Profitability, efficiency and growth in the private banking industry: Evidence from Switzerland and Liechtenstein

Abstract

This paper examines the overall performance of banks from Switzerland and Liechtenstein that are specialized in wealth management. More profitable and cost-efficient institutions feature comparatively low levels of assets managed per employee and high salaries paid. The ability to attract new money seems to be fostered by signalling competence via high margins rather than through the importance of bank-own funds or management mandates. Superior investment performance is associated with a higher share in mandated assets, higher salaries and less assets managed per employee. There is, however, no persistence found for the investment performance of wealth managers. No effects emerge from bank size and only little heterogeneity in overall performance can be traced back to the structure of assets, the size and location of the bank’s network, and differences in ownership.

Full text

“Profitability, efficiency and growth in the private banking industry: evidence from
Switzerland and Liechtenstein”
AUTH ORS Johann Burgstaller
Teodoro D. Cocca
ARTICLE INFO
Johann Burgstaller and Teodoro D. Cocca (2010). Profitability, efficiency and
growth in the private banking industry: evidence from Switzerland and
Liechtenstein. Banks and Bank Systems, 5(4)
JOURNAL "Banks and Bank Systems"
FOUNDER LLC “Consulting Publishing Company “Business Perspectives”
NUMBER OF REFERENCES
0
NUMBER OF FIGURES
0
NUMBER OF TABLES
0
© The author(s) 2019. This publication is an open access article.
businessperspectives.org

Banks and Bank Systems, Volume 5, Issue 4, 2010
10
Johann Burgstaller (Austria), Teodoro D. Cocca (Austria)
Profitability, efficiency and growth in the private banking industry:
evidence from Switzerland and Liechtenstein
Abstract
This paper examines the overall performance of banks from Switzerland and Liechtenstein that are specialized in
wealth management. More profitable and cost-efficient institutions feature comparatively low levels of assets managed
per employee and high salaries paid. The ability to attract new money seems to be fostered by signalling competence
via high margins rather than through the importance of bank-own funds or management mandates. Superior investment
performance is associated with a higher share in mandated assets, higher salaries and less assets managed per em-
ployee. There is, however, no persistence found for the investment performance of wealth managers. No effects emerge
from bank size and only little heterogeneity in overall performance can be traced back to the structure of assets, the size
and location of the bank’s network, and differences in ownership.
Keywords: private banking, profitability, efficiency, growth.
JEL classification: G21, L11.
Introduction
Economic performance in the private banking and
wealth management industry is rarely analyzed due
to a general lack of data (Cocca, 2008b). The studies
at hand, including their analysis of the available
data, are mainly descriptive in nature. In this paper
we, therefore, seek to provide a thorough examina-
tion of key operating figures and their determinants.
Measures of the profitability, efficiency and growth
of private banks1 in Switzerland and Liechtenstein
shall be applied in this respect.
Most banks operating in Switzerland offer financial
services to wealthy private and to institutional in-
vestors (Swiss Bankers Association, 2009). Swiss
banks are estimated to administer about one tenth of
the global assets under management (Swiss Bankers
Association, 2009). Switzerland’s market share in
the private assets that are managed off-shore is
about 30%, and approximately 60% of the assets
managed by Swiss banks are of foreign origin (Gei-
ger and Hürzeler, 2003; Swiss Bankers Association,
2009). The importance of wealth management for
the Swiss economy is stressed by Geiger and Hürze-
ler (2003), who argue that private banking activities
account for about half of the banking sector’s total
contribution. Switzerland and Liechtenstein are
strongly interlinked financial centers and institutions
can be considered as homogeneous in terms of the
market and the regulatory environment. Private
banks in Switzerland and Liechtenstein have rather
homogeneous salary policies, business and pricing
models, as well as a similar product offering.
The structure of this study is as follows. Section 1
describes the data gathered from financial reports to
Johann Burgstaller, Teodoro D. Cocca, 2010.
1 The term “private bank” is used to describe a bank that is specialized
in private banking and wealth management, not to discriminate “pri-
vate” banks from publicly-owned or incorporated banks.
analyze bank performance. The key figures to be
explained are the margin with fee-based activity, the
cost-income ratio, and the growth in managed assets
due to both the inflow of new money and invest-
ment performance. The hypotheses to be tested can
be found in Section 2, which establishes relation-
ships to the previous literature as well. As research
specific to institutions specialized in private banking
and wealth management is thin, we revert to studies
on banking in general. The literature survey, there-
fore, is fairly restricted with a focus on relatively
recent work related to European banks (see Goddard
et al., 2007, for a survey). Section 3 illustrates the
methods used, while the results are presented in Sec-
tion 4. The last Section summarizes and concludes.
1. Data
Basically, the data set used is made up by the fig-
ures gathered for and processed in “The interna-
tional private banking study 2007” (Cocca and
Geiger, 2007). In principle, this edition of the study
features banks from 15 countries. The data, how-
ever, are the most extensive for Switzerland and
Liechtenstein2. A bank has to fulfill the following
criteria to be considered: a share of income from
fees and commissions in total income of, at least,
one third and a documented strategic orientation
towards private banking activities. At the outset,
the sample consists of 94 Swiss and Liechtenstein
banks, but only 79 of them are useable for estima-
tion3. Ten institutions, however, are neglected be-
cause of their size (the two big banks) or an outly-
2 For these two countries, the figures for 2007 were amended in the
meantime.
3 Data for, at least, three years are necessary to calculate growth rates
and lagged growth rates for certain variables. Referring to 2007, 68 of
the 141 Swiss wealth managers which are classified as global, large,
medium and small players (micro players, defined as having AUM
below 1 billion CHF, are not considered) in Swiss Bankers Association
(2009) are in our sample. Their market coverage with respect to the
assets under management is about 75%.

