Real Activity Forecasts Using Loan Portfolio Information

• Published date: 01 June 2016
• DOI: http://doi.org/10.1111/1475-679X.12110
• Date: 01 June 2016

DOI: 10.1111/1475-679X.12110

Journal of Accounting Research

Vol. 54 No. 3 June 2016

Printed in U.S.A.

Real Activity Forecasts Using Loan

Portfolio Information

UROOJ KHAN

∗AND N. BUGRA OZEL

†

Received 2 September 2014; accepted 28 January 2016

ABSTRACT

Toextend and monitor loans, banks collect detailed and proprietar y informa-

tion about the financial prospects of their customers, many of whom are local

businesses and households. Therefore, banks’ loan portfolios contain poten-

tially useful information about local economic conditions. We investigate the

association between information in loan portfolios and local economic condi-

tions. Using a sample of U.S. commercial banks from 1990:Q1 to 2013:Q4, we

document that information in loan portfolios aggregated to the state level is

associated with current and future changes in statewide economic conditions.

Furthermore, the provision for loan and lease losses contains information in-

cremental to leading indicators of state-level economic activity and recessions.

Loan portfolio information also helps to improve predictions of economic

conditions at more granular levels, such as at the commuting zone level. We

discuss the relevance of these findings for economic analysis and forecasting,

∗Columbia Business School, Columbia University; †N. Jindal School of Management, Uni-

versity of Texas at Dallas.

Accepted by Douglas Skinner. We thank an anonymous referee, David Aboody,

B. W. Baer, Charles Calomiris, Judson Caskey, John Donaldson, Itay Goldstein (discussant),

Umit Gurun, Trevor Harris, Gur Huberman, Jack Hughes, Hanno Lustig, Sharon Katz, Sta-

nimir Markov, Emi Nakamura, Suresh Nallareddy, Doron Nissim, Stephen Penman, Shiva

Rajgopal, Dushyant Vyas, participants of the 14th Annual FDIC Bank Research Conference,

HKUST Accounting Symposium, Burton Workshop at Columbia University, and account-

ing seminars at the University of Texas at Dallas, and Claremont McKenna for useful dis-

cussions and helpful comments. We are grateful to Jinhwan Kim and Sangsoon Koh for

excellent research assistance. An Internet Appendix to this paper can be downloaded at

http://research.chicagobooth.edu/arc/journal-of-accounting-research/online-supplements.

895

Copyright C, University of Chicago on behalf of the Accounting Research Center,2016

896 U.KHAN AND N.B.OZEL

and the relation of our study to prior work on the informativeness of account-

ing information about the macroeconomy.

JEL codes: D82; E02; E32; G01; G21; M41

Keywords: banking; coincident index; economic growth; forecasting; loans

1. Introduction

Economics literature has long recognized the link between the banking

system and economic development (Bagehot [1873], Schumpeter [1912]).

Many studies have examined the informativeness of the level of financial

intermediation about productivity at the national (e.g., King and Levine

[1993], Levine and Zervos [1998]) and state-level economies (Samolyk

[1994], Driscoll [2004]). In deciding which opportunities to fund, banks

collect detailed, and often proprietary, information about their borrow-

ers and monitor them until the loan is repaid. The information about

the health and prospects of borrowers is summarized in banks’ disclosures

about loan portfolios. Since bank lending is concentrated in local markets,

by aggregating information about the financial condition of their borrow-

ers, banks can provide insights about local economic activity. Weinvestigate

this conjecture and find that banks’ loan portfolio information is predictive

of state-level economic growth. Importantly, the information embedded in

the provision for loan and lease losses is incremental to information con-

tained in other leading state-level economic indicators used for forecasting

economic growth.

