Journal of Financial Services Research 19:2/3 147±162, 2001 # 2001 Kluwer Academic Publishers. Manufactured in The Netherlands. Technological and Environmental Differences in the European Banking Industries MOHAMED E. CHAFFAI Universite de Sfax, Tunisia MICHEL DIETSCH CEPF-Institut d'eâtudes politiques, Universite Robert Schuman de Strasbourg, France ANA LOZANO-VIVAS Department of Economics, Universidad de MaÂlaga, Spain Abstract This paper analyzes the productive differences of banking among countries. It proposes a Malmquist type index that allows intercountry productivity differences to be broken down into pure technological differences and differences due to environmental effects. The most relevant feature of this index is its symmetry, since it avoids the problem of measurements being sensitive to the choice of the benchmark country. This index is used to explain the productivity gaps of banking industries across four major countries in Europe as well as the productivity gains that banks could obtain using alternative technologies or with different environments. An output distance function is de ned and the stochastic frontier approach used to carry out the comparison. Key words: Malmquist index, productivity, technology, environmental conditions, banking 1. Introduction Aconcern of particular importance in Europe is the analysis of the differences in banking performance among countries. The existence of such differences could help to explain crucial issues, such as the speed of convergence of European banking industries or the probability of future cross-border mergers and acquisitions. In the existing literature, most papers use cost or pro t frontiers methodology and measure the differences in bank ef ciency between countries by setting separate or common frontiers for all banking institutions. This assumes that any difference in ef ciency can be explained by countryspeci c banking technology (Fecher and Pestieau, 1993; Berg et al., 1993; Berg et al., 1995; Bergendahl, 1995; Allen and Rai, 1996; Pastor et al., 1997b; Dietsch et al., 1997. However, these studies make no adjustment for country-speci c conditions or norms. Recently, Dietsch and Lozano-Vivas (2000) analyzed banking sectors in France and Spain and de ned a common frontier that incorporates country-speci c conditions. The authors point out that the standard approach could misstate the relative ef ciency of rms from different countries, because it does not account for cross-country differences in demographic, regulatory, and economic conditions beyond the control of rm managers.
148 CHAFFAI ET AL. They also show how ef ciency scores obtained by the standard approach are arti cially low (high) for rms that operate under bad (good) home country conditions. The importance of the role played by environmental conditions in banking performance also has been analyzed across states in several studies of U.S. banking (Berger and Humphrey, 1991; Mester, 1997; DeYoung, 1998). From our perspective, to nd out how European integration determines banking performance in different countries, two main issues had to be considered: how different are the underlying domestic banking technologies and which particular environmental and regulatory conditions characterize the banking markets? So, in this paper, we attempt to contribute to the intercountry banking comparison literature by proposing a methodology that breaks down the intercountry performance differences into pure technological differences and differences due to environmental effects. This enlarges on the methodologies proposed by the previous papers in the cases where banking technologies are different. Based on this idea, a Malmquist type index is de ned with two components: one dealing with pure technological differences and the other re ecting the environmental effect, which affects the technology of each country's banking industry. The breakdown of the index is based on homothetic distance functions. Unlike the index proposed by Pastor et al. (1997b), the Malmquist index is symmetric and not sensitive to which country is taken as a benchmark reference. Moreover, because, in the analysis, controls also are set up for the likely in uence country-speci c environmental conditions exercise over productivity differences, the Malmquist type of index presented in this paper can measure properly the differences in a country's technology. The methodology is applied to compare the productivity differences of four European countriesðfrance, Germany, Italy, and SpainÐusing the econometric approach. In particular, the output distance function is de ned and the stochastic frontier approach is used to carry out the comparison. Each country's banking technology is represented by a homothetic quasi- exible functional form. In addition, the breakdown of productivity differences between countries may be sensitive to the choice of functional form. For that reason, the robustness of the results were tested using, alternatively, the Cobb-Douglas functional form. The results show that the environment exercises an important role in explaining the differences in intercountry banking productivity. In particular, it was found that productivity gaps between countries are very sensitive to environmental conditions. Even if the banking industry in a country uses better technology, compared to other industries, it can, in the end, be less productive due to a hostile environment. 2. Existing literature on international comparisons of banking industries The most relevant issue addressed in the existing literature on international banking comparisons is the measurement of ef ciency differences among the banking industries of different countries. With this common goal, a set of papers de nes national frontiers as well as common frontiers by pooling all ``cross-country'' banks. An overview of the existing literature dealing with international banking comparisons is shown in table 1. Berg et al. (1993) relied on data envelopment analysis (DEA) to establish the productive
