«Kenneth N. Kuttner Federal Reserve Bank of New York Darius N. Palia Rutgers University Are There Bank Effects in Borrowers’ Costs of Funds? ...»
R. Glenn Hubbard
Columbia University and National Bureau of Economic Research
Kenneth N. Kuttner
Federal Reserve Bank of New York
Darius N. Palia
Are There Bank Effects in
Borrowers’ Costs of Funds?
Evidence from a Matched Sample
of Borrowers and Banks*
I. Introduction We use a matched sample
of individual loans, bor-
rowers, and banks to in-
Empirical researchers in macroeconomics and corporate
vestigate the effect of ﬁnance have long been interested in effects of changes banks’ ﬁnancial health on in bank loan supply on borrowers’ costs of funds and the cost of loans, control- hence on a variety of investment decisions by borrowers ling for borrower risk and information costs.
(see, e.g., Roosa 1951). This interest has come to the Our principal ﬁnding is forefront in policy discussions of the credit crunch in that low-capital banks the United States in 1991 and the capital crunch for tend to charge higher Japanese banks in 1998. To the extent that a borrower loan rates than well- faces switching costs in a relationship with an individual capitalized banks. This effect is primarily associ- bank, bank-speciﬁc ﬁnancial health might affect a bor- ated with ﬁrms for which information costs are * We are grateful to David Beim, Charles Calomiris, Douglas likely to be important, Diamond, John Driscoll, Charles Himmelberg, Robert Hodrick, and, when borrowing Mitchell Petersen, Richard Roll, Eric Rosengren, Matthew Shapiro, from weak banks, these Jacky So, Phil Strahan, Suresh Sundaresan, and seminar participants at University of Chicago, Columbia University, University of Flor- ﬁrms tend to hold more ida, New York University, Northwestern University, University of cash. The results indicate California, Los Angeles, the Federal Reserve Bank of New York, that many ﬁrms face sig- the NBER Summer Institute Programs in Monetary Economics and niﬁcant costs in swi
rower’s cost of funds, even when observable characteristics relating to borrower risk are controlled for. And to the extent that certain borrowers face differentially costly external ﬁnancing from nonbank as opposed to bank lenders, shifts in the ability or willingness of banks to lend can affect these borrowers’ cost of funds and investment.
Sources of a special role for banks in the credit allocation process have been widely explored. Indeed, the existence of banklike ﬁnancial intermediaries is generally explained by informational asymmetries that lead to costly frictions in the allocation of capital (see, e.g., Diamond 1984, 1989, 1991;
Ramakrishnan and Thakor 1984). In this line of inquiry, the relative importance of private ﬁnancing for ﬁrms depends on the magnitude of information costs in acquiring external ﬁnancing.1 That is, the role for ﬁnancial intermediaries in ﬁnancing investment is most pronounced when high information costs create a signiﬁcant wedge between the costs of internal and external ﬁnancing (see, e.g., Bernanke 1983; Fama 1985). While there are signiﬁcant bodies of research on effects of ﬁrms’ balance sheet positions on ﬁrms’ investment decisions and on effects of banks’ balance sheet positions on banks’ lending decisions, empirical work linking bank and borrower variables has been much more limited.
One strand of research offers indirect evidence on the real decisions of bank-dependent borrowers. Using ﬁrm-level data for Japan, Hoshi, Kashyap, and Scharfstein (1993) concluded that investment is less sensitive to cash ﬂow for ﬁrms that are members of a keiretsu. Also using Japanese data, Gibson (1995) found that ﬁrm investment is sensitive to the ﬁnancial health of the ﬁrm’s main bank, holding constant Q and cash ﬂow (as proxies for investment opportunities and costly external ﬁnancing). Using data on small U.S. ﬁrms, Petersen and Rajan (1994) and Berger and Udell (1995) estimated that a close bank relationship increases credit availability for small borrowers. Using data on larger, publicly traded U.S. ﬁrms, Houston and James (1995) found that ﬁrms that rely on a single bank lender have a much greater sensitivity of investment to cash ﬂow than do ﬁrms that have multiple bank relationships or that borrow in public debt markets. They also estimate that ﬁrm-level sensitivity of investment to cash ﬂow increases with a ﬁrm’s reliance on banks for debt ﬁnancing.
