Journalof

BANKING &
ELSEVIER Journal of Banking & Finance 19 (1995) 1175-1189 FINANCE

Scale economies and cost complementarities in

commercial banks: On-and off-balance-sheet
activities
Julapa Jagtiani a,*, Alli Nathan b, Gordon Sick b
a Baruch College, City University of New York, 17 Lexington Avenue, New York, NY 10010, USA
b Faculty of Management, The University of Calgary, 2500 University Drive N.W., Calgary, AB
T2N-1N4, Canada,
Received June 1992; final version received February 1994

Abstract
Bank off-balance-sheet (OBS) activities have grown dramatically, and have become
significantly important especially at large banks. We investigate whether failure to incorpo-
rate OBS products may lead to a misspecification problem. Models that exclude deposits
from the output specification suggest economies of scale, which vanish when deposits are
included. We find that inclusion of OBS products has little effect on the scale economies
measures. With only a few exceptions, the results provide no evidence of cost complemen-
tarities in joint production. The tremendous growth of bank OBS activities in the 1980s may
be explained by the very small (approximately zero) pecuniary cost of these activities.

JEL classification: G21; C31

Keywords: Banks; Cost efficiencies;Off-balancesheet activities;Scale economies;Cost complementar-

ities

* Corresponding author: 278 Stamford Avenue, Stamford, CT 06902, U.S.A. Tel.: 203-325-4620,
Fax: 203-325-1939.

0378-4266/95/$09.50 © 1995 Elsevier Science B.V. All fights reserved

SSDI 0378-4266(94)00078-6
1176 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189

1. Introduction

This paper examines the impact of including off-balance-sheet (OBS) products

in measures of bank cost efficiency. Several studies have analyzed cost efficiency
in the banking industry 1, but, none have investigated the impact of OBS products,
which have grown dramatically over the past decade 2. At year-end 1989, OBS
outputs of all US banks amounted to $5,700 billion, compared to on-balance-sheet
activities of $3,200 billion.
The new risk-based capital requirements (Basle Accord 1988) could be viewed
as a regulatory tax that could cause banks to shrink their OBS production - with
potential impact on bank cost efficiency. Previous studies such as Jagtiani et al.
(1995), Jagtiani (1996), and Pavel and Phillis (1987) suggest that in the flat-rate
capital requirement period, OBS activities may have been used as a means to
avoid tying up capital. Indeed, we find that OBS activities have little impact on
bank costs.
The results of this paper are significant for three reasons. First, we find that
efficiency measures are sensitive to the model specification - ignoring deposits
and, to a lesser extent, OBS products as outputs may result in a misspecification of
the cost function. Second, we find no economies of scope in production, except
between the following outputs: OBS guarantees and OBS foreign currency prod-
ucts; OBS foreign currency products and deposits; and, aggregate OBS items and
earning assets. Third, we find that the growth of bank OBS activities is likely to
continue unless there are non-pecuniary costs of these activities, or the net
revenues from these activities become negligible.
The rest of the paper is organized as follows. Section 2 reviews the theoretical
concepts and the empirical methods used in the cost structure analysis. Section 3
describes the data and the classification of output and input price variables.
Section 4 discusses the empirical results, and Section 5 presents some conclusions
and policy implications.

1 Previous studies can be grouped into four categories based upon the institutions examined -(a)
small banks: Benston et ai., 1982, Berger et al., 1987, Gilligan and Smirlock, 1984, Gilligan et al.,
1984, Gropper, 1991, Kolari and Zardkoohi, 1987, Lawrence and Shay, 1986, and Shaffer, 1984; (b)
large banks: Hunter and Timme, 1986, Hunter and Timme, 1991, Hunter et al., 1990, Noulas et al.,
1990, Shaffer and David, 1991 and Shaffer, 1993; (c) large and small banks: Berger and Humphrey,
1991, Nathan, 1990, and Nathan and Neave, 1989, Nathan and Neave, 1992; and (d) non-bank financial
institutions: Goldberg et al., 1991, Le Compte and Smith, 1990, Mester, 1987, Mester, 1990, Mester,
1991, and Murray and White, 1983. Also see Gilbert, 1984, and Clark, 1988, for a survey.
2 Hunter, Timme and Yang (1990) include a proxy for some OBS items (letters of credit and
securitization activities) in their study, but do not specifically analyze the effects on cost efficiency.
Mester (1990) incorporates loan sales as a non-traditional output, but does not take into consideration
other OBS products.
 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189 1177

