Banks liquidity and cost of liquidity for corporations

Transcription

1 Banks liquidity and cost of liquidity for corporations Vitaly M. Bord Federal Reserve Bank of New York João A. C. Santos Federal Reserve Bank of New York and NOVA School of Business and Economics February 14, 2012 JEL classification: G21, G32. Key words: Bank liquidity; Credit commitments; All-in-undrawn spreads; Commitment fees. The authors thank Phil Strahan and seminar participants at Faculdade Economia da Universidade do Porto for useful comments. The views stated herein are those of the authors and are not necessarily the views of the Federal Reserve Bank of New York or the Federal Reserve System.

2 Banks liquidity and the cost of liquidity to corporations Abstract Banks are important to corporations not only because of their role as providers of credit, but also because of their role as providers of liquidity. We consider the liquidity shock on banks resulting from the collapse of the asset-backed security market in the fall of 2007 to investigate whether banks liquidity conditions hamper their ability to provide liquidity to corporations. We find that banks that increased their use of funding from the Federal Home Loan Bank system or the Fed s discount window following that liquidity shock charged higher fees to grant new lines of credit to corporations. This increase is more prevalent among credit lines that pose more liquidity risk to banks and affects predominantly bank dependent borrowers. We also show that this increase in undrawn fees was driven by the liquidity shock banks experienced following the collapse of the market for asset backed securities. Our findings, therefore, show that the crisis which affected the banking system had a negative effect not only on the availability and cost of credit to corporations, but also on the cost corporations pay to be guaranteed access to liquidity.

3 1 Introduction Historically, there has been a great deal of interest in the importance of banks financial conditions to the availability and cost of credit to corporations. However, banks are also important for corporations because of their role as providers of liquidity and their financial condition will likely affect this liquidity role. In this paper, we attempt to narrow this gap in the literature by investigating whether shocks to banks liquidity hamper their ability to perform their liquidity provision role. Specifically, we investigate whether banks increase the price they charge corporations to provide them access to liquidity via credit commitments in response to shocks to banks liquidity conditions. It is well established that banks funding choices affect their ability to provide credit to corporations. Diamond and Rajan (2001), for instance, show that deposit funding is important because it makes the bank fragile, giving it incentives to commit to monitoring borrowers. Diamond and Rajan (2000), in turn, show that bank capital affects loan policies because banks with low capital are focused on obtaining cash flow quickly; thus they charge more to borrowers with low cash flow, but give big discounts to borrowers with high cash flow. Lastly, Stein (1998) shows that when banks limit their access to the CD market in order to signal their quality, they are better off cutting down on lending when there is a monetary contraction. It is also well established that banks have a unique ability to provide liquidity to corporations. Holmström and Tirole (1998), for instance, show that banks are better positioned than financial markets to insure firms against liquidity shocks that disrupt their investments because they can commit to finance firms in the future (through credit lines). Kashyap, Rajan and Stein (2002), in turn, show that as long as the demand for liquidity from depositors and the demand for liquidity from borrowers are not highly correlated, banks should offer both of these functions because this saves on their need to hold costly liquid assets the buffer against unexpected deposit withdrawals and loan take downs. 1 However, and in contrast with banks credit provision role, we know very little about the importance of banks funding choices for their liquidity provision role to corporations. A notable exception is Gatev and Strahan (2006), who document that when market liquidity dries up and commercial paper rates rise banks experience deposit funding inflows giving them the opportunity to offer new backup up lines of credit at lower prices. The difference in the attention devoted to the credit and liquidity roles of banks to 1 Gatev, Schuermann, and Strahan (2006) report supporting evidence for banks liquidity role to corporations by showing that bank-stock return volatility increases with unused commitments, but the increase is smaller for banks with high levels of transactions deposits. Pennacchi (2006) also documents the existence of synergies between loan commitments and deposit taking, but finds that such synergies only hold after the creation of FDIC deposit insurance. 1

