Professional Documents
Culture Documents
Jijun Niu∗
Beedie School of Business, Simon Fraser University, 8888 University Drive, Burnaby, BC, Canada V5A 1S6
Foos, Norden, & Weber, 2010). In such cases, faster loan growth
needs not lead to higher valuations.
articleinfo In this paper, we empirically examine the relation between loan
growth and bank valuations. Our sample contains quarterly
Article history: Received 21 May 2015 Received in revised form 24 January 2016 observations on a large number of publicly traded bank holding
Available online 10 February 2016 companies (“banks”) in the US from 2002:Q1 to 2013:Q4. We use
Tobin’s q as a measure of bank valuations (e.g., Laeven & Levine,
JEL classification: G21 2007). To measure loan growth, we use the quarterly growth rate
Keywords: Bank Loan growth Valuation of total loans and leases adjusted for unearned income (e.g.,
abstract Kupiec, Lee, & Rosenfeld, 2014).
This paper examines the relation between loan growth and bank
valuations. Using publicly-traded bank holding companies in the US from
2002:Q1 to 2013:Q4, we find that faster loan growth is associated with ∗ Tel.: +1 778 782 4491.
higher bank valuations. This finding holds both in normal times and
E-mail address: jijun niu@sfu.ca 1 These numbers are calculated using data from
during the financial crisis of 2007–2009. When we divide banks into
several size groups, we find that faster loan growth is associated with the Statistics on Depository Insti- tutions,
available on the Federal Deposit
higher valuations at small and medium banks, but not at large banks. Insurance Corporation website.
Further analyses show that large banks (1) have a lower ratio of loans to Using the full sample, we find a positive relation between loan
total earning assets, (2) have a higher ratio of nonperforming loans to total growth and bank valuations. This positive relation is economically
loans, and (3) are more likely to engage in securitization activity. These
large and statistically significant and holds after controlling for
characteristics help explain why the relation between loan growth and
bank valuations differs at small and large banks.
var- ious bank characteristics such as size, capital, asset
diversification, and profitability, as well as time fixed effects.
© 2016 Board of Trustees of the University of Illinois. Published by
Elsevier Inc. All rights reserved. We perform a variety of robustness checks. For example, we (1)
use alternative measures of valuations and loan growth, (2) con-
trol for bank fixed effects, and (3) use an instrumental variable
approach to address the concern that loan growth and bank valua-
tions could both be affected by an omitted variable. Throughout,
we find a positive relation between loan growth and bank
1. Introduction
valuations.
Loans are the main earning assets for banks, and the interest rates Calomiris and Nissim (2014) show that the relations between
many bank characteristics and bank valuations have changed dur-
on loans are usually higher than those on securities. For exam- ple,
ing the financial crisis of 2007–2009. To see whether the positive
loans accounted for 52% of all the assets held by US banks at
relation between loan growth and bank valuations holds under
year-end 2013. The average interest rate on loans was 4.7%, while
different market conditions, we run separate regressions for the
the average interest rate on securities held by banks was only
periods before, during, and after the financial crisis of 2007–2009.
2.3%1. One may thus conjecture that, if a bank is able to grow its
We find that faster loan growth is associated with higher
loan portfolio at a faster pace, its valuations will increase.
valuations in each period.
But loan growth can occur for a variety of reasons. For example, a
Small and large banks differ along many dimensions (see, e.g.,
bank seeking to grow its loan portfolio may maintain a liberal
Demsetz & Strahan, 1997; Berger & Udell, 2002; Berger &
credit policy: reducing collateral requirements, weakening
Bouwman, 2013; Bertay, Demirguc-Kunt, & Huizinga, 2013;
covenants, and providing loans to borrowers rejected by other
Zemel, 2015). To see whether the relation between loan growth
banks (see, e.g., Rajan, 1994; Dell’Ariccia & Marquez, 2006;
and bank valuations differs at small and large banks, we divide banks, but not at large banks.
banks in our sample into several size groups, and run regressions To understand the reasons behind this result, we first show that
separately for each group. Interestingly, we find that faster loan large banks differ from small and medium banks in the following
growth is associated with higher valuations at small and medium
http://dx.doi.org/10.1016/j.qref.2016.02.001 1062-9769/© 2016 Board of Trustees of the University of Illinois. Published by Elsevier Inc. All rights reserved.