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 149 Table 1. International banking comparison literature Authors Country Sample Control for Managerial Differences Control for Technology Differences Control for Environmental Differences Approaches Berg et al. (1993) 3 Scandinavian Yes No No DEA countries Fecher and Pestieau 11 OECD countries Yes No No DFA (1993) Berg et al. (1995) 4 Scandinavian Yes No No DEA countries Bergendahl (1995) 4 Scandinavian countries Yes No No Nonparametric Allen and Rai (1996) 15 developing Yes No No SFA, DFA countries Dietsch and Lozano- 2 European countries Yes Yes Yes DFA Vivas (2000) Pastor et al. (1997b) 8 developing countries Yes Yes No Distance function Pastor, Lozano-Vivas, 10 European countries Yes No Yes DEA and Pastor (1997a) Dietsch and Weill 12 European countries Yes No No DFA, DEA (2000) Chaffai, Dietsch, and Lozano-Vivas (1998) 4 Mediterranean countries Yes Yes No Distance function Maudos, Pastor, Perez, and Quesada (1998) 10 European countries Yes No No DFA, SFA, DEA banking ef ciency differences among three Scandinavian countries. First, separate frontiers for each country were de ned and then pair-wise comparisons were made of the countries using separate frontiers. Acommon frontier was de ned and again results were compared between countries. Berg et al. (1995) did a follow-up study, adding Denmark to the sample. The same four countries were investigated by Bergendahl (1995), using mixed optimal strategy. Fecher and Pestieau (1993) and Allen and Rai (1996) used the distribution free approach (DFA) and the stochastic frontier approach (SFA) for a comparison of cost ef ciency differences among 11 OECD countries and 15 developed countries, respectively, de ning a common frontier. Maudos et al., (1998) measured pro t and cost ef ciency differences as well as technical progress differences, also using a common frontier, for 10 European countries. Three main caveats might be pointed out with regard to the papers just cited: (1) some of these papers use separate frontiers for ef ciency comparison purposes. However, separate frontiers cannot be used to compare differences in ef ciency between countries because they do not allow a comparison to be made of the banks of each country with respect to the same standard. (2) The studies that use common frontiers for determining ef ciency differences between countries measure these differences without controlling for differences in technology; that is, they assume that the underlying technology of the
150 CHAFFAI ET AL. banking industry is the same in all the countries. To de ne a common frontier properly, rst it is necessary to determine whether or not the underlying banking technology in all the countries is similar. If the technology used by the banks of different countries is the same, then it should be possible to compare the ef ciency levels of these banking industries directly by building a production or a cost common frontier, pooling the banks of all the countries together. Otherwise, these ef ciency measurements derived from the estimation of a common frontier cannot be compared directly. (3) None of these papers accounts for differences in environmental conditions among countries. However, recently Dietsch and Lozano-Vivas (2000) and Pastor et al. (1997a), using DFAand DEA, respectively, show that ef ciency differences are strongly explained by country-speci c differences. The authors conclude that the neglect of country-speci c environmental conditions induces an important misspeci cation of the common frontier and an overestimation of the inef ciency. In the existing literature, Pastor et al. (1997b) attempt to analyze the differences in technology among the banking industries of different countries. They present a comparison of banking technical ef ciency and technological differences for eight developed countries. These authors use the distance function and extend the Malmquist productivity index to break down differences in productivity of different banking systems into levels of ef ciency (catching up) and distances between the frontiers themselves. To control for differences in technology, they propose an index that measures the distance between frontiers. However, the use of such an index requires choosing the banking technology of one of the countries as a benchmark. In that paper, the Spanish banking industry serves as such a benchmark. This implies that the technological differences depend on the technology of the country chosen as the benchmark country. Consequently, the Malmquist index used by these authors is asymmetric in the sense that the differences in technology are sensitive to the choice of the benchmark country. After controlling for differences in technology, the authors measure differences in ef ciency among countries by using separate country frontiers. As pointed out before, ef ciency difference comparisons between countries have to be measured from a common country frontier, once homogeneity among technologies is controlled for. Separate frontiers for ef ciency comparison purposes can be used only in two cases: when the objective is to compare banking performances within each country or when the goal is to measure the rise in ef ciency that one banking industry could bene t from by using the technology of more developed banking industries (Chaffai et al., 1998). Moreover, Pastor et al. (1997b) assume that the differences among country frontiers are attributed totally to differences in technology, neglecting the likely in uence that country-speci c environmental conditions can exercise over such differences. 3. Methodology Knowledge of the differences in banking performance among countries is becoming a crucial policy issue. In the case of Europe, in particular, this knowledge could help predict the effects of an expected increase in cross-border competition or the speed of convergence of European banking industries. However, to get a complete understanding