4 corporations is surprising. Historically banks have been the main source of credit to corporations, but they have also provided corporations liquidity insurance by extending them lines of credit and loan commitments (throughout the paper we use these two terms interchangeably). Further, recent work shows that bank credit lines have become an important source of funding for firms (Sufi 2009). 2 The reduced attention devoted to banks liquidity role to corporations is also surprising because increasingly banks have relied on sources of funding other than deposits, such as the commercial paper market and the repo market, and they have extended their liquidity provision role, most notably by providing liquidity guarantees to SPVs and conduits. 3 While these changes facilitate banks liquidity function, they also expose banks to new sources of liquidity shocks possibly hindering their ability to provide liquidity to corporations. We build on the collapse of the market for asset-backed commercial paper securities (ABCP) in the fall of 2007 to investigate whether there is a link between banks liquidity conditions and the price they charge corporations to grant them access to liquidity. The collapse of the market for asset-backed commercial paper provides us with a good opportunity to investigate that link because it exposed banks to an important liquidity shock resulting from the backup guarantees they extended to off-balance sheet vehicles they had created to securitize mortgages and other loans they underwrote. Concerns about subprime residential mortgages and structured finance products began in early 2007, but the crisis really took hold in the summer of In June and July, two Bear Sterns hedge funds required assistance; Countrywide, one of the largest subprime mortgage originators, announced large losses; and IKB announced its conduit Rhinebridge Plc was unable to roll over asset-backed commercial paper and as a result it was unable to provide its conduit with the promised credit line. In August, BNP Paribas froze redemptions for three investment funds, citing its inability to value structured products. As these events unraveled, concerns about how to value structured products grew, confidence in the reliability of ratings declined, and the market for asset backed commercial paper began to dry up (Brunnermeir 2009). This resulted in large liquidity pressures on banks that were big issuers of these vehicles. For several years, banks that had been moving pools of loans into structured investment vehicles (SIVs) financed with short-term commercial paper. To make these vehicles more attractive to investors, banks offered credit enhancements to reduce the risk to investors in the event of unexpected losses and provided liquidity backstops to insure against refinancing 2 Firms value credit lines because they protect them against changes in interest rates (Thakor and Udell 1987), or signal their true quality (James 1981), or reduce credit rationing (Thakor 2005), or alleviate moral hazard problems (Boot, Thakor and Udell 1987, and Boot, Grennbaum and Thakor 1993). 3 According to Berger and Bauwman (2005), banks liquidity production role has been increasing steadily and about half of the liquidity they now create is done off the balance sheet via loan commitments. 2

5 risk in the asset-backed commercial paper. As these vehicles accumulated losses and investors lost confidence in them, calls on banks liquidity and credit guarantees started to mount. At the same time, the market for newly issued securitized credit declined sharply. Consequently, throughout the second half of 2007 banks experienced both a decrease in their ability to sell loans in securitized markets and faced a greater demand for funding arising from their prior commitments to supply funds to their asset-backed vehicles. While these developments were unfolding, banks were experiencing a sharp increase in their cost of short-term funding. As uncertainty about the securities values grew larger so did investors demands on the haircuts on the collateral that they accepted, increasing the cost to raise funding in the repo market. 4 According to the repo-haircut index put together by Gorton and Metric (2009), the average haircut demanded by market participants over the period between January and July of 2007 was 0.02%. During the August-November period that index went up to 3.73%. 5 That uncertainty, together with banks repeated announcements of large losses, increased the cost banks had to pay to raise funding in the commercial paper market and the interbank market as well. The average 3-month commercial paper spread (over the 3-month Treasury) issued by financials went up from 24 bps over the period between January and July of 2007, to 116 bps points during the August-November time period, while the 3-month LIBOR-OIS went from 8.1 bps to 67.3 bps. We consider the liquidity shock that banks experienced in the fall of 2007 to investigate whether banks liquidity conditions affect their ability to provide liquidity to corporations. Since we do not have detailed information on U.S. banks liquidity guarantees to SIVs, we proxy the shock to each bank resulting from the collapse of the market for ABCP by the bank s use in the fall of 2007 of those sources of liquidity that they resort to when they are unable to meet their liquidity needs in the markets, namely the Federal Home Loan Bank (FHLB) system and the Fed s discount window. The Fed is believed to be the lender of last resort, because banks tend to approach the discount window only when they are unable to meet their liquidity needs elsewhere. Banks that are unable to meet their liquidity needs in the markets can also consider meeting them at the FHLB system, which according to Ashcraft, Bech and Frame (2008) acts as the lender of the next-to-last resort in the U.S. 6 Banks may prefer to borrow from the FHLBs over the Fed 4 Repurchase agreements, or repos allow market participants to obtain collateralized funding by selling their securities and agreeing to repurchase them when the loan matures. 5 These statistics refer to transactions between two high quality dealer banks, and haircuts depend on the risk of the two parties and on the nature of the collateral. The authors thank Gary Gorton for making these statistics available to them. 6 The FHLB System is a US government-sponsored enterprise that was created during the Great Depression. 3