2. Theoretical concepts and empirical methods

We estimate multi-output cost functions of large banks in order to study scale

economies, product cost complementarities, and the sensitivity of costs to various
outputs. We use a translog cost function (Christensen et al., 1973) with the
corresponding cost share equations:
In C = I + Eiotiln Yi + ~,j/3yln wj + 1//2~',iEk/x/kin yiln Yk

+ 1/2F.jF, h3,jhln wjln w h + EiEjt3ijln yiln wj + e, (1)

Sj = flj + Eh3%lnw h + E,~ijln y, + Ej, (2)

where C = total costs; Sj = cost share of input j ( j = 1, 2 . . . . m); Yi = output

level of product i (i = 1, 2, ... n); wy = price of input j ( j = 1, 2., ... m); and e,
~j = normal disturbance terms with zero means ( j = 1, 2, . . . m). Symmetry
requires that /xik =/xk~ for all i, k; and, T~h = Thj for all j, h. Linear homogeneity
in prices requires that E j /3j = 1; 7~h "/jh = 0 for all j; ~2j ~ij = 0 for all i. We
estimate the system with these restrictions. We use the Diewert and Wales (1987)
measures of scale economies (SCALE) and cost complementarities between
outputs i and k (COMPik) , specified as follows:

SCALE = E~(01n C( Y , W ) /OIn Yi) = Ei( Oli "[- ~-~k/x~kln Yk + EjS, jln wj),
(3)
COMP/k = it2C(Y,W)/OyiOyk ( i , k = 1,2 . . . . n)

ot [/xik + ( a i + ~t/x,tln Yl + ~2jaijln w~)

X ( otk + ~,l I"t'tkIn Yl + Ej6kj In wj)]. (4)

Both SCALE and COMP/k are local measures. If S C A L E < 1, then there are
economies of scale. If COMPik < 0, then an increase in the level of output k will
reduce the marginal cost of producing output i and there is a cost complementarity
between outputs i and k. To evaluate these measures, we use the sample mean
values of input prices and output levels. We also examine S C A L E at means of the
size quartiles.
Mester (1987) proposes a global measure of economies of scope: SCOPE---
•i[ C(0 . . . . . Yi, "" . 0 ) - C ( Y ) ] / C ( Y ) , where C(0 . . . . . Yi, . . . 0 ) is used as a
proxy for the production costs of the single output y~. To use. this measure, one
must assume that cost structures of single-and multi-product firms are comparable.
In addition, the translog cost function is undefined for zero output levels; so,
Mester and others use ad hoc values of 0.01 or 0.001 as proxies for zero output
levels. The conclusions drawn from these papers can be sensitive to the choice of
these ad hoc parameters, because the cost function is steeply sloped near the
1178 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189

singularity at zero. 3 Furthermore, the zero output levels require an extrapolation

beyond the data set. Our local measure of cost complementarities is not subject to
this concern since it has been shown that the translog functional forms perform
well locally (R611er, 1990).

3. Data and classification of input and output variables

Our analysis is based on quarterly Call Report data as of December 31, 1988. 4
This period of study starts immediately after the approval of the risk-based capital
requirement in July 1988, but prior to its actual implementation in December
1990. From the population of approximately 14,000 banks, our sample includes
the 134 banks that simultaneously produce all OBS outputs - guarantees, foreign
currency products and interest rate products. Across the sample of banks, total
assets range from $355 million to $154 billion (average $12 billion), and total
OBS outputs range from $30 million to $766 billion (average $28 billion).
We use two different approaches to define on-balance-sheet outputs. Under the
intermediation approach, earning assets (loans and investments) are considered to
be bank outputs. Deposits are regarded as intermediate outputs, which are com-
bined with labour and physical capital to produce loans and investments. Under
the value-added approach, both earning assets and deposits are considered as bank
outputs. Our OBS outputs are divided into three categories: guarantees, foreign
currency transactions, and interest rate transactions. We estimate eight different
models, which differ only in the specification of the output vector. Each model is
estimated separately and independently, and inferences from each model are
compared. All the models include labour, physical capital and financial capital as
inputs.