Banks and Bank Systems, Volume 5, Issue 4, 2010
11
ing growth rate of assets related to mergers and
acquisitions. The final estimation sample consists
of 69 wealth managers, with five of them operating
in Liechtenstein. Most of the observations are for
the 2005-2007 period, one bank has data back to
2002. Except indicated otherwise, all monetary
figures are in million Swiss francs (CHF), and in
real 2006 terms. Two of the main variables in the
data set are the assets under management (AUM)
and the net new money inflow (NNM). Client
assets include managed fund, discretionary and
advisory assets, as well as fiduciary, savings and
time deposits. Assets held for custody and trans-
actional purposes are not included. New money
(newly acquired minus the withdrawn client as-
sets) does not include interest and dividend in-
come from the managed assets, as well as the
change in assets which is due to market and cur-
rency fluctuations.
The measures to be explained in this study are the
fee margin, the cost-income ratio, and two growth
indicators. Profitability in asset management is de-
scribed by a margin which is calculated as the ratio
of the net income from fees and commissions to
AUM (in basis points). The cost-income ratio (CIR)
is obtained by dividing total administrative expenses
by total net income. Depreciation is included with
the operating costs to account for the fact that not all
banks may own property in phyiscal capital, and
therefore have higher expenses due to renting and
leasing. See the next Section for a discussion of the
pros and cons of using the CIR as an efficiency
measure. As growth in private banking is under-
stood in terms of increasing the level of managed
client assets, the first growth variable we apply is
the net inflow of new money relative to AUM at the
beginning of the year. This is interpretable as the
growth in AUM due to the ability to raise new
money. The remaining increase in client assets is
attributed to investment performance.
Among the potential determinants of profitability,
efficiency and growth are bank size, specialization,
the composition of managed assets, the amount of
assets managed per employee, salaries paid and
capital strength. Size is measured by the logarithm
of total assets managed1. The performance of wealth
managers is supposed to differ also with respect to
their specialization in private banking activities. A
readily available indicator of specialization (diversi-
fication) is the share of net income from fee-based
activities in total net income. As different types of
products and business fields are not equally intense
1 The number of employees or the income from fees and commissions
could be applied as size indicators as well. As these, however, are highly
correlated with AUM, similar results are obtained as with total assets.
in terms of costs and revenues, Welch (2006) sug-
gests that the activity mix is one of the main influ-
ences on the efficiency of financial institutions2.
Two additional variables capture the structure of the
managed assets. Firstly, we consider the share of
AUM invested in own funds (in “own administered
collective investment schemes”) and, secondly, the
part of assets administered under discretionary man-
agement mandates. Remarkably, the two associated
fractions are hardly correlated. Other, non-discre-
tionary assets make up the remaining assets (most of
them by far). Wealth managers perceive investment
in own funds as a growth driver (being related to net
new money growth) and may use it to signal exclu-
sivity and superior investment skills (Cocca and
Geiger, 2007). In the private banking industry, dis-
cretionary mandates, as well as assets in own funds,
are valued as a profit-generating device. Both types
of funds are associated with relatively higher mar-
gins (Cocca and Geiger, 2007). However, “in-
house” investing and offering sophisticated products
for mandated assets may well lead to higher in-
comes, but such strategies might also be more in-
tense in terms of effort and costs.
The volume of assets managed per employee (in
million CHF) and personnel costs per capita (in
1000 CHF) are further potential determinants of
profitability, efficiency and growth. Additionally,
capital strength is used as an indicator for bank sta-
bility and risk (aversion). As the Bank for Interna-
tional Settlements (BIS) tier 1 capital ratio is not
available for most institutions in the sample, the
balance sheet capitization is applied. Table 1 (see
Appendix A) provides descriptive statistics for all
the determinants mentioned above (with size in mil-
lion CHF). Some further attributes are represented
by binary variables capturing geographical and
ownership differences across private banks. First,
we employ a dummy variable taking on the value 1
(and zero otherwise) if the bank operates in Liech-
tenstein, another one indicates foreign-controlled
banks or subsidiaries of foreign institutions. Further
dummies indicate whether the bank has more than
one location within Switzerland (or Liechtenstein)
and whether there are offices abroad. Domestic
banks are additionally divided into independent
banks and subsidiaries of a larger financial group. It
seems worth mentioning that almost all independent
Swiss banks in the sample are family-owned (with
family property of 50% and beyond). With the cor-
responding binary variable indicating domestic, but
dependent banks, the base group finally consists of
2 A siliar argument is put forward by DeYoung (1997). He argues that,
with respect to efficiency comparisons, peers should feature a similar
product mix.

Banks and Bank Systems, Volume 5, Issue 4, 2010
12
independent Swiss banks with one domestic location
and no foreign representations. Of the 69 banks in
the estimation sample, 5 operate in Liechtenstein, 40
are foreign-controlled and 8 are domestic but part of
a larger financial group. Multiple domestic locations
are indicated for 40 banks, 23 have offices abroad.
2. Hypotheses and relevant literature
2.1. Profitability.The first set of hypotheses shall
be composed for the fee margin earned on managed
assets. Larger banks might be more profitable be-
cause of economies of scale or the possibility to
extract rents due to market power. As it is defined,
however, the fee margin, rather than a measure of
overall profitability, is a price variable. Conse-
quently, its heterogeneity at the bank level may not
reflect direct effects of scale1. Cocca (2008a) argues
that smaller private banks pursue business strategies
that earn higher margins, which is supported by a
significantly negative correlation between the fee
margin and bank size within our sample. If the size
differential vanishes after controlling for cost factors
and efficiency, there might be an indication that the
associated strategical focus is a response to scale
diseconomies of smaller banks.
Pricing decisions are to be influenced by specializa-
tion (diversification) as well. One hypothesis is that
banks earning a larger share of their income through
fee-based activities are also able to achieve a higher
margin. On the other hand, diversification with re-
spect to income sources is presumed to result in
increased profitability (evidence for European banks
is provided by Beckmann, 2007; and Carbó Val-
verde and Rodríguez Fernández, 2007). As a conse-
quence, wealth managers with substantial retail
banking activities could use the corresponding sur-
pluses to engage in price competition. Fee margins
of Swiss and Liechtenstein banks, however, are
strongly increasing with the degree of specialization
in wealth management.
Another important determinant of bank profitability
is capital strength. A common view in the retail
banking literature is that a higher capitalization in-
dicates lower default risk, which enables banks to
refinance at a lower cost (Pasiouras and Kosmidou,
2007)2. On the other hand, a negative relation of the
1 For example Pasiouras and Kosmidou (2007) find that, in the Euro-
pean Union, smaller banks are more profitable. They argue that this
result is consistent with previous studies that observed economies of
scale mainly for smaller banks. The issue of scale economies in bank-
ing, however, remains controversial (Vander Vennet, 2002).
2 Additionally, Pasiouras and Kosmidou (2007) argue that banks with
higher capital levels also need less external funding and, therefore, have
higher profits. Another reason for a positive capital-profits relation is that
holding equity capital is relatively costly compared to debt (because of tax
reasons and the dilution of control), so banks seek to recover some of
these costs through higher margins (Saunders and Schumacher, 2000).
equity to assets ratio and bank performance might
be deduced from the conventional risk-return hy-
pothesis (Pasiouras and Kosmidou, 2007). Most
studies, however, observe the profitability of well-
capitalized banks to be relatively higher. As the
relations mentioned above are of lesser significance
for banks which are highly specialized in wealth
management services, we propose another role of
the capitalization variable. As it is argued by God-
dard et al. (2004a), a high capital-assets ratio might
be interpreted as signalling “that a bank is operating
over-cautiously and ignoring potentially profitable
diversification or other opportunities” (Goddard et
al., 2004a, p. 1073). If this is the case for wealth
managers as well, we might presume capital strength
to be negatively related to asset growth as well as
the fees that can be charged.
There is much research in favor of a significant rela-
tion between the cost-income ratio (CIR) and bank
margins as well as profits. With respect to European
banks, Vander Vennet (2002) suggests that “opera-
tional efficiency has become the major determinant
of bank profitability”. While Pasiouras and Kosmi-
dou (2007) view a high CIR primarily as a measure
of poor expenses management, Maudos and Fernán-
dez de Guevara (2004) argue that the CIR depicts
the quality of the management in selecting highly
profitable assets and low-cost liabilities. A man-
agement quality interpretation of the CIR is also
employed in this paper. To find out whether it is
represented in differential asset management mar-
gins, structural factors, such as the average level of
salaries and the assets managed per employee, are
controlled for. Cocca and Geiger (2007) observe that
lower fees per unit of AUM are charged by banks
with more AUM per employee. One possible expla-
nation for this is that with an increasing volume of
funds to be managed, the less intense the relation
between the client advisor and the client can be. Such
a limited attention to the client probably leads to a
“smaller penetration of the client base with products
and services” (Cocca and Geiger, 2007, p. 62).
Another question to be studied is whether there is
persistence of profits (POP) also in the private bank-
ing industry. Margins not converging to the industry
average over time may indicate the presence of
market power or the ability to deter entry (Berger et
al., 2000; Goddard et al., 2004a). Evidence on per-
sistence effects with the profitability of retail banks
is found by Berger et al. (2000), Goddard et al.
(2004a,b) and Carbó Valverde and Rodríguez
Fernández (2007). Additional hypotheses are related
to factors which are specific to the wealth manage-
ment context. Firstly, the fee margin should be
higher for banks with larger shares of assets being
mandated or invested in own funds (Cocca and Gei-
ger, 2007). With respect to our estimation sample,