6 in order to avert the stigma that arises with such borrowing. 7 Indeed, during the second half of 2007, the Fed saw only a modest increase in the demand for primary credit through its discount window, this even after the Fed reduced from 100 to 50 bps the premium on the primary credit or discount rate over the federal funds target, and allowed eligible institutions using the primary credit program to borrow funds for up to thirty days, with the possibility of renewal. The FHLB System, in contrast, increased its advance lending by $235 billion to $875 billion by the end of that year (a 36.7% increase). The FHLB system, however, is not a perfect substitute for the Fed s discount window. It does not have the same unlimited capability to provide liquidity as the Fed does. It only provides funding to banks that are members of the FHLB system. Finally, it only lends against collateral made of mortgages, while the Fed accepts as collateral a much wider set of assets. To ascertain whether the liquidity shock arising from the collapse of the ABCP market affected banks ability to provide liquidity to corporations, we investigate whether banks that relied more heavily on the FHLB system and the Fed s discount window in the fall of 2007 charged higher prices to provide liquidity to corporations. We focus on the fall of 2007 because as we noted above the turmoil in the ABCP market began in the summer of We end our analysis in November 2007, because in December of that year the Fed reacted to the turmoil in the money market by creating the Term Auction Facility (TAF) to sell a fixed quantity of one month (and later longer term) credit in a competitive auction. These actions reduced borrowing costs temporarily, with spreads falling in January and February of 2008 (Armantier, Krieger, and McAndrews 2008). We consider the liquidity banks provide corporations via credit lines. The pricing structure of a credit line includes both an undrawn fee and an all-in-drawn spread. According to Dealscan, the former price equals the fees (commitment fee and annual fee) that the borrower must pay its bank for funds committed under the credit line but not taken down. 8 That price compensates the bank for the liquidity risk it incurs by guaranteeing the firm access to funding at its discretion over the life of the credit line and up to the total commitment amount. In contrast, the all-in-drawn spread, which is defined over LIBOR and equals the annual cost to a borrower for drawn funds, compensates the bank for the credit risk it incurs when the borrower draws down on its credit line. Given that we are interested in the price banks charge It consists of 12 cooperatively owned wholesale banks that act as a general source of liquidity to financial institutions. It provides liquidity through advances or (over) collateralized lending to its members. 7 There is a generalized belief that banks avoid borrowing from the discount window because this sends a negative signal about their financial condition. 8 Deaslcan uses the wording all-in-undrawn spread when referring to the cost firms pay on undrawn commitments, but in reality that cost is not a spread because the fees are not markups over market interest rates. 4

7 corporations for providing them access to liquidity, we focus on the undrawn fee banks charged on new credit lines during the fall of However, we also investigate the all-in-drawn spread that banks charged on the same credit lines both to ascertain if banks liquidity pressures also resulted in an increase in the credit spread they charged corporations and to ensure that the effect we identify on undrawn fees is indeed driven by the bank s liquidity conditions. Our results show that the liquidity shock resulting from the collapse of the asset-backed security market led to an increase in the cost banks charge corporations to grant them access to liquidity. We find that banks that increased their use of funding from the FHLB system and from the Fed s discount window in the fall of 2007 charged higher undrawn fees to grant new lines of credit to corporate borrowers when compared to the undrawn fees they charged before the crisis. We also find that this result is driven by corporations that pose more liquidity risk to banks, and that banks differentiate borrowers according to their dependency on them for external funding, targeting bank dependent borrowers. The increase in undrawn fees that we detect is likely driven by the banks liquidity shock resulting from the collapse of the ABCP market and not, for example, by an increase in the demand for credit lines by corporations. A reason is that, as we noted above, we find this increase to be more pronounced among credit lines that pose more liquidity risk to banks. Another reason is that we continue to find these results both when we instrument for the banks demand for funding from the FHLB system and the Fed s discount window by the bank s liquidity exposure to ABCP vehicles, and when we instrument by the bank s liquidity exposure to ABCP vehicles backed by MBS. The results derived with the latter instrument are particularly important because the turmoil in the mortgage market in the second half of 2007 had a strong impact on MBS-backed ABCP conduits. Yet another reason why our results are likely driven by the liquidity shock that affected banks is that we do not find a similar increase on the credit spreads that banks charge on the credit lines. While we find some evidence that banks also increased credit spreads on credit lines, we do not find this increase to be as closely aligned with the liquidity risk of the credit line as in the case of undrawn fees. Also, we find that our results on undrawn fees are robust to controlling for the credit spread the bank s charge on the credit line. Our findings are important because they show that banks ability to perform their function of liquidity providers to corporations is closely linked to their liquidity conditions, notwithstanding the existence of a lender of last resort. Our findings are also important because they show that the subprime crisis which affected the banking system had an effect not only on the availability and cost of credit to corporations, but also on the price corporations had to pay to guarantee access to liquidity. 9 9 See Santos (2010) for evidence showing that banks passed onto their corporate borrowers part of the losses 5

8 Our paper is related to the literature that has investigated the role of banks as liquidity providers to corporations. 10 The focus of this literature is to identify synergies between deposit funding and extension of loan commitments. Our focus instead is to investigate whether banks liquidity provision role depends on their liquidity risk. In this regard, our paper is related to Acharya, Almeida and Campello (2009) who argue that an increase in aggregate risk creates liquidity risk for banks that are exposed to undrawn credit lines leading them to increase the cost of credit lines. In contrast to them, we focus on the liquidity risk arising from banks funding choices, and we identify the effect that banks liquidity risk has on the prices they charge corporations for granting them access to liquidity via credit lines. 11 This part of our paper is related to the studies on pricing of loan commitments, including Shockley and Thakor (1997), Chava and Jarrow (2008) and Gatev and Strahan (2006). Our focus is on the importance of bank liquidity conditions on the pricing of these commitments, an issue absent from existing literature. The remainder of our paper is organized as follows. The next section presents our methodology and data, and characterizes our sample. Section 3 investigates whether banks liquidity conditions in the second half of 2007 affected the price they charged corporations for granting them access to liquidity via loan commitments. Section 4 investigates whether banks liquidity conditions also affected the credit spreads banks charged on these loan commitments. Section 5 concludes the paper. 2 Data, methodology, and sample characterization 2.1 Data The data for this project come from several sources, including the Loan Pricing Corporation s Dealscan database (LPC), the Center for Research on Securities Prices s stock prices database (CRSP), Compustat, the Federal Reserve s Y9C reports, and from the Shared National Credit (SNC) program run by the Federal Deposit Insurance Corporation, the Federal Reserve Board, the Office of the Comptroller of the Currency, and the Office of Thrift Supervision. We use LPC s Dealscan database to identify the firms that took out credit lines from they incurred during the subprime crisis by charging them higher loan interest rates. 10 See Kashyap, Rajan and Stein (2002), Gatev, Schuermann, and Strahan (2006), Pennacchi (2006), and Gatev and Strahan (2006). 11 Acharya, Almeida and Campello (2009) investigate the prices borrowers pay on their credit lines but they do not consider the potential influence that the lending bank may have on those prices. They test their theory instead by investigating whether there is a negative relationship between the fraction of the borrower s liquidity that is provided by lines of credit and the beta of the lending bank. 6

