Bank Capital and Lending Relationships

• Date: 01 April 2018
• Author: MICHAEL SCHWERT
• Published date: 01 April 2018
• DOI: http://doi.org/10.1111/jofi.12604

THE JOURNAL OF FINANCE •VOL. LXXIII, NO. 2 •APRIL 2018

Bank Capital and Lending Relationships

MICHAEL SCHWERT∗

ABSTRACT

This paper investigates the mechanisms behind the matching of banks and firms in

the loan market and the implications of this matching for lending relationships, bank

capital, and credit provision. I find that bank-dependent firms borrow from well-

capitalized banks, while firms with access to the bond market borrow from banks

with less capital. This matching of bank-dependent firms with stable banks smooths

cyclicality in aggregate credit provision and mitigates the effects of bank shocks on

the real economy.

BANKING RELATIONSHIPS ARE AN IMPORTANT SOURCE of external finance for many

firms. An extensive literature highlights the effects of banking relationships on

firms’ cost of capital and nonfinancial outcomes.1An important segment of this

literature shows that informational frictions make it costly for firms to switch

lenders, leading to a “credit channel” that transmits shocks from banks to their

borrowers.2While the effects of banking relationships are well understood,

little evidence exists on the determinants of these relationships. What drives

a firm to borrow from one bank instead of another? Do banks specialize in

lending to certain types of borrowers? And if endogenous matching occurs in

the loan market, what are the consequences for credit provision? I address

these questions in this paper.

∗Michael Schwert is at Fisher College of Business, The Ohio State University. This paper is

based on a chapter of my PhD dissertation at Stanford GSB. I am grateful to Ilya Strebulaev for

our frequent discussions and his thoughtful advice. I also thank Michael Roberts (the Editor),

an anonymous referee, Anat Admati, Shai Bernstein, Harry DeAngelo, Darrell Duffie, Isil Erel,

Marco Giacoletti, Will Gornall, Larry Harris, Dirk Jenter, Stephen Karolyi,Peter Koudijs, Arvind

Krishnamurthy, Mike Ostrovsky, Francisco Perez-Gonzalez, Mitchell Petersen, Josh Rauh, Bill

Schwert, Amit Seru, Ren´

e Stulz, Victoria Vanasco, Victor Westrupp, and seminar participants at

Carnegie Mellon, Harvard Business School, Northwestern, Ohio State, Stanford GSB, University of

Southern California, University of Washington, WashingtonUniversity in St. Louis, and Wharton

for helpful comments and suggestions. I have read the Journal of Finance’s disclosure policy and

have no conflicts of interest to disclose.

1Early theories on this subject include Leland and Pyle (1977), Diamond (1984,1991), Sharpe

(1990), and Rajan (1992). For empirical work, see James (1987), Petersen and Rajan (1995), Berger

and Udell (1995), Berlin and Mester (1999), Hubbard, Kuttner, and Palia (2002), Berger et al.

(2005), Bharath et al. (2007), Schenone (2010), and Bolton et al. (2016).

2Theoretical work in this area includes Bernanke (1983) and Holmstrom and Tirole (1997). For

empirical evidence, see Slovin, Sushka, and Polonchek (1993), Gertler and Gilchrist (1994), Kang

and Stulz (2000), Khwaja and Mian (2008), Leary (2009), Chava and Purnanandam (2011), Lin

and Paravisini (2011), Chernenko and Sunderam (2014), and Chodorow-Reich (2014).

DOI: 10.1111/jofi.12604

787

788 The Journal of Finance R

I begin by showing that bank-dependent firms borrow from well-capitalized

banks, while firms with access to the public debt markets borrow from banks

with less equity capital. Firms without a public debt rating, which I define

as bank-dependent, borrow from banks that have 1.8% higher equity capital

than banks that lend to rated firms. This difference is economically significant,

equivalent to 14% of the mean and 27% of the standard deviation of the bank

capital ratio. This sorting is evident nearly every year between 1987 and 2012

and is robust to the empirical approach used to identify it. My paper is the first

to uncover this robust stylized fact on the matching of banks and firms.