Another body of research has concluded that replacing banking relationships is costly. James (1987) found that, on announcement of a bank loan, ﬁrms earn positive abnormal returns.2 Similar in spirit to this article is that of Slovin,
1. Perhaps less well understood are costs associated with reliance on banks, including regulatory taxes (Fama 1985), information monopoly power (Sharpe 1990; Rajan 1992), and costs of lender control (Diamond 1994).
2. While James (1987) found that all bank loans earn positive abnormal returns, Lummer and McConnell (1989) found that only loan renewals earn positive abnormal returns and that loan initiations do not. However, Slovin, Johnson, and Glascock (1992) showed that differentiating between loan initiations and loan renewals is unnecessary, because both types of loans earn positive abnormal returns (only in the case of small ﬁrms, not in the case of large ﬁrms).
Accordingly, we control for ﬁrm size but not for whether the loan is a renewal or an initiation.
Bank Effects 561 Sushka, and Polonchek (1993), who studied the effects of the de facto failure of Continental Illinois Bank and its subsequent rescue by the Federal Deposit Insurance Corporation (FDIC) during 1984 on the share prices of the bank’s loan customers. In particular, they concluded that the impending failure led to negative excess returns for ﬁrms with a lending relationship with Continental (especially for those lacking a relationship with another bank), while the rescue led to positive excess returns for those ﬁrms. We employ a larger sample of banks and ﬁrms than do Slovin, Sushka, and Polonchek, and, more important, we control for ﬁrm characteristics related to borrower-speciﬁc operating risk and scope for moral hazard.
We attempt to bridge the gap in existing research by matching data on the terms of individual loans with information on the borrower and bank lender in the transaction. This matching allows us to investigate whether, holding constant proxies for borrower risk and information costs, bank liquidity or capital affects terms of lending. In particular, we focus on measuring the effects of borrower and bank characteristics on the interest rate charged to the borrower.
Our principal ﬁndings are ﬁve. First, even after controlling for proxies for borrower risk and information costs, the cost of borrowing from low-capital banks is higher than the cost of borrowing from well-capitalized banks. Second, this cost difference is traceable to borrowers for which information costs and incentive problems are a priori important. Third, estimated “weak-bank” effects remain even after controlling for unobserved heterogeneity in the matching of borrowers and banks. Fourth, weak-bank effects are quantitatively important only for high-information-cost borrowers, consistent with models of switching costs in bank-borrower relationships and with the underpinnings of the bank lending channel of monetary policy. Fifth, when we investigate determinants of cash holdings of borrowing ﬁrms, we ﬁnd that ﬁrms facing high information costs hold more cash than other ﬁrms, all else being equal, and those ﬁrms (and only those ﬁrms) have higher cash holdings when they are loan customers of weak banks. These results suggest that declines in banks’ ﬁnancial health can lead to precautionary saving by some ﬁrms, a response that may affect their investment spending.
The article is organized as follows. Section II describes the data sets we use to match loan, bank, and borrower characteristics. Our empirical tests are reported in Section III. Section IV concludes and discusses broader implications of our ﬁndings.