9 banks and when they do so. We further use this database to obtain information on each credit line, including the undrawn fee, all-in-drawn spread, maturity, seniority status, and purpose; some information on the borrower, including its sector of activity, and its legal status (private or public firm); and finally information on the lending syndicate, including the identity and role of the banks in the syndicate. We rely on Compustat to get balance sheet data for firms with credit lines in LPC. Even though LPC contains loans from both privately-held firms and publicly-listed firms, we limit our tests to publicly listed firms because Compustat has balance sheet data only for publicly listed firms. We rely on the CRSP database to link companies and subsidiaries that are part of the same firm, and to link companies over time that went through mergers, acquisitions or name changes. We then use these links to merge the LPC and Compustat databases in order to find out the financial condition of firms at the time they take out a credit line from a bank. 12 We also use CRSP to determine our measures of return and volatility of the firm s stock price. We use the Reports of Condition and Income to obtain bank data, including information on bank liquidity, its use of the FHLB system, its capital-to-asset ratio, size, profitability and risk, for the lead bank(s) in each credit line syndicate. Whenever possible we get this data at the bank holding company level using Y9C Reports. If these reports are not available then we rely on Call Reports which have data at the bank level. We use information from a proprietary Federal Reserve database on daily discount window borrowing, which includes information on the borrower, amount borrowed, available collateral, and interest rate. 13 We use the 3-month LIBOR to account for the cost to accessing the FHLB system, and we proxy for the cost to borrowing from the Fed s discount window by the discount window rate set by the Fed. Finally, we use the SNC data to get information on each borrower s drawdown history. At the end of each year, the SNC program gathers confidential information on all credits that exceed $20 million and are held by three or more federally supervised institutions. 14 Like the LPC Dealscan database, the SNC data is dominated by the syndicated loan market. In contrast with the LPC data, which has only information at the time of the credit origination, the SNC database tracks credits over time. This gives us the opportunity to identify which 12 The CRSP data was first used to obtain CUSIPs for the companies in LPC where this information was missing through a name-matching procedure. With a CUSIP, LPC could then be linked to Compustat, which is a CUSIP-based dataset. We proceed by using the PERMCO variable from CRSP to group companies across CUSIP, since that variable tracks the same company across CUSIPs and ticker changes. We adopted a conservative criteria and dropped companies that could not be reasonably linked. 13 The discount window data were processed solely within the Federal Reserve for the analysis presented in this paper. 14 The SNC data were processed solely within the Federal Reserve for the analysis presented in this paper. 7

10 borrowers draw down on their credit lines and when they do it. Linking the information from these databases, we are able to determine the financial condition of the firm at the time it takes out a credit line, and to ascertain its draw down history on previous credit lines. We are also able to find out how much the firm pays the bank for the right to withdraw at its discretion, the undrawn fee, and the spread that it will pay when it draws down on the credit line, the all-in-drawn spread. In addition, we are able to determine the financial condition of the bank at that time, including its liquidity status, and how much it has borrowed from the FHLB system and from the Fed s discount window. Using this information, we can investigate whether banks liquidity conditions play a role in the cost they charge firms to grant them access to liquidity. We describe next the methodology we follow to investigate this question. 2.2 Methodology In order to determine whether banks liquidity conditions affect the price they charge corporate borrowers to guarantee them access to liquidity, we investigate whether banks that were under more liquidity strains during the crisis of 2007 charged higher fees to extend new credit lines to corporations, controlling for firm-, bank-, and credit line-specific characteristics. To that end, we estimate the following model on a set of new credit lines extended by banks: UNDRAWN FEE f,l,b,t = c + α BK LIQ SHOCK b,t 1 + β CRISIS t + γ CRISIS t BK LIQ SHOCK b,t 1 I J K + η i X i,l,t + δ j Y j,f,t 1 + ζ k,t Z k,b,t 1 + ɛ f,t. (1) i=1 j=1 k=1 UNDRAWN FEE f,l,b,t is the all-in-undrawn fee on the credit line l of firm f from bank b at issue date t. According to Dealscan, our source of credit-line data, the all-in-undrawn fee, which usually includes both a one-time commitment fee and an annual fee the borrower pays the bank, is a measure of the cost the bank charges the firm for granting it access to liquidity via a credit line. The all-in-undrawn fee, therefore, compensates the bank for the liquidity risk it incurs by guaranteeing the firm access to liquidity at its discretion and up to the total commitment amount. BK LIQ SHOCK is our proxy for the bank s liquidity shock. Since we do not have information on the backup guarantees banks extended to off-balance vehicles they had created to securitize mortgages and other loans they underwrote, we proxy for the liquidity pressures that banks were under by the amount of funding they raised from those sources that banks revert to when they are unable to meet their liquidity needs in the markets the Federal Home 8