This endogenous matching has important implications for the effects of bank

shocks on the real economy and the overall cyclicality of credit provision. In-

formational frictions prevent bank-dependent borrowers from offsetting re-

ductions in loan supply from their relationship banks. Bank health is closely

related to lending activity, with well-capitalized banks better able to absorb

shocks.3Figure 1presents a concrete example, from the 2008 financial crisis,

of the association between bank capital and lending activity following a macroe-

conomic shock. Consider two salient points on this plot, JPMorgan Chase (JPM)

and Wells Fargo (WFC). JPMorgan Chase had a market equity ratio of 11.9%

prior to the crisis and cut lending by 75% during the crisis, while Wells Fargo

had a market equity ratio of 21.6% and cut lending by 48%. This anecdotal ev-

idence suggests that obtaining credit from Wells Fargo would have been easier

than obtaining credit from JPMorgan Chase during the financial crisis.

The intuitive takeaway from this plot is that bank-dependent firms benefit

from having relationships with better capitalized banks, like Wells Fargo, that

cut lending less in a downturn.4Based on a counterfactual exercise from my

empirical matching model, I find that, in the period surrounding the financial

crisis, bank-dependent firms faced 6.6% less loan supply shrinkage from their

precrisis relationship banks relative to the reverse matching assignment, that

is, bank-dependent firms borrowing from low-capital banks. This represents a

significant improvement in these firms’ position in the cross section of banks,

equivalent to 42% of the standard deviation of bank lending growth during the

crisis. The implication of this finding is that matching mitigates the impact

of financial sector shocks on the real economy. Prior findings on the effects

of bank shocks on firms’ ability to invest in production (Gertler and Gilchrist

(1994)) and employment (Chodorow-Reich (2014)) would be even stronger if

banks were homogeneous or the matching of firms and banks were random.

3Empirical evidence on the association between bank capitalization and lending activity in-

cludes Gambacorta and Mistrulli (2004),Cornettetal.(2011), Beltratti and Stulz (2012), Berger

and Bouwman (2013), and Carlson, Shan, and Warusawitharana (2013).

4Table IX provides evidence that cross-sectional differences in lending cutbacks in the crisis

were driven by supply at the bank level rather than borrower demand. The Internet Appendix,

which is available in the online version of the article on the Journal of Finance website, shows that

capital is the strongest predictor of lending reductions in the sample of commercial banks, while

other bank characteristics, such as exposure to mortgage-backed securities and deposit funding,

are poor predictors.

Bank Capital and Lending Relationships 789

Figure 1. Changes in loan volume around the financial crisis as a function of bank

capital. This figure reports percentage changes in loan volume between the periods October 2004

to June 2007 and October 2008 to June 2009 for each bank in my sample, plotted against the bank’s

capital in the fourth quarter of 2006. The precrisis period excludes the third quarter of each year,

to avoid issues with seasonality. The sample includes only banks that lent at least $100 million

in the precrisis period. Total volume is annualized for comparability across periods of different

length. Loan volume is measured as the dollar amount of new loan commitments by each bank to

firms in the sample. The commitment amount is the product of the facility amount and the bank’s

allocation, as either lead arranger or participant. When the bank’s allocation is missing in the data,

it is estimated as the fitted value from a Tobit regression of bank allocation on log loan amount,

the ratio of loan amount to lender assets, the ratio of loan amount to borrower assets, the number

of lead arrangers, the number of participants, and quarter fixed effects. Bank capital is defined as

the market equity ratio. The regression line has β=3.55, t=4.20, and R2=0.54. (Color figure

can be viewed at wileyonlinelibrary.com)

The matching of bank-dependent firms with well-capitalized banks is also

beneficial in the aggregate because firms with access to public debt markets

can partially offset reductions in loan supply by issuing bonds. Figure 2shows

that the borrowers of poorly capitalized banks were substantially more likely to

issue bonds in the period immediately following the financial crisis. This result

extends Becker and Ivashina (2014), who find evidence of aggregate substitu-

tion from loans to bonds during this period, by showing that such substitution

concentrated among borrowers of banks that reduced lending more during the

crisis.5If the matching of firms and banks were such that more firms with

access to public debt markets had relationships with well-capitalized banks,

5The cluster of points at zero in Figure 2indicates that there are two distinct types of lending

clientele in the sample. Specifically, the banks in the cluster at zero are smaller and have fewer

borrowers with bond market access than the banks whose borrowers issued bonds during the crisis.