Liquidity Commonality in the Secondary Corporate Loan Market. Newcastle Business School, The University of Newcastle, Australia

Transcription

1 Liquidity Commonality in the Secondary Corporate Loan Market John Anthony *, Paul Docherty, Doowon Lee, Abul Shamsuddin Newcastle Business School, The University of Newcastle, Australia Abstract Despite the economic importance of the secondary corporate loan market, the extant literature has not examined the measurement or impact of liquidity in this market. Using principal component analysis and canonical correlations, we examine the performance of liquidity proxies and identify the bid-ask spread as a parsimonious measure of liquidity. Using this measure, liquidity commonality (a measure of liquidity risk) is shown to increase 18-fold during the financial crisis of from its pre-crisis level. A relatively large increase in liquidity commonality is consistent with the nature of the loan market, being a relatively illiquid over-the-counter market dominated by banks and diversified institutional investors. A doubling in the cross-sectional economic significance of credit quality during the crisis suggests that both supply and demand factors may explain this liquidity risk. Analyzing supply and demand drivers using novel trading liquidity data from Markit, we find that liquidity commonality is significantly related to banking sector funding liquidity. Keywords: Liquidity, liquidity commonality, bank funding liquidity, financial crisis, OTC markets, loan market JEL codes: C23, G01, G12, G14 * Corresponding Author. Department of Accounting and Finance, Newcastle Business School, University of Newcastle, University Drive, Callaghan, NSW Australia Tel: address: john.anthony@newcastle.edu.au 1

2 1. Introduction Recent empirical literature has explored liquidity commonality in various markets. For example Karolyi, Lee, and van Dijk (2012) in the equity market, Friewald, Jankowitsch, and Subrahmanyam (2012) in the bond market, and Corò, Dufour, and Varotto (2013) in the credit default swap market, and Marshall, Nguyen, and Visaltanachoti (2013) in commodities markets. Liquidity commonality in the secondary corporate loan market (herein the loan market ), indeed liquidity more broadly, has received very little attention, despite U.S. trading volumes reaching $628 billion in 2014 (Thomson Reuters LPC, 2015). 1 This may be due to data limitations, a common limitation in over-the-counter markets. We overcome this limitation using a novel trading liquidity dataset from Markit. 2 Trading liquidity data provides additional insight into the measurement and cross-sectional characteristics of the level of liquidity, and allows us to examine the drivers of liquidity commonality in the loan market. 3 The extent of commonality and the drivers of commonality in the loan market may be different to other markets given the relatively important role played by banks. The accurate measurement of liquidity is important to market participants that require liquidity data for mark-to-market pricing or investment decisions. Liquidity encompasses search and bargaining costs, that is, the ability to find and negotiate with willing buyers in a timely manner, as well as transaction costs. These costs may be more pronounced in overthe-counter markets (Duffie, Gârleanu, & Pedersen, 2005). Examining several popular measures used in the recent literature to evaluate liquidity in other asset classes, together with trading liquidity measures from Markit, we find that the bid-ask spread parsimoniously measures the level of liquidity in the loan market. 1 Trading volume data is obtained from surveys of 11 of the largest sell-side institutions and 8 of the largest buy-side institutions. 2 Markit data is used extensively in the extant literature. For example Pu, Wang, and Wu (2011) and Fan (2006) analyse credit default swap data from Markit and Jankowitsch, Nashikkar, and Subrahmanyam (2011) use bond valuation data from Markit to calculate price dispersion. 3 The loan market is sometimes referred to as the leveraged loan market, given the preponderance of low grade leveraged credits. Refer to Ehsani and Beyhaghi (2015) for a discussion on the unique properties of loans, drivers of loan market growth and comparisons with other debt markets. 2

3 Once the most effective measures of liquidity are identified, these measures can be used to test for the presence and drivers of liquidity risk in the loan market. Liquidity risk is the risk that liquidity will decrease precisely at the time it is most needed (Beber, Brandt, & Kavajecz, 2009). Vayanos (2004) models a flight to liquidity during more volatile periods. Increased risk aversion leads investors towards more liquid assets. Such a flight to liquidity should be reflected in higher risk premiums, that is, an increase in liquidity risk during volatile periods. Such a flight to liquidity is observed by Longstaff (2004) in the treasury bond market and Beber et al. (2009) in the European sovereign debt market. Lin, Wang, and Wu (2011) find that the average return on bonds with high sensitivities to market liquidity is 4% greater than those with low sensitivity. 4 To date, flights to liquidity have not been examined in the context of loan markets. The apparent significance of liquidity risk makes the topic of interest to investors, policy makers, regulators and exchanges (Coughenour & Saad, 2004). One channel through which liquidity risk affects prices is liquidity commonality, the relationship between individual loan liquidity and broader market liquidity. 5 The extant literature considers the extent of liquidity commonality in various markets, particularly during crisis periods. Friewald et al. (2012) and Karolyi et al. (2012) present evidence of increases in liquidity commonality during the Global Financial Crisis ( GFC ) in the bond market and equity market respectively. We investigate the time-varying properties of liquidity commonality in the loan market, paying particularly attention to the impact of the GFC. We also contribute to the debate as to the relative importance of supply and demand factors. 6 4 There is debate as to the relative importance of liquidity risk and credit risk (Beber et al., 2009). 5 Acharya and Pedersen (2005) model three different channels through which liquidity affects prices, one of which is liquidity commonality. 6 The loan market is subject to extensive analysis in the corporate finance and banking literature. For example, Altman, Gande, and Saunders (2010) find that bank loans Granger cause bond returns prior to a loan default, implying that bank loans have an information advantage over bonds. Bushman, Smith, and Wittenberg- Moerman (2010) provide evidence that institutional investors exploit information obtained from loan investments for equity trading. Drucker and Puri (2009) investigate differences between loans that are traded in 3