11 Loan Bank system and the Fed s discount window. Specifically, we proxy for the liquidity pressures affecting banks by the amount of funding they raised from the FHLB system in the quarter prior to the credit line, FHLB, and by the amount of funding they borrowed from the Fed s discount window, DW, also in the quarter prior to the credit line, and both scaled by the bank s assets. 15 CRISIS t is a dummy variable equal to one for the crisis period. Our crisis period starts in August 2007, when BNP Paribas announced it could not value assets in three of its investment funds which had a profound effect on the asset-backed commercial paper market. Our crisis period ends in November 2007 because in December the Fed introduced TAF which helped banks deal with the liquidly problems they were facing. Since during the crisis it became more expensive for banks to meet their liquidity needs either by raising funding in the commercial paper market or by borrowing in the repo market or the interbank market, we expect β to be positive. Further, since banks resort to the FHLB system, and in particular to the Fed s discount window, only when they are unable to meet their liquidity needs in the markets, we expect γ to be positive. In other words, banks that raised funding from these two sources during the crisis were likely under more liquidity strains and consequently were under more pressure to charge higher fees to their borrowers for guaranteeing them access to liquidity via credit lines. We investigate these effects controlling for a set of factors that are specific to each credit line, Y i,l,t, and a set of factors that are borrower specific, X j,f,t, which are expected to affect the liquidity risk the bank incurs when it extends the credit line. We also control for a set of bank-specific factors Z k,b,t which might affect the rate at which the bank is willing to guarantee firms access to liquidity. We describe these sets of variables next. We begin by discussing the firm-specific variables that we use. Several of these variables are proxies for the firm s liquidity needs. Others proxy for the firm s risk of failure. Since both of these increase the likelihood that the firm will draw down its line of credit, we expect them to increase the all-in-undrawn fee the bank charges on the credit line. LAGE is the log of the firm s 15 To be more precise FHLB is computed as the sum of all FHLB borrowing outstanding as of end of the previous quarter scaled by assets. Throughout the analysis, for the regressions involving the FHLB variable we exclude the loans associated with banks that are not members of the FHLB system. DW is computed as the median of the Discount window loans taken out over the previous three months, also scaled by assets. We use the median to prevent double counting since discount window loans are often rolled over. Using the maximum over the same period does not qualitatively change the results. According to the Federal Reserve, discount window loans come in three types, and at three rates. Primary credit loans are available to sound depository institutions. Secondary credit loans are available to those that do not qualify for primary credit loans, and seasonal credit loans are available to small institutions in order to meet seasonal funding needs. All banks in our sample take out only primary credit loans. 9

12 age in years. 16 LASSETS is the log of the firm s assets in hundreds of millions dollars. Older and larger firms are typically better established better diversified across customers, suppliers, and regions, and so less risky. They also tend to have access to more funding sources, so we expect this variable to have a negative effect on the undrawn fee. The next set of firm variables attempt to capture the firm s needs for external liquidity. CASH is the firm s cash divided by total assets; this measures the firms liquid asset base. CURRENT RAT IO is the firm s current assets divided by its current liabilities; this measures the firm s liquidity that is free. Since firms with more current assets are less likely to draw down on their credit lines, we expect these variables to be negative. P ROF MARGIN is the firm s profit margin (net income divided by sales). LINTCOV is the log of 1 plus the interest coverage ratio (i.e., earnings before interest, taxes, depreciation, and amortization (EBITDA) divided by interest expense); more profitable firms as well as firms with more interest coverage have a greater cushion for servicing debt. These firms are less risky and they should have less demand for external liquidity, so they should pay lower undrawn fees on their credit lines. In contrast, firms with more growth opportunities as determined by their market to book ratio, MKTTOBOOK, and firms that are growing faster according to the growth rate of their assets, ASSETS GROWTH, are more likely to use their credit lines and so we expect them to pay higher undrawn fees. Our final set of firm controls attempts to capture the firm s risk of default. LEV ERAGE is the firm s leverage ratio (debt over total assets); higher leverage suggests a greater chance of default. We complement this set of firm controls with two variables linked to the firm s stock market price. STOCK RET is the firm s excess stock return (relative to the overall market) over the last twelve months. To the extent a firm outperforms the market s required return, it should have more cushion against default and thus pay lower undrawn fees. STOCK V OL is the standard deviation of the firm s stock return over the past twelve months. Higher volatility indicates greater risk, and thus a higher probability of default. We expect both of these variables to have a positive impact on undrawn fees. T ANGIBLES is the firm s tangible assets inventories plus plant, property, and equipment as a fraction of total assets. Firms with more tangible assets are less prone to moral hazard problems and so everything else equal these firms are less likely to draw down on their lines of credit and thus pay lower undrawn fees. We complement this set of risk controls with dummy variables for the credit rating of the borrower and dummy variables for single-digit SIC industry groups. Credit rating agencies claim that they have access to private information on firms that is not captured by our public Compustat data. Also, firms with a credit rating are more likely to have access to bond market funding. Likewise, a given industry may face additional risk factors and liquidity needs that 16 To compute the firm age we proxy the firm s year of birth by the year of its equity IPO. 10

