Description
The purpose of this paper is to investigate use of efficiency analysis as a technique for
investigating bank safely and soundness
Journal of Financial Economic Policy
Using DEA to investigate bank safety and soundness – which approach works best?
David Tripe
Article information:
To cite this document:
David Tripe, (2010),"Using DEA to investigate bank safety and soundness – which approach works best?",
J ournal of Financial Economic Policy, Vol. 2 Iss 3 pp. 237 - 250
Permanent link to this document:
http://dx.doi.org/10.1108/17576381011085449
Downloaded on: 24 January 2016, At: 21:38 (PT)
References: this document contains references to 19 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 365 times since 2010*
Users who downloaded this article also downloaded:
Abdelkader Boudriga, Neila Boulila Taktak, Sana J ellouli, (2009),"Banking supervision and nonperforming
loans: a cross-country analysis", J ournal of Financial Economic Policy, Vol. 1 Iss 4 pp. 286-318 http://
dx.doi.org/10.1108/17576380911050043
Sven Blank, J onas Dovern, (2010),"What macroeconomic shocks affect the German banking system?:
Analysis in an integrated micro-macro model", J ournal of Financial Economic Policy, Vol. 2 Iss 2 pp.
126-148 http://dx.doi.org/10.1108/17576381011070193
Irina Barakova, Ajay Palvia, (2010),"Limits to relative performance evaluation: evidence from
bank executive turnover", J ournal of Financial Economic Policy, Vol. 2 Iss 3 pp. 214-236 http://
dx.doi.org/10.1108/17576381011085430
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Using DEA to investigate bank
safety and soundness – which
approach works best?
David Tripe
Centre for Banking Studies, Massey University, Palmerston North, New Zealand
Abstract
Purpose – The purpose of this paper is to investigate use of ef?ciency analysis as a technique for
investigating bank safely and soundness.
Design/methodology/approach – Three different data envelopment analysis (DEA) models were
applied to set of data for the major New Zealand banks over a ten-quarter period – a CCR model,
a pro?t ef?ciency model and a non-oriented slacks-based approach.
Findings – Most useful results are obtained using the slacks-based approach.
Research limitations/implications – The period covered by the study was from late 2005
until early 2008, prior to the global ?nancial crisis having major impacts on the New Zealand
banking sector.
Practical implications – The study is of particular value in the New Zealand context where there
has historically not been any bank deposit insurance, obliging depositors to make their own
assessments of bank safety and soundness.
Originality/value – The paper makes a contribution to very small literature which uses ef?ciency
analysis to explore bank safety and soundness. It also makes use of the slacks-based DEA approach,
which has not yet been widely used in the banking literature.
Keywords Banking, Data analysis, New Zealand
Paper type Research paper
1. Introduction
Studies of bank and ?nancial institution ef?ciency have been undertaken for a variety
of purposes. Ef?ciency in ?nancial institutions is and should be a matter of public
concern, as not only can more ef?cient ?nancial institutions be expected to be more
pro?table, but one should also expect ?nancial institution ef?ciency to lead to greater
amounts of funds being intermediated, and better service at lower prices for
consumers. Other things being equal, more ef?cient ?nancial institutions should
exhibit greater safety and soundness (Berger et al., 1993, pp. 221-2), while also showing
better credit quality in the loan portfolio (Berger and DeYoung, 1997). These issues also
provide a basis for looking at the ef?ciency implications of both bank mergers and
acquisitions and government policy initiatives.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
The author would like to thank Knox Lovell for asking a question and suggesting the idea that
underpins this paper. The author would also like to thank participants at the eleventh EWEPA at
Pisa, the IDEAS Conference in Philadelphia and the fourteenth FINSIA/Melbourne Centre for
Financial Studies conference for comments on earlier versions of this paper. Particular thanks
are due to Mette Asmiild and Joe Paradi.
No blame should be attributed to anyone other than the author for any de?ciencies in the
paper.
Bank safety
and soundness
237
Journal of Financial Economic Policy
Vol. 2 No. 3, 2010
pp. 237-250
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381011085449
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This study focuses on bank safety and soundness, which has been of particular
importance in the New Zealand context because, prior to October 2008, there was no
general deposit protection scheme for bank customers. This meant that the onus of
assessing the soundness of the banks with which they dealt fell upon the banks’
customers: banks are requiredto make disclosures to the public on a quarterlybasis, and it
is this information which is also provided to the Reserve Bank of New Zealand as the
prudential supervisor of the banking system. The principle has been that if the Reserve
Bank had no better information thanwas made available to the general public, it could not
be liable to recompense depositors if a bank failed[1].
Banks’ quarterly disclosures include a year-to-date income statement and balance
sheet, which allow an analyst to construct a quarterly income statement, and thus
make a detailed assessment of ?nancial performance. Such an assessment of ?nancial
performance is likely to focus on key risk factors, and thus credit risk will be assessed
by levels of impaired and past due loans and bad and doubtful debt expense, interest
rate risk by the level of and volatility in a bank’s net interest income, operational risk
by the bank’s non-interest expense, and solvency by the ratio of capital to assets and
the ?ow of pro?ts to sustain that capital level. Relevant accounting ratios have been
compared between banks, and through time, in an effort to understand pro?tability
change from quarter to quarter, and to monitor exceptional items which might give
grounds for concern.
Grifell-Tatje´ and Lovell (1999) noted that a change in a ?rm’s pro?ts may derive from
a number of sources, including a change in input or output prices. The other sources of
pro?t change are likely to be measured using the techniques of ef?ciency analysis,
including technical change leading to an increase in output without any increase in
resource utilisation, animprovement or decline inoperating ef?ciency(X-ef?ciency), or a
change in output proportionately greater or less than input utilisation, re?ecting
economies or diseconomies of scale. Further, sources of pro?tability improvement
include changes in product or resource mix, associated with economies of scope and
allocative ef?ciency (Lovell, 2008). Ef?ciency analysis allows us to explore trade-offs
between the different inputs and outputs used, something which is inherently
problematic if one is constrained to use of ratios (Golany and Storbeck, 1999).
Against this background, it is considered appropriate to use the methods of
ef?ciency analysis to study the pro?t performance of New Zealand banks from quarter
to quarter, as part of the process of assessing their safety and soundness. We tested
three different approaches, and ?nd that the clearest distinctions are made between
individual bank performance when we use a slacks-based, data envelopment analysis
(DEA) method.
The rest of the paper proceeds as follows. In the next section, we outline the
approach that will be followed against the background of the key literature that is
relevant to this research. In Section 3, we look at the New Zealand banking system, and
the data that are generated from the disclosures to permit this research. Section 4
reports results that are obtained, while Section 5 provides a summary and conclusion.
2. The method for this research, and relevant prior research on bank
ef?ciency
Berger and Humphrey (1997) reported on 130 studies of bank ef?ciency, and a great
many more articles have been published since that review. These studies have used
JFEP
2,3
238
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
both parametric methods –stochastic frontier analysis (SFA), the distribution free
approach and the thick frontier approach – and non-parametric methods – DEA and
the free disposal hull. The key differences between the approaches are that the
non-parametric methods do not take account of random error in observations, while the
parametric approaches require speci?cation of a functional form. The complexity and
diversity of production relations in banking mean that the speci?cation of the
functional form is not self-evident: the imposition of an inappropriate functional form
can effectively distort the ?ndings of any analysis.
This study uses DEA, a number of forms of which have been developed since the
original constant returns to scale (CCR) model was developed by Charnes et al. (1978).
Banker et al. (1984) introduced a variable returns to scale (BCC) model, use of which
allows a more extensive exploration of scale economies. Further, subsequent variants
include the cost, revenue and pro?t models, which take account of input and output
prices, and which can thus provide estimates of allocative ef?ciency[2]. More recently,
again we have seen the introduction of slacks-based models (Tone, 2001), which
identify a broader range of inef?ciencies through consideration of non-radial slacks[3].
Another distinction between DEA and the parametric approaches is in the focus on
the individual decision-making unit (DMU – bank in this case). DEA both generates
individual ef?ciency scores for each bank for each time period, and identi?es
which inputs have been over-utilised and/or outputs underutilised to engender that
inef?ciency. Of the parametric approaches, only SFA provides an observation-speci?c
ef?ciency score, but DEA identi?es a peer group of more ef?cient units against which
an individual unit’s ef?ciency can be compared. Since we are looking to understand the
factors behind differences in individual bank ef?ciency and pro?tability, DEA is a
more appropriate method for this study.
Another factor that sometimes in?uences the choice between DEA and one of the
parametric approaches is the number of DMUs to be included in the analysis.
Parametric methods generally require larger data sets, not so much in that DEA is
better with a smaller data set, but rather because the greater ?exibility of DEA does
not preclude its use for smaller data sets. In this case, with a cross-section of no more
than six comparable banks for the period studied, DEA would be a necessity. Even
here, though, we need a greater number of observations to allow meaningful
differences in ef?ciency to be observed: we achieve this by studying our cross section
of banks over consecutive time periods, in a panel. We would wish to analyse the
data set as a panel in any case, however, because to understand pro?tability changes,
we need to be looking at previous accounting periods alongside the most recent ones.
The standard approach to looking at ef?ciency change between two different
periods is by use of the Malmquist index, which allows a decomposition of productivity
change into the effects of technological change and change in ef?ciency. This is not
what we are looking for in this study, where we expect pro?tability change to be
impacted more by changes in prices and quantities, and in the ef?ciency with which
inputs are used and outputs created. We are also looking at a relatively short-time
period, which means that the effects of any technological change that might invalidate
the use of panel data would not be expected to be major[4].