4 Various theoretical models posit a relationship between supply-side factors and liquidity commonality. Brunnermeier and Pedersen (2009) link liquidity commonality with variations in funding liquidity, that is, the ability of traders to fund positions in the market. Traders could be intermediaries or fund managers with leveraged positions. Traders avoid taking positions when funding liquidity is tight, leading to a mutually reinforcing process or liquidity spiral of funding liquidity, market liquidity and price declines. Increased liquidity commonality during a crisis is evidence of a liquidity spiral, as a crisis will simultaneously affect numerous loans and many intermediaries. In Kyle and Xiong (2001) financial intermediaries are convergence traders that benefit from temporary movements away from fundamental value. In a crisis period trading losses reduce the ability of intermediaries to hold inventory, reducing liquidity across markets where they operate. As intermediaries in loan markets are typically banks, trading losses may be particularly relevant during the GFC, as this period was characterized by banking sector weakness. Further, the major investors in the loan market, such as collateralized loan obligation funds (CLOs) 7, are leveraged, and therefore face funding risk. Thus there is substantial theoretical basis for a strong link between the funding liquidity of banks and liquidity risk. Indeed empirical evidence from the equity markets indicates that declines in liquidity commonality are associated with large market declines, which are better explained by supply-side factors (Hameed, Kang, & Viswanathan, 2010). While the theory is less developed, there are demand-side explanations for liquidity commonality, in particular due to the trading activity of institutional investors. During volatile periods fund managers are forced to liquidate when performance declines (Vayanos, 2004). 8 To the extent that these investors are diversified, this results in correlated trading. In the secondary loan market and untraded loans. They find that traded loans have more restrictive covenant structures and that borrowers accept these structures to gain greater access to credit. Other studies consider the impact of secondary loan markets on relationship banking, balance sheet management and banking sector risk (Parlour & Plantin, 2008). In perhaps the first asset pricing study of the loan market, Ehsani and Beyhaghi (2015) find that cross-sectional differences in expected returns are explained by momentum. 7 CLOs are the dominant investor class in the loan market (Culp, 2013). 8 Favero, Pagano, and Ernst-Ludwig von (2010) present a model where investors are less concerned about liquidity when market volatility is high because of a decline in alternative investment opportunities. 4

5 the loan market the situation may be exacerbated by a lack of idiosyncratic liquidity information, leading to the sale of both liquid and illiquid loans. Indeed institutional investors with flexible mandates may avoid illiquid asset classes completely. 9 Empirical evidence by Koch, Ruenzi, and Starks (2010) and Karolyi et al. (2012) suggests demand-side factors are significant. Koch et al. (2010) show that the co-movement of stocks with high mutual fund ownership is twice that of low mutual fund ownership stocks, and turnover is related to both the liquidity position of mutual funds and liquidity commonality. Karolyi et al. (2012) also find that correlated trading by institutional investors as well as investor sentiment, another demand-related factor, explain a greater proportion of commonality compared to supply-side drivers in various global equity markets. Supply and demand drivers of liquidity commonality may have differing impacts in different markets. For example, Qian, Tam, and Zhang (2014) argue that supply-side factors are more relevant to the Chinese stock market due to capital controls and government interference. Similarly supply-side factors may be relatively important in the loan market which, despite the broad range of investors, remains heavily influenced by banks. 10 Not only are banks the dominant market-maker, 11 at times the secondary loan market represents an important outlet for loans originated by banks in the primary market (Culp, 2013). 12 When secondary loan prices trade below the prices held on bank balance sheets (effectively representing trading losses), the ability of banks to provide liquidity on other positions may reduce (Kyle & Xiong, 2001). Further, loans that trade in the loan market are at the riskier end of bank portfolios (Gatev & Strahan, 2009), and are therefore likely to have relatively high funding liquidity requirements. On the demand side, banks remain a significant investor class (Wells Fargo, 2014), and broad-based portfolio sales by banks may impact liquidity commonality. 9 Despite increases in trading volumes in the loan market, volumes remain substantially below those in the corporate bond market, which itself is considered a relatively illiquid market (Edwards, Harris, & Piwowar, 2007). 10 The modern loan market comprises a broad range investors including banks, non-bank financial institutions, securitization vehicles, hedge funds and mutual funds (Ivashina & Scharfstein, 2010). 11 According to Markit (2014) of the 63 loan market trading desks that provide activity data, 51 are banks. 12 During the GFC banks that were unable to distribute loans in the primary market broke with their syndicates and sold directly into the secondary market (Culp, 2013). 5

