Description
The purpose of this paper is to provide the first empirical study of external economies
(agglomeration economies) in the banking industry.
Journal of Financial Economic Policy
External economies in banking
Sherrill Shaffer
Article information:
To cite this document:
Sherrill Shaffer, (2012),"External economies in banking", J ournal of Financial Economic Policy, Vol. 4 Iss 4
pp. 354 - 365
Permanent link to this document:
http://dx.doi.org/10.1108/17576381211279316
Downloaded on: 24 January 2016, At: 21:45 (PT)
References: this document contains references to 45 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 233 times since 2012*
Users who downloaded this article also downloaded:
Wenling Lu, David A. Whidbee, (2013),"Bank structure and failure during the financial crisis", J ournal of
Financial Economic Policy, Vol. 5 Iss 3 pp. 281-299 http://dx.doi.org/10.1108/J FEP-02-2013-0006
Eric Osei-Assibey, Baimba Augustine Bockarie, (2013),"Bank risks, capital and loan supply: evidence from
Sierra Leone", J ournal of Financial Economic Policy, Vol. 5 Iss 3 pp. 256-271 http://dx.doi.org/10.1108/
J FEP-09-2012-0041
J ames R. Barth, Gerard Caprio, Ross Levine, (2013),"Bank regulation and supervision in 180
countries from 1999 to 2011", J ournal of Financial Economic Policy, Vol. 5 Iss 2 pp. 111-219 http://
dx.doi.org/10.1108/17576381311329661
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
External economies in banking
Sherrill Shaffer
Department of Economics and Finance, University of Wyoming,
Laramie, Wyoming, USA
Abstract
Purpose – The purpose of this paper is to provide the ?rst empirical study of external economies
(agglomeration economies) in the banking industry.
Design/methodology/approach – The author extends a standard speci?cation of banking costs to
control for community and market characteristics.
Findings – Banks’ costs are a decreasing function of the number of rival banks and an increasing
function of market population. Estimated magnitudes of these effects, modest at the bank level, are
large in aggregate. Multimarket operation of rival banks is also important.
Originality/value – These ?ndings suggest a previously unrecognized cost-side bene?t of
structure-based antitrust policies, and have additional implications for public policy toward
banking structure, as well as calling for a re-interpretation of previous studies of scale economies, cost
ef?ciency, and price-cost margins in banking.
Keywords Agglomeration, Localization, Urbanization, Banking, Antitrust, Costs
Paper type Research paper
1. Introduction
It has long been recognized that productivity and costs may vary systematically not
only as a function of the technology used by the individual ?rm but also as a function of
the size and other characteristics of the community in which the ?rmoperates. The latter
effects have been termed “agglomeration economies” or external economies, which can
be further decomposed into “urbanization economies” (the size of the city itself) and
“localization economies” (the size of the ?rm’s industry within the city). This paper
presents an exploratory test of economies of agglomeration in the US commercial
banking industry, using ?rm-level data from a nationwide sample of single-market US
banks operating inmetropolitan areas. Afurther contribution is to point out a previously
unrecognized policy implication of agglomeration economies: structure-based antitrust
regulations can potentially improve welfare by limiting excessive consolidation that
raises aggregate costs.
Although the empirical model differs only in straightforward ways from many
previous studies of banking costs, the results shed important new light on a number of
issues. Economies of agglomeration have been studied previously, but only at the
aggregate level or within the manufacturing sector (see Eberts and McMillen, 1998, for
a review). By contrast, this study uses micro-level data from a single major service
industry, one in which community characteristics may affect the ?rm’s cost structure
in at least two direct ways. One commonly recognized source of external economies is
information spillovers (Glaeser, 1998). Information about ?nancial risk is fundamental
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
JEL classi?cation – G21, L10
The author is grateful for helpful comments on earlier versions of this paper from Allen Berger,
Robert Collender, Beth Cooperman, Bob Hunt, Ken Kopecky, James McAndrews and Tara Rice.
JFEP
4,4
354
Journal of Financial Economic Policy
Vol. 4 No. 4, 2012
pp. 354-365
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381211279316
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
to the banking industry, suggesting that spillovers of such information may have
signi?cant cost effects[1]. A novel aspect here is that such spillovers among banks may
be either positive, reducing costs as discussed in previous non-banking agglomeration
studies; or negative, as discussed for banks by Broecker (1990), Nakamura (1993),
Riordan (1993), Shaffer (1998) and Marquez (2002), though without previously being
linked to external economies[2]. In addition, access to a pool of employees with
specialized skill has also been recognized as a possible source of external economies.
This factor likely applies to the banking industry and is consistent with observed
concentrations of banks in Switzerland or of bank headquarters in Frankfurt.
1.1 Previous studies
Empirical evidence of agglomeration economies has been found for urban
manufacturing industries using data as early as the 1950s. Carlino (1979) estimates a
decomposition of overall economies of scale into internal economies, localization
economies, and urbanization economies. Urbanization economies are found in aggregate
by Carlino (1982) and Moomaw(1983), while Henderson (1986) and Moomaw(1988) ?nd
that localization economies are dominant in urban manufacturing. Nakamura (1985)
?nds that urbanization economies are more important for ?rms in light industries but
localization economies are more important for ?rms in heavy industries. Fogarty and
Garofalo (1988) and Calem and Carlino (1991) estimate agglomeration economies in
urban manufacturing in the presence of technical change.
These studies focus on urban manufacturing rather than service industries, and
have typically been hampered by the level of detail of available data. Similar studies by
Beeson (1987) and Beeson and Husted (1989) used state-level data to circumvent this
problem. By contrast, the banking industry is an important service industry with
abundant ?rm-level data, and as such is an ideal candidate for extending the empirical
literature on agglomeration economies.
The cost structure of the banking industry has been extensively studied over the past
30 years. A few studies have explored regional differences in the cost structure of US
banks (Evanoff and Israilevich, 1991; Hamid and Verma, 1994). To the author’s
knowledge, no previously published study has attempted to distinguish between internal
and external sources of economies or diseconomies of scale in the banking industry.
2. The empirical model
I estimate a standard empirical banking cost model to facilitate comparison with
previous studies. Prior research and common regulatory practice alike have used
metropolitan statistical areas (MSAs) (or counties in rural areas) as a standard proxy
for geographic banking markets (Dick, 2008; Ho and Ishii, 2011). If a measurable
localization cost effect can be observed within a typical MSA, any additional network
effects involving broader regions would only serve to strengthen the overall impact of
inter?rm interactions on banks’ costs.
The de?nition of inputs and outputs follows the intermediation model of Klein (1971)
and Sealey and Lindley (1977), in which labor and capital are inputs used both to
obtain deposits and, in conjunction with deposits, to originate loans and other earning
assets. The interest rate paid on deposits is an input price, as is the average wage rate
paid to bank employees; bank output is measured as the dollar value of earning assets,
decomposed into:
External
economies
in banking
355
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
.
loans and leases; and
.
securities.
Fee income is treated as an additional output, re?ecting the growing importance of
off-balance-sheet activities for banking in recent years; alternate estimates omitting fee
income (not reported) did not materially change the signs, signi?cance, or magnitudes
of the various agglomeration measures. This model is supported by empirical evidence
suggesting that aggregate deposits behave primarily as inputs (Gilligan and Smirlock,
1984; Hughes and Mester, 1993) and has been widely employed in previous studies of
banking costs (Berger and Mester, 1997). A ?ner disaggregation of outputs could be
made, but at the expense of more severe multicollinearity[3]. Alternate dependent
variables are total costs, interest expenses, and noninterest expenses.
Agglomeration effects are measured by a vector of ?ve variables. Localization effects
are captured by the number of banks operating in the local market, as reported in the
FDIC’s summary of deposits. Aseparate aspect of localization is measured by the ratio of
single-market banks to total banks inthe market[4]. Market size and congestion effects are
incorporated through the natural logarithmof the market’s population. Asecondaspect of
congestion is measured by the market’s population density per square mile, following
Andersson et al. (2004), Carlino et al. (2005) and Strumsky et al. (2005). The quality of labor
force education, a measure of cumulative human capital, is measured as the percentage of
population above 25 years of age that had completed four years of college as of 2000. This
control variable is commonly included in empirical studies of agglomeration effects.
The sample is drawn from all US commercial banks operating exclusively within a
single MSA as of year-end 2000. As in many previous studies, banks less than ?ve
years old are excluded from the sample because such banks typically have portfolio
compositions and cost structures quite different from those of more mature banks. The
sample year is chosen to coincide with the availability of contemporaneous
demographic (census) data in a non-crisis year. Previous research has found that the
choice of year does not affect the estimates of scale economies in banking (Humphrey,
1990). Table I summarizes the sample statistics, drawn from a ?nal set of 1,075
commercial banks.