II. The Data Our interest in isolating effects of borrower and bank characteristics on the cost of funds for investment creates a high data hurdle. We require information on loans, borrowers, and banks for each transaction. Our basic source of data is a sample of 11,621 loan agreements with principal amounts totaling $1,895 billion (with an average loan size of $164 million), covering about 4,840 562 Journal of Business business ﬁrms in the United States. The data are taken from the 1993 release of the Dealscan database supplied by the Loan Pricing Corporation (LPC) and cover the period from 1987 to 1992.3 For a given loan, the LPC data record the identity and location of the borrower; the purpose, contract date, type, and amount of the loan;4 the identities of the lenders (for our purposes, U.S. banks) party to the loan at origination; and price and some nonprice terms. Almost all (97%) of the loans are ﬂoating-rate. To obtain more information about borrower characteristics, we matched the ﬁrms in LPC with those in the Compustat database. To obtain more information about bank characteristics, we match the banks in LPC (i.e., the lead bank for a given loan) with data from the Reports of Condition and Income (Call Reports) compiled by the FDIC, the Comptroller of the Currency, and the Federal Reserve System.5 We use as a measure of the cost of funds the “drawn all-in spread,” or AIS, reported by LPC.6 The AIS is intended to provide a standard measure of the overall cost of the loan, expressed as a spread over the benchmark London interbank offering rate (LIBOR) and taking into account both one-time and recurring fees associated with the loan. The AIS is deﬁned accordingly as the coupon spread, plus any annual fee, plus any up-front fee divided by the maturity of the loan. For loans not based on LIBOR, the LPC converts the coupon spread into LIBOR terms by adding or subtracting a constant differential reﬂecting the historical averages of the relevant spreads.7 Before investigating empirically the effects of borrower and bank characteristics on the cost of funds, we begin by documenting patterns for loan rates (measured by the AIS), loan maturity, bank size, ﬁrm leverage, use of colOther studies using the LPC data for different purposes include Carey (1995a, 1995b); Beim (1996); and Carey, Post, and Sharpe (1998). In general, the loan agreements in the Dealscan database cover a signiﬁcant fraction of the dollar value of outstanding consumer and industrial loans (see Carey et al. 1998). According to LPC, the great majority of the data were collected from letters of commitment and credit agreements drawn from ﬁlings with the Securities and Exchange Commission. (Registered ﬁrms are required to disclose information about any ﬁnancing in excess of 10% of their total assets and, while not required to do so, often choose to include the complete text of the credit agreement as an attachment to their ﬁling.) Especially in the more recent years of our sample, some data were collected from news reports or through LPC’s relationships with major banks.
4. Some of the loan packages, or “deals,” incorporated multiple “facilities” originated by the borrower on that date. Our empirical analysis is at the level of the facility because loan packages with more than one lender do not necessarily involve all lenders in all facilities and because the spread depends on facility-speciﬁc attributes.
5. We lose observations when LPC does not report the loan spread or whether the loan is secured and when we cannot match the loan transaction data to the Call Report data or Compustat.
6. An “undrawn all-in spread” on undrawn lines of credit is also reported, but we do not use it in the analysis.
7. The differentials used in the AIS reported in the LPC data set are as follows: 205 basis points for the prime rate, 19 basis points for the commercial paper rate, 125 basis points for the Treasury-bill rate, 25 basis points for the federal funds rate, 12 basis points for the bankers’ acceptance rate, and 9 basis points for the rate on negotiable certiﬁcates of deposit.
Carey (1995a) found the loan spread, as measured by the AIS, to be comparable to bond spreads, controlling for differences in maturity and collateral. Replacing these constants with time-varying differentials based on year-speciﬁc average spreads has a minimal effect on the results.
Bank Effects 563
lateral, and bank capital-asset ratios across borrower-size groupings (measured by sales). As table 1 shows, smaller borrowers on average pay a higher AIS, obtain shorter-term loans, are more likely to rely on secured ﬁnancing, and have somewhat lower leverage than larger borrowers. In addition, smaller borrowers tend to be the loan customers of smaller banks; these small banks, in turn, tend to be better capitalized. Through their common dependence on borrower size, therefore, the AIS would appear to be an increasing function of bank capital. Detecting a link between bank ﬁnancial weakness and terms of lending will therefore require controls for borrower and bank characteristics.
III. Borrower Characteristics, Bank Characteristics, and the Cost of Funds Absent informational frictions, in a competitive loan market, the loan interest rate charged by a bank to a borrower should reﬂect the bank’s cost of funds and the risk characteristics of the borrower. Changes in borrower risk will affect the risk premium in the loan rate. Bank-speciﬁc increases in the cost of funds would not be passed on to loan customers in the absence of informational or competitive frictions; borrowers could simply switch banks. With informational frictions, this simple loan-pricing story changes in three ways.
First, borrower information costs and incentive problems may inﬂuence the cost of funds to the borrower. Second, to the extent that the bank-borrower relationship reduces information and incentive costs relative to other forms of ﬁnancing, borrowers face switching costs in changing lenders; hence an idiosyncratic increase in the bank’s cost of funds (say, from a decrease in capital or balance sheet liquidity) could increase the cost of funds to borrowers.