Another key issue in studying the ef?ciency of ?nancial institutions is in selecting a
model of the banking ?rm. The two main methods are the production and
intermediation approaches, although the production approach is more often used for
Bank safety
and soundness
239
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
measurement of the ef?ciency of individual branches, rather than for banks as a whole.
The intermediation approach exists in a number of different forms, but it is essentially
based on the process of converting depositors’ funds into loans, and it is the
predominant method for studying banks as a whole. The approach we use in this study
is broadly consistent with the intermediation approach.
Dyson et al. (2001) remind us that the important point in selecting inputs and
outputs for DEA studies (and, by implication, other ef?ciency studies as well) is that
we should be looking for input variables which include as much as possible of the
resources that DMUs use. In turn, we should be looking at output variables which
include as much as possible of the relevant outputs, but which also identify the key
success factors for the entities whose ef?ciency is being studied.
Relevant inputs and outputs can thus be identi?ed from looking at the way banks
generate pro?ts. They incur interest expense, and use non-interest expense and
shareholders’ equity to generate interest and non-interest income. A side effect of
this production process is bad and doubtful debt expense, which reduces pro?tability,
and which might thus be regarded as a negative output (and which can thus be
processed within DEA as an input)[5]. The interest income and expense can be
decomposed into prices (interest rates) and quantities, although such a decomposition
cannot meaningfully be conducted for the other inputs and outputs used.
Use of shareholders’ equity as an input is not universal in the prior literature, but it
can be justi?ed. In the ?rst place, shareholders’ equity is a source of funding, and if a
bank has more equity, it should need less funding (and thus less interest expense),
while the higher proportion of equity should reduce the cost of those funds. More
importantly, however, as discussed by Berger and Mester (1997), equity gives a bank a
greater cushion against risk, which allows it to increase the size of its loan portfolio
(which might also otherwise be constrained by capital regulation), and to take on
riskier loans which should be expected to generate greater interest revenues.
This contrasts with our decision to omit physical capital or staff as inputs, with
prior researchers often having calculated prices for these based on relevant
expenditure. In our view, a major problem with attempting to include physical capital
is that, in many cases, items of physical capital will be leased rather than owned. Even
if they were owned, there might be problems with their valuation[6]. In practical terms,
neither the costs associated with physical capital or staff costs are consistently
disclosed quarterly, and as staff numbers are not reported either, neither of these
measures which have commonly been used in previous research are available to us.
We are therefore obliged to use aggregate non-interest expense as an input[7].
Having identi?ed the method that should be used for his research, we now look at
the New Zealand banking system and the data that it provides for our study. In the
course of our discussion, we will also address some further methodological issues.
3. New Zealand banking data
We noted in the introduction that New Zealand lacked any formal deposit protection
scheme (prior to October 2008), and that supervision of the banking system was based
on public disclosure. The key elements of this are the quarterly disclosure statements,
which include year-to-date ?nancial statements. Even if this were not the basis for the
prudential supervision of the banks, the disclosure statements are a boon for
researchers in terms of the data they provide.
JFEP
2,3
240
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
One of the reasons why this approach to banking supervision is viable in NewZealand
is that the number of banks undertaking retail business is relatively small. As at 31 March
2008, the four largest banks accounted for 89.4 per cent of total banking system assets.
Two further smaller banks that also emphasized retail business accountedfor a further 2.9
per cent of banking system assets[8].
A further consideration in the Reserve Bank of New Zealand’s lighter approach
to banking supervision is the extent of foreign ownership of the banks, with more than
97 per cent of bank assets under foreign ownership at 31 March 2008. The four major
banks, ANZNational Bank(ANZ, ownedbythe ANZBanking Group Limited), ASBBank
(ASB, owned by the Commonwealth Bank of Australia (CBA)), Bank of New Zealand
(BNZ, owned by the National Australia Bank), and the Westpac Banking Corporation
(Westpac) are all Australianowned, withtheir parents all amongthe tenlargest companies
on the Australian Stock Exchange. The New Zealand-owned banks are Kiwibank
(government-owned through New Zealand Post) and TSB Bank (TSB, owned by a local
community trust), which have a combined market share of 2.9 per cent.
Our sample frame for this study thus comprises six banks, the four Australian-owned
majors and the two NewZealand-owned banks, which together dominate retail banking.
A limited amount of retail banking is provided by the New Zealand branches of HSBC
and (Korean-owned) Kookmin Bank, but these two entities show balance sheets and
income statements which are quite different from those included in the study – it is
argued that the necessary condition of relative homogeneity to allow reasonable
comparisons would not apply to these banks. Rabo NewZealand Limited is also omitted
from the study: although it is New Zealand incorporated, it specialises in lending to the
rural sector and would appear to rely to at least some extent on the Rabobank Nederland
New Zealand branch for its functioning.
During the period covered by this study, the CBA and Westpac each had two
registered banks operating in the New Zealand market. In the CBA’s case, the
overwhelming majority of its business is in its subsidiary, ASB, with this apparently
capable of surviving independently (which it did for a long time up to its parent’s
registration of a branch). In the Westpac case, however, the New Zealand subsidiary was
established primarily to comply with New Zealand regulatory requirements, and it is
doubtful that it could survive in its own right. Moreover, its relatively recent
establishment (in late 2006) means that it is not clear that it yet has a stable pattern of
operations. For both of these reasons, we have chosen to use the Westpac branch in the
study, rather than the subsidiary.
The initial analysis in this study covers the period from the December quarter
2005 through to the March quarter 2008 – ten quarters for six banks, giving us
60 observations in total, which should be ample relative to the combined total of six inputs
and outputs. A key factor in the choice of start date was the adoption of international
?nancial reporting standards (IFRS) by the major Australian banks (and thus for their
NewZealand operations as well), for their ?nancial years ending in 2006. IFRS will have
had a particular effect on bad and doubtful debt expense (in terms of removing banks’
discretion to record additional provisions without objective evidence of impairment),
on interest income and expense where fees and costs have to be amortised over the life
of the assets and liabilities to which they apply, and because of the way derivative
assets and liabilities have to be accounted for, with the revaluations impacting
particularly on non-interest income[9]. Following the adoption of IFRS, less of the
Bank safety
and soundness
241
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
variability in pro?tability ought to be able to be attributed to variation in accounting
practices[10].
Data were checked using a slacks-based super-ef?ciency model (Tone, 2002) and no
cases were identi?ed which suggested any problems[11]. A review of the correlations
between the inputs, the outputs and between the inputs and outputs also failed to
identify any concerns in respect of the data being used for the study.
A minor adjustment was required to the data for bad and doubtful debt expense,
to remove negative values and replace these with zeros. This was because the negative
values arti?cially shifted the ef?cient frontier to the extent that a negative ef?ciency
score was generated for one DMU in the slacks-based model. Negative values for bad
and doubtful debt expense would indicate an impact of recoveries of write-offs from
earlier accounting periods: there is thus an argument that a value of zero would more
fairly re?ect current performance. Signi?cant changes in ef?ciency scores were
generated only for one DMU, which continued to show signi?cant inef?ciency (but no
longer a negative score).
As a further preliminary check on our data, recognizing the range of bank size from
NZ$ 3 billion[12] for TSBup to NZ$ 100 billion for ANZ, we checked for scale effects. The
strongest indication of scale inef?ciency was found for ANZ, with mean scale ef?ciency
of 0.96, but which was nonetheless shown as being fully scale ef?cient in one quarter.
Noting the tendency of observations of different size to showas scale inef?cient because
of the relative dearth of other observations in that size bracket (Dyson et al., 2001), which
would be the case for ANZ, we have preferred to use constant returns to scale models so
as to increase the models’ sensitivity (by reporting numerically lower ef?ciency scores).
Moreover, having regard to the objectives of our study, we note that we want to identify
sources of inef?ciency regardless of their cause: if they are scale related, they are still
important for identifying issues of banks’ safety and soundness.
4. Results
The three models around which this analysis is undertaken are a basic CCR model,
a pro?t ef?ciency model (to look at the effects of decomposing monetary amounts and
the interest rates applying to them), and a slacks-based model (where we are expecting
the greatest inef?ciencies). We will deal with each of these in turn, with a particular
focus on the observations for the March 2008 quarter, in terms of understanding the
changes in pro?tability. These can be highlighted by looking at the changes in banks’
return on assets (ROA) in the March 2008 quarter relative to the previous, December
2007 quarter, as shown in Figure 1.
Table I shows us the ef?ciency scores for each bank for each quarter, from the
CCR model.
The differences between pro?ts (as measured by ROA) and the ef?ciency scores are
striking, but we also note that the range of ef?ciency scores in the CCR model is not great,
with a mean ef?ciency score across all 60 DMUs of 0.979. We can select a couple of
cases for more detailed examination: let us look at BNZ, which showed an improvement
in pro?tability but a deterioration in ef?ciency, and at Kiwibank, which showed
an improvement in ef?ciency but a deterioration in pro?tability, in the ?nal quarter of
the study.
In the BNZ’s case, we ?nd that it showed an increase in bad and doubtful debt
expense in the March 2008 quarter, and although this may not have had a big impact
JFEP
2,3
242
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
on pro?tability, it has had a signi?cant impact on ef?ciency. For the study of bank
safety and soundness, this is a useful warning ?ag.
In Kiwibank’s case, we ?nd evidence for excess non-interest expense in particular
in the December 2007 quarter, but the relationship is not strong. Part of the problem
would appear that, for the CCR model, we are using an input-oriented model,
and although this is logical for studies of banks as a whole (where they can change
their inputs but not the aggregate volume of their outputs, which are determined
by the market), it is not especially helpful for understanding pro?tability change
(where pro?tability is determined by both inputs and outputs)[13].