4. Empirical results

Table 1 defines inputs, outputs and price variables. Table 2 specifies the eight
models (a)-(h), with various combinations of outputs: on-balance-sheet: earning

3 The expansion path economies introduced by Berger, Hanweck and Humphrey (1987) help to
mitigate these problems. They analyze scale and scope economies by focusing on an expansion path
between two output bundles with different product mixes, which do not involve zero output quantities.
Their measures, however, allow product mixes to vary as outputs expand.
4 In response to concerns regarding possible window-dressing effects on year-end data, we also
perform the same analysis on September 1988 data. The results are similar in both quarters. The results
based on September 1988 data are available from the authors. Using year-end output data (rather than
the average over the year) does not affect our results, assuming that the output level differentials across
banks do not fluctuate widely during the year.
 J. Jagtiani et al. / Journal of Banking & Finance 19 (1995) 1175-1189 1179

Table 1
Description of output and input price variables
Variable Definition
Earning assets $ amounts of loans and investments
Deposits $ amounts of total deposits
OBS guarantees $ amounts of standby letters of credit,
commercial letters of credit,
and loan commitments
OBS foreign currency $ amounts of foreign currency swaps,
foreign currency options, and foreign currency
futures and forward contracts
OBS interest rate $ amounts of interest rate swaps,
interest rate options, and interest rate
futures and forward contracts
Aggregate OBS items $ amounts of OBS Guarantees,
OBS Foreign Currency Items,
and OBS Interest Rate Items
Price of labour Ratio of $ amounts of wages and salaries
to total number of employees
Price of physical capital Ratio of $ amounts of premises
and other expenses to $ amounts
of total assets
Price of financial capital Ratio of $ amounts of interest
expenses to $ amounts of total deposits

assets and deposits; and O B S : guarantees, f o r e i g n c u r r e n c y and interest rate

products. O B S outputs are e x a m i n e d both in a g g r e g a t e and individually. T a b l e 2
also s u m m a r i z e s our f i n d i n g s o f scale e c o n o m i e s and cost c o m p l e m e n t a r i t i e s .
T a b l e 3 presents the e s t i m a t e d cost c o m p l e m e n t a r i t i e s ( C O M P i k ) and scale
e c o n o m i e s ( S C A L E ) for the full sample. W i t h i n e a c h m o d e l w e test the null
h y p o t h e s e s S C A L E = 1 and COMPik = 0. T a b l e 4 p r e s e n t s the e s t i m a t e d S C A L E
v a l u e s for f o u r size-quartiles ranked by total assets. 5

4,1. Economies o f scale

W h e n only earning assets and O B S items are i n c l u d e d as outputs, as in M o d e l s

(c) and (d) o f T a b l e 3, the S C A L E index is less than unity. T h i s is consistent w i t h
other studies that identify significant scale e c o n o m i e s for large banks, and thereby

5 The estimated cost function satisfies the concavity conditions if the matrix [ F - (DIAG(S) -
ssT}] is negative semidefinite, where r = ['Yjh] is the matrix of second-order partial derivatives of the
transiog cost function with respect to the input prices wj; S is the vector of cost shares sfi and,
DIACAS) is the diagonal matrix formed from S (Diewert and Wales, 1987). All the models estimated in
our analysis satisfy the concavity conditions. The' results are available from the authors.
Table 2
Specifications of the models and summary of the results
Model Model classification Output variables Results
t~

Scale economies Cost complementarities

(at sample mean)
Model (a) Intermediation Earning assets No
Model (b) Value added Earning assets, deposits No No