The functional form is the translog, perhaps the most widely used form for
empirical cost studies in recent years. The translog cost function, augmented by
agglomeration and market structure terms, is de?ned as:
ln C ¼ a
o
þ
X
4
i¼1
a
i
Q
i
þ
X
3
j¼1
b
j
P
j
þ
1
2
X
4
i¼1
X
4
k¼1
d
ik
Q
i
Q
k
þ
1
2
X
3
j¼1
X
3
k¼1
g
jk
P
j
P
k
þ
X
4
i¼1
X
3
j¼1
r
ij
Q
i
P
j
þ
X
3
h¼1
h
h
X
h
ð1Þ
where:
C ¼ total costs, including both interest and noninterest expenses (alternate
speci?cations replace this variable with interest and noninterest expenses,
respectively).
Q
i
¼ ln q
i
: quantity of output i:
JFEP
4,4
356
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
q
1
¼ total loans and leases.
q
2
¼ securities plus federal funds sold.
q
3
¼ fee income.
P
j
¼ ln p
j
: price of input j:
p
1
¼ price of labor, calculated as annual wage and bene?t expenses per employee.
p
2
¼ price of funding, calculated as average interest costs per dollar of all deposits.
p
3
¼ price of physical capital, calculated as annual expenses on premises and
equipment divided by the stock of these items.
x
1
¼ number of banks in the MSA.
x
2
¼ fraction of banks in the MSA that do not have of?ces in other markets.
x
3
¼ natural logarithm of population of MSA.
x
4
¼ population density per square mile of MSA.
x
5
¼ percentage of adult population completing four years of college as of 2000.
The usual restrictions of symmetry and linear homogeneity in factor prices are
imposed:
d
ik
¼d
ki
; g
jk
¼g
kj
;
X
b
j
¼1;
j
X
r
ij
¼0 for all i; and
j
X
g
jk
¼0 for all k:
Estimates are ?tted by OLS with heteroscedastic-consistent standard errors
(White, 1980). A Lagrange multiplier test failed to reject the null of no
Variable Mean SD
log (total cost) 10.10 1.29
log (interest expense) 9.40 1.37
log (noninterest expense) 9.35 1.26
Number of banks 64.02 47.01
% 1-MSA banks 0.451 0.149
log (population) 13.62 1.10
Population density 135.8 87.3
Education 25.78 6.18
log (total loans) 11.01 1.12
log (total securities) 9.90 1.21
log (noninterest income) 6.52 1.36
average deposit interest rate 0.036 0.008
average wage rate 42.44 11.92
price of physical capital 0.388 1.187
Source: Bank-level data from regulatory Call Reports; number of banks in MSA derived from the
FDIC’s Annual Summary of Deposits database; MSA-level data from US Census Bureau (available in,
for example, County and City Extra: Annual Metro, City and County Data Book, Bernan Press,
Lanham, MD)
Table I.
Summary statistics
External
economies
in banking
357
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
heteroscedasticity at conventional levels in all but one speci?cation, so no additional
correction was made for heteroscedasticity. Mishkin (1990) has shown that adjustments
for heteroscedasticity such as weighted least squares can actually degrade statistical
inference. Alternate estimates using least absolute deviation instead of OLS yielded
similar results, not reported here.
Share equations implied by Shephard’s lemma were not ?tted, following Berger and
Mester’s (1997) objection that Shephard’s lemma imposes the undesirable assumption of
perfect allocative ef?ciency. Share equations for individual components of cost (interest
expenses and noninterest expenses) would lack appropriate theoretical underpinnings.
Moreover, the cross-equation restrictions implied by any appropriate share equations do
not involve any of the market variables that are our primary interest in this study, and so
would not directly improve the ef?ciency of coef?cient estimates on those variables.
The ?exible Fourier form, used in several recent studies, has been shown to provide
better global approximations than the translog. However, the purpose of the output
quantities and input prices in the model here is not to characterize the total cost function
with complete precision, but rather to control for the cost effects of those variables
suf?ciently to permit a test of whether market variables (structural and demographic)
exert an independent in?uence on costs. Our adjusted R
2
is very high (0.96-0.99),
indicating that the translog form is fully adequate for the present purpose. Moreover,
Altunbas and Chakravarty (2001) have shown that the slightly better goodness-of-?t
typically afforded by the ?exible Fourier form does not reliably translate into improved
forecasting of bank costs; one might suspect that this shortcoming may be due in part to
over?tting and to the more severe multicollinearity introduced by additional terms all
formed from the same few output variables (themselves highly correlated, as noted in
[3]), degrading the precision of coef?cient estimates. Further, Berger and Mester (1997)
have reported that the ef?ciency rankings of banks are highly correlated between
translog and ?exible Fourier estimates.
In the speci?cation here, a negative coef?cient h
1
on the number of banks in the
market would be predicted by a combination of the structure-conduct-performance
hypothesis and the expense-preference hypothesis. To the extent that banks enjoy
greater market power in more concentrated markets, and to the extent that they choose
to dissipate at least some monopoly rents in excessive costs (through higher executive
salaries, plush of?ces, and larger staffs, as ?rst postulated by Edwards (1977)), then
banks’ costs would be higher in more concentrated markets. Previous literature has
identi?ed other factors such as information spillovers and labor force enrichment that
could also cause localization economies.
The ratio of single-market banks to all banks operating in a market may re?ect
several factors. Multimarket banks have access to a larger and more diverse labor pool
and informational resources than single-market banks, and may thus contribute more
strongly to informational spillovers and improved job matching, with effects on all
banks in their markets. This reasoning suggests that, if the coef?cient h
1
on the number
of banks in the market is negative, then we would expect the coef?cient h
2
to be positive.
Conversely, to the extent that multimarket operation requires additional layers of
management and coordination costs, then it is possible that h
2
could be negative.
The ratio of single-market banks to all banks in the market is also related to the ratio of
locally owned banks to all banks in the market, found by Collender and Shaffer (2003) to
be signi?cantly associated with local economic growth rates.
JFEP
4,4
358
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Urbanization economies would yield a negative coef?cient h
3
on population, but
congestion costs could reverse that sign. Superior job matching in a more densely
populated labor market could yield a negative coef?cient h
4
on population density.
A negative coef?cient h
5
on the education variable could re?ect higher productivity
(as in Rauch (1993)) whereas a positive coef?cient might re?ect superior bargaining
effectiveness of a more educated workforce.
3. Estimates
Table II summarizes regression coef?cients on the agglomeration variables. The
adjusted R
2
range from 0.958 to 0.991. For brevity, the table does not report all
coef?cients, but only those on the structural and demographic variables associated
with agglomeration effects. F-tests on the vector of structural variables x
1
through x
5
indicate joint signi?cance of these variables at the 0.01 level in all speci?cations[5].
The coef?cient onthe number of banks per market is signi?cantly negative at the 0.01
level in all regressions except that for interest expenses, where the point estimate is
negative and signi?cant at the 0.10 level. Thus, bank costs are lower where more banks
operate, consistent with economics of localization[6]. When costs are disaggregated into
interest and noninterest components, the localization effects appear to be concentrated
primarily in the noninterest dimension, in terms of both the magnitude and signi?cance
of the estimated coef?cient.
Although the model and data cannot distinguish among alternative causes of
localization economies, the ?ndings here are consistent with the theoretical predictions of
generalized informational spillovers. An alternative explanation of localization economies
is the joint expense preference and structure-conduct-performance hypotheses, such that
banks inless concentrated markets mayface stronger price competition onthe output side
along with associated pressure to minimize costs. Previous tests of the expense-preference
hypothesis (Edwards, 1977; Hannan, 1979; Hannan and Mavinga, 1980) have relied on an
assumption that the structure-conduct-performance hypothesis is valid, while most
tests of competitive conduct in banking have been interpreted under the assumption
of cost-minimizing behavior (Shaffer, 2004); more sophisticated theory and empirics are
needed to provide a valid joint test of these two hypotheses, but in any case the newbasis
for antitrust policies suggested by our ?ndings remains valid.
OLS estimates
Dependent variable Total cost Interest expenses Noninterest expenses
Number of banks in MSA 20.000826 (24.21)
*
20.000268 (21.69)
* * *
20.001359 (24.21)
*
% 1-MSA banks 20.1668 (24.27)
*
20.09968 (23.04)
*
20.2616 (24.03)
*
log (Population) 0.05514 (5.26)
*
0.01004 (1.26) 0.1079 (6.21)
*
Population density 20.000147 (21.41) 20.000098 (21.05) 20.000223 (21.34)
Education 20.002088 (22.14)
* *
20.000109 (20.13) 20.004132 (22.47)
* *
Adjusted R
2
0.985 0.991 0.959
Notes: Signi?cant at:
*
0.01,
* *
0.05, and
* * *
0.10 level; t-statistics in parentheses; these statistics are
based on standard errors that are heteroscedastic-consistent (white); education – percentage of adults
who completed at least four years of college; outputs are de?ned as total loans, total securities, and
noninterest income (a proxy for off-balance-sheet operations); input prices are the average deposit
interest rate, average wage rate, and average price of physical capital; sample size ¼ 1,075
Table II.