Table II shows us the ef?ciency scores from the pro?t ef?ciency model.
Ef?ciency scores are now signi?cantly lower on aggregate, with a mean ef?ciency
score across all DMUs of 0.943, which means that this model is more appropriately
identifying inef?ciencies. This re?ects the aspects of the pro?t model, that it combines
Figure 1.
Comparison of banks’
ROA (de?ned as net pro?t
after tax divided by
average total assets)
between December quarter
2007 and March quarter
2008
1.8
1.6
1.4
1.2
1.0
P
e
r
c
e
n
t
a
g
e
0.8
0.6
0.4
0.2
0.0
ANZ ASB
Note: Figures are annualized
BNZ Westpac TSB Kiwibank
12/07
3/08
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.988 0.967 0.958 1.000 1.000 1.000
March 2006 0.931 0.962 0.954 1.000 1.000 1.000
June 2006 1.000 0.981 1.000 0.975 0.987 0.976
September 2006 0.966 0.974 0.961 0.984 0.991 1.000
December 2006 0.939 1.000 1.000 0.976 0.985 0.981
March 2007 0.978 1.000 0.948 0.981 0.973 1.000
June 2007 0.929 0.997 0.990 0.981 1.000 1.000
September 2007 0.967 0.982 0.964 0.958 1.000 1.000
December 2007 0.959 1.000 0.972 0.950 0.973 0.962
March 2008 0.979 1.000 0.944 1.000 0.960 1.000
Average 0.964 0.986 0.969 0.981 0.987 0.992
Table I.
Ef?ciency scores for each
bank for each quarter,
from the CCR model
Bank safety
and soundness
243
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
allocative and technical ef?ciency, that it identi?es both cost and revenue inef?ciencies
(and the reported pro?t inef?ciency thus allows for both of these), and that it is not
oriented (in the way that the CCR model is). This is consistent with the de?nition of
pro?t ef?ciency as:
Ep ¼
p
0
y
0
2c
0
x
0
p
0
y* 2c
0
x*
where x
*
and y
*
represent the vectors of the optimal quantities of inputs and outputs,
x
0
and y
0
are the vectors of the observed quantities of the inputs and outputs,
and p
0
and c
0
the prices (Cooper et al., 2000; Paradi, 2006).
If we look at the BNZ, we ?nd that the major driver for (the now much more
pronounced) inef?ciency in the December 2007 and March 2008 quarters is bad and
doubtful debt expense. This contrasts with the simple ratios, where BNZ shows as
having lower levels of bad and doubtful debt expense than either ANZ or Westpac[14],
but the outcomes observed are a consequence of looking at all the inputs and outputs
together, on a multivariate basis. BNZ also shows some inef?ciency as deriving from
interest revenues, with insuf?cient interest bearing assets (and this is despite what
appears to be a relatively good net interest income performance in this quarter and the
previous ones).
ANZ’s inef?ciency also derives from bad and doubtful debt expense, but also from
an overuse of equity (re?ecting its need to hold a signi?cantly larger quantity of equity
because of the goodwill on its balance sheet from the surplus over book value paid
when it acquired the National Bank of New Zealand in 2003). These two factors
account for most of the ANZ’s inef?ciency throughout the period studied.
TSB’s inef?ciency in the March 2008 quarter derives from higher levels of bad and
doubtful debt expense (which appears to re?ect a standard adjustment in the March
quarter, which is the ?nal quarter of its ?nancial year), and a de?ciency in non-interest
income (which TSB generally earns less of in any case).
Kiwibank’s inef?ciency appears to be caused by higher levels of non-interest
expense, and, to a much lesser extent, from higher levels of interest expense.
Kiwibank’s non-interest expenses have been consistently greater than for other banks,
but the much stronger identi?cation of inef?ciency in this case may be a consequence
of revenues being lower than in previous quarters. What we are ?nding overall is that
the pro?t model is giving us more useful results than the CCR model, in terms of
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 1.000 0.928 0.862 1.000 0.998 1.000
March 2006 0.888 0.890 0.810 1.000 0.982 1.000
June 2006 1.000 0.926 1.000 0.947 0.987 0.961
September 2006 0.856 0.994 0.907 0.936 0.896 1.000
December 2006 0.876 1.000 1.000 0.917 0.990 0.930
March 2007 0.933 1.000 0.857 0.972 0.983 1.000
June 2007 0.912 1.000 0.973 0.949 1.000 0.764
September 2007 0.895 0.940 0.883 0.944 1.000 1.000
December 2007 0.971 1.000 0.832 0.889 1.000 0.887
March 2008 0.893 1.000 0.838 1.000 0.963 0.800
Average 0.922 0.968 0.896 0.955 0.980 0.934
Table II.
Ef?ciency scores for each
bank for each quarter,
from the pro?t ef?ciency
model
JFEP
2,3
244
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
understanding differences in pro?tability from quarter to quarter, with a stronger
relationship between pro?tability change and ef?ciency change.
The slacks-based model (which was run on a non-oriented basis), the results from
which are shown in Table III, shows signi?cantly greater levels of inef?ciency, with a
mean ef?ciency score across all DMUs of 0.797.
A major reason for the lower scores is that the slacks-based model also takes
account of non-radial slacks[15].
ANZ now shows as being particularly inef?cient in all but one of the ten quarters in
our analysis. The major contributors to this are excessive levels of equity and higher
levels of bad and doubtful debt expense, as previously discussed in respect of the pro?t
ef?ciency model.
ASB once again shows as relatively ef?cient, which is interesting in view of its
much lower level of gross income (de?ned as net interest income plus non-interest
income), as shown in Figure 2. The relatively high-ef?ciency re?ects the relatively
low usage of resources to generate that gross income. Where ASB was inef?cient,
it appears that this could be mainly attributed to bad and doubtful debt expense, with
some shortages of non-interest income.
BNZ’s inef?ciency in the March 2008 quarter appears to be largely attributable to
excess bad and doubtful debt expense, although it also shows as using excess equity
and interest expense, and as generating insuf?cient non-interest income. Although
Westpac showed as fully ef?cient in the March 2008 quarter, its inef?ciency in
previous quarters appears to be able to be attributed mainly to excess interest expense
and bad and doubtful debt expense.
TSB showed a dramatic drop in ef?ciency in the March 2008 quarter. The major
contributor to its inef?ciency is low levels of non-interest income, with lesser impacts
from high levels of bad and doubtful debt expense and equity. This would not be too
much of a surprise to analysts who have studied the bank’s ratios: the bank is rather
more strongly capitalised than others, largely to compensate for its small size and lack
of access to external capital, while the nature of its business means that its non-interest
income should be expected to be low.
Kiwibank also shows as relatively ef?cient overall, which is likely to re?ect its
higher levels of gross income, as evident in Figure 2. Once again, where there was
signi?cant inef?ciency, this appeared to relate mainly to higher levels of bad and
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.847 0.740 0.727 1.000 1.000 1.000
March 2006 0.605 0.677 0.626 1.000 1.000 1.000
June 2006 1.000 0.888 1.000 0.769 0.818 0.828
September 2006 0.581 0.719 0.735 0.862 0.842 1.000
December 2006 0.578 1.000 1.000 0.727 0.648 0.844
March 2007 0.684 1.000 0.673 0.820 0.538 1.000
June 2007 0.598 0.948 0.839 0.874 1.000 1.000
September 2007 0.694 0.728 0.767 0.873 1.000 1.000
December 2007 0.687 1.000 0.739 0.770 0.566 0.784
March 2008 0.689 1.000 0.690 1.000 0.273 1.000
Average 0.696 0.870 0.780 0.870 0.769 0.946
Table III.
Ef?ciency scores for each
bank for each quarter,
from the slacks-based
model
Bank safety
and soundness
245
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doubtful debt expense (even though bad and doubtful debt expense for Kiwibank, like
TSB, has been low). We note, however, that inef?cient Kiwibank DMUs show only
Kiwibank in their reference sets, while Kiwibank only plays a very small part in the
reference sets for their inef?cient DMUs[16]. There is thus a question as to whether
Kiwibank shares a common frontier with the other banks in the study.
If we look at the inef?ciencies across the data set as a whole, we ?nd that bad and
doubtful debt expense is the most prevalent. This is likely to re?ect its variability in
terms of its contribution to bank pro?tability, and the fact that the absolute values of
the numbers are sometimes very low, which means that the relative change in bad and
doubtful debt expense can be very large. It can be expected to be more signi?cant in its
impact on pro?ts in coming quarters, with higher levels of bad and doubtful debt
expense associated with the economic downturn. Our results may also be a
consequence of bad and doubtful debt expense being a negative output, which we are
treating as an input, and it is therefore considered appropriate to re-run the analysis
with this variable omitted. The economic meaning for this model would be to relate to
banks’ underlying pro?tability (i.e pro?t before bad and doubtful debt expense,
extraordinary items or tax), which also re?ects the fact that bad and doubtful debt
expense will generally relate to the bank’s performance in quarters prior to the one that
is being studied. It is this underlying pro?tability which is going to be more important
for a bank’s long-term pro?tability and soundness.
Results for the (non-oriented) slacks-based model, without bad and doubtful debts
as an input, are reported in Table IV. The mean ef?ciency across all DMUs is 0.812,
which is only marginally higher than for the previous model.
If we look at individual banks, we ?nd that ANZ shows as consistently relatively
inef?cient. The major contributor to this continues to be an excessive level of equity
(as discussed previously), although the main subsidiary reason is nowan insuf?ciency of
Figure 2.