Model (c) Intermediation Earning assets, aggregate OBS items Yes No e,

Model (d) Intermediation Earning assets, OBS guarantees, Yes No
OBS foreign currency, OBS interest rate
Model (e) Deposits, aggregate OBS items No No
Model (0 Deposits, OBS guarantees, OBS foreign currency, No No (except between OBS
e~
OBS interest rate guarantees and OBS foreign ~t

currency; and, OBS foreign

currency and deposits)
Model (g) Value added Earning assets, deposits, aggregate OBS items No No (except between aggregate
OBS items and earning assets)
Model (h) Value added Earning assets, deposits, OBS guarantees, No No (except between OBS
OBS foreign currency, OBS interest rate guarantees and OBS foreign currency)
L
 J. Jagtiani et al./Journal of Banking & Finance 19 (1995) 1175-1189 1181

infer that regulatory constraints such as interstate branching restrictions may be

limiting efficiencies in the banking system.
However, the SCALE index is not significantly different from unity in all the
other models: Model (a) only earning assets; Model (b) earning assets and
deposits; Model (e) deposits and aggregate OBS items; Model (f) deposits and
individual OBS items; Model (g) earning assets, deposits, and aggregate OBS
items; and Model (h) earning assets, deposits, and individual OBS items. In
particular, the SCALE index is insignificant in all the models that include deposits
as a relevant output. Thus, the scale economies identified in Models (c) and (d)
may be the result of output misspecification. The measures of scale economies are
sensitive to output definition, and care must be taken before drawing normative
inferences.
Table 4 shows the SCALE measure of each model for each size quartile. The
shape of the cost function is also sensitive to output specification. In Models (c)
and (d), where SCALE values are significantly different from one at the sample
mean, the quartile values indicate that the SCALE values are significant for the
smaller banks and for the mid-sized banks. Thus, the average cost curve is
L-shaped for these models.
In all other models, the SCALE values are not significantly different from one.
In Models (a) and (b), we observe that the rate at which the scale economies arise
increases with bank size. Thus, the average cost curves for these models have the
shape of an inverted J. On the other hand, Models (e), (f), and (h) exhibit average
cost curves that are L shaped, while Model (g) has a cost curve that is J shaped.

4.2. Cost complementarities

In Table 3, the estimates of COMPik are not significantly different from zero,
except for the following joint production pairs: OBS foreign currency products and
OBS guarantees in models (f) and (h); OBS foreign currency products and
deposits in model (f); and aggregate OBS products and earning assets in model
(g).
The lack of cost complementarities can be interpreted in two ways. First,
suppose that the objective of the regulator is to influence banks to move to a
socially efficient minimum-marginal-cost production function. Then, if COMPIk >
0, allowing the banks to expand output i results in higher marginal cost of output
k. This might require that the regulator constrain the banks' level of output i. If
COMP;k < 0, then the marginal cost of k would fall if any constraint on i is
loosened. Whenever COMPik = 0, there is no need for regulatory action, because
the optimizing behaviour of the banks results in a socially efficient optimum. The
second interpretation is that there may exist no cost complementarities at all -at
least for many of the different output combinations. Banks may be offering OBS
products for other reasons, such as regulatory avoidance of capital adequacy
Table 3
Cost complementarifies and scale economies measures. Cost complementarities measure (COMPi~) is reported for each pair of outputs in each model. Scale
economies measure (SCALE) is reported for each model. Input variables are the same in all models -include labour, physical capital and financial capital.
Both cost complementarities and scale economies measures are evaluated at the sample mean, based on quarterly data ending December 31, 1988. Standard
errors are reported in parentheses. Significance at the 5% level for H0: SCALE = 1 and H0: COMPI~ = 0 is indicated by ° and * * respectively. X 2 represents
the test statistic for no contemporaneous correlation among the residuals. Log(L) is the logarithm of the likelihood function of the maximum likelihood
estimation
Earning assets Deposits Aggregate OBS item OBS guarantees OBS foreign currency OBS interest rate SCALE
Model (a):
Faming assets __ 0.978
(0.012)
X 2 = 66; log(L)= 567
Model (b):
Earning assets - 0.020 0.999 e~