Banks and Bank Systems, Volume 5, Issue 4, 2010
13
the correlation between margins and asset composi-
tion is much stronger for the latter group of assets.
Secondly, we also test whether pricing policies dif-
fer with respect to past growth and investment per-
formance.
As with efficiency and growth below, heterogeneity
in fee margins may be expected to emerge as well
with respect to the location of operations (Switzer-
land or Liechtenstein), ownership (foreign banks,
domestic banks being part of a larger financial
group) and the number of locations (multiple do-
mestic ones and/or locations abroad).
2.2. Efficiency. The literature of efficiency in finan-
cial institutions is vast and fragmented with respect
to the efficiency concepts and the measurement con-
cepts used (see, amongst others, Berger and Mester,
1997; Bauer et al., 1998; or Goddard et al., 2007). As
with profitability before, we will not refer to the
cross-country aspects of efficiency comparisons (for
which Pasiouras (2008), provides an overview).
The cost-income ratio (or efficiency ratio) is a very
popular measure of efficiency as it is easily calcu-
lated and readily available. It is seen as the key indi-
cator of efficiency within the banking community,
as well as by analysts and regulators (Vander Ven-
net, 2002; Forster and Shaffer, 2005; Beccalli et al.,
2006). However, the CIR has some drawbacks.
Firstly, a strict measurement of productive effi-
ciency is diluted as both revenues and costs depend
on the development of prices. A high value for the
CIR is both compatible with poor cost management
and fierce competition in output markets. Because
of this interpretation ambiguity (Bikker and Bos,
2004; DeYoung and Rice, 2004), the CIR rather is a
profitability measure. Secondly, the calculation of
simple accounting ratios, such as the CIR, does not
automatically ensure comparisons to adequate peers.
The researcher has to take care of the fact that effi-
ciency ratios are not easily comparable across firms
with differences in size and business models (DeY-
oung, 1997; Welch, 2006). Therefore, frontier effi-
ciency measures of managerial best practice are
generally favored over traditional accounting ratios
(Berger and Humphrey, 1997)1. Nevertheless, “the
measured efficiencies from all of the useful ap-
proaches should be reasonably consistent with stan-
dard non-frontier performance measures, such as
return on assets or the cost/revenue ratio” (Bauer et
al., 1998, p. 87). Bikker and Bos (2004) argue that
such a consistency is thwarted in case there are dif-
1 Non-parametric frontier methods, for example, use optimization
techniques and are viewed as producing more objective comparisons
and efficiency rankings. Efficiency considerations based on several
inputs and outputs can be condensed to a single measure. At the same
time, it is possible to locate the sources of input overuse and output
underproduction.
ferential effects of competitional forces on revenues.
As a consequence, a robustness analysis with re-
spect to frontier efficiencies shall accompany our
results generated from using the CIR.
At the outset, the relation between bank size and
efficiency seems ambiguous. On the one hand, lar-
ger banks might benefit from scale economies and
market power (see Forster and Shaffer (2005) for
details on the potential relation between size and
efficiency in retail banking)2. Larger banks might
also be able to hire managers with increased cost-
management skills but, on the other hand, to run a
large bank is more complex as well (DeYoung,
1997). The potential for scale effects in private
banking seems to be limited due to the services na-
ture of wealth management activities3. Possibly, a
size effect can only be achieved “by advising fewer
clients with larger assets” (Cocca, 2008b). In our
data sample, the correlation between the CIR and
size is low, but the cost-income ratio seems to be
strongly related to the assets managed per employee.
Specialization and diversification are presumed to
have a two-edged influence on efficiency as well. A
higher productivity with increased specialization
might be opposed by higher (personnel) costs. Addi-
tionally, economies of scope might arise in inte-
grated business models (Cocca, 2008a). The evi-
dence on the relation of profitability to efficiency is
mixed. Whereas Berger (1995) and Berger and
Mester (1997) find a positive relation, Casu and
Molyneux (2003) and Pasiouras (2008) do not.
Weill (2004) reports that cost efficiency measures
are positively correlated with bank profits for most
countries under study, but not for Switzerland. He
takes the observed negative connection as an indica-
tion of weak competition with profitable banks be-
having inefficiently (enjoying the so-called “quiet
life”). Well-capitalized banks are found to be more
efficient by Pasiouras (2008), a relation that is not
confirmed by Casu and Molyneux (2003).
In addition to the hypotheses related to the potential
determinants of the cost-income ratio mentioned
above, we will also test whether there is persistence in
efficiency ratios. Berger and Humphrey (1997) report
that, across the retail banking literature, there is appar-
ent evidence of persistence of relative efficiencies
across firms over time. Finally, it is presumed that asset
growth and efficiency are connected as expansionary
strategies might be costly, at least in the short run.
2 Pasiouras and Kosmidou (2007) find a significantly positive relation
between bank size and efficiency. Berger and Humphrey (1997) do as
well, but argue that it is less clear whether this is due to a benefit from
scale economies.
3 One cost factor for which there are marked differences between small
and large private banks are the costs to be incurred to meet regulatory
standards and requirements. Bührer et al. (2005) argue that the “regula-
tory burden” is much larger for smaller institutions.