Comparison of banks’
gross incomes (de?ned as
net interest income plus
non-interest income,
divided by average total
assets) between December
quarter 2007 and March
quarter 2008
4.5
4.0
3.5
3.0
2.5
2.0
P
e
r
c
e
n
t
a
g
e
1.5
1.0
0.5
0.0
ANZ
Note: Figures are annualized
ASB BNZ Westpac TSB Kiwibank
12/07
3/08
JFEP
2,3
246
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
non-interest income. ASB and BNZ also show, in most cases, a surplus of equity and a
shortage of non-interest income, although there are a number of quarters where ASB
shows a small excess of interest expense. ASB continues to show as fully ef?cient in the
March quarter 2008.
Westpac also shows as fully ef?cient in the March 2008 quarter: for the earlier
quarters in which it showed as less than fully ef?cient, the inef?ciency can be related to
excess equity and insuf?cient non-interest income. TSB’s inef?ciency derives from the
same two factors, although the shortage of non-interest income is much more
important in this case.
Kiwibank’s inef?ciency is different, with no shortage of non-interest income, except
in the March 2008 quarter when it was much less than usual. In Kiwibank’s case,
the major sources of inef?ciency are excess non-interest expense (ratio analysis shows
much higher relative costs than for the other banks) and insuf?cient interest income.
5. Summary and conclusion
This paper has sought to investigate the changes in pro?tability from quarter to
quarter for New Zealand’s banks with signi?cant retail business, with a particular
focus on the changes in pro?tability between the December quarter 2007 and the
March quarter 2008. We tested three different DEA methodologies, and found that
the slacks-based approach was most useful, although there was also value in using the
pro?t ef?ciency approach. The CCR model showed very little variation in ef?ciency
between the different banks, and thus was less effective at identifying the causes of
pro?tability change.
There was no clear indication that scale mattered, although the largest bank,
ANZ-National, showed as least ef?cient, and it is possible that some of this
inef?ciency may be a consequence of scale (although it is noted that the input that
consistently contributed to inef?ciency was equity, which ought to show positive
bene?ts from increased scale). ANZ is carrying extra equity, however, because of the
(intangible) goodwill associated with the acquisition of the National Bank of
New Zealand in 2003.
The success of the approach followed in this paper ought to be of assistance to
analysts in assessing the safety and soundness of New Zealand’s banks, a process
which is of particular importance in the New Zealand market because of the historic
lack of deposit insurance or other guarantees for depositors. It is not altogether clear
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.723 0.723 0.847 1.000 1.000 1.000
March 2006 0.700 0.674 0.678 1.000 1.000 0.967
June 2006 0.631 0.721 1.000 0.887 0.867 0.939
September 2006 0.708 0.657 0.843 0.890 0.889 0.928
December 2006 0.665 0.739 1.000 0.826 0.744 0.929
March 2007 0.710 1.000 0.781 0.869 0.663 1.000
June 2007 0.636 0.799 0.850 0.798 1.000 0.815
September 2007 0.681 0.630 0.777 0.797 1.000 0.905
December 2007 0.684 0.775 0.711 0.719 0.581 0.877
March 2008 0.693 1.000 0.735 1.000 0.342 0.702
Average 0.683 0.772 0.822 0.879 0.809 0.906
Table IV.
Ef?ciency scores for each
bank for each quarter,
from the Slacks-based
model, without bad
and doubtful debt
expense as an input
Bank safety
and soundness
247
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
that inclusion of bad and doubtful debts as a negative output (and thus as an input) is
helpful but we should be able to make further assessments of this as we look at the
results for later quarters when the global economic downturn has its effect in
increasing individual banks’ levels of bad and doubtful debt expense. There may be
merit in looking for other ways of dealing with bad and doubtful debt expense as an
undesirable output in our DEA models[17]. It will also be interesting to investigate the
use of new DEA models as these become more readily available as part of standard
software packages.
Notes
1. The introduction of a deposit protection scheme on 12 October 2008 has not undermined this
principle for wholesale deposits, and banks continue to provide quarterly disclosures to the
market. For more detail on the disclosure regime, see Mortlock (1996).
2. DEA has been criticised previously for only providing measures of technical ef?ciency. See,
for example, Berger and Mester (1997).
3. For a more detailed exposition of DEAmethodologies, see Cooper et al. (2000) or Avkiran (2006).
4. This is consistent with the approach outlined by Tulkens and Vanden Eeckaut (1995).
In practice, no indication was found in the results from this research that there was any
general change in banks’ ef?ciency through time.
5. Re?ecting concerns about banks’ use of bad and doubtful debt expense to smooth pro?ts,
we also tried use of total impaired and past due assets as an alternative input. The resulting
analysis showed a narrower range of ef?ciency scores (less discriminatory power).
The Spearman’s rank order correlation between the two sets of results was 0.783 (signi?cant
at less than 1 per cent).
6. Beyond that, physical capital is more-or-less insigni?cant as an input for New Zealand
banks, with ?xed assets comprising an average of 0.2 per cent of assets.
7. It was suggested to the author by a conference participant that the input-output set used was
essentially equivalent to measurement of pro?t, to which the author countered by
highlighting the inclusion of shareholders’ equity as an input. In fact, there are no signi?cant
correlations between banks’ return on assets (de?ned as pro?t after tax divided by average
total assets) and ef?ciency scores.
8. There were a total of 18 registered banks as at 31 December 2008.
9. TSB and Kiwibank were not required to adopt IFRS until the ?nancial years commencing in
2007, but the simpler nature of their operations means that the change in accounting
standards has had less impact on results.
10. This thus mitigates the potential problem of random error in the use of DEA.
11. The highest super-ef?ciency score was 1.25, well belowthe guideline level of concern of 2 – see
Hartman et al. (2001).
12. One New Zealand dollar is equivalent to approximately 60 cents US.
13. In fact, for a CCR model, the results from input-oriented or output-oriented model will be
identical.
14. Detailed data are omitted for reasons of space and readability, but are available from the
author on request.
15. There was no slacks-based pro?t model available in the version of the DEA-Solver software
(5.0) used for this research.
JFEP
2,3
248
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
16. We checked this by examining the lambdas in a variable returns to scale model. For only
four DMUs did a lambda value for Kiwibank exceed 0.1, and the highest of these was 0.216.
17. See the discussion by Burley (2006).
References
Avkiran, N.K. (2006), Productivity Analysis in the Service Sector with Data Envelopment Analysis,
3rd ed., N K Avkiran, Ipswich.
Banker, R.D., Charnes, A.W. and Cooper, W.W. (1984), “Some models for estimating technical
and scale inef?ciencies in data envelopment analysis”, Management Science, Vol. 30 No. 9,
pp. 1078-92.
Berger, A.N. and DeYoung, R. (1997), “Problem loans and cost ef?ciency in commercial banks”,
Journal of Banking & Finance, Vol. 21, pp. 849-70.
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,
pp. 895-947.
Berger, A.N., Hunter, W.C. and Timme, S.G. (1993), “The ef?ciency of ?nancial institutions:
a review and preview of research past, present and future”, Journal of Banking & Finance,
Vol. 17, pp. 221-49.
Burley, H.T. (2006), “Accounting for the in?uence of the environment in a two-stage analysis –
public hospitals”, in Avkiran, N.K. (Ed.), Productivity Analysis in the Service Sector with
Data Envelopment Analysis, 3rd ed., Ch. 17, N K Avkiran, Ipswich, pp. 181-92.
Charnes, A., Cooper, W.W. and Rhodes, E. (1978), “Measuring the ef?ciency of decision making
units”, European Journal of Operational Research, Vol. 2, pp. 429-44.
Cooper, W.W., Seiford, L.M. and Tone, K. (2000), Data Envelopment Analysis, Kluwer Academic,
Boston, MA.
Dyson, R.G., Allen, R., Camanho, A.S., Podinovski, V.V., Sarrico, C.S. and Shale, E.A. (2001),
“Pitfalls and protocols in DEA”, European Journal of Operational Research, Vol. 132,
pp. 245-59.
Golany, B. and Storbeck, J.E. (1999), “A data envelopment analysis of the operational ef?ciency of
bank branches”, Interfaces, Vol. 29, May-June, pp. 14-26.
Grifell-Tatje´, E. and Lovell, C.A.K. (1999), “Pro?ts and productivity”, Management Science,
Vol. 45 No. 9, pp. 1177-93.
Hartman, T.E., Storbeck, J.E. and Byrnes, P. (2001), “Allocative ef?ciency in branch banking”,
European Journal of Operational Research, Vol. 134, pp. 232-42.
Lovell, C.A.K. (2008), “Productivity accounting”, paper presented at the Asia-Paci?c Productivity
Conference, Taiwan, July.
Mortlock, G. (1996), “New disclosure regime for registered banks”, Reserve Bank Bulletin, Vol. 59
No. 1, pp. 21-9.
Paradi, J. (2006), “Pro?t ef?ciency – health insurance”, in Avkiran, N.K. (Ed.), Productivity
Analysis in the Service Sector with Data Envelopment Analysis, 3rd ed., Ch. 10, N K
Avkiran, Ipswich, pp. 101-9.
Tone, K. (2001), “A slacks-based measure of ef?ciency in data envelopment analysis”, European
Journal of Operational Research, Vol. 130, pp. 498-509.
Bank safety
and soundness
249
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Tone, K. (2002), “A slacks-based measure of super-ef?ciency in data envelopment analysis”,
European Journal of Operational Research, Vol. 143, pp. 32-41.
Tulkens, H. and Vanden Eeckaut, P. (1995), “Non-parametric ef?ciency, progress and regress
measures for panel data: methodological aspects”, European Journal of Operational
Research, Vol. 80, pp. 474-99.
