discussion papers

FS IV 95 - 3 Com petition in the E uropean Banking Industry: An Aggregate Structural M odel of Com petition Damien Neven* Lars-Hendrik Roller**

*

University o f Lausanne

**

Wissenschaftszentrum Berlin fur Sozialforschung

January 1995

ISSN Nr. 0722-6 7 4 8

Forschungsschwerpunkt Marktprozeß und Unter­ nehmensentwicklung (HMV) Research Unit Market Processes and Corporate Development (HM)

Zitierweise/Citation: Damien Neven, Lars-Hendrik Roller, Competition in the European Banking Industry: An Aggregate Structural Model of Competition Discussion Paper FS IV 95 - 3, Wissenschaftszentrum Berlin, 1995. Wissenschaftszentrum Berlin für Sozialforschung gGmbH, Reichpietschufer 50, 10785 Berlin, Tel. (030) 2 54 91 - 0

ABSTRACT

Competition in the European Banking Industry: An Aggregate Structural Model of Competition*

The objective of this study is to analyze competition in the European banking industry in light of the different regulatory frameworks to which banks in various countries are exposed. We focus on the provision of mortgages to households and loans to the corpo­ rate sectors in seven European countries. We develop an aggregate, multiproduct model for the European banking industry and estimate the structural model econometrically with data on demand and costs for the period 1981-1989. In general we reject non-co­ operative Nash behavior in favour of more collusive cartel-like conduct. In particular, we find support for the hypothesis that the corporate market is more competitive than the household market. W e observe that the market for mortgages is substantially less competitive in those countries which have specialized institutions for the provision of these loans. We find that conduct in the market for loans to the corporate sector varies a great deal across countries and that lightly regulated countries are not necessarily those where competition is toughest. Finally, we find no evidence that the degree of coordination among banks has been reduced over the period.

This paper was written while the second author was visiting the Graduate School of Business at Stanford University. The authors acknowledge financial assistance from INSEAD's research grant. Carole Bonanni provided competent research support.

ZUSAMMENFASSUNG

Wettbewerb im europäischen Bankgewerbe: Ein aggregiertes Strukturgleichungsmodell des Wettbewerbs In dieser Studie wird der Wettbewerb im europäischen Bankgewerbe im Hinblick auf unterschiedliche Regulierungsformen untersucht, denen Banken in den verschiedenen Ländern unterhegen. Anhand der Daten aus sieben europäischen Ländern wird der Wettbewerb bei der Vergabe von Hypotheken an private Haushalte und der Vergabe von Krediten an Unternehmen analysiert. Es wird ein aggregiertes Mehrproduktmodell für das europäische Bankgewerbe entwickelt und als ökonometrisches Strukturgleichungs­ modell anhand von Daten über die Nachfrage und die Kosten für den Zeitraum 1981 bis 1989 geschätzt. Die Hypothese nichtkooperativen Nash-Verhaltens wird zu Gunsten eines mehr kollusiven und kartellähnlichen Verhaltens abgelehnt. Unterstützung findet die Hypothese, daß der Markt für Untemehmenskredite wettbewerbsintensiver ist als der für Hypotheken an private Haushalte. Der Markt für Hypotheken ist in jenen Ländern merklich weniger wettbewerbsintensiv, die über spezielle Institutionen zur Bereitstel­ lung dieser Kreditformen verfügen. Die Analyse zeigt auch, daß das Marktverhalten auf dem Gebiet der Untemehmenskredite sich in erheblichem Maße zwischen den einzelnen Ländern unterscheidet und daß die geringfügig regulierten Länder nicht notwendiger­ weise jene sind, in denen der Wettbewerb am intensivsten ist. Schließlich zeigt sich, daß es keine Anhaltspunkte dafür gibt, daß sich das Ausmaß an Koordination zwischen den Banken im Untersuchungszeitraum verringert hat.

3

I. Introduction The objective of this study is to analyze competition in the European banking industry and assess the extent to which coordination among banks can be associated with the characteristics of the regulatory framework in which they operate.