Banks and Bank Systems, Volume 5, Issue 4, 2010
14
2.3. Growth. Two indicators for private banks’ growth
are applied in this study: the increases in assets due
to both the net inflow of new money and the in-
vestment performance. The literature on retail banks
measures growth mainly with respect to the balance
sheet total. Nevertheless, several hypotheses can be
derived which are suitable in the wealth manage-
ment context as well. Tests on the so-called “law of
proportionate effect” (LPE), for example, deal with
the relationship between growth and size, as well as
with persistence effects. The LPE is based on the
presumption that the growth rates of firms may, on
average, be equal (and random). As a consequence,
the distribution of firm size would become more
skewed and concentrated over time. Tschoegl
(1983) derives three testable hypotheses for the
LPE. We test the first two of these for the wealth
management industry in Switzerland and Liechten-
stein: that growth is independent of initial size and
that there is no growth persistence.
A relation of growth with (initial) bank size may
well be presumed to emerge from efficiency advan-
tages of large firms, the exercise of market power,
or the ability to pursue entry-deterring strategies
(Goddard et al., 2004a). No size-growth connection
is found by Tschoegl (1983) in his early, interna-
tional study. The same result emerges for European
banks in Goddard et al. (2004a) and, with one ex-
ception, Wilson and Williams (2000). For the USA,
however, Goddard et al. (2002) and Janicki and
Prescott (2006) report that large banks seem to grow
faster. Tschoegl (1983) argues that current growth is
not a good predictor of growth rates in subsequent
periods. Goddard et al. (2004a) reject the second
LPE hypothesis by observing positive growth per-
sistence for a sample of European banks.
Goddard et al. (2004a) also find that banks with a
high capital-assets ratio tend to pursue relatively
cautious growth policies. Profitability seems to be
another important determinant, a prerequisite for
future growth (Goddard et al., 2004a). As private
banks increase their scale of operations by attracting
inflows of new money, it may be presumed that the
corresponding efforts are supported by signals re-
lated to high profits as well as investment skills.
Cocca and Geiger (2007) argue that wealth manag-
ers attract money by superior investment perform-
ance. In this respect, clients will recognize a bank’s
competence especially with funds increasing in
value that are invested ‘in-house’ or under manage-
ment mandates.
3. Methodological framework
In order to study the determinants of key figures for
the wealth management industry, we employ the
following estimation process. The variables to be
explained are the fee margin, the CIR, as well as the
growth in assets due to NNM and performance. The
estimated models are similar for the dependent vari-
ables and involve the explanatory factors described
in Section 1. As persistence effects are to be exam-
ined, a dynamic panel estimation of the form:
yit =
D
i +
U
yi,tí1 +
E
' xtí1 +
H
it
is utilized, where y is the respective dependent vari-
able and x is a vector of explanatory variables with
E
as the corresponding coefficient vector. As a lag
of the dependent variable is considered, the princi-
pal estimation method is dynamic one-step GMM
(Arellano and Bond, 1991) with robust standard
errors. Heterogeneity (bank-specific effects) thereby
is accounted for through taking first differences. The
lagged regressand, which necessarily is correlated
with the error term(s), is instrumented by its second
lag1 and the first differences of the explanatory vari-
ables. In the end, results from System GMM estima-
tion (Arellano and Bover, 1995; Blundell and Bond,
1998) are reported. In this setting, the differenced
equation is amended by a level equation with an
instrument set consisting of the lagged difference of
the dependent variable and the levels of all the ex-
ogenous ones (including a constant). Tests on auto-
correlation of orders one and two (Arellano and
Bond, 1991) are used to ensure that the model is not
misspecified2. Instrument validity is evaluated by
use of the Hansen (1982) J-test from the two-step
model, which is robust to heteroscedasticity but may
be weakened with many instruments (Roodman,
2009a,b). Potential endogeneity of explanatory vari-
ables is no issue in the model described above as the
first lags of the regressors are considered. That is,
the characteristics of the wealth managers at the
beginning of the period are used to examine the
banks’ performance over the year.
If the past of the dependent variable proves insig-
nificant (if there is no persistence), a non-dynamic
estimation is applied for robustness purposes.
Thereby, the primary choice is a random-effects
(RE) panel model. Fixed effects (FE) are not con-
sidered in these cases due to the following reasons.
Firstly, time-invariant information (such as binary
variables) cannot be used in FE estimation. Sec-
ondly, there is little within variation in our data due
to the short sample period. In such cases, the inclu-
sion of individual effects would leave no variation
to be explained by the applied regressors. The test
1 In principle, lagged levels of the dependent variable, dated t – 2 and
earlier, are eligible instruments. Using only the first of them follows the
advice to keep the list of instruments short, in order to avoid bias from
overfitting endogenous variables (see Roodman, 2009a).
2 If the error in the original model is assumed to be free of autocorrela-
tion, then the first-differenced errors are serially correlated. Second
order autocorrelation (in the first-differenced residuals) should be absent
to ensure consistency of the estimates.