Corresponding author
David Tripe can be contacted at: [email protected]
JFEP
2,3
250
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_588817132.pdf
The purpose of this paper is to investigate use of efficiency analysis as a technique for
investigating bank safely and soundness
Journal of Financial Economic Policy
Using DEA to investigate bank safety and soundness – which approach works best?
David Tripe
Article information:
To cite this document:
David Tripe, (2010),"Using DEA to investigate bank safety and soundness – which approach works best?",
J ournal of Financial Economic Policy, Vol. 2 Iss 3 pp. 237 - 250
Permanent link to this document:
http://dx.doi.org/10.1108/17576381011085449
Downloaded on: 24 January 2016, At: 21:38 (PT)
References: this document contains references to 19 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 365 times since 2010*
Users who downloaded this article also downloaded:
Abdelkader Boudriga, Neila Boulila Taktak, Sana J ellouli, (2009),"Banking supervision and nonperforming
loans: a cross-country analysis", J ournal of Financial Economic Policy, Vol. 1 Iss 4 pp. 286-318 http://
dx.doi.org/10.1108/17576380911050043
Sven Blank, J onas Dovern, (2010),"What macroeconomic shocks affect the German banking system?:
Analysis in an integrated micro-macro model", J ournal of Financial Economic Policy, Vol. 2 Iss 2 pp.
126-148 http://dx.doi.org/10.1108/17576381011070193
Irina Barakova, Ajay Palvia, (2010),"Limits to relative performance evaluation: evidence from
bank executive turnover", J ournal of Financial Economic Policy, Vol. 2 Iss 3 pp. 214-236 http://
dx.doi.org/10.1108/17576381011085430
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Using DEA to investigate bank
safety and soundness – which
approach works best?
David Tripe
Centre for Banking Studies, Massey University, Palmerston North, New Zealand
Abstract
Purpose – The purpose of this paper is to investigate use of ef?ciency analysis as a technique for
investigating bank safely and soundness.
Design/methodology/approach – Three different data envelopment analysis (DEA) models were
applied to set of data for the major New Zealand banks over a ten-quarter period – a CCR model,
a pro?t ef?ciency model and a non-oriented slacks-based approach.
Findings – Most useful results are obtained using the slacks-based approach.
Research limitations/implications – The period covered by the study was from late 2005
until early 2008, prior to the global ?nancial crisis having major impacts on the New Zealand
banking sector.
Practical implications – The study is of particular value in the New Zealand context where there
has historically not been any bank deposit insurance, obliging depositors to make their own
assessments of bank safety and soundness.
Originality/value – The paper makes a contribution to very small literature which uses ef?ciency
analysis to explore bank safety and soundness. It also makes use of the slacks-based DEA approach,
which has not yet been widely used in the banking literature.
Keywords Banking, Data analysis, New Zealand
Paper type Research paper
1. Introduction
Studies of bank and ?nancial institution ef?ciency have been undertaken for a variety
of purposes. Ef?ciency in ?nancial institutions is and should be a matter of public
concern, as not only can more ef?cient ?nancial institutions be expected to be more
pro?table, but one should also expect ?nancial institution ef?ciency to lead to greater
amounts of funds being intermediated, and better service at lower prices for
consumers. Other things being equal, more ef?cient ?nancial institutions should
exhibit greater safety and soundness (Berger et al., 1993, pp. 221-2), while also showing
better credit quality in the loan portfolio (Berger and DeYoung, 1997). These issues also
provide a basis for looking at the ef?ciency implications of both bank mergers and
acquisitions and government policy initiatives.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
The author would like to thank Knox Lovell for asking a question and suggesting the idea that
underpins this paper. The author would also like to thank participants at the eleventh EWEPA at
Pisa, the IDEAS Conference in Philadelphia and the fourteenth FINSIA/Melbourne Centre for
Financial Studies conference for comments on earlier versions of this paper. Particular thanks
are due to Mette Asmiild and Joe Paradi.
No blame should be attributed to anyone other than the author for any de?ciencies in the
paper.
Bank safety
and soundness
237
Journal of Financial Economic Policy
Vol. 2 No. 3, 2010
pp. 237-250
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381011085449
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This study focuses on bank safety and soundness, which has been of particular
importance in the New Zealand context because, prior to October 2008, there was no
general deposit protection scheme for bank customers. This meant that the onus of
assessing the soundness of the banks with which they dealt fell upon the banks’
customers: banks are requiredto make disclosures to the public on a quarterlybasis, and it
is this information which is also provided to the Reserve Bank of New Zealand as the
prudential supervisor of the banking system. The principle has been that if the Reserve
Bank had no better information thanwas made available to the general public, it could not
be liable to recompense depositors if a bank failed[1].
Banks’ quarterly disclosures include a year-to-date income statement and balance
sheet, which allow an analyst to construct a quarterly income statement, and thus
make a detailed assessment of ?nancial performance. Such an assessment of ?nancial
performance is likely to focus on key risk factors, and thus credit risk will be assessed
by levels of impaired and past due loans and bad and doubtful debt expense, interest
rate risk by the level of and volatility in a bank’s net interest income, operational risk
by the bank’s non-interest expense, and solvency by the ratio of capital to assets and
the ?ow of pro?ts to sustain that capital level. Relevant accounting ratios have been
compared between banks, and through time, in an effort to understand pro?tability
change from quarter to quarter, and to monitor exceptional items which might give
grounds for concern.
Grifell-Tatje´ and Lovell (1999) noted that a change in a ?rm’s pro?ts may derive from
a number of sources, including a change in input or output prices. The other sources of
pro?t change are likely to be measured using the techniques of ef?ciency analysis,
including technical change leading to an increase in output without any increase in
resource utilisation, animprovement or decline inoperating ef?ciency(X-ef?ciency), or a
change in output proportionately greater or less than input utilisation, re?ecting
economies or diseconomies of scale. Further, sources of pro?tability improvement
include changes in product or resource mix, associated with economies of scope and
allocative ef?ciency (Lovell, 2008). Ef?ciency analysis allows us to explore trade-offs
between the different inputs and outputs used, something which is inherently
problematic if one is constrained to use of ratios (Golany and Storbeck, 1999).
Against this background, it is considered appropriate to use the methods of
ef?ciency analysis to study the pro?t performance of New Zealand banks from quarter
to quarter, as part of the process of assessing their safety and soundness. We tested
three different approaches, and ?nd that the clearest distinctions are made between
individual bank performance when we use a slacks-based, data envelopment analysis
(DEA) method.
The rest of the paper proceeds as follows. In the next section, we outline the
approach that will be followed against the background of the key literature that is
relevant to this research. In Section 3, we look at the New Zealand banking system, and
the data that are generated from the disclosures to permit this research. Section 4
reports results that are obtained, while Section 5 provides a summary and conclusion.
2. The method for this research, and relevant prior research on bank
ef?ciency
Berger and Humphrey (1997) reported on 130 studies of bank ef?ciency, and a great
many more articles have been published since that review. These studies have used
JFEP
2,3
238
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
both parametric methods –stochastic frontier analysis (SFA), the distribution free
approach and the thick frontier approach – and non-parametric methods – DEA and
the free disposal hull. The key differences between the approaches are that the
non-parametric methods do not take account of random error in observations, while the
parametric approaches require speci?cation of a functional form. The complexity and
diversity of production relations in banking mean that the speci?cation of the
functional form is not self-evident: the imposition of an inappropriate functional form
can effectively distort the ?ndings of any analysis.
This study uses DEA, a number of forms of which have been developed since the
original constant returns to scale (CCR) model was developed by Charnes et al. (1978).
Banker et al. (1984) introduced a variable returns to scale (BCC) model, use of which
allows a more extensive exploration of scale economies. Further, subsequent variants
include the cost, revenue and pro?t models, which take account of input and output
prices, and which can thus provide estimates of allocative ef?ciency[2]. More recently,
again we have seen the introduction of slacks-based models (Tone, 2001), which
identify a broader range of inef?ciencies through consideration of non-radial slacks[3].
Another distinction between DEA and the parametric approaches is in the focus on
the individual decision-making unit (DMU – bank in this case). DEA both generates
individual ef?ciency scores for each bank for each time period, and identi?es
which inputs have been over-utilised and/or outputs underutilised to engender that
inef?ciency. Of the parametric approaches, only SFA provides an observation-speci?c
ef?ciency score, but DEA identi?es a peer group of more ef?cient units against which
an individual unit’s ef?ciency can be compared. Since we are looking to understand the
factors behind differences in individual bank ef?ciency and pro?tability, DEA is a
more appropriate method for this study.
Another factor that sometimes in?uences the choice between DEA and one of the
parametric approaches is the number of DMUs to be included in the analysis.
Parametric methods generally require larger data sets, not so much in that DEA is
better with a smaller data set, but rather because the greater ?exibility of DEA does
not preclude its use for smaller data sets. In this case, with a cross-section of no more
than six comparable banks for the period studied, DEA would be a necessity. Even
here, though, we need a greater number of observations to allow meaningful
differences in ef?ciency to be observed: we achieve this by studying our cross section
of banks over consecutive time periods, in a panel. We would wish to analyse the
data set as a panel in any case, however, because to understand pro?tability changes,
we need to be looking at previous accounting periods alongside the most recent ones.
The standard approach to looking at ef?ciency change between two different
periods is by use of the Malmquist index, which allows a decomposition of productivity
change into the effects of technological change and change in ef?ciency. This is not
what we are looking for in this study, where we expect pro?tability change to be
impacted more by changes in prices and quantities, and in the ef?ciency with which
inputs are used and outputs created. We are also looking at a relatively short-time
period, which means that the effects of any technological change that might invalidate
the use of panel data would not be expected to be major[4].