Banks collect a variety of information about their borrowers throughout

the lending relationships. This includes “hard” information that borrowers

provide, such as financial statements and tax returns, hard information that

banks gather, such as data about borrowers’ supply chains and account ac-

tivities, and “soft” information that banks gather through interactions with

borrowers and the local community, such as credibility and commitment of

the borrowers. Arguably, most of the information banks collect is not eas-

ily accessible by the public because the typical borrower is an individual or

a privately held firm. For example, Minnis and Sutherland [2015] report

that banks receive nonpublic tax return data from businesses they lend

to. Relatedly, Norden and Weber [2010] document that banks use confi-

dential information about credit line usage, limit violations, and checking

account activities to adjust lending terms and loan loss provisioning. More-

over, banks use soft information, which is inherently hard to measure objec-

tively and is not readily publicly available, in conjunction with hard infor-

mation to better evaluate and monitor borrowers (e.g., Petersen and Rajan

[1994], Grunert, Norden, and Weber [2005]).

Bank lending, by its nature, is geographically segmented into local mar-

kets as loan monitoring costs increase with distance from the borrower

(e.g., Laderman, Schmidt, and Zimmerman [1991], Morgan and Samolyk

[2003]). Petersen and Rajan [2002] point out the local nature of soft

REAL ACTIVITY FORECASTS USING LOAN PORTFOLIO INFORMATION 897

information necessary for credit decisions. Along the same lines, Agarwal

and Hauswald [2010] find that, in small business lending, a lender’s ability

to collect proprietary intelligence erodes with its distance from the bor-

rower. Thus, to the extent that the hard and soft information banks collect

shapes loan portfolios, the condition of loan portfolios can provide insights

about local economic activity by aggregating information about borrowers.

Accordingly, we predict that information in loan portfolios will be associ-

ated with contemporaneous and future growth in local economies. We also

predict that, because banks rely significantly on nonpublic information in

lending decisions, information in loan portfolios would be incremental to

leading economic indicators—which are typically based on publicly avail-

able hard information—in predicting local economic activity.

We define local economies as U.S. states since most banks operate in a

single state (Morgan and Samolyk [2003]).1At the same time, understand-

ing and forecasting economic trends in U.S. states is important because

they have a bearing on a wide array of matters ranging from corporate de-

cisions about the location of manufacturing units (Bartik [1985]) to the

outcome of political elections (Niemi, Stanley, and Vogel [1995]) and tax

policies (Cornia and Nelson [2010]). Moreover, because of the costs asso-

ciated with collecting and processing data, most economic indicators are

estimated, rather than measured directly, beyond the state level.2

We focus on three aspects of loan portfolios that can impound banks’ ex-

pectations regarding their borrowers’ financial prospects: estimated credit

losses, the risk premium on loans, and loan growth. Estimated credit losses

reflect information regarding borrowers’ repayment ability. Prior research

shows that banks are able to assess credit losses ahead of their realization to

some extent (e.g., Beatty and Liao [2011], Balasubramanyan, Zaman, and

Thomson [2014], Harris, Khan, and Nissim [2015]) and that banks real-

ize credit losses at higher rates during economic downturns (Laeven and

Majnoni [2003], Harris, Khan, and Nissim [2015]).3Therefore, higher esti-

mated credit losses can be indicative of weaker future economic growth. We

measure expected credit losses using the provision for loan and lease losses

and the change in nonperforming loans. Banks also price protect against

greater credit risk by charging a higher interest rate on loans (Morgan and

Ashcraft [2003], Harris, Khan, and Nissim [2015]). Therefore, changes in

the risk premium on loans can be indicative of trends in future economic

activity. Finally, banks are more willing to extend loans during expansionary

1In our sample, the average commercial bank collects over 94% of its deposits from a single

state.

2See Meyer and Yeager [2001] for a discussion of issues in the estimation of labor and

income at the county level.

3Prior research provides some evidence consistent with bank managers using loan quality

metrics for earnings, capital, and tax management purposes, which may reduce the association

between loan quality metrics of the manipulating bank and the economic activity. However,

such manipulations are idiosyncratic in nature and, when aggregated in a sufficiently large

sample, they should average out.