Regression estimates
for market variables
External
economies
in banking
359
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The coef?cient on the fraction of single-market banks is signi?cantly negative at the
0.01 level in all regressions. Hence, banks on average experience lower interest
expenses, noninterest expenses, and total costs in markets where most of the banks
operate exclusively in the home market. Based on the discussion above, this ?nding
suggests that multimarket access to workers and information is more than offset by
additional costs (such as due to administration and coordination), and does not confer a
net cost bene?t to banks.
The coef?cient on the natural logarithm of population is signi?cantly positive at the
0.01 level in all except the interest expense equation. This means that banks on average
face higher noninterest expenses and total costs in more populous markets. These
effects comprise diseconomies of urbanization. The coef?cient on population density
exhibits a consistently negative point estimate, but is not signi?cant.
The coef?cient on education is signi?cantly negative in the regressions on total
cost and noninterest expenses, but not in the regressions on interest expenses. One
interpretation of this ?nding is that a better-educated workforce may be more productive,
allowing banks to conduct their operations at lower overhead and labor-related costs.
At the same time, the absence of a strong association between education and interest
expense makes it dif?cult to distinguish whether noninterest cost differentials are passed
on to the banks’ customers or not.
3.1 Magnitudes
The magnitude of the estimated association between banking structure and costs is
modest at the individual bank level. If one bank enters an average market without
changing the value of any other variables in the regression, each incumbent bank’s total
cost would be lower by approximately 0.08 percent (8 basis points), or just under $20,000
annually on average, based on the OLS estimates reported in Table II. However, if the
entrant operates in only one market (so that the variable x
2
also rises as x
1
rises), then the
net estimated effect on costs involves both h
1
and h
2
, and amounts to a reduction in cost
of 22 basis points or about $53,000 annuallyfor eachincumbent onaverage. If the entrant
operates in other markets, the cost effect of the associated change in x
2
slightly more
than offsets that of the change in x
1
and implies a small increment in the incumbents’
total costs. Conversely, if a single-market bank is merged into (or acquired by) an
incumbent multimarket bank, the average bank in the market is estimated to suffer an
increment in total costs of 22 basis points or $53,000 annually. The acquisition of a
single-market bank by a bank not previously operating in the same market would be
associated with an increment in total costs of 26 basis points or over $63,000 annually.
The acquisition of a multimarket bank by another multimarket bank operating in the
same market would yield an increase in costs of 8 basis points or $20,000 annually for the
average bank in the market.
If we assume that an entrant would fail to expand the aggregate demand for banking
services in the market, so that each incumbent’s loss of market share to the entrant
implies a slight reduction in scale, total costs (though not average costs) would be an
additional 2.3 basis points lower on average (or $5,600 annually) due to the entry of one
bank. The much smaller magnitude of this scale effect indicates that, for practical
purposes, we can ignore the scale terms in calculating the structure-cost linkages.
If the number of banks remains constant but one local bank expands its operations
into other markets (or is merged into a multimarket bank), the estimates indicate that
JFEP
4,4
360
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the average bank in the original market will experience a 2.6 basis point higher total
cost, holding constant all output quantities, input prices, and local demographic
characteristics. This ?gure equates to an annual cost increment of $6,300 for the
average bank. Like the scale effects of redistributing market shares, this effect is small
in magnitude although statistically signi?cant.
Decomposing total costs into their interest and noninterest components, we ?nd that
most of the linkage with market structure is concentrated in noninterest expenses. This
conclusion is implied by both the lower statistical signi?cance and the smaller absolute
magnitude of the estimated coef?cient on interest expenses. One possible explanation
for this pattern might be the fact that noncore deposits are provided in regional or
national markets rather than in local markets, restricting the ability of any bank to
adjust its funding costs at the local level.
Aggregated across markets or across the industry, these modest estimates imply
substantial dollar totals. In an average MSA with 64 banks, a single-market entrant
would be associated with an annual cost savings of about $3.4 million for that market’s
banking services. Aggregating this ?gure across the more than 300 MSAs nationwide
implies an annual national cost savings of more than $1 billion if each MSA receives
one more single-market bank. On the other hand, the additional costs associated with
the entrant banks would alter these market totals. Conversely, the acquisition of one
single-market bank by a bank not previously operating in the same market would be
associated with an increase in marketwide banking costs of more than $4 million
annually (or a nationwide increase of more than $1 billion annually) without any
offsetting reduction[7]. More extensive consolidation would implicitly be associated
with even larger increases in costs unless offset by scale economies[8].
Such aggregation must be interpreted with caution, however, for several reasons.
We must keep in mind that these estimates re?ect entry that has occurred exogenously,
as the outcome of presumably rational business decisions by bankers, so that we
cannot conclude (for instance) that a particular change in costs can be reliably
generated by arti?cially introducing a new bank into a community[9]. Moreover, much
of the entry in recent years has involved multimarket banks rather than single-market
banks, and de novo banks (the most likely case of a single-market entrant) have been
found to exhibit systematically abnormal performance measures through their ?rst
nine years (DeYoung and Hasan, 1998; Shaffer, 1998). Therefore, even though the
general lessons seem clear, one must be cautious in drawing precise policy conclusions
from the magnitudes of these exploratory calculations.
The signi?cantly positive coef?cient on the natural logarithm of population
indicates that banks face higher costs in larger markets, ceteris paribus, evidence of
diseconomies of urbanization. The estimated coef?cient indicates that a 1 percent
increase in population is associated with higher total costs per bank of 5.5 basis points,
ceteris paribus, implying higher annual costs of just over $13,000 per average bank, or
more than $850,000 per MSA on average. Because market population is exogenous to a
given bank, conditional on its choice of market, there would seem to be no obvious
policy implications from this ?nding.
4. Discussion and conclusion
This study has examined the effects of agglomeration and urbanization on the cost
structure of US banking, using an established model of the banking ?rmand a standard
External
economies
in banking
361
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
functional formto facilitate comparison with previous studies. The sample year is 2000,
selected for the availability of concurrent demographic (census) data in a non-crisis
year[10]. No published study appears to have tested for agglomeration effects in the
banking industry previously.
Banks exhibit localization economies and urbanization diseconomies. The estimated
magnitude of localization economies is modest at the bank level but large in aggregate.
The localization economies are consistent with previous empirical ?ndings for the
manufacturing industry and with the predictions of established theories of information
spillovers transmitted through the labor market and other mechanisms. They are also
consistent with the negative coef?cient found on the number of within-market banks in
a nationwide cost study by DeYoung et al. (1998), though those authors did not
interpret that coef?cient in terms of external economies. Decomposing total costs into
their interest and noninterest components, I ?nd evidence that most of the association
between market structure and cost is concentrated in the noninterest component. It is
possible that regional or national markets for noncore funding might help explain the
relative insensitivity of funding costs to local market structure.
Such localization economies suggest a previously unrecognized bene?t of
structure-based antitrust policies, since the externality implies that equilibrium
market structure may be more concentrated than the least-cost structure. Further,
localization economies introduce a confounding effect in pro?t-concentration studies
unless adequately controlled for, and might also help explain the high degree of
estimated cost inef?ciency found in prior studies (as surveyed by, e.g. Berger and
Humphrey, 1997). The empirical correlation between bank size and market size implies
an additional bias in traditional studies that could potentially affect estimates of scale
economies or diseconomies[11]. These important policy-relevant effects may warrant
further study, along with the stability of external banking economies over time and any
factors explaining variations in external economies.
Notes
1. Research on information spillovers in nonbanking ?rms has often focused on information
related to R&D, rather than information related to ?nancial risk.
2. The argument for negative informational spillovers recognizes that imperfectly correlated
assessments by different banks of a given borrower’s credit risk implies that the average
default rate is an increasing function of the number of potential lenders in a market, a pattern
empirically supported in Shaffer (1998). This effect renders the net impact of agglomeration
on overall costs an open empirical question.
3. The pairwise correlation coef?cients are 0.69 between loans and securities, 0.78 between
loans and noninterest income, and 0.62 between securities and noninterest income. Quadratic
and cross-product terms in a translog function typically exhibit even higher correlations.
4. Although the sample contains only banks that operate within a single MSA, most MSAs
contain additional banks that operate in multiple geographic markets, and their presence is
quanti?ed by this variable. As noted below, the importance of this variable is motivated by
the ?nding of Collender and Shaffer (2003) that it is signi?cantly associated with local
economic growth rates.
5. The test statistic is 15.88 for total costs, 4.91 for interest expenses and 18.54 for noninterest
expenses. The 0.01 critical value for F with 5 and 1,049 degrees of freedom is 3.04.
JFEP
4,4
362
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
6. DeYoung et al. (1998) similarly controlled for the number of banks per MSA in a cost study
using a 1992 nationwide sample of single-market banks and, although they did not interpret
this coef?cient in terms of external economies, their ?ndings are consistent with those
obtained here.
7. These ?gures omit the many rural banking markets across the USA, but our MSA-generated
estimates cannot reliably be extended to calculate structural effects in rural markets.
8. The empirical literature on banking cost functions does not support the hypothesis that scale
economies in the banking industry would be strong enough to offset this amount of
additional costs even if consolidation results in larger average bank sizes.
