movements in secondary Treasury markets. The underlying mechanism has two ingredients: primary dealers’ limited risk-bearing capacity and end-investors’ imperfect capital mobility. Primary dealers are expected to participate actively and submit competitive bids in all Treasury auctions. They tend to hedge the risk they are about to acquire in Treasury auctions by short selling similar securities in the secondary market before these auctions, thus exerting downward price pressure in the secondary market.3 Consistent with this interpretation, we find that the price impact in the secondary market is more pronounced precisely when the auction size is larger, when dealers are more capital constrained, or when interest rates are more volatile. Further, primary dealers’ short selling would not have a price impact were there enough end- investors or arbitrageurs to supply liquidity. For one thing, arbitrageurs are unlikely to be able to absorb the entire supply from Treasury auctions, which is in the magnitude of tens of billions of dollars. In addition, major end-investors, including foreign investors, state and local governments, insurance companies, and bond mutual funds are passive 2 In a recent study, Fleckenstein, Longstaff, and Lustig (2010) provide another piece of evidence on the imperfection of the Treasury markets: the price of a Treasury bond is significantly higher than an inflation-swapped TIPS issue that replicates the cash flows of the Treasury bond. 3 Primary dealers can also take short positions in the when-issued market. But this can also depress spot Treasury prices, if the counterparties with long positions in the when-issued market hedge their exposures in the secondary Treasury market. 3 investors, many of whom do not have the intention, or resources, to engage in short- term arbitrage trades. This interpretation has a number of further predictions. First, it implies that repo rates should be lower before Treasury auctions compared to those after. This is because short positions in the secondary Treasury market are usually established through reverse repo transactions. Specifically, primary dealers lend cash to their counterparties and take Treasury securities as collateral, which they then short-sell in secondary markets. Primary dealers’ strong demand for these transactions before auctions implies that they would be willing to accept lower interest rates on their lending, leading to lower repo rates. Consistent with this prediction, we find that the average overnight special repo rate backed by 2-/5-/10-year Treasury notes during the 5 days before auctions of a 2-/5-/10-year notes is, on average, 25/23/32 basis points lower than that during the post-auction 5-day period. Second, auctions of Treasury securities with one maturity should also affect prices of other maturities, and the impact should be stronger if the maturity differential is smaller. Consistent with this prediction, we find a similar pattern in 10-year note returns around both 2-year and 5-year note auctions, even when there is no 10-year note auction in surrounding days. Moreover, relative to 2-year note auctions, 5-year note auctions have a much stronger price impact on 10-year notes. Our paper is related to the growing literature on the price impact of supply and demand shocks.4 The contribution of our paper is to provide clean evidence in arguably the most liquid financial markets where the relatively small supply shocks are both well 4 See, for example, Shleifer (1986), Kaul, Mehrotra, and Morck (2000), Wurgler and Zhuravskaya (2002), Newman and Rierson (2003), Mitchell, Pulvino, and Stafford (2004), Coval and Stafford (2007), Frazzini and Lamont (2008), and Lou (2012). 4 anticipated and frequently repeated. Our paper is closely related to the findings in Fleming and Rosenberg (2007) that primary dealers take short positions shortly before Treasury auctions, and that the newly issued securities tend to have high returns after auctions. Our paper is also related to studies that examine the permanent impact of bond supply on Treasury yields and spreads, and changes in yields around policy announcements.5 Our paper differs from these studies in that it examines the temporary price impact around auctions, as well as the spillover effects across maturities and across markets. Our findings are broadly consistent with Vayanos and Vila (2009), where local demand and supply can distort Treasury yields due to market segmentation. Our results are distinct from the on-the-run premium phenomenon. Motivated by Duffie (1996), several recent studies analyze the on-the-run premium and find that the specialness of an on-the-run security decreases when the next issuance approaches.6 While this mechanism is consistent with the price decrease of the on-the-run security before the next auction of the same maturity, it cannot explain the post-auction price increase, or the yield/return pattern for off-the-run securities and the spillover effects to other maturities and markets. 5 See, for example, Simon (1991), Duffee (1996), Krishnamurthy (2002), Kuttner (2006), Greenwood and Vayanos (2010), and Krishnamurthy and Vissing-Jorgensen (2010), Han, Longstaff and Merrill, (2007), Garbade and Rutherford (2007), Gagnon, Raskin, Remache and Sack (2010), D’Amico and King (2011), Krishnamurthy and Vissing-Jorgensen (2011), and Swanson (2011). 6 See, for example, Jordan and Jordan (1997), Krishnamurthy (2002), Meli (2002), and Goldreich, Hanke, and Nath (2005). 