The Quarterly Review of Economics and Finance 61 (2016) 185–191
losses several years later (e.g., Clair, 1992; Keeton, 1999; Salas & Saurina, 2002;
loan growth is a form of investment (Megginson et al., 1995;
Jimenez & Saurina, 2006; Hess et al., 2009; Foos et al., 2010; Amador et al., 2013).
Houston & James, 1998). Thus, the agency costs of free cash flow
J. Niu / The Quarterly Review of Economics and Finance 61 (2016) 185–191 187
banks tend to allocate a higher proportion of their assets to small
multiple interactions with the borrower, the borrower’s suppliers business loans (e.g., Berger, Kashyap, Scalise, Gertler, &
and customers, and other members in the local community. The Friedman, 1995; DeYoung, Hunter, & Udell, 2004; McNulty,
bank evaluates the profitability of servicing the borrower over time Murdock, & Richie, 2013). Such loans often require the use of soft
and across products (Boot, 2000; Boot & Thakor, 2000; Berger & information. Scott (2004) and Uchida, Udell, and Yamori (2012)
Udell, 2002). In contrast, transaction lending is mainly based on find that small banks are more active in producing soft
quantitative and verifiable information (e.g., financial ratios, credit information. Cole, Goldberg, and White (2004) find that large
scores), and each transaction is evaluated on a stand-alone basis. banks base loan approval decisions pri- marily on verifiable
Theory suggests that small banks have an advantage over large financial data of the borrowers, whereas small banks rely more on
banks in delivering relationship lending. Berger and Udell (2002) pre-existing relationships with the borrow- ers. Berger, Miller,
argue that a bank offering relationship lending must delegate more Petersen, Rajan, and Stein (2005) present several pieces of
lending authority to its loan officers, because they have greater evidence consistent with the notion that small banks are more
access to relevant information. This delegation of authority, how- likely to engage in relationship lending.
ever, creates agency problems within the bank. Small banks have
fewer layers of management, and thus are better able to deal with
these agency problems. Brickley, Linck, and Smith (2003) find 3. Sample and variables
that small banks have few branches and normally concentrate in
one local market area, and local office managers and investors We begin with a list of publicly-traded bank holding companies
from the local community own almost all the stock in such banks.
(“banks”) in the US4. For each bank on the list, we obtain
The authors argue that such an ownership structure provides small
quarterly accounting data from the Federal Reserve’s Y-9C
banks with an advantage in dealing with the agency problems
reports, and stock data from the Center for Research in Security
associated with relationship lending. Stein (2002) argues that the
Prices (CRSP). Our sample period goes from 2002:Q1 to 2013:Q4.
organizational structures of small banks strengthen the incentives
The sample starts in 2002:Q1 because several variables used in our
of loan officers to produce “soft” information that is necessary for
analyses became available in that quarter.
relationship lending3.
A bank can grow its loan portfolio mainly in two ways. First, the
A large body of literature finds that small banks engage more in bank can increase lending to existing borrowers or find new
relationship lending, whereas large banks engage more in trans- borrowers (“internal growth”). Second, the bank can engage in
action lending. For example, a number of studies find that small mergers and acquisitions (“external growth”). Previous studies
find that these two types of growth have different impact on bank before and after the merger or acquisition are included in the
performance (e.g., Clair, 1992; Foos et al., 2010). Because theories sample. Our final sample includes over 15,000 quarterly
(e.g., Rajan, 1994; Dell’Ariccia & Marquez, 2006) focus on observations on 632 banks.
internal growth, if a bank is involved in a merger or acquisition in Following the literature (e.g., Caprio, Laeven, & Levine, 2007;
a given quarter, the corresponding bank-quarter observation is Laeven & Levine, 2007), we use Tobin’s q as a measure of bank
excluded val- uations. Tobin’s q is calculated as the ratio of the market
value of equity plus the book value of liabilities to the book value
of assets. Later in robustness checks we employ an alternative
mea- sure of bank valuations. Following Kupiec et al. (2014), we
3
measure loan growth as the quarterly growth rate of total loans and
Stein (2002) defines soft information as “information that cannot be directly
verified by anyone other than the agent who produces it.” leases adjusted for unearned income. We employ an alternative
measure of loan growth in robustness checks.