(0.427) (0.006)
Deposits
X 2 = 206; log(L) = 835
Model (c)f
Earning assets -0.017 0.958"
(0.269) (0.017)
Aggregate OBS items
X 2 = 54; log(L) = 570
Model (d):
Faming assets - 0.027 - 0.003 - 0.002 0.956 *
(0.402) (0.027) (0.056) (0.018) L
OBS guarantees 0.011 - 0.025
(0.011) (0.018)
OBS foreign currency 0.004
(0.006)
OBS interest rate
X 2 ffi 58; log(L) = 576
Model (e):
Deposits 0.015 0.998
(0.038) (0.008)
Aggregate OBS item
X 2 = 207; log(L) = 821
Model 09:
- 0.006 0.012 * * -0.002 0.987
Deposits
(0.047) (o.o04) (0.009) (0.007)
OBS guarantees -0.005 * * 0.003
(0.001) (0.002)
OBS foreign currency - 0.0002
¢....,
(0.001)
OBS interest rate
X z = 193; log(L) = 844
Model (g):
Earning assets - 0.019 - 0.039 * * 1.oo5
(0.424) (0.018) (0.OO7)
Deposits 0.035
(0.030)
Aggregate OBS items
X 2 = 197; log(L) = 847
Model (h):
Earning assets 0.030 - 0.034 -0.0004 - 0.006 0.995
(0.488) (0.018) (0.004) (0.007) (0.007)
0.004 0.011 - 0.001
Deposits
(0.038) (0.006) (0.009)
L
OBS guarantees - 0.004' * 0.003
(0.001) (0.002)
OBS foreign currency 0.0001
(0.0005)
OBS interest rate
X 2 = 193; log(L) = 876
 1184 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189

Table 4
Scale economies measure by asset-size-based quartiles. The estimated of scale economies measures
(SCALE) are evaluated in each model based on each asset-size-quartiles and the overall sample mean
output levels. Sample period is quarterly data ending December 31, 1988. Standard errors are reported
in parentheses. Significance at the 5 percent level (H0: SCALE = 1) is indicated by *.
Model Quartile 1 Quartile 2 Quartile 3 Quartile 4 Sample mean
specification (smallest) (largest)
Model (a) 0.996 0.981 0.974 0.961 0.978
(0.0219) (0.0127) (0.0135) (0.0224) (0.0124)
Model (b) 1.002 1.0001 0.999 0.998 0.999
(0.0057) (0.0056) (0.0057) (0.0063) (0.0055)
Model (c) 0.952 0.957" 0.959" 0.964 0.958 '
(0.0309) (0.0202) (0.0176) (0.0225) (0.0168)
Model (d) 0.935 * 0.959 0.961 0.968 0.956 '
(0.0323) (0.0212) (0.0221) (0.0236) (0.0183)
Model (e) 0.995 0.996 0.997 1.003 0.998
(0.0083) (0.0078) (0.0077) (0.0077) (0.0076)
Model (f) 0.987 0.986 0.985 * 0.989 0.987
(0.0079) (0.0076) (0.0076) (0.0074) (0.0074)
Model (g) 1.002 1.004 1.005 1.008 1.005
(0.0080) (0.0076) (0.0075) (0.0074) (0.0073)
Model (h) 0.995 0.996 0.994 0.996 0.995
(0.0073) (0.0072) (0.0074) (0.0071) (0.0070)

requirements, and/or strategic objectives related to customer demand, revenue

advantages, and risk management.
The complementarity values, COMPik, are significantly less than zero for OBS
foreign currency products and OBS guarantees in Models (f) and (h), and for
aggregate OBS items and earning assets in Model (g). This is consistent with the
possibility of synergies arising from the use of shared information on customers,
as the products are often offered to those existing borrowers who have a reputable
credit relationship with the banks. In addition, some of the foreign currency
products may benefit from insurance-related links, if they are being offered as
hedging instruments.
On the contrary, we observe diseconomies in the joint production of deposits
and OBS foreign currency items, in Model (f). This is surprising, given the