13 are not captured by our controls, so this dummy allows us to capture such risk at a very broad level. In addition, we distinguish borrowers that have a commercial paper program since these borrowers have access to an additional source of funding. We now discuss our credit line-specific variables. We control for the size of the credit commitment (measured by LAMOUNT,the log of loan amount in hundreds of millions dollars) and for the maturity of the commitment (measured by MATURITY,the commitment maturity in years). Larger and longer commitments represent more liquidity risk to the bank and so we expect both of these variables to have a positive effect on the undrawn fee. Because the purpose of the credit line is likely to contain information on the likelihood that the firm will draw down, we include dummy variables for credit lines taken out for corporate purposes, CORPORATE PURP, to repay debt, DEBT REPAY, for working capital purposes, WORKING CAP, to back up a commercial paper program, CP BACKUP, for mergers and acquisions, M&A, or to refinance existing debt, REFINANCE. We include dummy variables equal to one if the credit line has restrictions on paying dividends, DIV IDENDREST, is secured, SECURED, and if the borrower has a guarantor, GUARANT OR. 17 All else equal, any of these features should protect the bank in case the firm draws down on the credit line, but it is well known that lenders are more likely to require these features if they think the firm is riskier (see for example Berger and Udell (1990)), so the their effect on the undrawn fee may be reversed. We distinguish firms that have a lending relationship with the bank by including a dummy variable equal to one if they took out the previous credit line from the same bank as the current one, RELATIONSHIP. Banks will have more information about the liquidity needs of the firms they have a relationship with and so the liquidity risk of granting these firms access to liquidity should be lower. On the other hand, a relationship may also indicate greater information monopoly, so the effect of this variable on the undrawn fee is also ambiguous. We now turn our attention to the set of bank-specific variables that we use. We control for the various funding sources banks normally use, including insured deposits, INSUREDDEP, uninsured deposits, UNINSUREDDEP, subordinated debt, SUBDEBT, commercial paper, CP, repo market, REP O, and the interbank market, FEDFUNDS, (all scaled by the bank s assets). 18 We also control for the bank s holdings of cash and marketable securities as a fraction of total assets, BK LIQUIDITY. Banks with more liquid assets as well as banks with access to the public bond market, and banks with more deposits, in particular those with more 17 We do not control for the seniority of the credit line because all credit lines in the sample are senior. 18 In the case of funding in the interbank market our proxy captures only the portion of this funding that is overnight. Information on the term loans that banks take out in the Fed Funds market is lumped together with other variables in the call report. 11

14 insured deposits, should find it easier to extend loan commitments, so we expect these variables to have a negative effect on undrawn fees. With regard to the commercial paper market, the repo market and the interbank market, even though these are perceived to be inexpensive sources of liquidity for banks, since the cost to access these markets went up during the crisis we may find that banks which rely more extensively on these funding sources charge higher fees to extend lines of credit. We further control for the bank size as determined by the log of the bank s assets, BK LASSETS, and for its performance as determined by its return on assets, BK ROA. Larger banks may be better-diversified or have better access to funding markets, which may enable them to charge lower undrawn fees. Similarly, better performing banks may be able to extend liquidity at a lower cost. Conversely, indicators of bank risk such as the volatility of return on assets, BK ROAV OL, may mean that the bank faces a higher cost of funds and consequently charge higher fees to extend new credit lines. 19 As with the borrowers, we include dummy variables to account for the credit rating of the bank. We complement these sets of firm- credit-line and bank-specific controls with yearly dummy variables to control for the overall economic conditions, and we proxy the cost of borrowing from the FHLB system with the three-month LIBOR and the cost of borrowing from the Fed s discount window with the discount window rate. 2.3 Sample characterization Table 1 characterizes the sample of credit lines we use in our investigation on the importance of banks liquidity for the cost they charge corporations to grant them access to liquidity. Panel A compares the credit lines taken out by nonfinancial corporations before the crisis with the credit lines corporations take out during the crisis. Panel B compares the borrowers of these credit lines while panel C compares the banks that extend these sets of credit lines. Lastly, panel D compares the interest rates before and during the crisis on various credit markets. Our sample period goes from January 2005 through November We end in November of 2007 because in December of that year the Fed introduced TAF, which had an important effect on the liquidity available to banks. We begin in 2005, in order to have a reasonable number of years before the beginning of the financial crisis. As we noted above, we define the crisis as the period between August and November of A look at panel A of table 1 reveals that the undrawn fees decline in the crisis. This difference appears to be contrary to our priors that the cost banks charge to provide liquidity to corporations went up during the crisis. However, this decline derives from the fact that the 19 We use the volatility of ROA rather than stock return because a large number of the banks in the sample do not have publicly traded shares. 12