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 151 of the productivity difference effects, it is necessary to go beyond a global measurement of the productivity gaps between countries and distinguish the productivity differences arising from differences in technologies from those arising from differences in the environment. The following example illustrates the importance of such analysis. Let us assume that the global productivity gap between country Aand country B is equal to 50%. Consider a rst case, in which the effect of pure technological differences account for a large part of this gap, while the environmental effects account for a smaller part of it. The second case is the reverse: The effect of technology is low, while the environmental effect is high. Even if country A's banks dominate country B's banks in terms of total productivity in both cases, the policy implications are very different. In the rst case, country A's banks could attempt to operate in country B, because the former are equipped with better technology and so their costs would be lower. In the second case, they would not try to invest in the other country, because they would suffer too high costs and lose their competitive advantage mainly due to environmental conditions. An index that breaks down the differences between country frontiers into two components is speci ed. The rst component deals with pure technological differences and the second component re ects environmental effects that affect the banking technology of each country. The breakdown of the index proposed is based on homothetic distance functions. Here, the homotheticity assumption is used to break down the total productivity gap between countries into pure technological effects and environmental effects. This assumption easily could be relaxed, but in that case, only a global measurement of the productivity gap between countries (a mixture of technological and environmental effects) would be obtained and it would not be possible to get the speci c contribution of each of the two effects separately. The technology of each country's banking industry is represented by an output distance function: 1 D I O y I ; x I ; z I ˆmin f : yi f f [ LI x I ; z I 1 where y I ; x I ; z I is the output set of the banks of country I: The banks of country I produce the output vector y I, by using the input vector x I and they are operating in an environment represented by the vector z I. We assume that the technology of each country's banking industry is homothetic with respect to the environmental variable z. So, according to FaÈre and Primont (1995), the output distance function is given by D I O y I ; x I ; z I ˆDI O yi ; x I ; 1 G I z I 2 where G I z I is any positive real function such that the distance function in equation (1) veri es the usual necessary properties to represent the technology; that is, the distance function is nondecreasing, positive linearly homogeneous, concave in output, and decreasing in input. Moreover, for any production plan, we have 0 D I O y O ; x O ; z O 1 For instance, let us consider two countries, I and J. The technology of each country's
152 CHAFFAI ET AL. Figure 1. Technological index and output distance function. Shown is the speci c technology of country banking industries I and J assuming that each banking industry produces one output and uses one input. The geometric mean of two gaps, (OA 00 OA 0 ) and (OB 00 OB 0 ), measures the difference between the technology of the two banks, Aand B. banking industry is represented by a speci c output homothetic distance function. To illustrate this point, let us consider the simple technology in the case of an industry that produces a single output and uses a single input (see gure 1). Each country's banking industry has its speci c technology represented by the lines OI (for country I) and OJ (for country J). The difference between the technology of the two banks, A and B, is measured by the geometric mean of two gaps: OA 00 OA 0 and OB 00 OB 0. By using the geometric mean of these two gaps, we avoid the obligation to choose, arbitrarily, one of the country's banking technologies as a benchmark. The measurement of the productivity gap between the two banking industries, which can be attributed to differences in technologies, is r OA 0 OB 0 Geff I=J ˆ 3 OA 00 OB 00 In terms of the distance function, this ratio is s D J O Geff I=J ˆ yi ; x I ; z I D J O yj ; x J ; z J D I O yi ; x I ; z I D I 4 O yj ; x J ; z J If this ratio is greater than 1, that means that the technology used by country I's banking industry ``dominates'' the technology used by country J's banking industry, and vice versa. It is very simple to verify that the index (4) is symmetric, Geff I=J ˆ Geff J=I 1. Moreover, (4) is just a component of the Malmquist productivity index proposed by FaÈre et al. (1994). Indeed, if (4) is multiplied by the ef ciency change of the two banks, A and B, the Malmquist index is obtained. To break down the productivity gap, using the property of homotheticity of the distance function in (2), the index in (3) alternatively could be written as
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 153 s s G Geff I=J ˆ I z I G I z J D J O yi ; x I ; 1 D J O yj ; x J ; 1 G J z I G J z J D I O yi ; x I ; 1 D I O yj ; x J ; 1 5 The rst term in equation (5) measures the difference in technology due to the differences in the environment. If this term is greater than 1, that means that environmental conditions in country I help country I's banks perform better than the environmental conditions of country J could do. In other words, the speci c environmental conditions of country J will not allow a bank in country I to improve its productivity, if this bank would have to operate in an environment like that in country J. The second term in (5) measures the ``pure'' technological differences between the two industries; that is, the net technological differences without environmental effects. 2 The two ratios in (5) could move in opposite directions. For example, the rst component could be greater than 1, while the second component could be less than 1. If environmental effects are the same for the two industries, the index will just re ect the differences between the frontiers due to the technology. Given a sample of N I banks in country I and a sample of N J banks in country J, itis possible to measure the difference between the frontiers according to (4) or (5) for each pair of banks in the two countries. However, as the number of possible combinations increases with the sample size, it would be more convenient to compare the ratio of Geff for banks with comparable activities and of a comparable size or compare them at the sample mean. Separate frontiers for each country under investigation are de ned because the gap between these frontiers is being measured. 