4
The list is derived from the CRSP-FRB Link provided by the Federal Reserve
Laeven and Levine (2007) investigate the impact of diversifi-
Bank of
New York. cation on bank valuations. We use their empirical specification to
Table 1 Variable definitions.
select control variables. First, we control for bank size, measured
Variable Definition as the natural logarithm of total assets in thousands of constant
Tobin’s q The ratio of the market value of equity plus the December 2013 dollars6. Second, we control for the ratio of equity
book value of liabilities to the book value of assets Loan growth The quarterly to total assets. Third, we control for the ratio of net loans to total
growth rate of total loans and leases earning assets, where total earning assets include net loans, secu-
adjusted for unearned income Size The natural logarithm of total assets in rities, and investments. Fourth, we control for the ratio of total
thousands deposits to total liabilities. Finally, we control for asset diversity,
of constant December 2013 dollars Capital The ratio of equity to total assets Loans computed as 1 − |(net loans − other earning assets)/total earning
The ratio of net loans to total earning assets, where
assets|, where other earning assets include securities and invest-
total earning assets include net loans, securities, and investments Deposits The ratio
ments. Asset diversity takes value between zero and one, with
of total deposits to total liabilities Asset diversity 1 − |(net loans − other earning
assets)/total earning higher values indicating more diversification cross lending and
assets|, where other earning assets include securities and investments, and total non-lending activities.
earning assets include net loans, securities, and investments Market-to-book ratio Table 1 lists the definition of each variable used in this paper.
The ratio of the market value of equity to the book
Because most of the variables in our sample contain outliers, we
value of equity Abnormal loan growth A bank’s loan growth in a given quarter
minus the
median loan growth of all the banks in the same quarter Return on equity The ratio
of pre-tax profits to book value of equity Cost-income ratio The ratio of total
noninterest expense to total operating income Loan loss provision The ratio of loan 5
We identify mergers and acquisitions in the banking industry using the merger
loss provision to net interest files provided by the Federal Reserve Bank of Chicago.
income Nonperforming loans The ratio of nonperforming loans to total loans,
6
We convert dollar values to constant December 2013 dollars using the Consumer
where nonperforming loans are loans that are 90 days or more past due or have
Price Index obtained from the Bureau of Labor Statistics.
nonaccrual status Securitization An indicator variable that equals one if a bank
engages in securitization activity in a given quarter, and zero otherwise
Tobin’s q 1.043 0.068 0.993 1.037 1.085 17,533 Loan growth 0.017 0.043 −0.008 0.013 0.035 16,406 Size 14.680 1.510 13.622 14.307 15.337 17,533 Capital 0.094 0.026 0.077
0.091 0.106 17,533 Loans 0.741 0.141 0.671 0.765 0.839 17,510 Deposits 0.833 0.120 0.783 0.859 0.914 16,973 Asset diversity 0.485 0.225 0.315 0.461 0.644 17,510
Note: This table reports summary statistics for the main variables. Please see Table 1 for variable definitions.