Another key issue in studying the ef?ciency of ?nancial institutions is in selecting a
model of the banking ?rm. The two main methods are the production and
intermediation approaches, although the production approach is more often used for
Bank safety
and soundness
239
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
measurement of the ef?ciency of individual branches, rather than for banks as a whole.
The intermediation approach exists in a number of different forms, but it is essentially
based on the process of converting depositors’ funds into loans, and it is the
predominant method for studying banks as a whole. The approach we use in this study
is broadly consistent with the intermediation approach.
Dyson et al. (2001) remind us that the important point in selecting inputs and
outputs for DEA studies (and, by implication, other ef?ciency studies as well) is that
we should be looking for input variables which include as much as possible of the
resources that DMUs use. In turn, we should be looking at output variables which
include as much as possible of the relevant outputs, but which also identify the key
success factors for the entities whose ef?ciency is being studied.
Relevant inputs and outputs can thus be identi?ed from looking at the way banks
generate pro?ts. They incur interest expense, and use non-interest expense and
shareholders’ equity to generate interest and non-interest income. A side effect of
this production process is bad and doubtful debt expense, which reduces pro?tability,
and which might thus be regarded as a negative output (and which can thus be
processed within DEA as an input)[5]. The interest income and expense can be
decomposed into prices (interest rates) and quantities, although such a decomposition
cannot meaningfully be conducted for the other inputs and outputs used.
Use of shareholders’ equity as an input is not universal in the prior literature, but it
can be justi?ed. In the ?rst place, shareholders’ equity is a source of funding, and if a
bank has more equity, it should need less funding (and thus less interest expense),
while the higher proportion of equity should reduce the cost of those funds. More
importantly, however, as discussed by Berger and Mester (1997), equity gives a bank a
greater cushion against risk, which allows it to increase the size of its loan portfolio
(which might also otherwise be constrained by capital regulation), and to take on
riskier loans which should be expected to generate greater interest revenues.
This contrasts with our decision to omit physical capital or staff as inputs, with
prior researchers often having calculated prices for these based on relevant
expenditure. In our view, a major problem with attempting to include physical capital
is that, in many cases, items of physical capital will be leased rather than owned. Even
if they were owned, there might be problems with their valuation[6]. In practical terms,
neither the costs associated with physical capital or staff costs are consistently
disclosed quarterly, and as staff numbers are not reported either, neither of these
measures which have commonly been used in previous research are available to us.
We are therefore obliged to use aggregate non-interest expense as an input[7].
Having identi?ed the method that should be used for his research, we now look at
the New Zealand banking system and the data that it provides for our study. In the
course of our discussion, we will also address some further methodological issues.
3. New Zealand banking data
We noted in the introduction that New Zealand lacked any formal deposit protection
scheme (prior to October 2008), and that supervision of the banking system was based
on public disclosure. The key elements of this are the quarterly disclosure statements,
which include year-to-date ?nancial statements. Even if this were not the basis for the
prudential supervision of the banks, the disclosure statements are a boon for
researchers in terms of the data they provide.
JFEP
2,3
240
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
One of the reasons why this approach to banking supervision is viable in NewZealand
is that the number of banks undertaking retail business is relatively small. As at 31 March
2008, the four largest banks accounted for 89.4 per cent of total banking system assets.
Two further smaller banks that also emphasized retail business accountedfor a further 2.9
per cent of banking system assets[8].
A further consideration in the Reserve Bank of New Zealand’s lighter approach
to banking supervision is the extent of foreign ownership of the banks, with more than
97 per cent of bank assets under foreign ownership at 31 March 2008. The four major
banks, ANZNational Bank(ANZ, ownedbythe ANZBanking Group Limited), ASBBank
(ASB, owned by the Commonwealth Bank of Australia (CBA)), Bank of New Zealand
(BNZ, owned by the National Australia Bank), and the Westpac Banking Corporation
(Westpac) are all Australianowned, withtheir parents all amongthe tenlargest companies
on the Australian Stock Exchange. The New Zealand-owned banks are Kiwibank
(government-owned through New Zealand Post) and TSB Bank (TSB, owned by a local
community trust), which have a combined market share of 2.9 per cent.
Our sample frame for this study thus comprises six banks, the four Australian-owned
majors and the two NewZealand-owned banks, which together dominate retail banking.
A limited amount of retail banking is provided by the New Zealand branches of HSBC
and (Korean-owned) Kookmin Bank, but these two entities show balance sheets and
income statements which are quite different from those included in the study – it is
argued that the necessary condition of relative homogeneity to allow reasonable
comparisons would not apply to these banks. Rabo NewZealand Limited is also omitted
from the study: although it is New Zealand incorporated, it specialises in lending to the
rural sector and would appear to rely to at least some extent on the Rabobank Nederland
New Zealand branch for its functioning.
During the period covered by this study, the CBA and Westpac each had two
registered banks operating in the New Zealand market. In the CBA’s case, the
overwhelming majority of its business is in its subsidiary, ASB, with this apparently
capable of surviving independently (which it did for a long time up to its parent’s
registration of a branch). In the Westpac case, however, the New Zealand subsidiary was
established primarily to comply with New Zealand regulatory requirements, and it is
doubtful that it could survive in its own right. Moreover, its relatively recent
establishment (in late 2006) means that it is not clear that it yet has a stable pattern of
operations. For both of these reasons, we have chosen to use the Westpac branch in the
study, rather than the subsidiary.
The initial analysis in this study covers the period from the December quarter
2005 through to the March quarter 2008 – ten quarters for six banks, giving us
60 observations in total, which should be ample relative to the combined total of six inputs
and outputs. A key factor in the choice of start date was the adoption of international
?nancial reporting standards (IFRS) by the major Australian banks (and thus for their
NewZealand operations as well), for their ?nancial years ending in 2006. IFRS will have
had a particular effect on bad and doubtful debt expense (in terms of removing banks’
discretion to record additional provisions without objective evidence of impairment),
on interest income and expense where fees and costs have to be amortised over the life
of the assets and liabilities to which they apply, and because of the way derivative
assets and liabilities have to be accounted for, with the revaluations impacting
particularly on non-interest income[9]. Following the adoption of IFRS, less of the
Bank safety
and soundness
241
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
variability in pro?tability ought to be able to be attributed to variation in accounting
practices[10].
Data were checked using a slacks-based super-ef?ciency model (Tone, 2002) and no
cases were identi?ed which suggested any problems[11]. A review of the correlations
between the inputs, the outputs and between the inputs and outputs also failed to
identify any concerns in respect of the data being used for the study.
A minor adjustment was required to the data for bad and doubtful debt expense,
to remove negative values and replace these with zeros. This was because the negative
values arti?cially shifted the ef?cient frontier to the extent that a negative ef?ciency
score was generated for one DMU in the slacks-based model. Negative values for bad
and doubtful debt expense would indicate an impact of recoveries of write-offs from
earlier accounting periods: there is thus an argument that a value of zero would more
fairly re?ect current performance. Signi?cant changes in ef?ciency scores were
generated only for one DMU, which continued to show signi?cant inef?ciency (but no
longer a negative score).
As a further preliminary check on our data, recognizing the range of bank size from
NZ$ 3 billion[12] for TSBup to NZ$ 100 billion for ANZ, we checked for scale effects. The
strongest indication of scale inef?ciency was found for ANZ, with mean scale ef?ciency
of 0.96, but which was nonetheless shown as being fully scale ef?cient in one quarter.
Noting the tendency of observations of different size to showas scale inef?cient because
of the relative dearth of other observations in that size bracket (Dyson et al., 2001), which
would be the case for ANZ, we have preferred to use constant returns to scale models so
as to increase the models’ sensitivity (by reporting numerically lower ef?ciency scores).
Moreover, having regard to the objectives of our study, we note that we want to identify
sources of inef?ciency regardless of their cause: if they are scale related, they are still
important for identifying issues of banks’ safety and soundness.
4. Results
The three models around which this analysis is undertaken are a basic CCR model,
a pro?t ef?ciency model (to look at the effects of decomposing monetary amounts and
the interest rates applying to them), and a slacks-based model (where we are expecting
the greatest inef?ciencies). We will deal with each of these in turn, with a particular
focus on the observations for the March 2008 quarter, in terms of understanding the
changes in pro?tability. These can be highlighted by looking at the changes in banks’
return on assets (ROA) in the March 2008 quarter relative to the previous, December
2007 quarter, as shown in Figure 1.
Table I shows us the ef?ciency scores for each bank for each quarter, from the
CCR model.
The differences between pro?ts (as measured by ROA) and the ef?ciency scores are
striking, but we also note that the range of ef?ciency scores in the CCR model is not great,
with a mean ef?ciency score across all 60 DMUs of 0.979. We can select a couple of
cases for more detailed examination: let us look at BNZ, which showed an improvement
in pro?tability but a deterioration in ef?ciency, and at Kiwibank, which showed
an improvement in ef?ciency but a deterioration in pro?tability, in the ?nal quarter of
the study.
In the BNZ’s case, we ?nd that it showed an increase in bad and doubtful debt
expense in the March 2008 quarter, and although this may not have had a big impact
JFEP
2,3
242
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
on pro?tability, it has had a signi?cant impact on ef?ciency. For the study of bank
safety and soundness, this is a useful warning ?ag.
In Kiwibank’s case, we ?nd evidence for excess non-interest expense in particular
in the December 2007 quarter, but the relationship is not strong. Part of the problem
would appear that, for the CCR model, we are using an input-oriented model,
and although this is logical for studies of banks as a whole (where they can change
their inputs but not the aggregate volume of their outputs, which are determined
by the market), it is not especially helpful for understanding pro?tability change
(where pro?tability is determined by both inputs and outputs)[13].
Table II shows us the ef?ciency scores from the pro?t ef?ciency model.