9. Because banking structure is predetermined in our dataset, the long-run endogeneity of
structure does not cause an econometric problem in our estimates.
10. Because of the extreme inef?ciency of manually compiling a sample of single-market banks
using data from three sources, I did not attempt to extend the sample beyond a single year.
11. The correlation between total bank deposits in a metropolitan statistical area (MSA) and
total assets of the average bank headquartered in that MSA was 0.33 as of June 1992 across
the USA, signi?cantly positive at the 0.0001 level.
References
Altunbas, Y. and Chakravarty, S.P. (2001), “Frontier cost functions and bank ef?ciency”,
Economics Letters, Vol. 72 No. 2, pp. 233-40.
Andersson, F., Burgess, S. and Lane, J. (2004), “Cities, matching and productivity gains of
agglomeration”, CEPR Discussion Paper No. 4598.
Beeson, P.E. (1987), “Total factor productivity growth and agglomeration economies in
manufacturing, 1959-1973”, Journal of Regional Science, Vol. 27, pp. 183-90.
Beeson, P.E. and Husted, S. (1989), “Patterns and determinants of productive ef?ciency in state
manufacturing”, Journal of Regional Science, Vol. 29 No. 1, pp. 15-28.
Berger, A.N. and Humphrey, D.B. (1997), “Ef?ciency of ?nancial institutions: international
survey and directions for future research”, European Journal of Operational Research,
Vol. 98, pp. 175-212.
Berger, A.N. and Mester, L.J. (1997), “Inside the black box: what explains differences in the
ef?ciencies of ?nancial institutions?”, Journal of Banking & Finance, Vol. 21 No. 7,
pp. 895-947.
Broecker, T. (1990), “Credit-worthiness tests and interbank competition”, Econometrica, Vol. 58
No. 2, pp. 429-52.
Calem, P.S. and Carlino, G.A. (1991), “Urban agglomeration economies in the presence of
technical change”, Journal of Urban Economics, Vol. 29 No. 1, pp. 82-95.
Carlino, G., Chatterjee, S. and Hunt, R. (2005), “Matching and learning in cities: urban density and
the rate of innovation”, Working Paper Number 04-16/R, Federal Reserve Bank of
Philadelphia, Philadelphia, PA.
Carlino, G.A. (1979), “Increasing returns to scale in metropolitan manufacturing”, Journal of
Regional Science, Vol. 19, pp. 343-51.
Carlino, G.A. (1982), “Manufacturing agglomeration economies as returns to scale”, Papers of the
Regional Science Association, Vol. 50, pp. 95-108.
Collender, R.N. and Shaffer, S. (2003), “Local bank of?ce ownership, deposit control, market
structure, and economic growth”, Journal of Banking & Finance, Vol. 27 No. 1, pp. 27-57.
External
economies
in banking
363
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
DeYoung, R. and Hasan, I. (1998), “The performance of De Novo commercial banks: a pro?t
ef?ciency approach”, Journal of Banking & Finance, Vol. 22 No. 5, pp. 565-87.
DeYoung, R., Hasan, I. and Krichhoff, B. (1998), “The impact of out-of-state entry on the cost
ef?ciency of local commercial banks”, Journal of Economics and Business, Vol. 50 No. 2,
pp. 191-203.
Dick, A.A. (2008), “Demand estimation and consumer welfare in the banking industry”, Journal
of Banking & Finance, Vol. 32, pp. 1661-76.
Eberts, R.W. and McMillen, D.P. (1998), “Agglomeration economies and urban public
infrastructure”, in Cheshire, P. and Mills, E.S. (Eds), Handbook of Regional and Urban
Economics, Applied Urban Economics, Vol. 3, Elsevier, Amsterdam.
Edwards, F.R. (1977), “Managerial objectives in regulated industries: expense-preference
behavior in banking”, Journal of Political Economy, Vol. 85, pp. 147-62.
Evanoff, D.D. and Israilevich, P.R. (1991), “Regional differences in bank ef?ciency and
technology”, Annals of Regional Science, Vol. 25 No. 1, pp. 41-54.
Fogarty, M.S. and Garofalo, G.A. (1988), “Urban spatial structure and productivity growth in the
manufacturing sector of cites”, Journal of Urban Economics, Vol. 23, pp. 60-70.
Gilligan, T. and Smirlock, M. (1984), “An empirical study of joint production and scale economies
in commercial banking”, Journal of Banking & Finance, Vol. 8, March, pp. 67-77.
Glaeser, E.L. (1998), “Are cities dying?”, Journal of Economic Perspectives, Vol. 12 No. 2,
pp. 139-60.
Hamid, S.S. and Verma, S.K. (1994), “Regional variations in technological change and in potential
gains from consolidation among US commercial banks”, Journal of Regional Science,
Vol. 34 No. 3, pp. 411-23.
Hannan, T.H. (1979), “Expense-preference behavior in banking: a reexamination”, Journal of
Political Economy, Vol. 87 No. 4, pp. 891-5.
Hannan, T.H. and Mavinga, F. (1980), “Expense preference and managerial control: the case of
the banking ?rm”, Bell Journal of Economics, Vol. 11 No. 2, pp. 671-82.
Henderson, J.V. (1986), “The ef?ciency of resource usage and city size”, Journal of Urban
Economics, Vol. 19, pp. 47-70.
Ho, K. and Ishii, J. (2011), “Location and competition in retail banking”, International Journal of
Industrial Organization, Vol. 29, pp. 537-46.
Hughes, J.P. and Mester, L.J. (1993), “A quality and risk-adjusted cost function for banks:
evidence on the ‘too-big-to-fail’ doctrine”, Journal of Productivity Analysis, Vol. 4,
September, pp. 293-315.
Humphrey, D.B. (1990), “Why do estimates of bank scale economies differ?”, Federal Reserve
Bank of Richmond Economic Review, Vol. 76 No. 5, pp. 38-50.
Klein, M. (1971), “A theory of the banking ?rm”, Journal of Money, Credit, and Banking, Vol. 7
No. 3, pp. 205-18.
Marquez, R. (2002), “Competition, adverse selection, and information dispersion in the banking
industry”, Review of Financial Studies, Vol. 15 No. 3, pp. 901-26.
Mishkin, F.S. (1990), “Does correcting for heteroscedasticity help?”, Economics Letters, Vol. 34
No. 4, pp. 351-6.
Moomaw, R. (1983), “Spatial productivity variations in manufacturing: a critical survey of
cross-sectional analyses”, International Regional Science Review, Vol. 8, pp. 1-22.
Moomaw, R. (1988), “Agglomeration economies: localization or urbanization?”, Urban Studies,
Vol. 25 No. 2, pp. 150-61.
JFEP
4,4
364
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Nakamura, L.I. (1993), “Loan screening within and outside of customer relationships”,
Federal Reserve Bank of Philadelphia Working Paper No. 93-15.
Nakamura, R. (1985), “Agglomeration economies in urban manufacturing industries: a case of
Japanese cities”, Journal of Urban Economics, Vol. 17, pp. 108-24.
Rauch, J.E. (1993), “Productivity gains from geographic concentration of human capital: evidence
from the cities”, Journal of Urban Economics, Vol. 34 No. 3, pp. 380-400.
Riordan, M.H. (1993), “Competition and bank performance: a theoretical perspective”, in Mayer, C.
and Vives, X. (Eds), Capital Markets and Financial Intermediation, Cambridge University
Press, Cambridge, pp. 328-43.
Sealey, C. Jr and Lindley, J.T. (1977), “Inputs, outputs, and a theory of production and cost at
depository ?nancial institutions”, Journal of Finance, Vol. 32, pp. 1251-66.
Shaffer, S. (1998), “The winner’s curse in banking”, Journal of Financial Intermediation, Vol. 7
No. 4, pp. 359-92.
Shaffer, S. (2004), “Patterns of competition in banking”, Journal of Economics and Business,
Vol. 56 No. 4, pp. 287-313.
Strumsky, D., Lobo, J. and Fleming, L. (2005), “Metropolitan patenting, inventor agglomeration
and social networks: a tale of two effects”, Los Alamos National Laboratory Technical
Report LAUR-04-8798.
White, H. (1980), “A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity”, Econometrica, Vol. 48 No. 4, pp. 817-38.
Further reading
Jackson, W.E. (1992), “Is the market well de?ned in bank merger and acquisition analysis?”,
The Review of Economics and Statistics, Vol. 74, November, pp. 655-61.
Kwast, M., Starr-McCluer, M. and Wolken, J.D. (1997), “Market de?nition and the analysis of
antitrust in banking”, Antitrust Bulletin, Vol. 42 No. 4, pp. 973-95.
Whitehead, D.D. (1980), “Relevant geographic banking markets: how should they be de?ned?”,
Federal Reserve Bank of Atlanta Economic Review, January/February, pp. 20-8.
Corresponding author
Sherrill Shaffer can be contacted at: [email protected]
External
economies
in banking
365
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_675103274.pdf
The purpose of this paper is to provide the first empirical study of external economies
(agglomeration economies) in the banking industry.