Tobin’s q 1.0000 Loan growth 0.2593* 1.0000 Size 0.1093* −0.0312* 1.0000 Capital −0.0036 −0.0187 0.0934* 1.0000 Loans −0.1584* −0.0031 −0.2223* −0.0144 1.0000 Deposits
−0.0957* −0.0371* −0.4302* 0.0094 0.3694* 1.0000 Asset diversity 0.0971* −0.0370* 0.0782* −0.0191 −0.7751* −0.1072* 1.0000
Note: * indicates significance at the 1% level. Please see Table 1 for variable definitions.
assets is associated with lower valuations. The coefficients on
other control variables are insignificant.
winsorize all the continuous variables at the 1% and 99% levels to
mitigate the effect of outliers. We perform a series of robustness checks. In column (2) of Table
4, we use the market-to-book ratio as an alternative mea- sure of
Table 2 presents summary statistics for the main variables. Tobin’s
bank valuations. This ratio is computed as the market value of
q has a mean of 1.043, indicating that for an average bank in our
equity divided by the book value of equity (Caprio et al., 2007).
sample the market value of assets exceeds the book value of assets
The results indicate that loan growth is positively associated with
by 4.3%. Loan growth has a mean of 1.7%, which is compa- rable
the market-to-book ratio.
with the 1.4% reported by Kupiec et al. (2014). Loan growth
varies widely across banks and time, as indicated by the standard In column (3), we regress Tobin’s q on abnormal loan growth
deviation of 4.3%. On average, loans account for 74.1% of total instead of loan growth. Following the literature (e.g., Foos et al.,
earn- ing assets, and deposits account for 83.3% of total liabilities. 2010; Amador et al., 2013), abnormal loan growth is equal to a
Asset diversity has a mean of 0.485 with a standard deviation of bank’s loan growth in a given quarter minus the median loan
0.225, suggesting that banks in our sample differ considerably growth of all the banks in the same quarter. As shown, the results
with regard to the composition of their earning assets. This result are very similar to those reported in column (1).
is consistent with Laeven and Levine (2007). In column (4), we include additional controls in the regression.
Table 3 presents the correlation matrix of the main variables. We Return on equity is the ratio of pre-tax profits to book value of
note that the correlation between loan growth and Tobin’s q is equity. Cost-income ratio is the ratio of total noninterest expense
posi- tive and significant. This correlation result is informative but to total operating income. Loan loss provision is the ratio of loan
it does not control for other bank characteristics. We therefore loss provision to net interest income. We find that return on equity
estimate regressions. is positively associated with valuations, while cost-income ratio is
negatively associated with valuations. Importantly, the coefficient
on loan growth remains positive and significant.
4. Empirical results
In column (5), we include bank fixed effects in the regression to
control for unobserved, time-invariant differences across banks.
To examine the relation between loan growth and bank val- We continue to find that faster loan growth is associated with
uations, we regress Tobin’s q on loan growth and a set of bank higher valuations.
characteristics. We include quarter fixed effects to control for
Another concern about our results is that loan growth and bank
com- mon factors that affect the valuations of all the banks in a
valuations could both be affected by an omitted variable. If this is
given quarter. We estimate the regression using ordinary least
the case, OLS estimation will produce biased coefficient esti-
squares (OLS). Because the observations on a specific bank over
mates. To address this concern, we use an instrumental variable
time are not independent, the standard errors are clustered at the
approach. Specifically, we use deposit growth as an instrument for
bank level.
loan growth. The ex ante rationale for the use of the instrument is
The baseline regression results are presented in column (1) of as follows. On the one hand, theories (e.g., Kashyap, Rajan, &
Table 4. The coefficient on loan growth is positive and significant Stein, 2002; Song & Thakor, 2007) suggest a strong link between
at the 1% level. This result shows that faster loan growth is lend- ing and deposit-taking (instrument relevance). On the other
associated with higher valuations. The economic magnitude of the hand, we are not aware of any theory that suggests that deposit
coefficient is also significant. A one percentage point increase in growth directly affects bank valuations (instrument exogeneity).
loan growth is associated with an increase in Tobin’s q of 0.118,
Column (6) reports the second-stage regression results7. The
which amounts to 11% of the sample mean of Tobin’s q.
instrumental
Turning to control variables, we find that size is positively asso-
ciated with bank valuations. As Laeven and Levine (2007) point
out, size often affects valuations through economies of scale.