Note to Table 3:
At the 1 percent significance level, none of the SCALE values are significantly different from one,
indicating efficient scale of operations in all models irrespective of the choice of outputs. The results
regarding COMPu, remain the same as those reported based on the 5 percent level, except in Model (g)
where the significant cost complementarities observed between aggregate OBS items and earning assets
disappears at the 1 percent level, indicating efficient scope of operations. The results are also confirmed
based on the Bonferroni t-test, as discussed in Section 4.2.
 J. Jagtiani et al. / Journal of Banking & Finance 19 (1995) 1175-1189 1185

possibility of synergies arising from customer and insurance-related links dis-

cussed above. However, this may also be a result of model misspecification, as
Model (f) does not include earning assets in its output specification.
The above conclusions are based on non-zero values of COMP/k , each tested
separately at the 5% level, using standard t-tests. However, within each model, it
is possible to test for the joint hypothesis of non-zero values for all COMPik,
using the Bonferroni t-test (Dunn, 1961 and Savin, 1980). The inferences are the
same as those reported above - the only exception being Model (g), in which
COMPik becomes insignificant with Bonferroni t-test. Thus, we cannot reject the
joint hypothesis of no cost complementarities in all models except in Models (f)
and (h). Here, the joint hypothesis is rejected in favour of the cost advantages in
the joint production of OBS guarantees and OBS foreign currency items and the
cost disadvantages in the joint production of deposits and OBS foreign currency
items. 6

4.3. Graphic presentation of the cost function

Fig. 1 shows a graph of Model (g), which has cost as a function of the three
outputs: deposits, earning assets, and aggregate OBS activities. 7 In the top left
panel, cost is a function of deposits and earning assets, with OBS activities set at
the mean of the logarithm. Cost is much more sensitive to the volume of deposits
than it is to earning assets. The top right panel allows costs to vary with OBS
volume and earning assets, and it is clear that costs do not vary strongly with OBS
activities. From the bottom panel, when cost is allowed to vary with OBS volume
and deposits, the graph shows that cost is again not sensitive to OBS volume.
The low sensitivity of bank costs to OBS volume suggests that the tremendous
growth of OBS activities could continue into the future, if these activities continue
to provide benefits to banks in the form of better risk management or increased fee
income. However, the growth may not continue if there are some non-pecuniary
costs involved, or if there are no further gains from risk management (if banks are
already in a fully hedged position). The slight diseconomies between deposits and
OBS foreign currency items observed in Model (f) and the higher sensitivity of
costs to deposits may indicate the possibility of such cost disadvantages.
The approach we have used identifies the technical or operational efficiency of
the cost structure by focusing on the cost-minimizing behaviour of the banks. This

6 Similar conclusions are also indicated based on the standard t-test procedure at the 1% significant
level as noted in Table 3.
7 The outputs are allowed to vary across the range of values assumed by the banks in our sample.
Similar graphs were developed for other models, but the main characteristics appear in Model (g).
1186 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189

Cost Cost
($ooo) \ ($ooo)
\
~,I.00E+08 1.00E+08

~'~.I.00E+07 1.00E+07
,<
~1.00E+06 1.00E+06
Y
\
1.00E+05 1.00E+05
o ~'~ ~ ~ "~. ~
1.00E+04 ~ ~, "~, ~, ~ 1.00E+04

+o + ~ P, ~ ~,
~8uI ~
• ~ LU ,,+, ,,+, ,.: oo uj
~ 0 ~D 0 iO 0 iO 0 ~D 0
o. ~ Q ~ o. • q ~ q ~ q
Deposits ~ o5 ~ ¢6 OBS Itern~ ~ e5 ~ ~
Assets ($000) ($000) Assets ($000)
($000)