Journal of Financial Economic Policy
External economies in banking
Sherrill Shaffer
Article information:
To cite this document:
Sherrill Shaffer, (2012),"External economies in banking", J ournal of Financial Economic Policy, Vol. 4 Iss 4
pp. 354 - 365
Permanent link to this document:
http://dx.doi.org/10.1108/17576381211279316
Downloaded on: 24 January 2016, At: 21:45 (PT)
References: this document contains references to 45 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 233 times since 2012*
Users who downloaded this article also downloaded:
Wenling Lu, David A. Whidbee, (2013),"Bank structure and failure during the financial crisis", J ournal of
Financial Economic Policy, Vol. 5 Iss 3 pp. 281-299 http://dx.doi.org/10.1108/J FEP-02-2013-0006
Eric Osei-Assibey, Baimba Augustine Bockarie, (2013),"Bank risks, capital and loan supply: evidence from
Sierra Leone", J ournal of Financial Economic Policy, Vol. 5 Iss 3 pp. 256-271 http://dx.doi.org/10.1108/
J FEP-09-2012-0041
J ames R. Barth, Gerard Caprio, Ross Levine, (2013),"Bank regulation and supervision in 180
countries from 1999 to 2011", J ournal of Financial Economic Policy, Vol. 5 Iss 2 pp. 111-219 http://
dx.doi.org/10.1108/17576381311329661
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
External economies in banking
Sherrill Shaffer
Department of Economics and Finance, University of Wyoming,
Laramie, Wyoming, USA
Abstract
Purpose – The purpose of this paper is to provide the ?rst empirical study of external economies
(agglomeration economies) in the banking industry.
Design/methodology/approach – The author extends a standard speci?cation of banking costs to
control for community and market characteristics.
Findings – Banks’ costs are a decreasing function of the number of rival banks and an increasing
function of market population. Estimated magnitudes of these effects, modest at the bank level, are
large in aggregate. Multimarket operation of rival banks is also important.
Originality/value – These ?ndings suggest a previously unrecognized cost-side bene?t of
structure-based antitrust policies, and have additional implications for public policy toward
banking structure, as well as calling for a re-interpretation of previous studies of scale economies, cost
ef?ciency, and price-cost margins in banking.
Keywords Agglomeration, Localization, Urbanization, Banking, Antitrust, Costs
Paper type Research paper
1. Introduction
It has long been recognized that productivity and costs may vary systematically not
only as a function of the technology used by the individual ?rm but also as a function of
the size and other characteristics of the community in which the ?rmoperates. The latter
effects have been termed “agglomeration economies” or external economies, which can
be further decomposed into “urbanization economies” (the size of the city itself) and
“localization economies” (the size of the ?rm’s industry within the city). This paper
presents an exploratory test of economies of agglomeration in the US commercial
banking industry, using ?rm-level data from a nationwide sample of single-market US
banks operating inmetropolitan areas. Afurther contribution is to point out a previously
unrecognized policy implication of agglomeration economies: structure-based antitrust
regulations can potentially improve welfare by limiting excessive consolidation that
raises aggregate costs.
Although the empirical model differs only in straightforward ways from many
previous studies of banking costs, the results shed important new light on a number of
issues. Economies of agglomeration have been studied previously, but only at the
aggregate level or within the manufacturing sector (see Eberts and McMillen, 1998, for
a review). By contrast, this study uses micro-level data from a single major service
industry, one in which community characteristics may affect the ?rm’s cost structure
in at least two direct ways. One commonly recognized source of external economies is
information spillovers (Glaeser, 1998). Information about ?nancial risk is fundamental
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
JEL classi?cation – G21, L10
The author is grateful for helpful comments on earlier versions of this paper from Allen Berger,
Robert Collender, Beth Cooperman, Bob Hunt, Ken Kopecky, James McAndrews and Tara Rice.
JFEP
4,4
354
Journal of Financial Economic Policy
Vol. 4 No. 4, 2012
pp. 354-365
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381211279316
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
to the banking industry, suggesting that spillovers of such information may have
signi?cant cost effects[1]. A novel aspect here is that such spillovers among banks may
be either positive, reducing costs as discussed in previous non-banking agglomeration
studies; or negative, as discussed for banks by Broecker (1990), Nakamura (1993),
Riordan (1993), Shaffer (1998) and Marquez (2002), though without previously being
linked to external economies[2]. In addition, access to a pool of employees with
specialized skill has also been recognized as a possible source of external economies.
This factor likely applies to the banking industry and is consistent with observed
concentrations of banks in Switzerland or of bank headquarters in Frankfurt.
1.1 Previous studies
Empirical evidence of agglomeration economies has been found for urban
manufacturing industries using data as early as the 1950s. Carlino (1979) estimates a
decomposition of overall economies of scale into internal economies, localization
economies, and urbanization economies. Urbanization economies are found in aggregate
by Carlino (1982) and Moomaw(1983), while Henderson (1986) and Moomaw(1988) ?nd
that localization economies are dominant in urban manufacturing. Nakamura (1985)
?nds that urbanization economies are more important for ?rms in light industries but
localization economies are more important for ?rms in heavy industries. Fogarty and
Garofalo (1988) and Calem and Carlino (1991) estimate agglomeration economies in
urban manufacturing in the presence of technical change.
These studies focus on urban manufacturing rather than service industries, and
have typically been hampered by the level of detail of available data. Similar studies by
Beeson (1987) and Beeson and Husted (1989) used state-level data to circumvent this
problem. By contrast, the banking industry is an important service industry with
abundant ?rm-level data, and as such is an ideal candidate for extending the empirical
literature on agglomeration economies.
The cost structure of the banking industry has been extensively studied over the past
30 years. A few studies have explored regional differences in the cost structure of US
banks (Evanoff and Israilevich, 1991; Hamid and Verma, 1994). To the author’s
knowledge, no previously published study has attempted to distinguish between internal
and external sources of economies or diseconomies of scale in the banking industry.
2. The empirical model
I estimate a standard empirical banking cost model to facilitate comparison with
previous studies. Prior research and common regulatory practice alike have used
metropolitan statistical areas (MSAs) (or counties in rural areas) as a standard proxy
for geographic banking markets (Dick, 2008; Ho and Ishii, 2011). If a measurable
localization cost effect can be observed within a typical MSA, any additional network
effects involving broader regions would only serve to strengthen the overall impact of
inter?rm interactions on banks’ costs.
The de?nition of inputs and outputs follows the intermediation model of Klein (1971)
and Sealey and Lindley (1977), in which labor and capital are inputs used both to
obtain deposits and, in conjunction with deposits, to originate loans and other earning
assets. The interest rate paid on deposits is an input price, as is the average wage rate
paid to bank employees; bank output is measured as the dollar value of earning assets,
decomposed into:
External
economies
in banking
355
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
.
loans and leases; and
.
securities.
Fee income is treated as an additional output, re?ecting the growing importance of
off-balance-sheet activities for banking in recent years; alternate estimates omitting fee
income (not reported) did not materially change the signs, signi?cance, or magnitudes
of the various agglomeration measures. This model is supported by empirical evidence
suggesting that aggregate deposits behave primarily as inputs (Gilligan and Smirlock,
1984; Hughes and Mester, 1993) and has been widely employed in previous studies of
banking costs (Berger and Mester, 1997). A ?ner disaggregation of outputs could be
made, but at the expense of more severe multicollinearity[3]. Alternate dependent
variables are total costs, interest expenses, and noninterest expenses.
Agglomeration effects are measured by a vector of ?ve variables. Localization effects
are captured by the number of banks operating in the local market, as reported in the
FDIC’s summary of deposits. Aseparate aspect of localization is measured by the ratio of
single-market banks to total banks inthe market[4]. Market size and congestion effects are
incorporated through the natural logarithmof the market’s population. Asecondaspect of
congestion is measured by the market’s population density per square mile, following
Andersson et al. (2004), Carlino et al. (2005) and Strumsky et al. (2005). The quality of labor
force education, a measure of cumulative human capital, is measured as the percentage of
population above 25 years of age that had completed four years of college as of 2000. This
control variable is commonly included in empirical studies of agglomeration effects.
The sample is drawn from all US commercial banks operating exclusively within a
single MSA as of year-end 2000. As in many previous studies, banks less than ?ve
years old are excluded from the sample because such banks typically have portfolio
compositions and cost structures quite different from those of more mature banks. The
sample year is chosen to coincide with the availability of contemporaneous
demographic (census) data in a non-crisis year. Previous research has found that the
choice of year does not affect the estimates of scale economies in banking (Humphrey,
1990). Table I summarizes the sample statistics, drawn from a ?nal set of 1,075
commercial banks.