Capital is also positively associated with valuations, which is
In the first-stage regression (not reported), deposit growth enters signifi- cantly
7
at
consistent with the theory of Mehran and Thakor (2011). The the 1% level. Because the model is exactly identified (i.e., the number of
coefficient on loans is negative, indicating that a larger share of
loans in total earning
J. Niu / The Quarterly Review of Economics and Finance 61 (2016) 185–191 189
Table 4 Loan growth and bank valuations.(1) Baseline model (2) Market-to-book
ratio
(3) Abnormal loan growth
(4) Additional controls
(5) Bank fixed effects
(6) Instrumental variable
Loan growth 0.118*** (0.023) 1.820*** (0.251) 0.057*** (0.022) 0.045*** (0.012) 0.050** (0.023) Abnormal loan growth 0.117*** (0.023) Size 0.008*** (0.002) 0.101*** (0.017) 0.008***
(0.002) 0.007*** (0.002) −0.018*** (0.005) −0.018*** (0.005) Capital 0.188* (0.097) −1.185 (0.851) 0.188* (0.097) 0.090 (0.102) −0.266*** (0.057) −0.267*** (0.057) Loans −0.079*
(0.048) −0.808* (0.445) −0.079* (0.047) −0.081* (0.048) 0.022 (0.038) 0.025 (0.036) Deposits 0.046 (0.029) 0.620** (0.276) 0.046 (0.029) 0.045 (0.028) 0.023 (0.023) 0.024 (0.022)
Asset diversity −0.029 (0.026) −0.248 (0.247) −0.029 (0.026) −0.033 (0.026) 0.014 (0.020) 0.015 (0.019) Return on equity 0.183*** (0.042) 0.054*** (0.019) 0.054*** (0.019)
Cost-income ratio −0.024*** (0.009) −0.012*** (0.005) −0.012*** (0.005) Loan loss provision 0.006 (0.006) −0.009*** (0.003) −0.009*** (0.003) Constant 0.975*** (0.062) 0.695 (0.584)
0.977*** (0.062) 1.014*** (0.060) 1.317*** (0.080) – Quarter fixed effects Yes Yes Yes Yes Yes Yes Bank fixed effects No No No No Yes Yes Observations 15,887 15,887 15,887
15,886 15,886 15,823 F-test of excluded
instruments (p-value)
0.000
R-squared 0.454 0.482 0.454 0.488 0.663 0.663
Note: The dependent variable is Tobin’s q in every column except column (2) where the dependent variable is the market-to-book ratio. Standard errors are clustered at the bank level
and reported in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively. Please see Table 1 for variable definitions.
Table 5 Different sample periods.
(1) Before the crisis
Table 6 Sample split by size.
(2) During the
(3) After the
(1) Small banks (2) Medium crisis
crisis
banks
Loan growth 0.070*** (0.024) 0.210*** (0.039) 0.154*** (0.050) Size 0.013*** (0.002) 0.001 (0.002) 0.004* (0.002) Capital 0.219* (0.114) 0.072 (0.138) 0.169 (0.112) Loans −0.091**
(0.038)
(3) Large banks
Loan growth 0.102*** (0.023) 0.159*** (0.031) 0.035 (0.048) Size 0.015* (0.008) 0.013*** (0.004)
−0.008 (0.005)
−0.131* (0.076) −0.038 (0.073)
Capital 0.071 (0.105) 0.140 (0.109) 0.494* (0.256) Loans −0.042 (0.038) −0.079 (0.068) −0.075* (0.045) Deposits 0.114*** (0.028) −0.007 (0.047) −0.040 (0.051)
Deposits 0.110*** (0.035) 0.043 (0.040) −0.034 (0.061) Asset diversity −0.052**
Asset diversity −0.020 (0.019) −0.018 (0.037) −0.032 (0.023) (0.021)
Constant 0.784*** (0.113) 0.913*** (0.096) 1.313*** (0.117) Quarter fixed effects Yes Yes Yes Observations 5210 8331 2346 R-squared 0.501 0.485 0.469
Note: The dependent variable is Tobin’s q. The full sample is divided into three groups: small banks (total assets up to $1 billion), medium banks (total assets exceeding $1 billion and
up to $10 billion), and large banks (total assets exceeding $10 billion). Standard errors are clustered at the bank level and reported in paren- theses. ***, **
, and * indicate significance at
the 1%, 5%, and 10% level, respectively. Please see Table
1 for variable definitions.
and Levine (2007), who find a diversification discount in financial conglomerates.