Banks and Bank Systems, Volume 5, Issue 4, 2010
15
for individual effects with unbalanced panels of
Baltagi and Li (1990) is applied, which is a refine-
ment of the test of Breusch and Pagan (1980). In
case its null hypothesis of no random effects is ac-
cepted, a pooled model is estimated by ordinary
least squares (OLS). The absence of serial correla-
tion in the non-dynamic panel models is examined
by the test proposed by Wooldridge (2002).
4. Results
The estimation results from the dynamic panel mod-
els are presented in Table 2 (see Appendix A). The
equation for the fee margin over AUM features but
two significant factors. On the one hand, margins
are persistent to a large degree within the wealth
management industry, as it is frequently found for
profits in retail banking (see Section 2.1). The
lagged dependent variable shows a coefficient of
about 0.59. Such strong persistence indicates that
market forces do not exact a reversion of margins to
the industry average, at least in the short run. Mar-
ket power and entry deterrence abilities may also
enable individual private banks to maintain above-
average margins for a certain time. Among the other
potential determinants, only the previous growth by
an acquisition of new money is significant (at the
10% level). Private banks with success in attracting
new funds seem to obtain potential for increasing
profit margins as well.
Although many of the variables that describe size,
specialization, asset structure, efficiency and capital
strength are correlated to the fee margin individu-
ally1, none of them emerge to be significant in the
full model. After controlling for the respective other
bank characteristics, the individual effects seem to
cancel out as, with respect to heterogeneity of mar-
gins, two groups of banks with rather homogenous
characteristics seem to be present. The smaller
wealth managers are the ones that pursue strategies
which enable higher margins (Cocca, 2008a). Al-
though they are specialized in advising very wealthy
clients (Cocca and Geiger, 2007; Cocca, 2008b),
their investment in own funds, the assets managed
per employee and the salaries paid are comparably
low – three factors that are positively correlated to
margins across the entire sample. Additionally,
smaller banks are significantly better capitalized.
Rather surprisingly, having in mind the importance
of offering sophisticated products within the indus-
try, banks with high shares of mandated assets do
1 As mentioned in Section 2.1, margins are negatively correlated with
bank size and increasing with the degree of specialization. The fee
margin is also higher for banks with a larger share of assets in own
funds, higher personnel costs per employee and for banks which are
better capitalized. The correlation with AUM per employee, however, is
significantly negative.
not earn higher margins2. Also specialization and
past investment performance do not convey addi-
tional information on how to characterize wealth
managers with respect to their fee margins. The same
is true for geographical and ownership differences, as
well as for the number and site of locations.
Two observations from Table 2 lead us to estimate
also a non-dynamic model for the fee margin equa-
tion. Firstly, the p-value for the AR(1) test is above
0.05 and, secondly, the lagged margin is only sig-
nificant at the 10% level. As the lagrange multiplier
(LM) test on individual effects of Baltagi and Li
(1990) prefers the random effects (RE) model to
pooled OLS3, results from using RE are reported in
Table 3 (see Appendix A). Tests on non-signifi-
cance are based on robust standard errors. With the
non-dynamic estimation, several effects emerge
being statistically significant. Institutions that are
more specialized in private banking activities, as
well as banks with higher salaries, charge relatively
higher margins. The former result is, in principle,
consistent with the notion that more diversified
banks charge lower fees to gain market shares. The
success of such strategies, however, may be doubted
in view of the persistence effect found in the dy-
namic model. With more assets managed per em-
ployee, fee margins decrease. This result is in line
with limited client attention argument of Cocca and
Geiger (2007). As an additional outcome, also the
difference between the fees of banks from Switzer-
land and Liechtenstein is found to be different from
zero. Neither of the two growth variables seems to
be a discriminatory factor in this setting. Leaving
them out of the list of explanatory variables leads to
a gain in observations and degrees of freedom. By
doing so (the corresponding results are not shown in
tabular form), the share of mandated assets (posi-
tively) and the CIR (negatively) emerge as addi-
tional determinants. This suggests that management
quality and offering sophisticated products might as
well play a role in shaping profits. None the less, we
are inclined to not attach too much value to the re-
sults just described, as the misspecification of the
dynamic model is not very severe and the influence
of past on present margins is fairly strong. Once the
persistence effect is accounted for, the information
content of several factors individually correlated
with the fee margin vanishes. In neither model,
however, there is evidence on size effects or on a
significant association to the investment in own
funds, the capital ratio or past performance.
2 The fee margin and the share of assets under discretionary manage-
ment mandates are not even significantly correlated in our sample.
3 The p-value for the test with pooled OLS as the null hypothesis is
0.00. The serial correlation test of Wooldridge (2002) accepts the null of
no autocorrelation in the errors of a panel model with a p-value of 0.55.

Banks and Bank Systems, Volume 5, Issue 4, 2010
16
A strong persistence effect emerges for the cost-
income ratio (see Table 2), which is in line with
previous evidence on efficiency in the banking
industry (Berger and Humphrey, 1997). According
to the results from the autocorrelation and instru-
ment validity tests, the dynamic model is well
specified. The fact that no size effect on efficiency
is found coincides with the notion that scale effects
are little conceivable with the management of cli-
ent assets. Larger banks may well possess an im-
proved cost management, but also are inherently
more complex to run. Specialization and asset
structure cannot be used to distinguish efficient
from inefficient banks either. If a stronger focus on
private banking activities leads to respective pro-
ductivity gains, foregone economies of scope may
counter an improvement of efficiency measures.
Given personnel costs (which are negatively sig-
nificant), more assets managed per employee, ce-
teris paribus, are associated with higher cost-
income ratios. This is in line with the prior obser-
vation that fee margins are negatively correlated
with funds managed per staff member. Our results
do not confirm the observation of Weill (2004) that
profitable Swiss banks are relatively inefficient. At
least for the private banks in the applied sample
this can be negated, as the lagged fee margin is
insignificant. Additional influences are found pre-
sent for past growth through new money (linked to
lower values for the CIR) and the existence of mul-
tiple locations within the country (such banks have
a higher CIR).
Due to the drawbacks of the cost-income ratio as
a measure of efficiency (see Section 2.2), we also
apply inefficiency scores from non-parametric
data envelopment analysis (DEA) as an alterna-
tive. Details on the calculation of these figures
can be found in Appendix B. Using scores from
DEA analysis as a dependent variable, however,
leads to some complications with respect to esti-
mation (which are also described briefly in Ap-
pendix B). Suitable methods were derived by Si-
mar and Wilson (2007). In these models, how-
ever, none of the regressors is significant at con-
ventional levels, except for a persistence effect
also in these measures. Replacing the CIR with
DEA scores as an explanatory variable does not
affect the results from the three other equations.
This might be due to the fact that the CIR and the
DEA scores are related to a relatively large ex-
tent. The associated correlation coefficient for the
pooled sample is 0.28, which is rather high com-
pared to the correlations reported by Bauer et al.
(1998). For our sample of wealth managers it
might be concluded that cost-income ratios are
useful measures of managerial competence. Ambi-
guities in interpretation do not seem to be very
important as the impact of competitional forces on
the revenue side can be presumed relatively ho-
mogenous within our sample of banks.
The equation for the AUM growth due to new
money is well specified and the lagged NNM share
is positively significant at the 5% level. Thus, there
is also persistence in the part of growth that is due to
the inflow of new funds. Initial bank size, however,
is no indicator of a subsequently emerging, dispro-
portionate acquisition of money. Both results are in
line with those reported by Goddard et al. (2004a)
for European retail banks, and the latter finding
confirms the absence of a size-growth relationship
in private banking observed also by Cocca and Gei-
ger (2007) and Cocca (2008b). Simple preestimation
correlation analysis (associated results are not
shown in tabular form) reveals that growth is corre-
lated with the AUM share in own funds, personnel
costs per employee, and the fee margin. From these
factors, ceteris paribus, only the lagged margin is
statistically significant. Thus, if wealth managers are
able to attract funds by signalling competence with a
high share of assets invested in own funds, such an
effect is hidden behind signals associated with high
margins and profitability. High fees, for example,
may provide an indication of which banks offer
innovative products promising high returns. Supe-
rior past investment performance does not seem to
be rewarded by an increased inflow of new money,
which is against the argument provided by Cocca
and Geiger (2007). As well surprising is that our
results confirm the observation of Goddard et al.
(2004a), who found that well-capitalized banks
grow more slowly. Thus, it seems that the capital
strength also of wealth managers conveys informa-
tion about the coutiousness behind different busi-
ness models.
Investment performance makes up the final variable
to be analyzed. The first observation is that there is
no persistence in investment success, as the coeffi-
cient of lagged performance is not statistically dif-
ferent from zero. Therefore, the non-dynamic model
in Table 3 will be discussed. The associated test on
the pooled data does not reject the null of no random
effects (with a p-value of 0.89). As serial correlation
is present, we estimate a pooled OLS model with
Newey and West (1987) standard errors1. Also with
growth measured in the performance context, size
and specialization do not appear to be important
determinants. A superior investment performance is
evident for banks with a larger share of mandated
assets and wealth managers who pay higher salaries.
1 Regrettably, too few observations are present for several banks to
pursue the involved correction for serial correlation. Therefore, the
number of useable observations (banks) is reduced to 134 (50).