We undertake an econometric analysis

of margins in the provision of loans to the private sector for seven European countries (Belgium, Denmark, France, Germany, The Netherlands, Spain and the the U K ) in the period 1981-1989. We show that the level of coordination among banks varies across countries and that it can be associated with the characteristics of national regulations. In addition, we try to evaluate whether the various changes in regulation that have been implemented in national markets since the early eighties, had any effect on the degree of competition. These regulatory changes tend to liberalize the provision of services and increase competition. In addition, the new directives put forward by the Commission, which emphasize the integration of markets and cross border activities should also enhance competition. The directive on the liberalization of capital flows had already been implemented in most countries by the late eighties and its effects could in principle be traced out in our sample. The second banking directive has been implemented in the early nineties but a number of adjustments1 have already taken place in anticipation of this implementation, which could affect our data. Overall, there are therefore some important national and European regulatory changes which could presumably have affected the degree the coordination among banks. The paper is organized as follows. Section II will document the important differences in regulation across countries and describe the changes that have taken place since 1980. Section HI introduces an empirical model for the banking industry. Section IV addresses the implementation of the model. Section V discusses the findings and policy conclusions.

n . The regulation of the European Banking Industry Before describing current regulatory frameworks, it is worth recalling briefly the purpose of such regulation;

the traditional literature on banking regulation (see Diamond and Dybvik,

1983) emphasizes the asymmetric information between banks and investors as the market failure which justifies regulation; in this context, banks are prone to run and systemic risks lSee for instance, the various studies included in Dermine (1992)

4

can affect the stability of the financial system (Aghion, Bolton and Dewatripont, 1988). In turn, the protection of banks is warranted because the social value associated with some banking functions exceed their private value.

In particular, it is commonly emphasized that

the provision of liquidity and payments services are central to the functioning of the economy and that the interruption of these services is costly as the production process is severely affected and assets are liquidated prematurely. Hence, the costs of financial failure in terms of real resources underpins the wider social value of banks and justifies their protection. A number of alternative instruments have been used to protect the banking sector. Besides deposit insurance and lender of last resort facilities, the safety of the banking system can be enhanced by organizing the industry in such a way as to reduce risks and the associated probability of failures. By ensuring a high level of profitability, these regulations allow banks to accumulate reserves which act as a buffer in case of financial difficulty. By creating a high franchise value, these regulations will also give banks the incentive to avoid leaving or being expelled from the industry and, hence, to reduce risks. These regulations will however tend to give rise to allocative inefficiencies (to the extent power is exercized) and productive inefficiencies (to the extend that part the rents enjoyed by banks is dissipated in excessive costs). Banking regulation in Europe in the early eighties typically emphasized the protection of rents rather than deposit insurance (or lender of last resort facilities) to ensure the stability of the banking system. Regulatory instruments included structual elements such as the functional separations of institutions (for instance between commercial and investment banking), entry requirements and discriminatory rules regarding foreign banks and investors. These regulations can be expected to reduce entry and thereby to encourage collusion, to affect the presence of foreign firms, the size of banks, the frequency and type of mergers, acquisitions and participations and the scope of products that can be offered. Alternative instruments which restrict the behaviour of banks were also used, like direct restrictions on assets and liabilities (including prudential rules and rules on participations in non-banking firms), rules relating to information disclosure, credit ceilings, limitations on branching and the determination of fees commissions and rates on assets and liabilities (floors and ceilings, concerted practices). These regulations can be expected to provide banks with an incentive to overemphasize competitive tools which are not restricted. A detailed description of regulatory changes that have taken place in Europe since 1980 can be found in Gual and Neven (1992) and Financial Times (1991). In what follows, we summarizes the key characteristics of national regulatory systems outlined by these studies for

5 the countries under review namely France, Denmark, Germany, Spain, the UK, Belgium and the Netherlands. In very broad terms, the following pattern emerges; in the early eighties, Denmark, France and especially Spain had extensive direct restrictions on banks' conduct. These were direct restrictions on assets and liabilities as well as restrictive rules relating to competitive behaviour (regulated prices and credit quotas). Belgium had significant restrictions on competitive behaviour whereas restrictions on assets and liabilities were by and large confined to savings and public banks. Similarly, significant restrictions in the UK only concerned building societies. At the other end of the spectrum, the Netherlands had restrictions on competitive behaviour which were confined to fees and commissions and limited investment requirements; Germany had hardly any restriction, except that capital requirements were unusually large by European standards. In terms of structural restictions, the German regulation enforced a separation between banking and insurance. Importantly, it also confined the provision of mortgages to specialized institutions. Such specialization was also found in Denmark and to a large extent in the UK.