6 Banks are also highly leveraged investors, and may become forced sellers during a crisis (Diamond & Rajan, 2011). The risk management practices of banks may also have a more acute impact in over-the-counter markets. For example, value at risk calculations may reflect longer lead times on asset sales, exacerbating any liquidity spiral (Gârleanu & Pedersen, 2007). Thus reductions in bank funding liquidity may have a relatively large influence on the loan market. This paper presents evidence of increases in liquidity commonality in the loan market during the GFC, increases that appear to surpass those experienced in equity and bond markets. Consistent with evidence from other markets, lower grade loans have lower liquidity. 13 During the GFC the economic significance of credit quality doubles. This may be supplydriven, as market-makers avoid loans with higher funding liquidity requirements, or demand driven, as investors avoid high default risk assets during an economic downturn. We present evidence that, after controlling for market conditions, banking sector funding liquidity is a significant driver of liquidity commonality, evidence of a supply-driven liquidity spiral as in Brunnermeier and Pedersen (2009). The remainder of the paper is structured as follows. In Section 2, we describe the data that is available from Markit. In Section 3, we explore the measurement of liquidity in the loan market. In Section 4, we present cross-sectional differences in the level of liquidity in and out of the GFC. In Sections 5 and 6, we investigate the extent of liquidity commonality and its drivers. A brief conclusion follows. 2. Data Description and Summary Markit collects and validates intraday loan pricing data from 40 global loan trading institutions. The Thomson Reuters LPC database used in extant studies of the loan market includes information on bid-ask quotes and the number of quoting broker-dealers ( Depth ). From March 2008, Markit began publishing additional information relating to trading liquidity including the number of broker-dealers that move their daily quote ( Movers ), 13 Less liquid bonds have higher spreads and lower credit quality (Chen, Lesmond, & Wei, 2007), and less liquid stocks have higher returns (Pastor & Stambaugh, 2003). 6

7 total daily quotes ( Quotes Count ) and total daily Quoted Depth, which is the number of firm quotes multiplied by average quote size. Firm quotes are volume specific and are intended to be executed, providing increased insight into trading volumes and prices. Firm quotes typically represent 20% of all quotes. The sample begins on 28 September 2001, the earliest date that Markit makes loan market data available, and ends on 28 October The dataset includes 25,485,145 daily bid-ask quotes relating to 27,702 loans globally. We focus on U.S. issuers, reducing the sample by 40.3% to 16,534 loans. Potential errors are screened by eliminating any loans that include a daily pricing movement in excess of 50%, or that trade below 10% or above 150% of par. At least 20 days of data is required for each loan. The screens reduce the sample by 9.6% to 14,952 loans with 12,838,083 quotes. To avoid survivorship bias we include all loans that mature within the sample period. Moreover loans typically have relatively short tenors, and restricting the sample to long-life loans would severely restrict the sample size or require a reduction in the sample period. Markit provides loan characteristics such as age, size, issue date, maturity date and initial spread. Markit also maps Moody s credit rating data to individual loans from January 2007, giving us 3,717 rated loans. Thus for analysis that relies on ratings data we restrict the sample to the period from January 2007 through October Similarly, where we use trading liquidity data, the sample is restricted to the period from March 2008 through October For the purpose of analysis we split the sample into three sub periods. The GFC is defined as the 24-months from July 2007 through June The 24-month period before and after the GFC is the pre-gfc and post-gfc period respectively. During the GFC monthly average volatility, measured by MOVE, is more than one standard deviation above the historic mean in 16 of 24 months. Average monthly volatility in the pre-gfc period is more than one standard deviation below the mean in 19 of 24 months. The post GFC period is characterized by more normal levels of volatility; just two months (July and August 2009) are more than one standard deviation above the mean, and average volatility (98.0) is consistent with the 7

8 historic mean (98.9). 14 This definition of the GFC period is also consistent with recent bond market analysis by Dick-Nielsen, Feldhütter, and Lando (2012). [Insert Table 1] Annual summary statistics are shown in Table 1, and aggregated and distributional data for all years is shown in Table 2. Loans have relatively short tenors (average 5.6 years) and relatively short lives (average 1.5 years). According to practitioners this reflects a tendency for loans to be refinanced into longer tenor bonds. 15 Average loan age reaches a high point of 2.5 years at the end of the GFC, and a low point of 1.1 years just prior to the crisis, seemingly reflective of refinancing conditions. Average prices in the loan market are low during the GFC but are close to par in more normal times. 16 The average loan size is $392 million; although there is substantial right-tail skew reflecting the presence of several very large loans, such as the $16 billion loan to TXU in April The mean initial coupon over LIBOR is 360 basis points with positive skew, perhaps reflecting higher risk premiums during the GFC. Ratings are tightly clustered at Ba3/B1, reflecting the relatively low credit quality of loans that trade in the secondary loan market. Investment grade loans rarely trade in the secondary loan market. In contrast the secondary corporate bond market is dominated by investment grade debt. For example just 11% of corporate bonds analyzed by Chen et al. (2007) are below investment grade. The significant difference in the characteristics of loans and corporate bonds warrant an investigation of the most effective measure of liquidity in the loan market and the role of liquidity in this market. [Insert Table 2] 14 We measure historic means using the data period available from Datastream, being April 4, 1988 through March 27, Bonds tend to stay outstanding for longer. Bao, Pang and Wang (2011) report that the mean age of corporate bonds in their sample ranges from 2.73 years in 2003 to 7.23 years in Loans are quasi fixed and floating rate instruments. The floating portion is typically the London Interbank Offer Rate ( LIBOR ) and the fixed portion is the initial spread. Loans are typically callable and therefore rarely trade above par. 8