Banks and Bank Systems, Volume 5, Issue 4, 2010
17
The imposition of large AUM volumes per head
seems detrimental to growth via market success.
Poor cost management, depicted by high values for
the CIR, is an indicator for poor performance as
well. Ceteris paribus, also banks with locations out-
side their home country exhibit lower growth fig-
ures in this respect.
Conclusions
This paper systematically explores the performance
determinants for institutions that are specialized in
private banking activities. For the example of Swiss
and Liechtenstein banks, it is examined which at-
tributes characterize wealth managers with high
margins, superior cost efficiency and growth. Pre-
eminent features of banks that charge a higher fee
margin over mananged assets are a high degree of
specialization in wealth management, comparatively
low levels of assets managed per employee, and
higher salaries paid. In a dynamic model, however,
a fairly significant margin persistence effect masks
out most of the factors that are individually corre-
lated with profitability. In this setting, fee margins
are found to be higher for private banks with more
recent growth through the inflow of net new money.
Interestingly, wealth managers with less assets man-
aged and higher salaries paid per head turn out to
feature the lowest cost-income ratios. The attraction
of new money seems to be fostered by signalling
competence via high margins rather than through a
large importance of bank-own funds or management
mandates. Past performance is unrelated to new
money inflows as well, which is rather surprising.
Additionally, well-capitalized banks appear to pur-
sue more cautious growth policies. Superior invest-
ment performance is associated with a higher share
in mandated assets and higher salaries. The level of
assets managed per employee, as well as cost man-
agement abilities are reflected in market success too.
There is, however, no persistence found for the in-
vestment performance of wealth managers in Swit-
zerland and Liechtenstein. No effects emerge from
bank size and only little heterogeneity in overall
performance can be traced back to the structure of
assets, the size and location of the bank’s network,
and differences in ownership.
References
1. Arellano, M. and S. Bond (1991). Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Ap-
plication to Employment Equations, Review of Economic Studies, 58, pp. 277-297.
2. Arellano, M. and O. Bover (1995). Another look at the instrumental variable estimation of error-components mod-
els, Journal of Econometrics, 68, pp. 29-51.
3. Baltagi, B.H. and Q. Li (1990). A lagrange multiplier test for the error components model with incomplete panels,
Econometric Reviews, 9, pp. 103-107.
4. Banker, R.D., A. Charnes and W.W. Cooper (1984). Some models for estimating technical and scale inefficiencies
in data envelopment analysis, Management Science, 30, pp. 1078-1092.
5. Bauer, P.W., A.N. Berger, G.D. Ferrier and D.B. Humphrey (1998). Consistency Conditions for Regulatory
Analysis of Financial Institutions: a Comparison of Frontier Efficiency Methods, Journal of Economics and Busi-
ness, 50, pp. 85-114.
6. Beccalli, E., B. Casu and C. Girardone (2006). Efficiency and Stock Performance in European Banking, Journal of
Business Finance and Accounting, 33, pp. 245-262.
7. Beckmann, R. (2007). Profitability of Western European banking systems: panel evidence on structural and cycli-
cal determinants, Discussion Paper (Series 2: Banking and Financial Studies) No 17/2007, Deutsche Bundesbank,
Frankfurt.
8. Berger, A.N. (1995). The Profit-Structure Relationship in Banking: Tests of Market-Power and Efficient-Structure
Hypotheses, Journal of Money, Credit and Banking, 27, pp. 404-431.
9. Berger, A.N., S.D. Bonime, D.M. Covitz and D. Hancock (2000). Why are bank profits so persistent? The roles of
product market competition, informational opacity, and regional/macroeconomic shocks, Journal of Banking and
Finance, 24, pp. 1203-1235.
10. Berger, A.N. and D.B. Humphrey (1997). Efficiency of financial institutions: international survey and directions
for future research, European Journal of Operational Research, 98, pp. 175-212.
11. Berger, A.N. and L.J. Mester (1997). Inside the black box: what explains differences in the efficiencies of financial
institutions?, Journal of Banking and Finance, 21, pp. 895-947.
12. Bührer, C., I. Hubli and E. Marti (2005). The regulatory burden in the Swiss wealth management industry, Finan-
cial Markets and Portfolio Management, 19, pp. 99-108.
13. Bikker, J.A. and J.W.B. Bos (2004). Trends in Competition and Profitability in the Banking Industry: a Basic
Framework, Working Paper No. 18, De Nederlandsche Bank, Amsterdam.
14. Blundell, R. and S. Bond (1998). Initial conditions and moment restrictions in dynamic panel data models, Journal
of Econometrics, 87, pp. 115-143.
15. Breusch, T.S. and A.R. Pagan (1980). The LM Test and its Applications to Model Specification in Econometrics,
Review of Economic Studies, 47, pp. 239-253.
16. Carbó Valverde, S. and F. Rodríguez Fernández (2007). The determinants of bank margins in European banking,
Journal of Banking and Finance, 31, pp. 2043-2063.