4. The econometric model The stochastic frontier approach is used to represent the distance functions in each country. So the model takes into account possible noise in the data. However, it is known that the econometric approach is sensitive to the choice of functional forms, which represents the frontier. So, the robustness of the results to the choice of functional forms also is investigated. Two speci cations for the distance function, the Cobb-Douglas form and the translog form, are considered. The technology of each country is represented by a quasi- exible translog output distance function: 3 ln D O y it ; x it ; z t ˆa 0 Xp j ˆ t 0:5 XP j Xk h ˆ 1 b j ln y jit X R j 0 b jj 0 ln y jit ln y j 0 it g h ln x hit ln G z it e it where the b jj 0 verify the usual symmetry constraints. Under the homogeneity condition of 6
154 CHAFFAI ET AL. the output distance function with respect to the outputs, the following constraints also are introduced: X p j ˆ 1 b j ˆ 1; X L j 0 ˆ 1 b jj 0 ˆ 0 7 It should be mentioned here that the environmental variables are the same for all the banks in the same country. Consequently, z it ˆ z t in the model. Moreover, as G z is any positive real function, the distance function estimates and the breakdown of the Malmquist productivity index (5) may be sensitive to which speci cation is used for this function. To preserve the linearity of the model, the following simple functional form is considered:! G z t ˆexp l m z mt 8 X m Also, following FaÈre and Primont (1995), the units of measurements for the environmental variables are chosen so that G 1 ˆ1. So, the following constraint is added: X l m ˆ 0 9 m Moreover, the stochastic term e it is introduced into the model to account for possible noise in the data. As the values of the distance functions are not observed, the property of homogeneity can be used with respect to the outputs, and the function with respect to one of them, y 1 is normalized: ln y 1it ˆ a 0 Xp j ˆ 2 0:5 X j Xk h ˆ 1 b j ln y jit =y 1it X j 0 b jj 0 ln y jit =y 1it ln y j 0 it =y 1it g h ln x hit XM m ˆ 1 l m z mt u i e it 10 where u i ˆ ln D O y it ; x it ; z t. The assumption also is made that technical inef ciency is constant over time for all the banks in each country. This assumption is not really necessary, 4 but it is useful to break down the inef ciency component from the residual in equation (10) without imposing a speci c assumption on the distribution of the inef ciency component in the model. For each country, we estimate the frontier by the generalized least squares method. From the parameter estimates, we measure the technology differences and compute the two components of equation (5). 5. Data and variables In the empirical analysis, balanced panel data of four European banking industries is usedðfrance, Germany, Italy, and SpainÐover the 1993±1997 period. Data comes from
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 155 the Bankscope database of BVD-FITCH-IBCA. To make the inputs and outputs homogeneous and have a representative number of banks in each country, the analysis has been restricted to these four European countries. Databases include commercial banks as well as mutual or savings and loans banks. Despite the difference in their legal status, these banks compete in the same banking markets and are subject to the same regulations. As a result of checking the data carefully for consistency, a balanced sample of 595 banks per year over the period 1993±1997 was adhered to: 114 in France, 295 in Germany, 107 in Italy, and 79 in Spain. The environmental data comes from the OECD (Bank Pro tability and Main Economic Indicators) and Eurostat (Money and Finance). All the monetary data (i.e., banking outputs and inputs as well as environmental variables) were converted into Euros using the purchasing power parity hypothesis. 5.1. Output and input variables Three output variables are de ned (y 1 ˆ loans, y 2 ˆ other assets, and y 3 ˆ total deposits) and three input variables (x 1 ˆ labor, measured by expenses in labor inputs; 5 x 2 ˆ physical capital, measured by the book value of the banks' xed assets; and x 3 ˆ nancial inputs, measured by the interests paid by the banks). These de nitions of banking output and input are in agreement with the production approach. In particular, the transaction and liquidity services provided by banks through the supply of deposits have been taken into account, as well as the input costs that the deposits imply. Table 2 shows the average values of these variables. The average size of a loan portfolio has been observed to be very similar in each country, while the average size of deposits differs among countries. Table 2. Output, input, and environmental variables Average values of banking output include loans, other assets, and total deposits; banking input includes personal expenses (expenses in labor inputs, xed assets (book value of the bank's xed assets) and interest paid by the banks; environmental variables include population density ( population per km 2 ), per capita income (GDP per population), number of banks per population, number of branches per km 2. All the monetary variables are in millions of Euros. Spain France Germany Italy Loans 1.77 1.63 1.99 1.57 Other assets 1.58 1.38 1.50 1.49 Deposits 3.20 2.76 2.75 2.37 Personal expenses 0.048 0.048 0.033 0.065 Fixed assets 0.106 0.035 0.036 0.077 Average interests paid 0.30 0.16 0.16 0.17 Population/km 2 0.077 0.105 0.229 0.190 GDP/population 0.010 0.020 0.021 0.015 Number banks/population 0.0027 0.0038 0.0125 0.0031 Branches/km 2 0.067 0.084 0.062 0.067 Number of banks 79 114 295 107 Sources: BVD-FITCH-IBCA, OECD (Bank Pro tability and Main Economic indicators) and Eurostat (Money and Finance).