Small and large banks differ along many dimensions (see, e.g., Demsetz & Strahan, 1997; Berger & Udell, 2002; Berger & Bouwman, 2013;
Bertay et al., 2013; Zemel, 2015). To see whether the relation between loan growth and bank valuations differs at small and large banks, we divide
banks in our sample into three groups: small banks (total assets up to $1 billion), medium banks (total assets exceed- ing $1 billion and up to $10
billion), and large banks (total assets exceeding $10 billion). We then run regressions separately for each group.
Columns (1) to (3) of Table 6 present the regression results for small, medium, and large banks, respectively. The results indi- cate that faster loan
growth is associated with higher valuations at small and medium banks. In contrast, there is no such asso- ciation at large banks8. Interestingly,
these results are consistent with Zemel (2015), who finds that loan growth, in conjunction with
8
Our results are robust to how we define large banks. For example, we obtain qualitatively
similar results when we define large banks as those that have total assets exceeding $5
billion, or exceeding $50 billion. −0.007 (0.041) −0.006 (0.039)
Constant 0.860*** (0.047) 1.142*** (0.097) 0.983*** (0.092) Quarter fixed effects Yes Yes Yes Observations 8184 3287 4416 R-squared 0.177 0.270 0.175
Note: The dependent variable is Tobin’s q. The sample period is from 2002:Q1 to 2013:Q4. It is divided into three periods: before the crisis (2002:Q1–2007:Q2), dur- ing the crisis
(2007:Q3–2009:Q4), and after the crisis (2010:Q1–2013:Q4). Standard errors are clustered at the bank level and reported in parentheses. ***, **, and * indicate significance at the 1%,
5%, and 10% level, respectively. Please see Table 1 for variable definitions.
variable results confirm that faster loan growth is associated with higher valuations.
Calomiris and Nissim (2014) find that the relations between many bank characteristics and bank valuations have changed dur- ing the financial
crisis of 2007–2009. To see whether the positive relation between loan growth and valuations holds under differ- ent market conditions, we divide
the sample period into three periods: before the crisis (2002:Q1–2007:Q2), during the crisis (2007:Q3–2009:Q4), and after the crisis
(2010:Q1–2013:Q4). We then run regressions separately for each period. Our definition of the crisis period follows Berger and Bouwman (2013),
and the results are robust to alternative definitions.
Columns (1) to (3) of Table 5 present the regression results for the periods before, during, and after the crisis, respectively. The results indicate
that faster loan growth is associated with higher valuations in each of the three periods. Turning to control vari- ables, we find that asset diversity
is negatively associated with valuations before the crisis. This result is consistent with Laeven
instruments equals the number of endogenous regressors), we cannot perform a test of overidentifying restrictions.
190 J. Niu / The Quarterly Review of Economics and Finance 61 (2016) 185–191
Table 7 Comparison of banks of different size.
(1) Small banks Mean
(2) Medium banks
(5) Medium vs. Mean
large Difference in mean
Loans 0.758 0.750 0.675 0.084*** 0.075*** Nonperforming loans 0.016 0.019 0.020 −0.004*** −0.001** Securitization 0.031 0.063 0.470 −0.439*** −0.407***
Note: The first three columns report the mean values of selected variables for small, medium, and large banks. The last two columns test for significant differences in means using the
unequal variances t-test. Small banks have total assets up to $1 billion, medium banks have total assets exceeding $1 billion and up to $10 billion, and large banks have total assets
Note: The dependent variable is Tobin’s q. Standard errors are clustered at the bank level and reported in parentheses. ***, **
, and * indicate significance at the 1%, 5%, and 10% level,