Ef?ciency scores are now signi?cantly lower on aggregate, with a mean ef?ciency
score across all DMUs of 0.943, which means that this model is more appropriately
identifying inef?ciencies. This re?ects the aspects of the pro?t model, that it combines
Figure 1.
Comparison of banks’
ROA (de?ned as net pro?t
after tax divided by
average total assets)
between December quarter
2007 and March quarter
2008
1.8
1.6
1.4
1.2
1.0
P
e
r
c
e
n
t
a
g
e
0.8
0.6
0.4
0.2
0.0
ANZ ASB
Note: Figures are annualized
BNZ Westpac TSB Kiwibank
12/07
3/08
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.988 0.967 0.958 1.000 1.000 1.000
March 2006 0.931 0.962 0.954 1.000 1.000 1.000
June 2006 1.000 0.981 1.000 0.975 0.987 0.976
September 2006 0.966 0.974 0.961 0.984 0.991 1.000
December 2006 0.939 1.000 1.000 0.976 0.985 0.981
March 2007 0.978 1.000 0.948 0.981 0.973 1.000
June 2007 0.929 0.997 0.990 0.981 1.000 1.000
September 2007 0.967 0.982 0.964 0.958 1.000 1.000
December 2007 0.959 1.000 0.972 0.950 0.973 0.962
March 2008 0.979 1.000 0.944 1.000 0.960 1.000
Average 0.964 0.986 0.969 0.981 0.987 0.992
Table I.
Ef?ciency scores for each
bank for each quarter,
from the CCR model
Bank safety
and soundness
243
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
allocative and technical ef?ciency, that it identi?es both cost and revenue inef?ciencies
(and the reported pro?t inef?ciency thus allows for both of these), and that it is not
oriented (in the way that the CCR model is). This is consistent with the de?nition of
pro?t ef?ciency as:
Ep ¼
p
0
y
0
2c
0
x
0
p
0
y* 2c
0
x*
where x
*
and y
*
represent the vectors of the optimal quantities of inputs and outputs,
x
0
and y
0
are the vectors of the observed quantities of the inputs and outputs,
and p
0
and c
0
the prices (Cooper et al., 2000; Paradi, 2006).
If we look at the BNZ, we ?nd that the major driver for (the now much more
pronounced) inef?ciency in the December 2007 and March 2008 quarters is bad and
doubtful debt expense. This contrasts with the simple ratios, where BNZ shows as
having lower levels of bad and doubtful debt expense than either ANZ or Westpac[14],
but the outcomes observed are a consequence of looking at all the inputs and outputs
together, on a multivariate basis. BNZ also shows some inef?ciency as deriving from
interest revenues, with insuf?cient interest bearing assets (and this is despite what
appears to be a relatively good net interest income performance in this quarter and the
previous ones).
ANZ’s inef?ciency also derives from bad and doubtful debt expense, but also from
an overuse of equity (re?ecting its need to hold a signi?cantly larger quantity of equity
because of the goodwill on its balance sheet from the surplus over book value paid
when it acquired the National Bank of New Zealand in 2003). These two factors
account for most of the ANZ’s inef?ciency throughout the period studied.
TSB’s inef?ciency in the March 2008 quarter derives from higher levels of bad and
doubtful debt expense (which appears to re?ect a standard adjustment in the March
quarter, which is the ?nal quarter of its ?nancial year), and a de?ciency in non-interest
income (which TSB generally earns less of in any case).
Kiwibank’s inef?ciency appears to be caused by higher levels of non-interest
expense, and, to a much lesser extent, from higher levels of interest expense.
Kiwibank’s non-interest expenses have been consistently greater than for other banks,
but the much stronger identi?cation of inef?ciency in this case may be a consequence
of revenues being lower than in previous quarters. What we are ?nding overall is that
the pro?t model is giving us more useful results than the CCR model, in terms of
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 1.000 0.928 0.862 1.000 0.998 1.000
March 2006 0.888 0.890 0.810 1.000 0.982 1.000
June 2006 1.000 0.926 1.000 0.947 0.987 0.961
September 2006 0.856 0.994 0.907 0.936 0.896 1.000
December 2006 0.876 1.000 1.000 0.917 0.990 0.930
March 2007 0.933 1.000 0.857 0.972 0.983 1.000
June 2007 0.912 1.000 0.973 0.949 1.000 0.764
September 2007 0.895 0.940 0.883 0.944 1.000 1.000
December 2007 0.971 1.000 0.832 0.889 1.000 0.887
March 2008 0.893 1.000 0.838 1.000 0.963 0.800
Average 0.922 0.968 0.896 0.955 0.980 0.934
Table II.
Ef?ciency scores for each
bank for each quarter,
from the pro?t ef?ciency
model
JFEP
2,3
244
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
understanding differences in pro?tability from quarter to quarter, with a stronger
relationship between pro?tability change and ef?ciency change.
The slacks-based model (which was run on a non-oriented basis), the results from
which are shown in Table III, shows signi?cantly greater levels of inef?ciency, with a
mean ef?ciency score across all DMUs of 0.797.
A major reason for the lower scores is that the slacks-based model also takes
account of non-radial slacks[15].
ANZ now shows as being particularly inef?cient in all but one of the ten quarters in
our analysis. The major contributors to this are excessive levels of equity and higher
levels of bad and doubtful debt expense, as previously discussed in respect of the pro?t
ef?ciency model.
ASB once again shows as relatively ef?cient, which is interesting in view of its
much lower level of gross income (de?ned as net interest income plus non-interest
income), as shown in Figure 2. The relatively high-ef?ciency re?ects the relatively
low usage of resources to generate that gross income. Where ASB was inef?cient,
it appears that this could be mainly attributed to bad and doubtful debt expense, with
some shortages of non-interest income.
BNZ’s inef?ciency in the March 2008 quarter appears to be largely attributable to
excess bad and doubtful debt expense, although it also shows as using excess equity
and interest expense, and as generating insuf?cient non-interest income. Although
Westpac showed as fully ef?cient in the March 2008 quarter, its inef?ciency in
previous quarters appears to be able to be attributed mainly to excess interest expense
and bad and doubtful debt expense.
TSB showed a dramatic drop in ef?ciency in the March 2008 quarter. The major
contributor to its inef?ciency is low levels of non-interest income, with lesser impacts
from high levels of bad and doubtful debt expense and equity. This would not be too
much of a surprise to analysts who have studied the bank’s ratios: the bank is rather
more strongly capitalised than others, largely to compensate for its small size and lack
of access to external capital, while the nature of its business means that its non-interest
income should be expected to be low.
Kiwibank also shows as relatively ef?cient overall, which is likely to re?ect its
higher levels of gross income, as evident in Figure 2. Once again, where there was
signi?cant inef?ciency, this appeared to relate mainly to higher levels of bad and
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.847 0.740 0.727 1.000 1.000 1.000
March 2006 0.605 0.677 0.626 1.000 1.000 1.000
June 2006 1.000 0.888 1.000 0.769 0.818 0.828
September 2006 0.581 0.719 0.735 0.862 0.842 1.000
December 2006 0.578 1.000 1.000 0.727 0.648 0.844
March 2007 0.684 1.000 0.673 0.820 0.538 1.000
June 2007 0.598 0.948 0.839 0.874 1.000 1.000
September 2007 0.694 0.728 0.767 0.873 1.000 1.000
December 2007 0.687 1.000 0.739 0.770 0.566 0.784
March 2008 0.689 1.000 0.690 1.000 0.273 1.000
Average 0.696 0.870 0.780 0.870 0.769 0.946
Table III.
Ef?ciency scores for each
bank for each quarter,
from the slacks-based
model
Bank safety
and soundness
245
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doubtful debt expense (even though bad and doubtful debt expense for Kiwibank, like
TSB, has been low). We note, however, that inef?cient Kiwibank DMUs show only
Kiwibank in their reference sets, while Kiwibank only plays a very small part in the
reference sets for their inef?cient DMUs[16]. There is thus a question as to whether
Kiwibank shares a common frontier with the other banks in the study.
If we look at the inef?ciencies across the data set as a whole, we ?nd that bad and
doubtful debt expense is the most prevalent. This is likely to re?ect its variability in
terms of its contribution to bank pro?tability, and the fact that the absolute values of
the numbers are sometimes very low, which means that the relative change in bad and
doubtful debt expense can be very large. It can be expected to be more signi?cant in its
impact on pro?ts in coming quarters, with higher levels of bad and doubtful debt
expense associated with the economic downturn. Our results may also be a
consequence of bad and doubtful debt expense being a negative output, which we are
treating as an input, and it is therefore considered appropriate to re-run the analysis
with this variable omitted. The economic meaning for this model would be to relate to
banks’ underlying pro?tability (i.e pro?t before bad and doubtful debt expense,
extraordinary items or tax), which also re?ects the fact that bad and doubtful debt
expense will generally relate to the bank’s performance in quarters prior to the one that
is being studied. It is this underlying pro?tability which is going to be more important
for a bank’s long-term pro?tability and soundness.
Results for the (non-oriented) slacks-based model, without bad and doubtful debts
as an input, are reported in Table IV. The mean ef?ciency across all DMUs is 0.812,
which is only marginally higher than for the previous model.
If we look at individual banks, we ?nd that ANZ shows as consistently relatively
inef?cient. The major contributor to this continues to be an excessive level of equity
(as discussed previously), although the main subsidiary reason is nowan insuf?ciency of
Figure 2.