The functional form is the translog, perhaps the most widely used form for
empirical cost studies in recent years. The translog cost function, augmented by
agglomeration and market structure terms, is de?ned as:
ln C ¼ a
o
þ
X
4
i¼1
a
i
Q
i
þ
X
3
j¼1
b
j
P
j
þ
1
2
X
4
i¼1
X
4
k¼1
d
ik
Q
i
Q
k
þ
1
2
X
3
j¼1
X
3
k¼1
g
jk
P
j
P
k
þ
X
4
i¼1
X
3
j¼1
r
ij
Q
i
P
j
þ
X
3
h¼1
h
h
X
h
ð1Þ
where:
C ¼ total costs, including both interest and noninterest expenses (alternate
speci?cations replace this variable with interest and noninterest expenses,
respectively).
Q
i
¼ ln q
i
: quantity of output i:
JFEP
4,4
356
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
q
1
¼ total loans and leases.
q
2
¼ securities plus federal funds sold.
q
3
¼ fee income.
P
j
¼ ln p
j
: price of input j:
p
1
¼ price of labor, calculated as annual wage and bene?t expenses per employee.
p
2
¼ price of funding, calculated as average interest costs per dollar of all deposits.
p
3
¼ price of physical capital, calculated as annual expenses on premises and
equipment divided by the stock of these items.
x
1
¼ number of banks in the MSA.
x
2
¼ fraction of banks in the MSA that do not have of?ces in other markets.
x
3
¼ natural logarithm of population of MSA.
x
4
¼ population density per square mile of MSA.
x
5
¼ percentage of adult population completing four years of college as of 2000.
The usual restrictions of symmetry and linear homogeneity in factor prices are
imposed:
d
ik
¼d
ki
; g
jk
¼g
kj
;
X
b
j
¼1;
j
X
r
ij
¼0 for all i; and
j
X
g
jk
¼0 for all k:
Estimates are ?tted by OLS with heteroscedastic-consistent standard errors
(White, 1980). A Lagrange multiplier test failed to reject the null of no
Variable Mean SD
log (total cost) 10.10 1.29
log (interest expense) 9.40 1.37
log (noninterest expense) 9.35 1.26
Number of banks 64.02 47.01
% 1-MSA banks 0.451 0.149
log (population) 13.62 1.10
Population density 135.8 87.3
Education 25.78 6.18
log (total loans) 11.01 1.12
log (total securities) 9.90 1.21
log (noninterest income) 6.52 1.36
average deposit interest rate 0.036 0.008
average wage rate 42.44 11.92
price of physical capital 0.388 1.187
Source: Bank-level data from regulatory Call Reports; number of banks in MSA derived from the
FDIC’s Annual Summary of Deposits database; MSA-level data from US Census Bureau (available in,
for example, County and City Extra: Annual Metro, City and County Data Book, Bernan Press,
Lanham, MD)
Table I.
Summary statistics
External
economies
in banking
357
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
heteroscedasticity at conventional levels in all but one speci?cation, so no additional
correction was made for heteroscedasticity. Mishkin (1990) has shown that adjustments
for heteroscedasticity such as weighted least squares can actually degrade statistical
inference. Alternate estimates using least absolute deviation instead of OLS yielded
similar results, not reported here.
Share equations implied by Shephard’s lemma were not ?tted, following Berger and
Mester’s (1997) objection that Shephard’s lemma imposes the undesirable assumption of
perfect allocative ef?ciency. Share equations for individual components of cost (interest
expenses and noninterest expenses) would lack appropriate theoretical underpinnings.
Moreover, the cross-equation restrictions implied by any appropriate share equations do
not involve any of the market variables that are our primary interest in this study, and so
would not directly improve the ef?ciency of coef?cient estimates on those variables.
The ?exible Fourier form, used in several recent studies, has been shown to provide
better global approximations than the translog. However, the purpose of the output
quantities and input prices in the model here is not to characterize the total cost function
with complete precision, but rather to control for the cost effects of those variables
suf?ciently to permit a test of whether market variables (structural and demographic)
exert an independent in?uence on costs. Our adjusted R
2
is very high (0.96-0.99),
indicating that the translog form is fully adequate for the present purpose. Moreover,
Altunbas and Chakravarty (2001) have shown that the slightly better goodness-of-?t
typically afforded by the ?exible Fourier form does not reliably translate into improved
forecasting of bank costs; one might suspect that this shortcoming may be due in part to
over?tting and to the more severe multicollinearity introduced by additional terms all
formed from the same few output variables (themselves highly correlated, as noted in
[3]), degrading the precision of coef?cient estimates. Further, Berger and Mester (1997)
have reported that the ef?ciency rankings of banks are highly correlated between
translog and ?exible Fourier estimates.
In the speci?cation here, a negative coef?cient h
1
on the number of banks in the
market would be predicted by a combination of the structure-conduct-performance
hypothesis and the expense-preference hypothesis. To the extent that banks enjoy
greater market power in more concentrated markets, and to the extent that they choose
to dissipate at least some monopoly rents in excessive costs (through higher executive
salaries, plush of?ces, and larger staffs, as ?rst postulated by Edwards (1977)), then
banks’ costs would be higher in more concentrated markets. Previous literature has
identi?ed other factors such as information spillovers and labor force enrichment that
could also cause localization economies.
The ratio of single-market banks to all banks operating in a market may re?ect
several factors. Multimarket banks have access to a larger and more diverse labor pool
and informational resources than single-market banks, and may thus contribute more
strongly to informational spillovers and improved job matching, with effects on all
banks in their markets. This reasoning suggests that, if the coef?cient h
1
on the number
of banks in the market is negative, then we would expect the coef?cient h
2
to be positive.
Conversely, to the extent that multimarket operation requires additional layers of
management and coordination costs, then it is possible that h
2
could be negative.
The ratio of single-market banks to all banks in the market is also related to the ratio of
locally owned banks to all banks in the market, found by Collender and Shaffer (2003) to
be signi?cantly associated with local economic growth rates.
JFEP
4,4
358
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Urbanization economies would yield a negative coef?cient h
3
on population, but
congestion costs could reverse that sign. Superior job matching in a more densely
populated labor market could yield a negative coef?cient h
4
on population density.
A negative coef?cient h
5
on the education variable could re?ect higher productivity
(as in Rauch (1993)) whereas a positive coef?cient might re?ect superior bargaining
effectiveness of a more educated workforce.
3. Estimates
Table II summarizes regression coef?cients on the agglomeration variables. The
adjusted R
2
range from 0.958 to 0.991. For brevity, the table does not report all
coef?cients, but only those on the structural and demographic variables associated
with agglomeration effects. F-tests on the vector of structural variables x
1
through x
5
indicate joint signi?cance of these variables at the 0.01 level in all speci?cations[5].
The coef?cient onthe number of banks per market is signi?cantly negative at the 0.01
level in all regressions except that for interest expenses, where the point estimate is
negative and signi?cant at the 0.10 level. Thus, bank costs are lower where more banks
operate, consistent with economics of localization[6]. When costs are disaggregated into
interest and noninterest components, the localization effects appear to be concentrated
primarily in the noninterest dimension, in terms of both the magnitude and signi?cance
of the estimated coef?cient.
Although the model and data cannot distinguish among alternative causes of
localization economies, the ?ndings here are consistent with the theoretical predictions of
generalized informational spillovers. An alternative explanation of localization economies
is the joint expense preference and structure-conduct-performance hypotheses, such that
banks inless concentrated markets mayface stronger price competition onthe output side
along with associated pressure to minimize costs. Previous tests of the expense-preference
hypothesis (Edwards, 1977; Hannan, 1979; Hannan and Mavinga, 1980) have relied on an
assumption that the structure-conduct-performance hypothesis is valid, while most
tests of competitive conduct in banking have been interpreted under the assumption
of cost-minimizing behavior (Shaffer, 2004); more sophisticated theory and empirics are
needed to provide a valid joint test of these two hypotheses, but in any case the newbasis
for antitrust policies suggested by our ?ndings remains valid.
OLS estimates
Dependent variable Total cost Interest expenses Noninterest expenses
Number of banks in MSA 20.000826 (24.21)
*
20.000268 (21.69)
* * *
20.001359 (24.21)
*
% 1-MSA banks 20.1668 (24.27)
*
20.09968 (23.04)
*
20.2616 (24.03)
*
log (Population) 0.05514 (5.26)
*
0.01004 (1.26) 0.1079 (6.21)
*
Population density 20.000147 (21.41) 20.000098 (21.05) 20.000223 (21.34)
Education 20.002088 (22.14)
* *
20.000109 (20.13) 20.004132 (22.47)
* *
Adjusted R
2
0.985 0.991 0.959
Notes: Signi?cant at:
*
0.01,
* *
0.05, and
* * *
0.10 level; t-statistics in parentheses; these statistics are
based on standard errors that are heteroscedastic-consistent (white); education – percentage of adults
who completed at least four years of college; outputs are de?ned as total loans, total securities, and
noninterest income (a proxy for off-balance-sheet operations); input prices are the average deposit
interest rate, average wage rate, and average price of physical capital; sample size ¼ 1,075
Table II.