156 CHAFFAI ET AL. 5.2. Environmental variables In the empirical analysis, macroeconomic variables as well as variables measuring each country's banking industry structure are speci ed. These environmental variables may affect the performance of banks. Aset of four variables was obtained after checking a wider set of variables. So, it seems that the environmental variables signi cantly affect the banking technology of the countries under investigation. As in the paper by Dietsch and Lozano-Vivas (2000), the environmental variables take equal values for each bank in each country by year. Here, it must be emphasized that the goal is not to conduct a microlevel study but to compare average national performance levels. The set of the four environmental variables is presented in table 2. The table shows that the differences in environmental conditions are quite substantial during the period under review, which illustrates the lack of convergence among European countries. The two macroeconomic variables that signi cantly affect the technology of the banking industries under investigation are population density and per capita income (GDP per inhabitant). These variables measure the characteristics of the demand for banking products. For example, population density determines the strategies banks choose concerning the location of branches, and it affects the bank costs through the supply of banking services. Banking industries in a country with a lower density of population are likely to incur higher operating costs, which implies higher expenses for bank intermediation. On the other hand, per capita income affects numerous factors related to the demand for bank loans and deposits. Countries with a higher per capita income are assumed to have a banking system that operates in a mature environment, giving the banking system the chance to exert more activity. The variables that affect the banking technology of the countries analyzed and are related to banking structure and competition are the number of banks per inhabitant and the accessibility of banking services, measured by the density of branches. As shown in the empirical and theoretical literature, the higher the number of banks, the higher is the level of competition in the banking industry. On the other hand, branch density is an indicator of the space dimension for each national market. High numbers of branches also could indicate potential overbranching, which might affect the operating costs of the banking system. Table 2 suggests the existence of large environmental differences among countries. In particular, Spain appears to be the country with the lowest population density and the lowest per capita income, but it has a very high number of branches. So, apparently, banking production in Spain is more expensive than in the other countries. 6. Empirical results In a rst step of the comparative analysis, a case is considered where the differences in performance between countries are assumed to come only from differences in technology; that is, the speci c environmental variables that characterize each domestic banking market are ignored. So, rst, the distance function for each country is estimated without introducing the environmental variables. 6 From the parameter estimates of the distance
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 157 Table 3. Intercountry productivity comparisons without environmental effects Productivity gain indexes, Geff I=J are determined at the mean value of the sample by estimating the distance function for each country without introducing the environmental variables. The Geff I=J have been obtained using the translog (TL) and the Cobb-Douglas (CD) output distance function, respectively. Country J France Germany Italy Country I Year TL CD Year TL CD Year TL CD Spain 93 0.71 0.70 93 0.54 0.51 93 0.88 0.91 94 0.74 0.67 94 0.55 0.51 94 0.89 0.88 95 0.68 0.66 95 0.55 0.51 95 0.82 0.86 96 0.73 0.68 96 0.52 0.51 96 0.81 0.87 97 0.71 0.65 97 0.53 0.50 97 0.74 0.82 France 93 0.83 0.77 93 1.34 1.29 94 0.94 0.78 94 1.45 1.30 95 0.86 0.78 95 1.37 1.29 96 0.84 0.79 96 1.35 1.28 97 0.84 0.80 97 1.34 1.27 Germany 93 1.65 1.70 94 1.67 1.68 95 1.69 1.66 96 1.72 1.66 97 1.75 1.66 functions, the productivity gain indexes, Geff I=J, are computed following equation (4), at mean values of the samples. 