Spain had important functional separations, some specialization of institutions

and important entry restrictions. Belgium also had functional separation, a fairly extensive system of specialized institutions but a more liberal policy towards entry. France had a system of universal banking but extensive specialization. Little formal impediment of entry could be observed but the widespread public ownership which prevailed was probably a significant barrier to entry by acquisition. Regulatory changes that have taken place in the 1980s can be summarized as follows: Spain has undertaken the most significant deregulation; deregulation has focused on conduct where most rules have been relaxed (mainly after 1985-87). Significant structural deregulation has also taken place, particularly in terms of product scope and entry. Denmark has lifted most restrictions on credit ceilings and rates. France has undertaken some deregulation of conduct in particular regarding fees and commissions; some despecialization has occurred but structural impediments have by and large remained in place. Belgium has relaxed most structural rules. Basically, no deregulation has taken place in Germany and the Netherlands. As mentioned above, regulation was however not extensive to start with in these two countries. In the UK, which was also lightly regulated, structural restrictions have been progressively lifted and conduct regulations have all but disappeared.

6 HI. A Model of Competition for the European Banking Industry To assess the degree of competition across various European countries and how these margins are affected by the regulatory regime we derive an aggregate, structural model of oligopoly (see Bresnahan (1989) for survey of structural empirical models). By specifying an aggregate model we explicitly assume that some aggregation over banks can be undertaken, in any given country, so that the unit of observation is a country at a point in time2. Our data set, which is described in more detail below, consists of 7 countries from 1981-1989. The analysis below focuses on two segment of the banking industry, namely the provision of loans to the nonfinancial corporate sector (herein refered to as corporate segment) and the provision of mortgages to household (herein refered to as as the household segment).

Since most banks

are potentially in both segments there is a possibility of collusion through multimarket contact (see Bemheim and Whinston 1990). Our model will allow for such effects, but we will not explicitly test for them since we do not have data at the firm level. In addition to multimarket contact, we model the banking industry as having a joint (multiproduct) cost functions for the production of the two types of loans. To identify conduct in a structural model, we need to specify demand and cost conditions. We assume that banks face a country-level demand function for funds to non-financial institutions and households of the following type,

» - C .H

(1)

i=l

where M=C,H indicates corporate and household loans, s is the country subscript, t is the time period, qits are the loans given by bank i at a rate r, Ns are the number of banks in country s, and Zts is a vector of country-specific, exogenous fixed effects (see the appendix for a detailed description of the data). Note that the demand structures are independent across the two segments. Bank's cost structures are modelled by specifying the following firm-level, multiproduct cost function, + C "7

,0)„)

(2)

where F represents firm-specific fixed costs. These fixed costs, which are allowed to vary over time, represent costs of setting up the branch network for instance. Clearly such fixed

2Such aggregation implies some restrictions, which are discussed below.

7 costs vary significantly across countries and even banks. Alternatively, one could interpreted F as a measure of efficiency or productivity differential across firms, which again presumably are rather heterogeneous. Since the focus of this paper is not to address such issues of national or international productivity comparison3, we choose not to explicitly estimate or measure F. However, our cost structure is general enough to control for non-zero F. Variable costs (CVC) depend on output (loans) and a vector of country-specific factor prices (w), which include a wage index for the banking sector (WAGE), the interbank rate (INTER), and the number of branches (NBR). See the appendix for a detailed description of the data. Given the above cost and demand conditions, we can write the corresponding two first-order condition for bank i in each segment M as (we suppress the index for t and s to simplify notation),

(3)

MCa« () = o

where MC() is the marginal cost function which depends on output in both segments. The parameter X measures the degree of collusion. If X =0, prices equal marginal costs and the industry is perfectly competitive. X =1 is consistent with Nash behavior, whereas X 's larger than one imply collusive price setting, with monopoly pricing being present whenever X = NS. Since our data are at the industry level we proceed by summing (3) over firms, yielding the industry level conditions, XM 3rM{} oM+ Ns • r M() x äe« B

= 0, 2= 1

where

Q" = 5 X

(4)

1=1

For a monopoly (TV, = 1) this expression is obviously identical to (3). At this point we assume that the individual banks in a given country have symmetric marginal cost functions, which implies that q,lIS = qjts, j * i . Thus we can re-write (4) as,

dQ

Ns

,( ü t s ) = o

(5)

which is implementable with industry level data. Note that 0 M = \ M I N s. Therefore, perfect competition is consistent with 0 = 0. A 0 = 1 /V 5 is indicative of Nash; and finally monopoly is consistent with 0 = 1.

3Moreover, productivity and efficiency comparisons are more sensibly done with less aggregate data, for instance at the bank or even branch level.

8 One more observation might be in order. As long as the game played is a repeated game, our model will be able to detect the degree of collusion in the various market segments. However, we need to assume that the payoffs, that is the demand and marginal costs functions, are not misspecified.