Regression estimates
for market variables
External
economies
in banking
359
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The coef?cient on the fraction of single-market banks is signi?cantly negative at the
0.01 level in all regressions. Hence, banks on average experience lower interest
expenses, noninterest expenses, and total costs in markets where most of the banks
operate exclusively in the home market. Based on the discussion above, this ?nding
suggests that multimarket access to workers and information is more than offset by
additional costs (such as due to administration and coordination), and does not confer a
net cost bene?t to banks.
The coef?cient on the natural logarithm of population is signi?cantly positive at the
0.01 level in all except the interest expense equation. This means that banks on average
face higher noninterest expenses and total costs in more populous markets. These
effects comprise diseconomies of urbanization. The coef?cient on population density
exhibits a consistently negative point estimate, but is not signi?cant.
The coef?cient on education is signi?cantly negative in the regressions on total
cost and noninterest expenses, but not in the regressions on interest expenses. One
interpretation of this ?nding is that a better-educated workforce may be more productive,
allowing banks to conduct their operations at lower overhead and labor-related costs.
At the same time, the absence of a strong association between education and interest
expense makes it dif?cult to distinguish whether noninterest cost differentials are passed
on to the banks’ customers or not.
3.1 Magnitudes
The magnitude of the estimated association between banking structure and costs is
modest at the individual bank level. If one bank enters an average market without
changing the value of any other variables in the regression, each incumbent bank’s total
cost would be lower by approximately 0.08 percent (8 basis points), or just under $20,000
annually on average, based on the OLS estimates reported in Table II. However, if the
entrant operates in only one market (so that the variable x
2
also rises as x
1
rises), then the
net estimated effect on costs involves both h
1
and h
2
, and amounts to a reduction in cost
of 22 basis points or about $53,000 annuallyfor eachincumbent onaverage. If the entrant
operates in other markets, the cost effect of the associated change in x
2
slightly more
than offsets that of the change in x
1
and implies a small increment in the incumbents’
total costs. Conversely, if a single-market bank is merged into (or acquired by) an
incumbent multimarket bank, the average bank in the market is estimated to suffer an
increment in total costs of 22 basis points or $53,000 annually. The acquisition of a
single-market bank by a bank not previously operating in the same market would be
associated with an increment in total costs of 26 basis points or over $63,000 annually.
The acquisition of a multimarket bank by another multimarket bank operating in the
same market would yield an increase in costs of 8 basis points or $20,000 annually for the
average bank in the market.
If we assume that an entrant would fail to expand the aggregate demand for banking
services in the market, so that each incumbent’s loss of market share to the entrant
implies a slight reduction in scale, total costs (though not average costs) would be an
additional 2.3 basis points lower on average (or $5,600 annually) due to the entry of one
bank. The much smaller magnitude of this scale effect indicates that, for practical
purposes, we can ignore the scale terms in calculating the structure-cost linkages.
If the number of banks remains constant but one local bank expands its operations
into other markets (or is merged into a multimarket bank), the estimates indicate that
JFEP
4,4
360
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the average bank in the original market will experience a 2.6 basis point higher total
cost, holding constant all output quantities, input prices, and local demographic
characteristics. This ?gure equates to an annual cost increment of $6,300 for the
average bank. Like the scale effects of redistributing market shares, this effect is small
in magnitude although statistically signi?cant.
Decomposing total costs into their interest and noninterest components, we ?nd that
most of the linkage with market structure is concentrated in noninterest expenses. This
conclusion is implied by both the lower statistical signi?cance and the smaller absolute
magnitude of the estimated coef?cient on interest expenses. One possible explanation
for this pattern might be the fact that noncore deposits are provided in regional or
national markets rather than in local markets, restricting the ability of any bank to
adjust its funding costs at the local level.
Aggregated across markets or across the industry, these modest estimates imply
substantial dollar totals. In an average MSA with 64 banks, a single-market entrant
would be associated with an annual cost savings of about $3.4 million for that market’s
banking services. Aggregating this ?gure across the more than 300 MSAs nationwide
implies an annual national cost savings of more than $1 billion if each MSA receives
one more single-market bank. On the other hand, the additional costs associated with
the entrant banks would alter these market totals. Conversely, the acquisition of one
single-market bank by a bank not previously operating in the same market would be
associated with an increase in marketwide banking costs of more than $4 million
annually (or a nationwide increase of more than $1 billion annually) without any
offsetting reduction[7]. More extensive consolidation would implicitly be associated
with even larger increases in costs unless offset by scale economies[8].
Such aggregation must be interpreted with caution, however, for several reasons.
We must keep in mind that these estimates re?ect entry that has occurred exogenously,
as the outcome of presumably rational business decisions by bankers, so that we
cannot conclude (for instance) that a particular change in costs can be reliably
generated by arti?cially introducing a new bank into a community[9]. Moreover, much
of the entry in recent years has involved multimarket banks rather than single-market
banks, and de novo banks (the most likely case of a single-market entrant) have been
found to exhibit systematically abnormal performance measures through their ?rst
nine years (DeYoung and Hasan, 1998; Shaffer, 1998). Therefore, even though the
general lessons seem clear, one must be cautious in drawing precise policy conclusions
from the magnitudes of these exploratory calculations.
The signi?cantly positive coef?cient on the natural logarithm of population
indicates that banks face higher costs in larger markets, ceteris paribus, evidence of
diseconomies of urbanization. The estimated coef?cient indicates that a 1 percent
increase in population is associated with higher total costs per bank of 5.5 basis points,
ceteris paribus, implying higher annual costs of just over $13,000 per average bank, or
more than $850,000 per MSA on average. Because market population is exogenous to a
given bank, conditional on its choice of market, there would seem to be no obvious
policy implications from this ?nding.
4. Discussion and conclusion
This study has examined the effects of agglomeration and urbanization on the cost
structure of US banking, using an established model of the banking ?rmand a standard
External
economies
in banking
361
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
functional formto facilitate comparison with previous studies. The sample year is 2000,
selected for the availability of concurrent demographic (census) data in a non-crisis
year[10]. No published study appears to have tested for agglomeration effects in the
banking industry previously.
Banks exhibit localization economies and urbanization diseconomies. The estimated
magnitude of localization economies is modest at the bank level but large in aggregate.
The localization economies are consistent with previous empirical ?ndings for the
manufacturing industry and with the predictions of established theories of information
spillovers transmitted through the labor market and other mechanisms. They are also
consistent with the negative coef?cient found on the number of within-market banks in
a nationwide cost study by DeYoung et al. (1998), though those authors did not
interpret that coef?cient in terms of external economies. Decomposing total costs into
their interest and noninterest components, I ?nd evidence that most of the association
between market structure and cost is concentrated in the noninterest component. It is
possible that regional or national markets for noncore funding might help explain the
relative insensitivity of funding costs to local market structure.
Such localization economies suggest a previously unrecognized bene?t of
structure-based antitrust policies, since the externality implies that equilibrium
market structure may be more concentrated than the least-cost structure. Further,
localization economies introduce a confounding effect in pro?t-concentration studies
unless adequately controlled for, and might also help explain the high degree of
estimated cost inef?ciency found in prior studies (as surveyed by, e.g. Berger and
Humphrey, 1997). The empirical correlation between bank size and market size implies
an additional bias in traditional studies that could potentially affect estimates of scale
economies or diseconomies[11]. These important policy-relevant effects may warrant
further study, along with the stability of external banking economies over time and any
factors explaining variations in external economies.
Notes
1. Research on information spillovers in nonbanking ?rms has often focused on information
related to R&D, rather than information related to ?nancial risk.
2. The argument for negative informational spillovers recognizes that imperfectly correlated
assessments by different banks of a given borrower’s credit risk implies that the average
default rate is an increasing function of the number of potential lenders in a market, a pattern
empirically supported in Shaffer (1998). This effect renders the net impact of agglomeration
on overall costs an open empirical question.
3. The pairwise correlation coef?cients are 0.69 between loans and securities, 0.78 between
loans and noninterest income, and 0.62 between securities and noninterest income. Quadratic
and cross-product terms in a translog function typically exhibit even higher correlations.
4. Although the sample contains only banks that operate within a single MSA, most MSAs
contain additional banks that operate in multiple geographic markets, and their presence is
quanti?ed by this variable. As noted below, the importance of this variable is motivated by
the ?nding of Collender and Shaffer (2003) that it is signi?cantly associated with local
economic growth rates.
5. The test statistic is 15.88 for total costs, 4.91 for interest expenses and 18.54 for noninterest
expenses. The 0.01 critical value for F with 5 and 1,049 degrees of freedom is 3.04.
JFEP
4,4
362
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
6. DeYoung et al. (1998) similarly controlled for the number of banks per MSA in a cost study
using a 1992 nationwide sample of single-market banks and, although they did not interpret
this coef?cient in terms of external economies, their ?ndings are consistent with those
obtained here.
7. These ?gures omit the many rural banking markets across the USA, but our MSA-generated
estimates cannot reliably be extended to calculate structural effects in rural markets.
8. The empirical literature on banking cost functions does not support the hypothesis that scale
economies in the banking industry would be strong enough to offset this amount of
additional costs even if consolidation results in larger average bank sizes.