Cost
/ ($ooo)
/ 1.00E+08
/'y
f
~ .00E+07

< "~ .00E+06

1.00E+05
o ~ ~ ~ ~ ~

i i
1.00E+04
00
O
+
LLI
O

OBS I t e m s - ~ ¢6 ~
($000) Deposits ( $ 0 0 0 )

Fig. 1, Graphic presentation of bank cost as a function of deposits, earning assets, and ag~'cgat¢ O B S
activities, using Model (g), estimated on December 1988 data. Logarithms of non-plotted variables are
set at their means. All axes have a logarithmic scale; all axis labels are in dollars.

can yield normative implications for bank regulation. An alternative approach is to

examine the profit function of the banks, to account for the revenue incentives to
incur costs. This allows one to study the price or allocational efficiency of the
banks. Our objective was to examine how the introduction of OBS items affects
the inferences that have been drawn in other Studies about bank cost structure, so
we used the translog cost function, rather than a profit function.
 J. Jagtiani et al. /Journal o[Banking & Finance 19 (1995) 1175-1189 1187

5. Conclusions and policy implications

We find that empirical results on bank cost efficiencies are sensitive to output
specification. Both the mean measure of scale economies and the implied average
cost curve are sensitive to output definition - for example, models that exclude
deposits from the output specification suggest economies of scale, which vanish
when deposits are included. OBS products seem to have little or no significant
effect on the scale economies measures. The results in general suggest the lack of
scale economies. This may be due to the fiat-rate capital requirements (during the
period of study), which restricted on-balance-sheet outputs while encouraging
expansion of OBS outputs. This allowed the banks to move to an optimal output
level, by using the OBS items to proxy for the constrained items on the balance
sheet.
Further, we find that for most combinations of outputs, there is no evidence of
cost complementarity. The only exceptions involve either OBS foreign currency
products and guarantees, or aggregate OBS items and earning assets. In this case,
the regulators could consider encouraging joint production of these activities,
probably by imposing smaller capital requirements (smaller regulatory tax), in
order to increase output efficiency in the banking system. A certain amount of
caution, however, should be warranted given the diseconomies between deposits
and OBS foreign currency items.
Finally, the graphical representation of our results suggest that, unlike earning
assets and deposits, the volume of OBS activities has little or no impact on bank
costs. This suggests tl~at OBS activities could continue to grow as long as they
generate fee income, unless there are non-pecuniary costs of these activities such
as insolvency risk. Regulatory taxes on OBS activities, such as the risk-based
capital requirements, may be necessary in order to control the growth of these
activities, only if non-pecuniary social costs are found.

Acknowledgements

We thank Allen Berger, Curt Hunter, Douglas Evanoff, the anonymous referee,
and participants at the FMA and ASAC conferences for helpful comments,
suggestions, and discussions. Financial support from the Alberta Energy Company
and Social Sciences and Humanities Research Council of Canada are gratefully
acknowledged.

References

Benston, G.J., G.A. Hanweck and D.B. Humphrey, 1982, Scale economies in banking - A restructuring
and reassessment, Journal of Money, Credit and Banking 14, 435-456.
1188 J. Jagtiani et al. /Journal of Banking & Finance 19 (1995) 1175-1189