7 Table 3 shows the results for each pair of countries. Note that, in this table, the rows represents country I, while the columns represent country J,so a value of Geff I=J > 1 means that, at the mean point, country I's banks are more productive than country J's banks. In other words, the frontier of the country banking industry I envelops the frontier of the country banking industry J at that point. Without controlling for the effect of environmental differences, the results show that, on average, the productivity of the French, German, and Italian banks is higher than the productivity of the Spanish banks. For example, the results suggest that Spanish banks could increase their productivity by around 32%, 49%, or 13%, by using the technology of French, German, or Italian banks, respectively, rather than their own technology. Italian banks suffer a productivity loss of around 29% (66%) by using their own technology rather than French (or German) banking technology. Finally, French banks could improve their performance by around 22% if they decided to use German banking technology. The results seem to be very stable over time and independent of the choice of the functional form used to represent the distance function: the translog (TL) or the Cobb-Douglas (CD) distance function. 8 Moreover, the conclusions are similar to those obtained by Pastor et al. (1997b), except when Italian banks are compared with French and German banks. These differences could be because the index of the authors is not symmetric. However, as pointed out before, the differences between country frontiers might not be attributable totally to differences in banking technology. They also could be due to country-speci c environmental conditions. To control for environmental effects, the
158 CHAFFAI ET AL. Table 4. Intercountry productivity comparisons with environmental effects Productivity gain indexes, Geff I=J are determined at the mean value of the sample by estimating the distance function for each country introducing the environmental variables. The Geff I=J have been obtained using the translog (TL) and the Cobb-Douglas (CD) output distance function, respectively. Country J France Germany Italy Country I Year TL CD Year TL CD Year TL CD Spain 93 0.63 0.65 93 0.31 0.39 93 0.54 0.62 94 0.62 0.65 94 0.28 0.35 94 0.57 0.67 95 0.66 0.68 95 0.32 0.39 95 0.65 0.75 96 0.66 0.70 96 0.33 0.42 96 0.68 0.78 97 0.65 0.71 97 0.36 0.45 97 0.70 0.80 France 93 0.67 0.82 93 1.22 1.34 94 0.61 0.73 94 1.27 1.40 95 0.65 0.78 95 1.31 1.43 96 0.64 0.77 96 1.33 1.45 97 0.69 0.82 97 1.34 1.46 Germany 93 2.46 2.02 94 2.58 2.13 95 2.43 2.01 96 2.40 1.97 97 2.36 1.99 distance functions were estimated again by introducing environmental variables in the econometric model, according to equation (5). The differences in productivity resulting from pure technological effects and from environmental effects are shown in table 4. The results are very different from those in table 3 for all the pairs of countries. This proves that the environment has a large effect on the bank performance. For example, environmental conditions seem to widen the productivity gap between Spanish banks and German or Italian banks. On the other hand, environmental conditions slightly reduce the gap between French and Italian banks. Note that, on average, all the banks would bene t from the German environment. 9 Actually, if the countries are ranked in terms of the four in uential environmental variables used (see table 2), we observe that, on average, Germany has the best environmental conditions. Again, the results seem to be quite stable over time and independent of the choice of the functional form of the distance function. These results suggest that country-speci c environmental conditions help explain the productivity gaps between countries. The environment, in fact, could be considered a signi cant component of banking technology. These results are in agreement with those obtained by Dietsch and Lozano-Vivas (2000). The large differences between country frontiers shown in table 3 might not be attributable to pure technological differences, as Pastor et al. (1997a) conclude. Our results show that they could be due mainly to the consequences of environmental differences. Our breakdown of the productivity index con rms this assessment. Following equation (5), the productivity gains index is broken down into pure technological gains and environmental condition gains. Tables 5 and 6 show the results. Again, the indexes are computed at the sample mean.