15 pool of firms that obtain credit lines during the crisis is of better quality than the pool of firms in the pre-crisis period. Firms that get credit lines during the crisis appear to pose less liquidity risk to banks. They take out larger credit lines and they have more growth prospects as measured by their market-to-book ratios, but they also have larger cash holdings. In addition, there is a large decline during the crisis in the incidence of credit lines to backup commercial paper programs, which posed significant liquidity risk for the bank because of the disruptions in the commercial paper market at the time. Firms that get credit lines during the crisis also appear to pose less credit risk to banks. These firms have lower leverage. There is a higher portion of double-a rated firms and a lower portion of double-b rated firms among the pool of crisis borrowers. In addition, banks impose dividend restrictions less often on the credit lines they extend during the crisis. Recall that banks are more likely to impose this covenant on riskier borrowers than on safer ones. Panel C of table 1 confirms that banks were under large liquidity pressure during the crisis. Banks experienced significant declines in liquidity and deposits (including insured and uninsured deposits). They also experienced a significant decline in their repo funding. In response, they increased their use of subdebt and the Fed funds market. Banks, however, appear to not have been able to meet their liquidity needs in these markets because they increased significantly their use of funding from the FHLB system and from the Fed s discount window. This increase is important to us because we proxy banks liquidity pressures during the crisis by the extent of their use of these two funding sources. Lastly, panel D of table 1 confirms that during the crisis it became more expensive for banks to raise funding in the bond market, the commercial paper market, and the interbank market since the spreads in these markets, BBBSPREAD, CPSPREAD and LIBOR OIS, respectively, all went up during the crisis. It also likely became more expensive to raise funding in the FHLB system since banks usually pay a spread over LIBOR and this interest rate went up during the crisis. The discount window spread, DW SPREAD, declined during the crisis, but as we discussed in the introduction this spread does not capture the entire cost for banks from borrowing from the Fed s discount window because of the stigma costs that arise with this borrowing, among other things. Banks appear to have been able to raise funding during the crisis at a lower cost only on the short-term Fed funds market. 3 Bank liquidity and undrawn fees on banks credit lines Table 2 reports the first set of results of our investigation into the effect of banks liquidity conditions on the cost they charge corporations for granting them access to liquidity. Recall 13

16 that with the collapse of the market for asset-backed commercial paper in the fall of 2007, banks that had provided liquidity guarantees to SIVs experienced large calls for liquidity support. Banks that were unable to meet these calls with internal funds or through the markets had to revert to the FHLB system, and even the Fed s discount window to meet their liquidity needs. Model 1 of table 2 investigate whether the banks that raised more funding from the FHLB system during the fall of 2007 increased the undrawn fees they charged corporations that took out new lines of credit during that same period of time. Model 2 repeats this exercise but this time focusing on banks that borrowed from the Fed s discount window during the fall of The results of these models show that both CRISIS x FHLB and CRISIS x DW are positive and statistically significant. In other words, these models show that everything else equal, banks that had to rely on the FHLB system as well as banks that had to borrow from the Fed s discount window during the fall of 2007 charged higher undrawn fees on the new lines of credit that they extended to corporate borrowers. Since CRISIS is not statistically significant in either model, this indicates that banks which did not have to resort to these funding sources did not charge higher fees on the lines of credit they extended during the crisis when compared to their pre-crisis fees. Interestingly, these models also indicate that in the precrisis period while the use of funding from the FHLB system did not result in higher undrawn fees (FHLB is not statistically significant), banks that borrowed from the Fed s discount window did charge higher fees on their lines of credit (DW is positive and significant). This difference could arise because of the additional costs that arise with borrowing from the Fed s discount window or because of certain unique features of the banks that resort to this source of funding (more on this below). With regard to the controls we use in our model of undrawn fees, those that are statistically significant, are usually consistent with our expectations. Credit lines with a longer maturity pay higher fees, possibly because they pose higher liquidity risk to banks. This is also likely the reason why firms with low interest coverage and those with low current ratio pay higher fees on their credit lines, though the current ratio is only significant in model 1. Larger credit lines in turn pay lower fees, possibly because these are taken out by larger firms, which tend to be safer. Firms with higher risk of default as determined by their leverage, stock volatility, or ratings pay higher fees. Firms with more growth opportunities, as determined by their market-to-book ratio, pay lower fees which is contrary to our expectations since these firms are more likely to draw down on their credit lines. Also, contrary to our expectations we find that larger firms, as determined by their assets, pay larger fees on their credit lines. Among the bank controls other than the one already discussed above, only bank capital is significant and indicates that well capitalized banks charge lower fees on the credit lines they extend. 14