Comparison of banks’
gross incomes (de?ned as
net interest income plus
non-interest income,
divided by average total
assets) between December
quarter 2007 and March
quarter 2008
4.5
4.0
3.5
3.0
2.5
2.0
P
e
r
c
e
n
t
a
g
e
1.5
1.0
0.5
0.0
ANZ
Note: Figures are annualized
ASB BNZ Westpac TSB Kiwibank
12/07
3/08
JFEP
2,3
246
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
non-interest income. ASB and BNZ also show, in most cases, a surplus of equity and a
shortage of non-interest income, although there are a number of quarters where ASB
shows a small excess of interest expense. ASB continues to show as fully ef?cient in the
March quarter 2008.
Westpac also shows as fully ef?cient in the March 2008 quarter: for the earlier
quarters in which it showed as less than fully ef?cient, the inef?ciency can be related to
excess equity and insuf?cient non-interest income. TSB’s inef?ciency derives from the
same two factors, although the shortage of non-interest income is much more
important in this case.
Kiwibank’s inef?ciency is different, with no shortage of non-interest income, except
in the March 2008 quarter when it was much less than usual. In Kiwibank’s case,
the major sources of inef?ciency are excess non-interest expense (ratio analysis shows
much higher relative costs than for the other banks) and insuf?cient interest income.
5. Summary and conclusion
This paper has sought to investigate the changes in pro?tability from quarter to
quarter for New Zealand’s banks with signi?cant retail business, with a particular
focus on the changes in pro?tability between the December quarter 2007 and the
March quarter 2008. We tested three different DEA methodologies, and found that
the slacks-based approach was most useful, although there was also value in using the
pro?t ef?ciency approach. The CCR model showed very little variation in ef?ciency
between the different banks, and thus was less effective at identifying the causes of
pro?tability change.
There was no clear indication that scale mattered, although the largest bank,
ANZ-National, showed as least ef?cient, and it is possible that some of this
inef?ciency may be a consequence of scale (although it is noted that the input that
consistently contributed to inef?ciency was equity, which ought to show positive
bene?ts from increased scale). ANZ is carrying extra equity, however, because of the
(intangible) goodwill associated with the acquisition of the National Bank of
New Zealand in 2003.
The success of the approach followed in this paper ought to be of assistance to
analysts in assessing the safety and soundness of New Zealand’s banks, a process
which is of particular importance in the New Zealand market because of the historic
lack of deposit insurance or other guarantees for depositors. It is not altogether clear
Quarter ANZ ASB BNZ Westpac TSB Kiwibank
December 2005 0.723 0.723 0.847 1.000 1.000 1.000
March 2006 0.700 0.674 0.678 1.000 1.000 0.967
June 2006 0.631 0.721 1.000 0.887 0.867 0.939
September 2006 0.708 0.657 0.843 0.890 0.889 0.928
December 2006 0.665 0.739 1.000 0.826 0.744 0.929
March 2007 0.710 1.000 0.781 0.869 0.663 1.000
June 2007 0.636 0.799 0.850 0.798 1.000 0.815
September 2007 0.681 0.630 0.777 0.797 1.000 0.905
December 2007 0.684 0.775 0.711 0.719 0.581 0.877
March 2008 0.693 1.000 0.735 1.000 0.342 0.702
Average 0.683 0.772 0.822 0.879 0.809 0.906
Table IV.
Ef?ciency scores for each
bank for each quarter,
from the Slacks-based
model, without bad
and doubtful debt
expense as an input
Bank safety
and soundness
247
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
that inclusion of bad and doubtful debts as a negative output (and thus as an input) is
helpful but we should be able to make further assessments of this as we look at the
results for later quarters when the global economic downturn has its effect in
increasing individual banks’ levels of bad and doubtful debt expense. There may be
merit in looking for other ways of dealing with bad and doubtful debt expense as an
undesirable output in our DEA models[17]. It will also be interesting to investigate the
use of new DEA models as these become more readily available as part of standard
software packages.
Notes
1. The introduction of a deposit protection scheme on 12 October 2008 has not undermined this
principle for wholesale deposits, and banks continue to provide quarterly disclosures to the
market. For more detail on the disclosure regime, see Mortlock (1996).
2. DEA has been criticised previously for only providing measures of technical ef?ciency. See,
for example, Berger and Mester (1997).
3. For a more detailed exposition of DEAmethodologies, see Cooper et al. (2000) or Avkiran (2006).
4. This is consistent with the approach outlined by Tulkens and Vanden Eeckaut (1995).
In practice, no indication was found in the results from this research that there was any
general change in banks’ ef?ciency through time.
5. Re?ecting concerns about banks’ use of bad and doubtful debt expense to smooth pro?ts,
we also tried use of total impaired and past due assets as an alternative input. The resulting
analysis showed a narrower range of ef?ciency scores (less discriminatory power).
The Spearman’s rank order correlation between the two sets of results was 0.783 (signi?cant
at less than 1 per cent).
6. Beyond that, physical capital is more-or-less insigni?cant as an input for New Zealand
banks, with ?xed assets comprising an average of 0.2 per cent of assets.
7. It was suggested to the author by a conference participant that the input-output set used was
essentially equivalent to measurement of pro?t, to which the author countered by
highlighting the inclusion of shareholders’ equity as an input. In fact, there are no signi?cant
correlations between banks’ return on assets (de?ned as pro?t after tax divided by average
total assets) and ef?ciency scores.
8. There were a total of 18 registered banks as at 31 December 2008.
9. TSB and Kiwibank were not required to adopt IFRS until the ?nancial years commencing in
2007, but the simpler nature of their operations means that the change in accounting
standards has had less impact on results.
10. This thus mitigates the potential problem of random error in the use of DEA.
11. The highest super-ef?ciency score was 1.25, well belowthe guideline level of concern of 2 – see
Hartman et al. (2001).
12. One New Zealand dollar is equivalent to approximately 60 cents US.
13. In fact, for a CCR model, the results from input-oriented or output-oriented model will be
identical.
14. Detailed data are omitted for reasons of space and readability, but are available from the
author on request.
15. There was no slacks-based pro?t model available in the version of the DEA-Solver software
(5.0) used for this research.
JFEP
2,3
248
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
16. We checked this by examining the lambdas in a variable returns to scale model. For only
four DMUs did a lambda value for Kiwibank exceed 0.1, and the highest of these was 0.216.
17. See the discussion by Burley (2006).
References
Avkiran, N.K. (2006), Productivity Analysis in the Service Sector with Data Envelopment Analysis,
3rd ed., N K Avkiran, Ipswich.
Banker, R.D., Charnes, A.W. and Cooper, W.W. (1984), “Some models for estimating technical
and scale inef?ciencies in data envelopment analysis”, Management Science, Vol. 30 No. 9,
pp. 1078-92.
Berger, A.N. and DeYoung, R. (1997), “Problem loans and cost ef?ciency in commercial banks”,
Journal of Banking & Finance, Vol. 21, pp. 849-70.
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,
pp. 895-947.
Berger, A.N., Hunter, W.C. and Timme, S.G. (1993), “The ef?ciency of ?nancial institutions:
a review and preview of research past, present and future”, Journal of Banking & Finance,
Vol. 17, pp. 221-49.
Burley, H.T. (2006), “Accounting for the in?uence of the environment in a two-stage analysis –
public hospitals”, in Avkiran, N.K. (Ed.), Productivity Analysis in the Service Sector with
Data Envelopment Analysis, 3rd ed., Ch. 17, N K Avkiran, Ipswich, pp. 181-92.
Charnes, A., Cooper, W.W. and Rhodes, E. (1978), “Measuring the ef?ciency of decision making
units”, European Journal of Operational Research, Vol. 2, pp. 429-44.
Cooper, W.W., Seiford, L.M. and Tone, K. (2000), Data Envelopment Analysis, Kluwer Academic,
Boston, MA.
Dyson, R.G., Allen, R., Camanho, A.S., Podinovski, V.V., Sarrico, C.S. and Shale, E.A. (2001),
“Pitfalls and protocols in DEA”, European Journal of Operational Research, Vol. 132,
pp. 245-59.
Golany, B. and Storbeck, J.E. (1999), “A data envelopment analysis of the operational ef?ciency of
bank branches”, Interfaces, Vol. 29, May-June, pp. 14-26.
Grifell-Tatje´, E. and Lovell, C.A.K. (1999), “Pro?ts and productivity”, Management Science,
Vol. 45 No. 9, pp. 1177-93.
Hartman, T.E., Storbeck, J.E. and Byrnes, P. (2001), “Allocative ef?ciency in branch banking”,
European Journal of Operational Research, Vol. 134, pp. 232-42.
Lovell, C.A.K. (2008), “Productivity accounting”, paper presented at the Asia-Paci?c Productivity
Conference, Taiwan, July.
Mortlock, G. (1996), “New disclosure regime for registered banks”, Reserve Bank Bulletin, Vol. 59
No. 1, pp. 21-9.
Paradi, J. (2006), “Pro?t ef?ciency – health insurance”, in Avkiran, N.K. (Ed.), Productivity
Analysis in the Service Sector with Data Envelopment Analysis, 3rd ed., Ch. 10, N K
Avkiran, Ipswich, pp. 101-9.
Tone, K. (2001), “A slacks-based measure of ef?ciency in data envelopment analysis”, European
Journal of Operational Research, Vol. 130, pp. 498-509.
Bank safety
and soundness
249
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Tone, K. (2002), “A slacks-based measure of super-ef?ciency in data envelopment analysis”,
European Journal of Operational Research, Vol. 143, pp. 32-41.
Tulkens, H. and Vanden Eeckaut, P. (1995), “Non-parametric ef?ciency, progress and regress
measures for panel data: methodological aspects”, European Journal of Operational
Research, Vol. 80, pp. 474-99.
Corresponding author
David Tripe can be contacted at: [email protected]
JFEP
2,3
250
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
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_588817132.pdf