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 159 Table 5. Intercountry productivity comparisons: Pure technological effects Productivity gaps between countries are assumed to be due to pure technological differences; that is, the net technological differences without environment effects, at the mean value of the sample using the translog (TL) and the Cobb-Douglas (CD) output distance function, respectively. Country J France Germany Italy Country I Year TL CD Year TL CD Year TL CD Spain 93 0.87 1.17 93 1.11 1.00 93 1.26 0.95 94 0.83 1.15 94 1.01 0.93 94 1.20 0.95 95 0.85 1.16 95 1.02 0.95 95 1.21 0.94 96 0.83 1.14 96 0.99 0.92 96 1.19 0.94 97 0.80 1.14 97 0.93 0.87 97 1.14 0.89 France 93 1.08 0.74 93 1.42 0.79 94 1.03 0.70 94 1.43 0.80 95 1.05 0.71 95 1.39 0.79 96 0.98 0.68 96 1.38 0.79 97 0.98 0.67 97 1.34 0.77 Germany 93 1.41 1.08 94 1.50 1.15 95 1.49 1.14 96 1.49 1.14 97 1.53 1.19 Table 6. Intercountry productivity decomposition: Environmental effects Productivity gaps between countries are assumed to be due to environmental differences, at the mean value of the sample, using the translog (TL) and the Cobb-Douglas (CD) output distance function, respectively. Country J France Germany Italy Country I Year TL CD Year TL CD Year TL CD Spain 93 0.73 0.55 93 0.27 0.39 93 0.43 0.65 94 0.74 0.56 94 0.27 0.38 94 0.47 0.71 95 0.77 0.59 95 0.31 0.42 95 0.54 0.79 96 0.79 0.60 96 0.34 0.46 96 0.57 0.83 97 0.81 0.63 97 0.38 0.51 97 0.62 0.89 France 93 0.62 0.74 93 0.85 1.69 94 0.59 0.70 94 0.88 1.74 95 0.62 0.71 95 0.94 1.81 96 0.65 0.68 96 0.96 1.84 97 0.71 0.67 97 1.00 1.89 Germany 93 1.74 1.87 94 1.71 1.84 95 1.64 1.77 96 1.61 1.73 97 1.54 1.67
160 CHAFFAI ET AL. Looking at pure technological effect results (table 5), we see that, when the environmental component of the banking technology is ignored, the technological differences between countries on average are quite low. The technological gap is lower than 20% for the four pairs of countries, whatever the functional form. Note that, in terms of pure technology, the Spanish banks are close to the banks of the other countries. Spanish banks suffer no technological disadvantage with respect to France, Germany, or Italy. However, signi cant differences in terms of technology seem to exist between Italian banks, on the one hand, and French and German banks, on the other. In addition, the German banking industry does not always ``dominate'' other countries, except for Italy, as was seen in the previous results. Indeed, table 6 shows that the environmental differences explain most of the productivity gaps between countries. Comparing tables 5 and 6, we observe that the gap in productivity between countries is larger in terms of the environmental component than in terms of the technological component. These results con rm the assessment that the large differences between country frontiers are attributable mostly to environmental differences instead of pure technological differences. The results obtained by including environmental effects (table 4) show that the German banking industry always ``dominates'' other countries and the Spanish banking industry always is ``dominated'' by other countries. However, this is not the case when only pure technological gains are considered (table 5). This is further evidence that environmental conditions are an important component of banking technology. Overall, it is worth noting two important results: (1) the inclusion of environmental conditions diminishes differences in banking technology and (2) the differences due to environmental conditions always are larger than the differences in banking technology among the European banking industries. 7. Conclusion This paper proposes a Malmquist type index that allows productivity gaps among banking industries in different countries to be measured and the difference to be broken down into differences due to pure technological effects and differences due to environmental effects. The breakdown of the index is based on homothetic distance functions. The most relevant feature of the index is its symmetry. Therefore, it is not sensitive to the choice of benchmark country. The index takes into account the domestic environmental conditions in which the banks operate. This index is used to explain the productivity gaps of banking industries among four major countries of the Euroland (France, Germany, Italy, and Spain). The results show that environmental conditions play a major role in this explanation. On average, the differences due to environmental conditions always are larger than the differences in banking technology among the European banking industries. So, ignoring the environmental conditions could lead to erroneous conclusions when important issues for the future of the European banking industry are considered, such as the competitiveness of banking markets, the opportunities for ``cross-border'' consolidation, and the speed of future convergence of the different European banking industries.