Banks and Bank Systems, Volume 5, Issue 4, 2010
18
17. Casu, B. and P. Molyneux (2003). A comparative study of efficiency in European banking, Applied Economics,
35, pp. 1865-1876.
18. Cocca, T.D. (2008a). Modernes Private Banking: Zehn Thesen zu Entwicklung und Perspektiven der Vermögens-
verwaltung, in: Pernsteiner, H. (ed.), Finanzmanagement aktuell, Linde, Vienna, pp. 681-712.
19. Cocca, T.D. (2008b). Size effects and integrated business models in private banking, Journal of Financial Trans-
formation, Issue 23, pp. 26-30.
20. Cocca, T.D. and H. Geiger (2007). The International Private Banking Study 2007, Swiss Banking Institute, Uni-
versity of Zurich.
21. DeYoung, R. (1997). Measuring Bank Cost Efficiency: Don’t Count on Accounting Ratios, Financial Practice
and Education, 7, pp. 20-31.
22. DeYoung, R. and T. Rice (2004). How do banks make money? The fallacies of fee income, Federal Reserve Bank
of Chicago Economic Perspectives, 28, pp. 34-51.
23. Forster, J. and S. Shaffer (2005). Bank efficiency ratios in Latin America, Applied Economics Letters, 12, pp. 529-532.
24. Geiger, H. and H. Hürzeler (2003). The transformation of the Swiss private banking market, Journal of Financial
Transformation, Issue 9, pp. 93-103.
25. Goddard, J., D.G. McKillop and J.O.S. Wilson (2002). The growth of US credit unions, Journal of Banking and
Finance, 26, pp. 2327-2356.
26. Goddard, J., P. Molyneux and J.O.S. Wilson (2004a). Dynamics of Growth and Profitability in Banking, Journal
of Money, Credit and Banking, 36, pp. 1069-1090.
27. Goddard, J., P. Molyneux and J.O.S. Wilson (2004b). The profitability of European banks: a cross-sectional and
dynamic panel analysis, Manchester School, 72, pp. 363-381.
28. Goddard, J., P. Molyneux, J.O.S. Wilson and M. Tavakoli (2007). European banking: an overview, Journal of
Banking and Finance, 31, pp. 1911-1935.
29. Hansen, L.P. (1982). Large Sample Properties of Generalized Method of Moments Estimators, Econometrica, 50,
pp. 1029-1054.
30. Janicki, H.P. and E.S. Prescott (2006). Changes in the Size Distribution of U.S. Banks: 1960-2005, Federal Re-
serve Bank of Richmond Economic Quarterly, 92, pp. 291-316.
31. Maudos, J. and J. Fernández de Guevara (2004). Factors explaining the interest margin in the banking sectors of
the European Union, Journal of Banking and Finance, 28, pp. 2259-2281.
32. Newey, W.K. and K.D. West (1987). A Simple, Positive Semi-Definite, Heteroscedastic and Autocorrelation Con-
sistent Covariance Matrix, Econometrica, 55, pp. 703-708.
33. Pasiouras, F. (2008). International evidence on the impact of regulations and supervision on banks’ technical effi-
ciency: an application of two-stage data envelopment analysis, Review of Quantitative Finance and Accounting,
30, pp. 187-223.
34. Pasiouras, F. and K. Kosmidou (2007). Factors influencing the profitability of domestic and foreign commercial
banks in the European Union, Research in International Business and Finance, 21, pp. 222-237.
35. Roodman, D. (2009a). How to do xtabond2: an introduction to difference and system GMM in Stata, Stata Jour-
nal, 9, pp. 86-136.
36. Roodman, D. (2009b). A Note on the Theme of Too Many Instruments, Oxford Bulletin of Economics and Statis-
tics, 71, pp. 135-158.
37. Saunders, A. and L. Schumacher (2000). The determinants of bank interest rate margins: an international study,
Journal of International Money and Finance, 19, pp. 813-832.
38. Simar, L. and P.W. Wilson (2007). Estimation and inference in two-stage, semi-parametric models of production
processes, Journal of Econometrics, 136, pp. 31-64.
39. Swiss Bankers Association (2009). Wealth Management in Switzerland, Basel.
40. Tschoegl, A.E. (1983). Size, Growth, and Transnationality Among theWorld’s Largest Banks, Journal of Business,
56, pp. 187-201.
41. Vander Vennet, R. (2002). Cost and Profit Efficiency of Financial Conglomerates and Universal Banks in Europe,
Journal of Money, Credit and Banking, 34, pp. 254-282.
42. Weill, L. (2004), On the Relationship Between Competition and Efficiency in the EU Banking Sectors, Kredit und
Kapital, 37, pp. 329-352.
43. Welch, P. (2006). Measuring efficiency in the finance factory: time for a rethink, Journal of Financial Transfor-
mation, Issue 18, pp. 51-60.
44. Wilson, J.O.S. and J.M. Williams (2000). The size and growth of banks: evidence from four European countries,
Applied Economics, 32, pp. 1101-1109.
45. Wooldridge, J.M. (2002). Econometric Analysis of Cross Section and Panel Data, MIT Press, Cambridge.