9 3. Liquidity Measures There is no consensus on the most appropriate measure of liquidity (Dick-Nielsen et al., 2012). We compare the effectiveness of several proxies for liquidity in the loan market, including: the bid-ask spread ( Bid-Ask ); a measure similar to the negative covariance measure of Bao, Pan, and Wang (2011) which we label γ; a Price Dispersion measure similar to Jankowitsch et al. (2011) and Houweling, Mentink, and Vorst (2005); a measure similar to Percentage Zeros formulated using firm quote data ( Zero Trading ). Other than the trading liquidity variables and Zero Trading, all the liquidity measures can be calculated for the entire sample period. 17 Daily measures are converted to monthly averages to facilitate comparisons with monthly measures. Following is additional detail on liquidity measures. Bid-Ask Bid-Ask is perhaps the most common measure of liquidity in the literature (Chen et al., 2007), and is the most effective measure of liquidity in certain classes of assets, such as Treasury Bonds (Fleming, 2003). Chen et al. (2007) use the proportional bid-ask spread to analyze bond market liquidity, that is, the bid-ask spread relative to its price. We adopt a similar approach, dividing the bid-ask spread by the mid-quote. Zero Trading The Percentage Zeros measure was first proposed by Lesmond, Ogden, and Trzcinka (1999), who posit that a loan with low liquidity is less likely to experience movements in returns. Percentage Zeros is the percentage of days in a month without a price movement. Kang and Zhang (2014) use a similar Zero Volume measure relating to zero trading volumes in emerging markets, noting a high correlation between Percentage Zeros and Zero Volume. Zero Trading is the percentage of monthly trading days without a firm quote. Adjusted Serial Covariance (γ) 17 To facilitate comparisons with measures that use firm quote data, when comparing alternative measures we restrict our analysis to the period from March

10 Short-term deviations from fundamental value may represent market frictions such as search and bargaining costs (Bao et al., 2011). Motivated by Bao et al. (2011) and Roll (1984) we take the square root of the negative covariance,,,,, (1) where, is liquidity in month m for loan i, and,,, is the serial covariance in daily average mid-quote changes. Following Dick-Nielsen et al. (2012) we discard positive covariance results as the calculation is undefined. We calculate, for each month and formulate a market-weighted average. At least 15 observations are required. Market values are estimated by the mid-quote multiplied by issuance. 18 Dispersion Jankowitsch et al. (2011) calculate dispersion as the volume-weighted difference between individual broker-dealer quotes and market consensus valuations provided by Markit. Less liquid loans will display greater dispersion. Houweling et al. (2005) measure liquidity as the standard deviation of quoted yield differences amongst broker-dealers relative to their mean. We formulate a similar measure to Houweling et al. (2005) replacing yield with offer quote. 19 Price dispersion for loan l on day t is, Dispersion, 1, 1,,,, (2) where, is the number of quotes for loan l on day t,,, is the offer quote from brokerdealer s and, is the average offer quote across all broker-dealers. To calculate dispersion we require at least two quotes on any day. To the extent that loans with a single quote are more likely to be illiquid we may understate market illiquidity. We may overstate illiquidity to the extent that intraday quotes are nonsynchronous. 18 We ignore amortization thus our market values should be considered approximations. Loans typically amortize at 1% per annum (Ehsani & Beyhaghi, 2015) so results should be similar. 19 Yield data adjusts for amortization however yield data is only available on an aggregated daily basis. Given that spreads will move inversely with prices the results should be almost identical. 10

11 [Insert Table 3] Annual market-weighted liquidity measures are shown in Table 3, together with pair-wise correlations between each measure. Also included is volatility, measured as the standard deviation of the mid-quote. Bid-Ask, Gamma and Dispersion show a pattern of declining liquidity during the crisis and subsequent recovery. Other than Quotes Count, each measure is positively correlated with volatility. This may reflect the impact of volatility on the funding liquidity of intermediaries as in Brunnermeier and Pedersen (2009), or the impact of volatility on investor activity. Trading liquidity increases in the latter stages of the GFC and continues to increase beyond the GFC, only decreasing in the latter part of the sample. Interestingly, twice annualized γ (136 basis points), equivalent to Roll s (1984) effective bidask measure, closely approximates the average bid-ask spread over the sample period (129 basis points). 20 To understand whether the liquidity measures are driven by one or more common factors, we extract components using the principal component analysis method (see Table 4). We restrict the analysis to two components as the eigenvalue of the third principal component is less than one. The sign of each correlation coefficient is as expected, with the exception of Movers. 21 Thus we exclude Movers from our analysis, although we note that results are similar when this measure is included. We analyze changes in each liquidity measure because each measure is non-stationary. The first principal component explains 54.4% of variance and loads heavily on several measures, providing evidence that several measures address a similar facet of liquidity. A varimax (orthogonal) rotation produces a component 20 Roll (1984) used actual transaction prices. γ will estimate Roll (1984) exactly to the extent that mid-quotes reflect actual traded prices. 21 For example, if there is an increase in the number of broker-dealers that move their quote, the liquidity of a loan should increase. However an increase in Movers is associated with a decrease in liquidity for all other measures. 11