5 II. Data A. Institutional background of Treasury auctions In our sample period of 1980-2008, the U.S. Treasury Department auctioned, on average, $2.35 trillion worth of securities each year; the total amount auctioned was $6.7 trillion in 2008. The participants of these auctions include primary dealers, institutional investors, foreign central banks, the Federal Reserve, and a small number of individual investors. The Treasury Department publishes tentative auction schedules several months in advance and announces the auction size several days before each auction. Immediately after the announcements, dealers and investors start to trading forward contracts on the soon-to-be-issued Treasury security in the “when-issued” market. In each auction, primary dealers and other competitive bidders submit sealed bids of rate-quantity pairs that specify the amount to be purchased at each minimum yield. Two auction mechanisms have been employed in Treasury auctions: multiple-price and single-price auctions. Under both mechanisms, the clearing price is identified by equating the aggregate demand submitted by competitive bidders to the total issue amount minus the total demand from noncompetitive bidders (i.e., those who submit market orders). The difference between the two mechanisms lies in that, while in multiple-price auctions, competitive bidders pay for their allocated shares at their submitted prices, in single-price auctions, all winning bidders pay the same price. While almost all Treasury auctions in the 1980s were multiple-price auctions, the single-price mechanism is the dominant form in the more recent two decades.7 7 For a more detailed discussion of these two auction mechanisms see, e.g., Goldreich (2007). 6 B. Data sample From the U.S. Treasury Department website, we collect detailed information regarding auctions for Treasury notes, with maturities ranging from 2 to 10 years. Such information includes the auction date, issue date, auction mechanism, bids submitted, total tender amount received, total tender amount accepted, lowest and highest winning rates, etc. Our sample spans from January 1980 to June 2008, during which period 2- year Treasury notes are issued on a monthly basis and 10-year notes are issued on a quarterly basis and are frequently reopened in the following months. Our sample also includes these reopened issues. The issuing frequency of 5-year notes varied a few times in this period. In total, we have 332 2-year note auctions, 210 5-year note auctions, and 132 10-year note auctions.8 We then match our auction data with the CRSP daily U.S. Treasury database to obtain daily Treasury security prices and accrued interest, from which we compute Treasury yields and daily returns. Throughout our analysis, we use the average of the bid and ask prices reported by CRSP as our measure of the security price. The repo market in the U.S. is very large, with an estimated size of over $10 trillion (Gorton and Metrick (2010)). We obtain daily overnight special repo rates backed by 6-month T-bill, 2-, 5-, and 10-year Treasury notes from Datastream and GovPX. We also obtain daily general collateral (GC) repo rates (repo contracts backed by government general collaterals) and overnight Libor rates from Bloomberg, and AA- rated nonfinancial overnight commercial paper rates and Federal Funds rates from the website of the Federal Reserve Board. From Mueller, Vedolin, and Yen (2011), we 8 The Treasury also issued 3-, 4-, and 7-year notes during part of our sample period. Their issuance, however, was interrupted in the period of 1998-2003 for 3-year notes, 1990-2008 for 4-year notes, and 1993-2008 for 7-year notes. 7 obtain the daily model-free estimates of maturity-weighted implied volatility of nearest- to-expiry at-the-money options on the 30-year, 10-year, and 5-year Treasury securities. We obtain the CBOE Volatility Index, VIX, from CBOE website. Finally, following Adrian, Etula, and Muir (2012), we construct the quarterly aggregate leverage growth for broker-dealers from the Federal Reserve Flow of Funds data. Table I presents some summary statistics for our sample. Panel A shows that the average size of Treasury auctions is well over $10 billion. Panel B shows that the average daily Treasury note return ranges from 1.96 (for 2-year notes) to 2.31 (for 10- year notes) basis points. Panel C summarizes other main variables. The average option- implied Treasury return volatility is around 8%; the average VIX is about 20%; the average growth rate of the leverage ratio (asset/equity) of broker-dealers is roughly 11% in our sample. The average daily special repo rates backed by 6-month, 2-year, 5-year, and 10-year Treasury securities are 2.88, 2.65, 2.49, and 2.29 basis points, respectively. III. Price Impact on Secondary Markets To analyze the impact of Treasury auctions on the secondary Treasury market, we first examine yields-to-maturity of 2-year notes, the most frequently issued Treasury notes, around 2-year note auctions. For each 2-year note auction, we calculate daily yields to maturity of the on-the-run 2-year note during both the 10 days before and 10 days after the auction and compare them with the yield on the auction day. That is, we track the same 2-year note throughout the 20-day window around each auction, and this security is on-the-run before the auction and becomes off-the-run after the auction. Since there is a 2-year note auction almost every month in our sample, these 20-day event windows around auctions cover virtually all trading days in our sample period. 8