Banks and Bank Systems, Volume 5, Issue 4, 2010
19
Appendix A
Table 1. Descriptive statistics
Characteristic Mean St. dev. Minimum Maximum
AUM 21547.64 32173.78 452.00 187759.20
Fee income share 67.11 11.85 41.51 98.76
AUM in own funds 7.81 11.07 0.00 46.67
AUM under DMM 25.30 14.40 0.27 69.81
AUM per employee 61.14 27.72 20.92 182.72
Personnel costs per employee 240.39 95.11 114.92 912.40
Capital ratio 18.62 10.47 3.69 53.42
Fee margin 73.61 23.11 18.44 169.32
Cost-income ratio 60.96 11.53 34.54 88.54
AUM growth through NNM 6.28 11.03 -13.58 51.56
Performance 6.19 7.06 -19.62 26.13
Notes: This Table presents descriptive statistics on the largest estimation sample (154 observations on 69 banks) for the following
bank characteristics: the level of assets under management (AUM, million CHF), the share of income from fees and commissions in
total income (%), the shares of assets in own funds and under discretionary management mandates (%), the AUM managed per em-
ployee (million CHF), the personnel costs per employee (1000 CHF), the share of equity capital in the balance sheet (%), net income
from fees and commissions relative to AUM (the fee margin, basis points), the cost-income ratio (%), net new money relative to AUM
of the previous period (AUM growth through NNM, %), and the AUM growth due to the investment performance (%).
Table 2. Results from dynamic panel estimation
Fee margin CIR NNM share Performance
Log(AUM) -0.652
(0.80)
-1.016
(0.28)
-0.358
(0.81)
1.213
(0.37)
Fee income share 0.202
(0.41)
-0.155
(0.22)
-0.014
(0.89)
0.065
(0.40)
AUM in own funds -0.021
(0.91)
0.060
(0.34)
-0.040
(0.66)
-0.038
(0.71)
AUM under DMM 0.029
(0.72)
-0.024
(0.69)
-0.035
(0.70)
0.170*
(0.09)
AUM per employee -0.185
(0.32)
0.152*
(0.07)
0.018
(0.84)
-0.329**
(0.00)
Personnel costs per employee 0.042
(0.35)
-0.045**
(0.01)
0.003
(0.94)
0.066**
(0.01)
Capital ratio 0.088
(0.63)
-0.043
(0.72)
-0.358**
(0.03)
0.008
(0.95)
Fee margin 0.587*
(0.06)
0.115
(0.17)
0.264**
(0.01)
-0.117*
(0.07)
CIR -0.033
(0.83)
0.750**
(0.00)
0.146
(0.11)
-0.171**
(0.03)
AUM growth through NNM 0.199*
(0.08)
-0.084*
(0.07)
0.332**
(0.03)
0.041
(0.53)
Performance 0.113
(0.18)
-0.006
(0.91)
0.084
(0.20)
0.046
(0.57)
Liechtenstein -3.670
(0.32)
-1.013
(0.77)
3.622
(0.22)
2.055
(0.65)
Foreign -0.864
(0.82)
-0.435
(0.80)
-3.648
(0.19)
-1.000
(0.66)
Dependent -3.729
(0.28)
-2.324
(0.43)
-2.604
(0.45)
1.711
(0.56)
Domestic locations -1.505
(0.74)
5.470**
(0.02)
-0.088
(0.98)
1.177
(0.69)
Locations abroad -1.156
(0.78)
-0.894
(0.64)
1.629
(0.54)
-5.786**
(0.04)
Constant 25.000
(0.45)
25.544
(0.13)
-12.637
(0.42)
10.785
(0.42)
Number of banks 69 69 66 66
Number of observations 154 154 150 150
Number of instruments 26 26 25 25
AR(1) test (p-value) 0.054 0.016 0.019 0.177
AR(2) test (p-value) 0.390 0.217 0.708 0.693
Hansen test (p-value) 0.169 0.313 0.655 0.661
Notes: This Table presents results on how the four key figures vary across banks with certain characteristics. The dependent variables
(see the column headers) are the fee margin, the cost-income ratio (CIR), the growth in AUM due to the net inflow of new money
(termed NNM share in the column header), and the growth in assets due to the investment performance. All models are estimated by
System GMM. The explanatory variables, apart from the dummies, are lagged once. The p-values for the t-test on non-significance are
given in parentheses. One asterisk is for statistical significance at the 10% level, two of them indicate significance at the 5% level.

Banks and Bank Systems, Volume 5, Issue 4, 2010
20
Table 3. Results from non-dynamic models
Fee margin Performance
Log(AUM) -2.491 (0.42) 1.321 (0.21)
Fee income share 0.549** (0.00) 0.047 (0.51)
AUM in own funds 0.326 (0.12) -0.025 (0.80)
AUM under DMM 0.022 (0.83) 0.116* (0.08)
AUM per employee -0.420** (0.00) -0.234** (0.05)
Personnel costs per employee 0.099** (0.01) 0.050* (0.07)
Capital ratio 0.200 (0.42) 0.053 (0.61)
Fee margin -0.080 (0.29)
CIR -0.192 (0.21) -0.145* (0.10)
AUM growth through NNM 0.116 (0.31) 0.108 (0.10)
Performance 0.137 (0.17)
Liechtenstein -8.217** (0.03) 1.618 (0.64)
Foreign 1.093 (0.80) -1.467 (0.42)
Dependent -5.511 (0.33) 1.493 (0.51)
Domestic locations -1.410 (0.84) 1.566 (0.51)
Locations abroad -1.085 (0.85) -5.695** (0.01)
Constant 66.945** (0.01) 5.079 (0.67)
Number of banks 69 50
Number of observations 154 134
Estimation method RE OLS
Notes: This Table presents results from non-dynamic models for the fee margin and the investment performance as dependent vari-
ables. The fee margin equation is estimated by use of a random-effects (RE) panel model with robust standard errors. OLS with
Newey and West (1987) standard errors and a presumed serial correlation of order 1 is applied for the performance equation. The
explanatory variables, apart from the dummies, are lagged once. The p-values for the t-test on non-significance are given in paren-
theses. One asterisk is for statistical significance at the 10% level, two of them indicate significance at the 5% level.
Appendix B. Inefficiency scores from data envelopment analysis
Data envelopment analysis (DEA) is a non-parametric method that uses linear programing techniques to calculate
(in)efficiency scores for so-called decision-making units (DMU). The obtained scores measure efficiency relative to a
best-practice benchmark which is identified from the dataset at hand. That is, the most efficient DMU make up the
estimated production frontier. In this paper, scores for the (pure) technical efficiency of private banks are estimated. As
it is hardly possible to obtain sensible price information for the wealth managers’ inputs and outputs, we refrain from
an assessment of economic (cost and profit) efficiency. The applied model features an input-minimization orientation
and variable returns to scale (Banker et al., 1984), and yields indicators of proportional input overuse (output under-
production). The net income from fees and commissions and the net inflow of new money are chosen as outputs, the
input vector consists of AUM, personnel expenditures and other operating costs. Scores are normalized to 1 (100%) for
efficient DMU, lower values are assigned to inefficient banks. A score of 0.8, for example, indicates that a 20% reduction
of all inputs (while maintaining the output level) would be needed to reach the efficient benchmark. To match the CIR
with respect to the direction of (in)efficiency, the reciprocal values of these scores are used.
Replacing the CIR for DEA scores in case differences in efficiency are to be explained is a more challenging task. Casu
and Molyneux (2003), for example, discuss several problems emerging in this context and possible solutions to these.
Tobit regression has been commonly applied to account for the censored nature of DEA efficiency scores. However,
Simar and Wilson (2007) show that tobit (as is OLS) is not the proper choice. Another important problem is that the
DEA efficiency scores are relative measures which gives rise to the so-called “dependency problem” (the fact that
efficiency scores are partly interdependent leads to a violation of the independence assumption in regression analysis),
which invalidates standard inference techniques (Casu and Molyneux, 2003). As Simar and Wilson (2007) argue, the
estimated DEA efficiency scores “are serially correlated, and in a complicated, unknown way”. Some previously pro-
posed bootstrap-based solutions to these complexities were considered inappropriate by Simar and Wilson (2007). The
algorithms they provide and recommend instead (based on truncated regression with suitably bootstrapped standard
errors) shall be applied in our empirical investigation. With algorithm # 2 in Simar and Wilson (2007), inefficiency
scores are additionally corrected for estimation bias.