12 that loads heavily on Bid-Ask and Dispersion. 22 However Dispersion is only calculated when at least two quotes exist, limiting its usefulness to situations where two or more quotes exist. We cannot ignore the second component, which explains 21.1% of variance. The varimax rotation indicates that the second component is a Quotes Count and a Quoted Depth measure. The difficultly with trading liquidity measures is separating information and liquidity components. For example, traded volume may increase due to information shocks (Darolles, Fol, & Mero, 2015). To further explore the relationship between variables we conduct a canonical correlation analysis, splitting Bid-Ask, Dispersion and γ from the variables more closely associated with trading liquidity. The analysis indicates that Bid-Ask and Dispersion explain a substantial portion of the common variance amongst the trading liquidity variables (loading and respectively on the first canonical variable). 23 This result complements the principal component analysis results. We conclude that Bid-Ask parsimoniously captures the level of liquidity in the loan market. [Insert Table 4] 4. Cross Sectional Differences in Liquidity In the corporate bond market, credit quality is positively correlated with the level of liquidity (Ericsson & Renault, 2006). Larger bonds also tend to be more liquid and liquidity declines as bonds age (Bao et al., 2011). Higher liquidity on short-dated securities may be due to deliberate market-making activity by broker-dealers or gradual absorption into buy-and-hold portfolios (Houweling et al., 2005). We test for cross-sectional differences in the level of liquidity using the following equation, 22 The varimax rotation provides for uncorrelated components relating to the fewest number of liquidity measures. 23 Results are available from the corresponding author on request. 12

13 ,.,.,., (3) Where, is Bid-Ask for loan i in month t, Size is the log of the amount issued, is loan age, and is the Moody s credit rating. Average coefficients and t- statistics from monthly cross-sectional OLS regressions across the sample period are presented in Table 5. [Insert Table 5] All three characteristics explain cross-sectional differences in liquidity. Consistent with corporate bond markets, lower grade loans have lower liquidity. 24 The economic significance of credit quality doubles during the GFC, while Age becomes insignificant. 25 Given that the annual distribution amongst different rating categories is stable, any temporal changes in cross-sectional differences in liquidity are unlikely due to temporal changes in the crosssectional distribution of credit quality. 26 [Insert Table 6] 5. Commonality in Liquidity During the GFC average spreads in the loan market reached 1699 basis points, compared to 244 points on average during 2006 (Standard & Poor's, 2013). By the fourth quarter of 2008 there was a 79% reduction in primary market lending compared to the second quarter of 24 For example, in the bond market, Chen et al. (2007) and Ericsson and Renault (2006). In the primary loan market Gupta, Singh, and Zebedee (2008) find that lower grade loans have higher liquidity, however their analysis considers primary loans that never trade in the secondary market. 25 The decline in significance of Age during the GFC may be due to the lack of primary market activity, as broker-dealers refrained from making secondary markets in newly issued loans. 26 Refer to the Appendix for a breakdown of ratings per category each year. A chi-square test of monthly ratings for each category rejects the null that the proportions in each category are not constant (test statistic ). 13

14 2007 (Ivashina & Scharfstein, 2010). An important question is the role that liquidity risk played during the crisis. Several studies beginning with Chordia, Roll, and Subrahmanyam (2000) and later Karolyi et al. (2012) and Rösch and Kaserer (2013), among others, document the presence of liquidity commonality in stock markets, and show that liquidity commonality is increasing in volatility. In the bond market Bao et al. (2011) point to an increase in the level of market illiquidity by 12 standard deviations as evidence of commonality during the crisis. To investigate liquidity commonality in the loan market we follow a similar method to Chordia et al. (2000). We consider changes rather than the level of liquidity, as we are also primarily interested in co-movement and levels are also nonstationary. Liquidity commonality is modelled as,,., (5) Where is the monthly change in Bid-Ask for each loan i in month t, is the monthly change in sample-wide average Bid-Ask weighted by market value, and β is the liquidity beta. [Insert Table 6] Liquidity commonality in the loan market is born out of the crisis (see Table 6). For the market weighted portfolio, average adjusted R 2 is 0.02 in the 24 months immediately preceding the crisis, increasing to 0.36 during the crisis. The percentage of loans with significant liquidity betas increases from 11% to 70% in the GFC period. 27 Significance levels and R 2 remain slightly elevated in the post GFC period, perhaps due to ongoing high volatility in the initial months of the post-gfc period or persistence in liquidity We use a one-tail test as, by definition, commonality is a one-way relationship between individual loan liquidity and market liquidity. 28 During the crisis lower grade loans have higher liquidity risk (p-value <0.0001), however during the pre- and post-gfc periods there is no significant correlation between loan quality and liquidity risk at the 5% level. However we cautiously emphasize these results as the commonality sample is restricted to those loans with at least 20 months of data. This increases the concentration of observations around the mean rating. Nevertheless, this result is consistent with evidence from bond markets. During crisis periods when liquidity is low, investors 14