Berger, A.N., G.A. Hanweck and D.B. Humphrey, 1987, Competitive viability in banking -Scale,
scope and product mix economies, Journal of Monetary Economics 20, 501-520.
Berger, A.N. and D.B. Humphrey, 1991, The dominance of inefficiencies over scale and product mix
economics in banking, Journal of Monetary Economics 28, 117-148.
Christensen, L.R., D.W. Jorgenson and L.J. Lau, 1973, Transcendental logarithmic production fron-
tiers, Review of Economics and Statistics 55, 28-45.
Diewert, W.E. and T.J. Wales, 1987, Flexible functional forms and global curvature conditions,
Econometrica 55, 43-68.
Dunn, O.J., 1961, Multiple comparisons among means, Journal of the American Statistical Association,
56, 52-64.
Gilligan, T. and M.L. Smirlock, 1984, An empirical study of joint production and scale economies in
commercial banking, Journal of Banking and Finance 8, 67-77.
Gilligan, T., M.L Smirlock and W. Marshall, 1984, Scale and scope economies in the multi-product
banking firm, Journal of Monetary Economics 13, 393-405.
Goldberg, L., G. Hanweck, M. Keenan and A. Young, 1991, Economics of scale and scope in the
securities industry, Journal of Banking and Finance 15, 91-107.
Grupper, D.M., 1991, An empirical investigation of changes in scale economics for the commercial
banking firm, 1979-1986, Journal of Money, Credit and Banking 23, 718-727.
Hunter, W.C. and S.G. Timme, 1986, Technical change, organizational form and the structure of bank
production, Journal of Money, Credit and Banking 18, 152-166.
Hunter, W.G. and S.G. Timme, 1991, Technological change in large U.S. commercial banks, Journal of
Business 64, 339-362.
Hunter, W.C., S.G. Timme and W.K. Yang, 1990, An examination of cost subadditivity and
multiproduct production in large U.S. banks, Journal of Money, Credit and Banking 22, 505-525.
Jagtiani (Rungkasiri), J., 1996, Characteristics of banks that are more active in the swaps market,
Journal of Financial Services Research 10(2), forthcoming.
Jagtiani, J., A. Saunders and G. Udell, 1995, The effect of bank capital requirements on bank
off-balance sheet financial innovations, Journal of Banking and Finance, 19, 647-658.
Kolari, J. and A. Zardkoohi, 1987, Bank costs, structure and performance (Lexington, MA).
Lawrence, C. and R.P. Shay, 1986, Technology and financial intermediation in multiproduct banking
firms - An econometric study of U.S. banks, In: C. Lawrence and R.P. Shay, eds., Technological
innovation, regulation and the monetary economy (Ballinger, Cambridge).
Le Compte, R.L.B. and R.D. Smith, 1990, Changes in the cost of intermediation: The case of savings
and loans, Journal of Finance 45, 1337-1346.
Mester, LJ., 1987, A multiproduct cost study of savings and loans, Journal of Finance 42, 423-445
Mester, L.J., 1990, Traditional and nontraditional banking: An information theoretic approach, working
paper (Federal Reserve Bank of Philadelphia).
Mester, LJ., 1991, Agency costs among savings and loans, Journal of Financial Intermediation 1,
257-278.
Murray, J.D. and.R.W. White, 1983, Economics of scale and economics of scope in multiproduct
financial institutions: A study of British Columbia credit unions, Journal of Finance 38, 887-902.
Nathan, A., 1990, Cost efficiency and competitiveness of Canadian banks and trust companies,
Unpublished Doctoral Dissertation (Queen's University).
Nathan, A. and E.H. Neave, 1989, Competition and contestability in Canada's financial system:
Empirical results, Canadian Journal of Economics 22, 576-594.
Nathan, A. and E.H. Neave, 1992, Operating efficiency of Canadian banks, Journal of Financial
Services Research 6, 265-276.
Noulas, A.G., S.R. Ray and S.M. Miller, 1990, Returns to scale and input substitution for large U.S.
banks, Journal of Money, Credit and Banking 22, 95-108.
Pavel, C. and D. Phillis, 1987, Why commercial banks sell loans: An empirical analysis economic
perspectives, Federal Reserve Bank of Chicago, 3-14.
 J. J agtiani et al. / Journal of Banking & Finance 19 (1995) 1175-1189 1189

R6ller, L.H., 1990, Modelling cost structure: The Bell system revisited, Applied Economics 22,
1661-674.
Savin, N.E., 1980, The Bonferroni and the Scheffe multiple comparison procedures, Review of
Economic Studies XLVII, 255-273.
Shaffer, S., 1984, Scale economics in multiproduct finns, Bulletin of Economics Research 36, 51-67.
Shaffer, S., 1993, A test of competition in Canadian banking, Journal of Money, Credit and Banking
25, 49-61.
Shaffer, S. and E. David, 1991, Economies of superscale in commercial banking, Applied Economics
23, 283-293.