IV. Empirical Implementation The empirical implementation of the above model proceeds by linearizing the system of equations given by (1) and (5). Using the approach in de Bandt and Jacquimot (1992) we choose the following explanatory variables to predict demand for corporate and household loans. In particular we specify (1) as, r w = X a offA 5 r c = a oc +

+ « r i° g ^ H)-,- a 2fflog(COV5) + e H (D INVPER + a 2c PIMPPER+ a 3lo g (ß c ) + a 4log(G V P* WAGESHR ) + ec

where rH and r c denote respectively the interest rate on corporate loans and household mortgages, D is a vector of country dummies, CONS is private consumption, INVPER is annual capital investment, PIMPPER is price index of import goods and services, WAGESHR is the share of the total wage in GNP (see the appendix for details on the data and its sources). Table 1 provides summary statistics for the data used. Note that (T) controls for fixed country effects in the demand for household loans. The corresponding FOC defined in (5) can now be written as, 0 M- a 1M+ r M( ) - M C ( ) + g = O,

M = H ,C

(5')

where we specify,

MC() = B,

EQN

+ B,

qc

EQN

+ B, —---- — + ß.WHGB + Q.INT + fVNBR EQN EQN

where WAGE is the total wage bill, INT is the interbank rate (the cost of funds), and NBR is the number of branches. One potential problem with equation (5') is that the aggregation assumes symmetry in the cost function of banks within the same country, which in turn implies symmetry in the size of the banks. This is clearly empirically violated, since the size dispersion of European banks is

9 indeed substantial. Therefore, our test for Nash behavior, 0 = 1/ N s, is biased downwards, since a 0 based on an asymmetric industry structure is higher than the 0 from a symmetric industry structure. To correct for this, we express equation (5) in terms of the number of symmetric equivalent firms. To construct this number of symmetric equivalent fims, we use a result from Sleuwagen and Dehandschutter (1986). They derive upper and lower bounds on the Herfindahl index from the C4 concentration index. Since we only have data on the C4 concentration ratios in the various countries, we implement this procedure by taking the average of the two bounds. W e then use the constructed Herfindahl index to compute the equivalent number of symmetric firms (EQN) that corresponds to the index. EQN is then used in (5') to adjust market power for the asymmetric market structure. It should also be noted that the above specification assumes that the marginal cost function are symmetric over the two products, i.e. we assume that dc(

, qc ) / do" = 3c( q" , qc ) / dQc.

Given the degrees of freedom in our data set, a more flexible (asymmetric) multi-product cost function could not be estimated. The empirical implementation involves the estimation of the four equations given by (T) and (5'). We estimate the system by non-linear 3SLS.

V. Findings and Interpretations As indicated above, we estimate a system of four simultaneous equations ; one equation for the demand and one for supply in each market (loans to the corporate and household sector). We assume that the firms operate with one cost function but allow for different competitive behaviour in the household and corporate markets. We first assume that firms behave as Cournot competitors in all countries and in both markets and find that this hypothesis is rejected by the data. Next, we assume that firms behave as a cartel (again in all countries and in both markets) and test this hypothesis which is accepted in both markets. In a second step, we specify the behavioural parameters further by allowing for fixed effects across time and countries. These fixed effects are first estimated jointly with the demand and marginal cost equations. With this approach, the number of degrees of freedom is significantly reduced and our estimates are rather imprecise. As an alternative, we adopt a two step procedure, in which we first estimate the model without imposing restrictions on behavioural parameters and subsequently regress predicted values for the behavioural parameters on a time trend and country dummies. This procedure enables us to test for differences in behavioural parameters across countries.

10

In what follows, we shall review these various estimates in turn. The specification of the demand equations and of the marginal cost remains identical across these various estimations and the parameters estimates for these equations do not vary much. Accordingly, we will comment on these parameters only for the first set of estimates. The first set of estimates test whether banks behave as Cournot competitors in each country and in both markets. In other words we substitute the following into (5'),

and estimate the s Mfor both markets. Whenever s M = o we get Cournot behavior cross all countries. When s M > o, firms behave on average more collusive than Cournot. The results for (T) and (5') subject to (6') are reported in Table 2. We first consider the (inverse) demand for household loans : the demand elasticity at the sample mean is equal to 2.51. The elasticity that would be faced by individual banks in the absence of coordination can be obtained as the product of this estimate and the number of equivalent firms : at the sample mean, the elasticity faced by individual banks would be as high as -25. It also appears that demand for loans is highly cyclical: our proxy for consumer spending plans (CONS) is positive and highly significant. The demand for corporate loans has a price elasticity at sample mean which is equal to -2.17, an estimate similar to that obtained for the household sector. This may come as a surprise as one could have expected the corporate sector to be more sensitive to prices than private individuals. It seems that investment plans (INV) have a marginal effect on the demand for loans. The main determinant of this demand seems to be the share of wages in valued added. As payments to labour increase and those to capital fall, there is a need for additional financing which is met by bank loans. Our estimate of marginal cost indicates that there are some economies of scale in the provisions of loans. As can be seen in Table 2, the first-order effects of either output on marginal costs are insignificant. However, the interactive term is significant and negative, indicating that as the output of either product increases, marginal cost falls. In addition, we observe that marginal cost (at any level of output) is heavily influenced by the interbank rate, which can be seen as a proxy for the cost of funds. By contrast, the wage rate and the density of bank networks do not seem to affect the marginal cost of providing loans.