17 Since models 1 and 2 of table 2 were estimated with a pooled regression one may wonder whether our findings showing that banks which borrowed more from the FHLB system or the Fed s discount window during the fall of 2007 charge higher fees on their credit lines derive from banks increased use of these funding sources or alternatively from differences across banks. To ascertain the validity of these alternative explanations for our findings, we reestimate both models but this time with bank-fixed effects. The new results, which are reported in models 3 and 4 of that table, show that our findings are not driven by differences across banks. As in models 1 and 2 we continue to find that CRISIS x FHLB and CRISIS x DW are positive and statistically significant confirming that banks which had to increase their use of funding provided by the FHLB system during the crisis as well as banks which had to borrow more from the discount window during the crisis increased the fees they charged corporations for granting them access to liquidity via credit lines. Also as before, we continue to find that CRISIS is not statistically significant, indicating that banks which did not resort to these funding sources did not alter the undrawn fees on their credit lines. In sum, our first set of results suggests that banks that were under more liquidity pressure in the fall of 2007 and which had to resort to funding provided by the FHLB system or the Fed s discount window increased the price they charged corporations for granting them access to liquidity. In other words, these banks appear to have responded to the liquidity pressures they were under by increasing the price they charged corporations for bearing the liquidity risk that they exposed themselves to when extending new credit lines. Implicit in this conclusion is the assumption that credit lines expose banks to liquidity risk and that it is costly for them to hedge this risk because there is an opportunity cost for their investments in liquidity. Credit lines expose banks to liquidity risk because borrowers buy the right to withdraw at their will and up to the amount in the credit line. Implicit in that conclusion is also our assumption that banks resorted to funding provided by the FHLB system or the Fed s discount because of the liquidity pressure they experienced following the collapse of the market for commercial-paper backed securities in the fall of 2007 and not because of an increase at the time in demand for credit lines by corporations. We investigate these assumptions next. 3.1 Credit lines liquidity risk and undrawn fees In order to investigate whether the link we unveiled between banks liquidity conditions and the undrawn fees they charged on new credit lines during the fall of 2007 is more pronounced among credit lines that pose a bigger liquidity risk to banks, we first need to proxy for the liquidity risk of each credit line. One way to do that is to consider the maturity of the credit line. Everything else equal, credit lines with a longer maturity will likely pose more liquidity risk to banks than shorter-term credit lines because they expose the bank to borrowers draw 15

18 down demands for a longer period of time. The result of table 3 showing that banks charge higher undrawn fees on credit lines with longer maturity is consistent with this assertion. Further, as we can see from model 1 of table 3, which investigates the importance of the credit line maturity before and during the crisis, this effect is due to the crisis period. In the precrisis period the maturity of the credit line was not an important driver of undrawn fees banks charged on credit lines. Note that MATURITY is not significant in model 1. In contrast, CRISIS x MATURITY comes out positive and significant therefore indicating that during the crisis banks began to charge higher fees to extend credit lines of longer maturity, possibly because of their added concerns with liquidity risk at the time. Another way to proxy for the liquidity risk of credit lines is to consider the borrower s draw down history. Using information from the Shared National Credit (SNC) program on borrowers draw downs, we define the following two proxies for the liquidity risk that each credit line poses the bank. The first proxy is the median drawdown share on the borrower s past credit lines. The second proxy is the standard deviation of the drawdown share on the borrower s past credit lines. Everything else equal, borrowers with a history of drawing down heavily on their credit lines as well as borrowers with a more volatile draw down history will be more costly for banks because they both will require banks to set aside more liquidity to meet their funding demands. Models 2 and 3 investigate whether these hypotheses hold true during the fall of Model 2 uses the median of the borrower s historic draw-down share to identify credit lines that pose more liquidity risk to originating banks. Model 3, in turn, uses the volatility of the borrower s draw down history to identify these credit lines. As in model 1, both models 2 and 3 suggest that liquidity risk did not play an important role in the way banks priced credit lines in the pre-crisis period. In contrast, and consistent with our findings based on the maturity criterion, the latest results indicate that during the crisis banks started to charge higher fees on credit lines that pose more liquidity risk to them. According to model 2, during the crisis banks charged higher fees on credit lines they extended to borrowers that tend to draw down on their lines of credit (CRISIS x DD MED is positive and significant), but not on credit lines they extended to borrowers that traditionally do not draw down on their lines of credit (CRISIS is not significant). According to model 3, during the crisis banks charged higher fees on credit lines they extended to borrowers that have a volatile draw-down history (CRISIS x DD V OL is positive and significant), but not on credit lines they extended to borrowers with a predictable draw-down pattern (CRISIS is not significant). Now that we have established that our three proxies for liquidity risk in credit lines are correlated with the fees banks charged during the fall of 2007, and therefore added support to 16