15 The contrast between crisis and non-crisis periods appears more pronounced in the loan market. Karolyi et al. (2012) report equity market R 2 in the 20-25% range in normal times, increasing to 25-40% during the GFC. 29 In the German equity market Rösch and Kaserer (2013) find that R 2 increases by 5.4 times during the GFC, compared to 18 times in the loan market, and average equal-weighted liquidity beta increases by 5 times, compared to 9 times in the loan market. There are several potential explanations for a relatively large increase in liquidity risk during the GFC. First, the loan market may be disproportionately affected by a banking crisis. Banking sector weakness is reflected in increased funding costs, which affects all loans, or loan sales by distressed banks as in Diamond and Rajan (2011). Second, the loan market is substantially more opaque and less liquid than the equity market, or for that matter the bond market. 30 All trading is over-the-counter, with no record of transaction prices or loan-specific trading volumes. The lack of liquidity is further evidenced by the number of active brokerdealers; less than one-third of loans in the sample are quoted by more than two dealers. Indeed Markit was responding to demand for more liquidity information when it launched its trading liquidity data in March Faced with a lack of liquidity information, investors may seek a broad sale of assets, even if their preference is to sell illiquid assets. In markets where liquidity frictions are severe, banks may have no incentive to accumulate inventories during a crisis (Lagos, Rocheteau, & Weill, 2011). Thirdly, diversified institutional investors such as collateralized loan obligation funds dominate the loan market investor base, and substantial liquidity requirements are likely to generate sales pressure across a broad range of loans. 31 Finally, Vayanos and Wang (2012) present a model wherein liquidity is increasing in information asymmetry. This may be significant in the loan market, given that issuers are prefer more liquid and less risky sovereign bonds (Beber et al., 2009), suggesting that lower quality bonds have higher liquidity risk. This is confirmed by Lin et al. (2011) in the corporate bond market. 29 R 2 becomes more volatile during the crisis, including one month above 40% (Karolyi et al., 2012). 30 Most trades in the bond market are also over-the-counter (Biais & Green, 2007), although transaction costs decreased following the introduction of the TRACE price reporting system (Edwards et al., 2007). 31 In 2013, 47% of investors were collateralized loan obligation funds, 26% were prime funds, 15% were banks, 4% were insurance companies, and 8% were other investor such as hedge funds, distressed funds or high-yield funds (Wells Fargo, 2014). 15

16 typically private companies. For example, investors do not have the benefit of issuer-specific equity information, an important input into structural models of default prediction. We investigate these demand and supply drivers of liquidity commonality in the following section. 6. Commonality Drivers Given evidence of liquidity commonality, we examine the drivers of commonality. Supplyside explanations focus on the funding liquidity of intermediaries, whereas demand-driven explanations focus on movements in investor funds. Supply-driven explanations imply that investors should maintain liquidity buffers and that central banks can address a liquidity crisis by boosting funding liquidity (Brunnermeier & Pedersen, 2009). However demandside explanations suggest a focus on investor property rights and transparency, or other measures aimed at avoiding investor panic (Karolyi et al., 2012). Following Karolyi et al. (2012) we measure monthly commonality as R 2 from the intramonth daily movements in commonality,,, (6).,,.,,.,,, Where,, is the daily change in Bid-Ask for loan i in month t,,, and,, are lead and lag variables respectively. Following Chordia et al. (2000) we include lead and lag variables to account for differing rates of adjustment to market movements. Following Karolyi et al. (2012) we take the log transformation ln, / 1, so that our R 2 is not bounded by 0 and 1, where, is the from the regression of loan i in month t. Most demand-driven explanations focus on correlated trading by diversified institutional investors. Karolyi et al. (2012) use commonality in turnover to measure correlated trading activity. The dominance of diversified institutional investors such as CLOs makes common movements in trading activity a suitable measure of demand-driven liquidity commonality in 16

17 the loan market. We proxy for trading activity using commonality in Quoted Depth. Demand and supply measures are regressed on liquidity commonality, _. _. (7). _ Where _ is the equal-weighted average for each loan from Equation 6 in month t and _ is the equal-weighted average from the commonality in Quoted Depth regression. To control for market conditions we include, the change in market-weighted average loan prices and, the change in market liquidity. The change in funding liquidity is measured using either (i) TED, the change in the difference between three month LIBOR, the rate at which banks are willing to lend to each other, and three month Treasury yields, or (ii) LIBOR_OIS, the change in the difference between three month LIBOR and the three month overnight index swap. Cornett, McNutt, Strahan, and Tehranian (2011) show that TED is closely related to banking sector liquidity. To control for general funding liquidity conditions we include _, the change in the spread between the three month overnight index swap and three month Treasury yields. Because the overnight interest rate swap is unaffected by counter-party risk but has, in theory, the same default risk, and because three-month Treasury bills are one of the most liquid instruments in the market, this represents a reasonable proxy for general movements in funding liquidity. 32 LIBOR_OIS is our preferred measure of funding liquidity as it is uncorrelated with _. 33 The Brunnermeier and Pedersen (2009) model posits high correlation between market returns, market liquidity and funding liquidity. Indeed we encounter collinearity issues, in particular due to the high correlation between liquidity commonality and market prices. 34 We avoid these collinearity issues by making market 32 Krishnamurthy (2010) analyses the flight to liquidity during the GFC using the spread between the overnight index swap and Treasury yields. The floating rate leg of a three month overnight index swap is based on the average overnight federal funds rate over the next three months (Krishnamurthy, 2010). 33 We include TED for comparison purposes however we note some minor collinearity with _ (correlation 0.32). 34 Similarly, volatility has high correlation with market price (coefficient 0.93) and was excluded due to collinearity issues (variance inflation factors in excess of 10). 17