9. Because banking structure is predetermined in our dataset, the long-run endogeneity of
structure does not cause an econometric problem in our estimates.
10. Because of the extreme inef?ciency of manually compiling a sample of single-market banks
using data from three sources, I did not attempt to extend the sample beyond a single year.
11. The correlation between total bank deposits in a metropolitan statistical area (MSA) and
total assets of the average bank headquartered in that MSA was 0.33 as of June 1992 across
the USA, signi?cantly positive at the 0.0001 level.
References
Altunbas, Y. and Chakravarty, S.P. (2001), “Frontier cost functions and bank ef?ciency”,
Economics Letters, Vol. 72 No. 2, pp. 233-40.
Andersson, F., Burgess, S. and Lane, J. (2004), “Cities, matching and productivity gains of
agglomeration”, CEPR Discussion Paper No. 4598.
Beeson, P.E. (1987), “Total factor productivity growth and agglomeration economies in
manufacturing, 1959-1973”, Journal of Regional Science, Vol. 27, pp. 183-90.
Beeson, P.E. and Husted, S. (1989), “Patterns and determinants of productive ef?ciency in state
manufacturing”, Journal of Regional Science, Vol. 29 No. 1, pp. 15-28.
Berger, A.N. and Humphrey, D.B. (1997), “Ef?ciency of ?nancial institutions: international
survey and directions for future research”, European Journal of Operational Research,
Vol. 98, pp. 175-212.
Berger, A.N. and Mester, L.J. (1997), “Inside the black box: what explains differences in the
ef?ciencies of ?nancial institutions?”, Journal of Banking & Finance, Vol. 21 No. 7,
pp. 895-947.
Broecker, T. (1990), “Credit-worthiness tests and interbank competition”, Econometrica, Vol. 58
No. 2, pp. 429-52.
Calem, P.S. and Carlino, G.A. (1991), “Urban agglomeration economies in the presence of
technical change”, Journal of Urban Economics, Vol. 29 No. 1, pp. 82-95.
Carlino, G., Chatterjee, S. and Hunt, R. (2005), “Matching and learning in cities: urban density and
the rate of innovation”, Working Paper Number 04-16/R, Federal Reserve Bank of
Philadelphia, Philadelphia, PA.
Carlino, G.A. (1979), “Increasing returns to scale in metropolitan manufacturing”, Journal of
Regional Science, Vol. 19, pp. 343-51.
Carlino, G.A. (1982), “Manufacturing agglomeration economies as returns to scale”, Papers of the
Regional Science Association, Vol. 50, pp. 95-108.
Collender, R.N. and Shaffer, S. (2003), “Local bank of?ce ownership, deposit control, market
structure, and economic growth”, Journal of Banking & Finance, Vol. 27 No. 1, pp. 27-57.
External
economies
in banking
363
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
DeYoung, R. and Hasan, I. (1998), “The performance of De Novo commercial banks: a pro?t
ef?ciency approach”, Journal of Banking & Finance, Vol. 22 No. 5, pp. 565-87.
DeYoung, R., Hasan, I. and Krichhoff, B. (1998), “The impact of out-of-state entry on the cost
ef?ciency of local commercial banks”, Journal of Economics and Business, Vol. 50 No. 2,
pp. 191-203.
Dick, A.A. (2008), “Demand estimation and consumer welfare in the banking industry”, Journal
of Banking & Finance, Vol. 32, pp. 1661-76.
Eberts, R.W. and McMillen, D.P. (1998), “Agglomeration economies and urban public
infrastructure”, in Cheshire, P. and Mills, E.S. (Eds), Handbook of Regional and Urban
Economics, Applied Urban Economics, Vol. 3, Elsevier, Amsterdam.
Edwards, F.R. (1977), “Managerial objectives in regulated industries: expense-preference
behavior in banking”, Journal of Political Economy, Vol. 85, pp. 147-62.
Evanoff, D.D. and Israilevich, P.R. (1991), “Regional differences in bank ef?ciency and
technology”, Annals of Regional Science, Vol. 25 No. 1, pp. 41-54.
Fogarty, M.S. and Garofalo, G.A. (1988), “Urban spatial structure and productivity growth in the
manufacturing sector of cites”, Journal of Urban Economics, Vol. 23, pp. 60-70.
Gilligan, T. and Smirlock, M. (1984), “An empirical study of joint production and scale economies
in commercial banking”, Journal of Banking & Finance, Vol. 8, March, pp. 67-77.
Glaeser, E.L. (1998), “Are cities dying?”, Journal of Economic Perspectives, Vol. 12 No. 2,
pp. 139-60.
Hamid, S.S. and Verma, S.K. (1994), “Regional variations in technological change and in potential
gains from consolidation among US commercial banks”, Journal of Regional Science,
Vol. 34 No. 3, pp. 411-23.
Hannan, T.H. (1979), “Expense-preference behavior in banking: a reexamination”, Journal of
Political Economy, Vol. 87 No. 4, pp. 891-5.
Hannan, T.H. and Mavinga, F. (1980), “Expense preference and managerial control: the case of
the banking ?rm”, Bell Journal of Economics, Vol. 11 No. 2, pp. 671-82.
Henderson, J.V. (1986), “The ef?ciency of resource usage and city size”, Journal of Urban
Economics, Vol. 19, pp. 47-70.
Ho, K. and Ishii, J. (2011), “Location and competition in retail banking”, International Journal of
Industrial Organization, Vol. 29, pp. 537-46.
Hughes, J.P. and Mester, L.J. (1993), “A quality and risk-adjusted cost function for banks:
evidence on the ‘too-big-to-fail’ doctrine”, Journal of Productivity Analysis, Vol. 4,
September, pp. 293-315.
Humphrey, D.B. (1990), “Why do estimates of bank scale economies differ?”, Federal Reserve
Bank of Richmond Economic Review, Vol. 76 No. 5, pp. 38-50.
Klein, M. (1971), “A theory of the banking ?rm”, Journal of Money, Credit, and Banking, Vol. 7
No. 3, pp. 205-18.
Marquez, R. (2002), “Competition, adverse selection, and information dispersion in the banking
industry”, Review of Financial Studies, Vol. 15 No. 3, pp. 901-26.
Mishkin, F.S. (1990), “Does correcting for heteroscedasticity help?”, Economics Letters, Vol. 34
No. 4, pp. 351-6.
Moomaw, R. (1983), “Spatial productivity variations in manufacturing: a critical survey of
cross-sectional analyses”, International Regional Science Review, Vol. 8, pp. 1-22.
Moomaw, R. (1988), “Agglomeration economies: localization or urbanization?”, Urban Studies,
Vol. 25 No. 2, pp. 150-61.
JFEP
4,4
364
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Nakamura, L.I. (1993), “Loan screening within and outside of customer relationships”,
Federal Reserve Bank of Philadelphia Working Paper No. 93-15.
Nakamura, R. (1985), “Agglomeration economies in urban manufacturing industries: a case of
Japanese cities”, Journal of Urban Economics, Vol. 17, pp. 108-24.
Rauch, J.E. (1993), “Productivity gains from geographic concentration of human capital: evidence
from the cities”, Journal of Urban Economics, Vol. 34 No. 3, pp. 380-400.
Riordan, M.H. (1993), “Competition and bank performance: a theoretical perspective”, in Mayer, C.
and Vives, X. (Eds), Capital Markets and Financial Intermediation, Cambridge University
Press, Cambridge, pp. 328-43.
Sealey, C. Jr and Lindley, J.T. (1977), “Inputs, outputs, and a theory of production and cost at
depository ?nancial institutions”, Journal of Finance, Vol. 32, pp. 1251-66.
Shaffer, S. (1998), “The winner’s curse in banking”, Journal of Financial Intermediation, Vol. 7
No. 4, pp. 359-92.
Shaffer, S. (2004), “Patterns of competition in banking”, Journal of Economics and Business,
Vol. 56 No. 4, pp. 287-313.
Strumsky, D., Lobo, J. and Fleming, L. (2005), “Metropolitan patenting, inventor agglomeration
and social networks: a tale of two effects”, Los Alamos National Laboratory Technical
Report LAUR-04-8798.
White, H. (1980), “A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity”, Econometrica, Vol. 48 No. 4, pp. 817-38.
Further reading
Jackson, W.E. (1992), “Is the market well de?ned in bank merger and acquisition analysis?”,
The Review of Economics and Statistics, Vol. 74, November, pp. 655-61.
Kwast, M., Starr-McCluer, M. and Wolken, J.D. (1997), “Market de?nition and the analysis of
antitrust in banking”, Antitrust Bulletin, Vol. 42 No. 4, pp. 973-95.
Whitehead, D.D. (1980), “Relevant geographic banking markets: how should they be de?ned?”,
Federal Reserve Bank of Atlanta Economic Review, January/February, pp. 20-8.
Corresponding author
Sherrill Shaffer can be contacted at: [email protected]
External
economies
in banking
365
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
4
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_675103274.pdf