TECHNOLOGICAL AND ENVIRONMENTAL DIFFERENCES IN EUROPEAN BANKING 161 Acknowledgments We thank Knox Lovell, Loretta Mester, seminar participants at the Workshop on Banking and Finance, Alicante (Spain) and at the Works on Ef ciency and Productivity (Denmark), and the anonymous referee for many suggestions, which greatly improved the paper. Financial support from CICYT, PB98-1408 and the Fondation Banque de France pour la Recherche are gratefully acknowledged. Notes 1. The same analysis could also be based on an input distance function; under the constant return to scale assumption the results are similar. 2. It is worth noting here that the breakdown of the index into pure technological effect and environmental effect is possible thanks to the homotheticity assumption. To relax such an assumption in the model would yield an index that would allow the productivity gaps between countries to be measured but not allow these gaps to be broken down into a pure technological and an environmental effect. 3. As the distance function cannot be estimated with a system of share equations, quasi-collinearity among the variables may be important. For this reason, a quasi- exible translog speci cation is taken for the distance function by ignoring the cross-product terms on the input side. 4. It also could be assumed that this component is time variable. In that case, it is necessary to add other assumptions concerning the parametric evolution of technical inef ciency over time and to impose a particular distribution for the technical inef ciency components in the model. However, in the empirical exercise, the veracity of the constant technical ef ciency assumption is tested. 5. The input level is measured by expenses. The main reason is that, in the BVD-IBCA``Bankscope'' le, the information about the number of employees is not available for a large part of the banks' sample, especially in Germany and Italy. However, it seems that the use of expenses instead of the number of employees to measure labor input would not signi cantly affect the ef ciency results. Indeed, it was possible to extract information about the number of employees from two other data sources: the Commission Bancaire for France and the Consejo Superior Bancario for Spain. So, the correlations between labor expenses and number of employees, labor price, have been computed. The computation of correlation coef cients gives the following results: Correlation Coef cient Spain France Between labor expenses and number of employees 0.979 0.981 Between labor expenses and labor price 0.143 0.129 The results show that the banks that face high labor expenses also have a high number of employees and labor price. According to the capital input, most of the empirical bank ef ciency studies use the book values of xed assets as a proxy to measure physical capital. We adopt the same convention in our work. 6. To save space, the estimated parameter of different distance functions are not presented. 7. It is worth noting that results evaluated at the median sample point are close and consistent with the results obtained at the mean point. 8. As pointed out in note 4, the constancy of the technical ef ciency was checked over time. The estimated results show that the level of technical ef ciency was 0.699 in 1993 vs. 0.687 in 1997 for Spanish banks, 0.659 vs. 0.653 for French banks, 0.70 vs. 0.70 for German banks, and 0.687 vs. 0.68 for Italian banks. So, the assumption of constant technical ef ciency over time could not be ruled out. 9. As a robustness check, the homotheticity assumption was relaxed with respect to environmental variables and the model was estimated again. The results obtained are consistent with the results obtained in table 4.
162 CHAFFAI ET AL. References Allen, L., and A. Rai. ``Operational Ef ciency in Banking: An International Comparison.'' Journal of Banking and Finance 20 (1996), 655±672. Berg, S. A., P. N. D. Bukh, and F. R. Forsund. ``Banking Ef ciency in the Nordic Countries: A Four-Country Malmquist Index Analysis.'' Working paper, University of Aarhus, Denmark (September 1995). Berg, S., F. Fùrsund, L. Hjalmarsson, and M. Suominen. ``Banking Ef ciency in the Nordic Countries.'' Journal of Banking and Finance 17 (1993), 371±388. Bergendahl, G. ``DEAand Benchmarks for Nordic Banks.'' Working paper, Gothenburg University, Gothenburg, Sweden (December 1995). Berger, A., and D. B. Humphrey. ``The Dominance of Inef ciencies over Scale and Product Mix Economies in Banking.'' Journal of Monetary Economics 29 (1991), 117±148. Chaffai, M, M. Dietsch, and A. Lozano-Vivas. ``How Emergent Countries' Banks Could Improve Their Performance by Adopting the Technology of Developed Countries: The Case of Tunisia.'' Working paper LARGE, University of Strasbourg (September 1998). DeYoung, R. ``X-Inef ciency and Management Quality in Commercial Banking Industry.'' Journal of Financial Services Research 13 (1998), 5±22. Dietsch, M., and A. Lozano-Vivas. ``How the Environment Determines Banking Ef ciency: A Comparison between French and Spanish Industries.'' Journal of Banking and Finance 24 (2000), 985±1004. Dietsch, M., and L. Weill. ``Banking Ef ciency and European Integration: Productivity, Cost and Pro t Approaches.'' Researchin Banking and Finance. JAI Press, 2000. FaÈre, R., and D. Primont. Multi-Output Production and Duality: Theory and Applications. Norwell, M. Kluwer Academic Publishers, 1995. FaÈre, R., S. Grosskopf, M. Norris, and Z. Zhang. ``Productivity Growth, Technical Progress and Ef ciency Change in Industrialized Countries.'' American Economic Review 84 (1994), 66±83. Fecher, F., and P. Pestieau. ``Ef ciency and Competition in O.E.C.D. Financial Services.'' In: H. O. Fried, C. A. K. Lovell, and S. S. Schmidt, eds., The Measurement of Productive Ef ciency: Techniques and Applications. Oxford University Press. 1993, pp. 374±385. Maudos, J., J. H. Pastor, F. PeÂrez, and J. Quesada. ``E ciencia en Costes y en Bene cios de los Sistemas Bancarios Europeos.'' Mimeo (1998). Mester, L. ``Measuring Ef ciency at U.S. Banks.'' European Journal of Operational Research 98, no. 2 (1997), 230±242. Pastor, J. T., A. Lozano-Vivas, and J. M. Pastor. ``Ef ciency of European Banking Systems: A Correction by Environmental Variables.'' WP-EC 97-12. IVIE (1997a). Pastor, J. M., F. PeÂrez, and J. Quesada. ``Ef ciency Analysis in Banking Firms: An International Comparison.'' European Journal of Operational Research 98, no. 2 (1997b), 395±407.