18 returns orthogonal to market liquidity and funding liquidity. At least ten intra-month Quoted Depth observations are required to form a meaningful R 2. The sample includes 2,950 loans and 29,741 months of data. To avoid bias the same sample is used for both liquidity commonality and the demand-side driver. A sample restricted to firm quotes is also more likely to reflect actual trading activity. [Insert Table 7] Partial and full regressions of Equation 7 are presented in Table 7, together with pairwise time-series correlations. High correlations between the variables suggest that commonalities are driven by common factors. Market condition variables are significant at the 1% level, while funding liquidity is significant at the 5% level. 35 The standardized coefficient on each of the market condition variables is more than twice that of either of the banking sector funding liquidity measures, however funding liquidity may be a critical link in the mutually reinforcing feedback loop described by Brunnermeier and Pedersen (2009). Changes in general supply conditions, as measured by the spread between the overnight index swap and Treasury yields, do not have a significant influence on liquidity commonality. The demandside driver is also not significant; the explanations explored by Karolyi et al. (2012) relating to investor property rights and transparency in various global markets are perhaps less significant in the U.S. loan market. For example, while loans issuers are typically private, loan agreements typically require the disclosure of quarterly financial results and covenant compliance certificates (Standard and Poor's, 2012). 7. Conclusion This paper investigates liquidity in the economically significant secondary corporate loan market. Using principal component analysis and canonical correlations to analyze several 35 We conservatively use a two-tail test as the direction of the relationship is not certain. For example Deuskar, Gupta, and Subrahmanyam (2011) find that less liquid options trade at higher prices in the options market. 18

19 conventional measures of liquidity, together with novel trading liquidity data from Markit, we find that the bid-ask spread parsimoniously measures the level of liquidity. Credit quality is a key cross-sectional characteristic of liquidity, particularly during crisis periods. Liquidity commonality increases 18-fold during the global financial crisis of compared to the pre-crisis period. In keeping with Brunnermeier and Pedersen (2009) and the heavy influence of banks on the loan market, changes in market liquidity, market pricing and banking sector funding liquidity determine commonality. This is consistent with most evidence from other U.S. capital markets. While the results indicate that funding liquidity is an important driver of liquidity commonality, the GFC was both a banking crisis and a recession. Demand-side influences may be more important in crisis periods not associated with banking sector weakness. Nevertheless, measures aimed at promoting funding liquidity in the banking sector may reduce the severity of loan market declines. 19

20 References Acharya, V. V., & Pedersen, L. H. (2005). Asset pricing with liquidity risk. Journal of Financial Economics, 77(2), doi: Altman, E., Gande, A., & Saunders, A. (2010). Bank Debt versus Bond Debt: Evidence from Secondary Market Prices. Journal of Money, Credit, and Banking, 42(4), doi: Bao, J., Pan, J. U. N., & Wang, J. (2011). The Illiquidity of Corporate Bonds. Journal of Finance, 66(3), doi: /j x Beber, A., Brandt, M. W., & Kavajecz, K. A. (2009). Flight-to-Quality or Flight-to- Liquidity? Evidence from the Euro-Area Bond Market. The Review of Financial Studies, 22(3), doi: / Biais, B., & Green, R. (2007). The Microstructure of the Bond Market in the 20th Century. Carnegie Mellon University. Retrieved from Brunnermeier, M. K., & Pedersen, L. H. (2009). Market Liquidity and Funding Liquidity. The Review of Financial Studies, 22(6), doi: / Bushman, R. M., Smith, A. J., & Wittenberg-Moerman, R. (2010). Price Discovery and Dissemination of Private Information by Loan Syndicate Participants. Journal of Accounting Research, 48(5), doi: Chen, L., Lesmond, D. A., & Wei, J. (2007). Corporate Yield Spreads and Bond Liquidity. The Journal of Finance, 62(1), doi: / Chordia, T., Roll, R., & Subrahmanyam, A. (2000). Commonality in liquidity. Journal of Financial Economics, 56(1), doi: Cornett, M. M., McNutt, J. J., Strahan, P. E., & Tehranian, H. (2011). Liquidity risk management and credit supply in the financial crisis. Journal of Financial Economics, 101(2), doi: Corò, F., Dufour, A., & Varotto, S. (2013). Credit and liquidity components of corporate CDS spreads. Journal of Banking & Finance, 37(12), doi: Coughenour, J. F., & Saad, M. M. (2004). Common market makers and commonality in liquidity. Journal of Financial Economics, 73(1), doi: Culp, C. L. (2013). Syndicated Leveraged Loans During and After the Crisis and the Role of the Shadow Banking System. Journal of Applied Corporate Finance, 25(2), doi: /jacf Darolles, S., Fol, G. L., & Mero, G. (2015). Measuring the liquidity part of volume. Journal of Banking & Finance, 50(0), doi: Deuskar, P., Gupta, A., & Subrahmanyam, M. G. (2011). Liquidity effect in OTC options markets: Premium or discount? Journal of Financial Markets, 14(1), doi: 20

21 Diamond, D. W., & Rajan, R. G. (2011). Fear of Fire Sales, Illiquidity Seeking, and Credit Freezes. Quarterly Journal of Economics, 126(2), doi: Dick-Nielsen, J., Feldhütter, P., & Lando, D. (2012). Corporate bond liquidity before and after the onset of the subprime crisis. Journal of Financial Economics, 103(3), doi: Drucker, S., & Puri, M. (2009). On Loan Sales, Loan Contracting, and Lending Relationships. Review of Financial Studies, 22(7), doi: Duffie, D., Gârleanu, N., & Pedersen, L. H. (2005). Over-the-Counter Markets. Econometrica, 73(6), doi: / Edwards, A. K., Harris, L. E., & Piwowar, M. S. (2007). Corporate Bond Market Transaction Costs and Transparency. The Journal of Finance, 62(3), doi: / Ehsani, S., & Beyhaghi, M. (2015). The Cross-Section of Expected Returns in the Secondary Corporate Loan Market. Paper presented at the 2015 Society for Financial Studies Finance Cavalcade. Expected.pdf Ericsson, J., & Renault, O. (2006). Liquidity and Credit Risk. Journal of Finance, 61(5), doi: /j x Fan, Y. (2006). How Profitable Is Capital Structure Arbitrage? Financial Analysts Journal, 62(5), Favero, C., Pagano, M., & Ernst-Ludwig von, T. (2010). How Does Liquidity Affect Government Bond Yields? The Journal of Financial and Quantitative Analysis, 45(1), doi: / Fleming, M. J. (2003). Measuring Treasury Market Liquidity. Federal Reserve Bank of New York Economic Policy Review, 9(3), doi: Friewald, N., Jankowitsch, R., & Subrahmanyam, M. G. (2012). Illiquidity or credit deterioration: A study of liquidity in the US corporate bond market during financial crises. Journal of Financial Economics, 105(1), doi: Gârleanu, N., & Pedersen, L. H. (2007). Liquidity and Risk Management. The American Economic Review, 97(2), doi: / Gatev, E., & Strahan, P. E. (2009). Liquidity risk and syndicate structure. Journal of Financial Economics, 93(3), doi: Gupta, A., Singh, A. K., & Zebedee, A. A. (2008). Liquidity in the pricing of syndicated loans. Journal of Financial Markets, 11(4), doi: Hameed, A., Kang, W., & Viswanathan, S. (2010). Stock Market Declines and Liquidity. The Journal of Finance, 65(1), doi: / Houweling, P., Mentink, A., & Vorst, T. (2005). Comparing possible proxies of corporate bond liquidity. Journal of Banking & Finance, 29(6), doi: 21

22 Ivashina, V., & Scharfstein, D. (2010). Bank lending during the financial crisis of Journal of Financial Economics, 97(3), doi: Jankowitsch, R., Nashikkar, A., & Subrahmanyam, M. G. (2011). Price dispersion in OTC markets: A new measure of liquidity. Journal of Banking & Finance, 35(2), doi: Kang, W., & Zhang, H. (2014). Measuring liquidity in emerging markets. Pacific-Basin Finance Journal, 27(0), doi: Karolyi, G. A., Lee, K.-H., & van Dijk, M. A. (2012). Understanding commonality in liquidity around the world. Journal of Financial Economics, 105(1), doi: Koch, A., Ruenzi, S., & Starks, L. (2010). Commonality in Liquidity: A Demand-Side Explanation. Unpublished working paper, University of Texas at Austin and university of Mannheim. Krishnamurthy, A. (2010). How Debt Markets Have Malfunctioned in the Crisis. Journal of Economic Perspectives, 24(1), doi: Kyle, A. S., & Xiong, W. (2001). Contagion as a Wealth Effect. The Journal of Finance, 56(4), doi: / Lagos, R., Rocheteau, G., & Weill, P.-O. (2011). Crises and liquidity in over-the-counter markets. Journal of Economic Theory, 146(6), doi: Lesmond, D. A., Ogden, J. P., & Trzcinka, C. A. (1999). A New Estimate of Transaction Costs. The Review of Financial Studies, 12(5), doi: / Lin, H., Wang, J., & Wu, C. (2011). Liquidity risk and expected corporate bond returns. Journal of Financial Economics, 99(3), doi: Longstaff, F. A. (2004). The Flight-to-Liquidity Premium in U.S. Treasury Bond Prices. Journal of Business, 77(3), Markit. (2014). Loan Market data available on request from Markit. Marshall, B. R., Nguyen, N. H., & Visaltanachoti, N. (2013). Liquidity commonality in commodities. Journal of Banking & Finance, 37(1), doi: Parlour, C. A., & Plantin, G. (2008). Loan Sales and Relationship Banking. The Journal of Finance, 63(3), doi: / Pastor, L., & Stambaugh, R. (2003). Liquidity Risk and Expected Stock Returns. Journal of Political Economy, 111(3), doi: / Pu, X., Wang, J., & Wu, C. (2011). Are Liquidity and Counterparty Risk Priced in the Credit Default Swap Market? The Journal of Fixed Income, 20(4), 59-79,54,56. Qian, X., Tam, L. H. K., & Zhang, B. (2014). Systematic liquidity and the funding liquidity hypothesis. Journal of Banking & Finance, 45(0), doi: Roll, R. (1984). A Simple Implicit Measure of the Effective Bid-Ask Spread in an Efficient Market. The Journal of Finance, 39(4), doi: / Rösch, C. G., & Kaserer, C. (2013). Market liquidity in the financial crisis: The role of liquidity commonality and flight-to-quality. Journal of Banking & Finance, 37(7), doi: 22