11 As indicated above, we first estimate the behavioral parameters (s" and sc) as deviations from the value that they would take if firms behaved as Cournot players. We observe positive and significant deviations from Cournot behavior in both markets. The estimated deviation from Cournot behaviour is also higher for the household market than for the corporate market, thereby suggesting that firms may collude more in the household m arket Having rejected the hypothesis that banks behave as Cournot players, we test the hypothesis that they behave collectively as a cartel. Accordingly, we estimate the behavioral parameters as deviations from cartel behavior, i.e. (6') becomes QM = l + s M. The demand and cost specifications are the same as those presented above. The results are presented in Table 3. As suggested above, the parameters of the demand and cost functions are not much affected. Regarding the conduct parameters, we find that the deviations from cartel behavior are not significantly different form zero for both markets. W e therefore cannot reject the hypothesis that banks behave collectively as a cartel in both household and corporate loan markets. So far, we have imposed the same type of competition (and associated behavioral parameters) on all countries. It may of course be that some countries are more competitive than others. Our discussion of the institutional developments in Section II would indeed suggest some variation across European markets. Accordingly, we check whether the deviations from Cournot behaviour vary across countries by introducing country dummies in the behavioral equations (5'). The simultaneous estimation of equations (T) and (5') with fixed effects for countries reduces the number of degrees of freedom by 12 (6 country dummies per market). We are thus left with only 20 degrees of freedom, leading to a dramatic drop in the precision of the estimation. As a consequence, Country dummies in the behavioral equations are not very significant.4 This inconclusive evidence regarding country differences is presumably due to the reduced degrees of freedom. Accordingly, we adopt an alternative approach in which we first compute (for both markets) the behavioral parameter for each year and country that are implied by our estimates of demand elasticities and marginal cost. We condition our results on the estimated marginal cost and demand elasticities generated by the first model presented above (in which the behavioral parameters are estimated as deviations from Cournot). In particular, using (5') we can contract a series of behavioral data for the household and the corporate sector that are conditioned on the estimates of Table 2 as follows,

4If the estimates of the demand equations are not much affected, some of the parameters of the cost functions lose their significance.

12

eM= [ MC0Ar a

Q],

m

=h ,c ,

(7')

M

where we use the predicted marginal costs given in Table 1. In a second step we regress the constructed conduct variables on a set of country dummies by running the following twoequation model,

eM = ---- + VT» + ,

EQN

M = H, C,

(8’)

where 0M is defined in (7') and the D's are the country dummy variables. In other words we test for systematic differences from coumot behavior across countries within the two sectors. Before we present these results we test for constant conduct in the European banking industry within the household and corporate secots. We thus estimate model (8') and compare the unrestricted estimates (where we allow for all country dummies) to the restricted estimation (where the behavioral variables are regressed on a constant)5. W e perform a standard F-test. The F-statistics is equal to 1.71 and 4.13 for the household and corporate markets, respectively. The critical value at the 5 % level is 2.25 and 3.12 at the 1% level. We therefore reject the restriction that the behavior is identical across countries for the corporate sector. For the household sector the statistical evidence against homogeneity in conduct is less significant. Interestingly, this is consistent with the finding that the household sector is more collusive than the corporate sector, as one would expect a higher degree of homogeneity in sectors that are more collusive. Having rejected constant conduct in the banking industry across our sample of European countries at least for the corporate sector, we now allow for the conduct parameter to vary in the household and corporate sectors. The results from the estimation of model (8') are presented in Table 4. All country dummies are highly significant rejecting that markets are perfectly competitive. Interestingly, the largest deviations are observed for Spain for both the corporate and household markets. This suggests that overall Spain is the most collusive market. More precisely, we have tested whether the remaining countries are significantly different from Spain. Using the standard errors implicit in Table 4, we observe that the behaviour of banks in the household market is significantly (at the 10% level) more competitive in Belgium, France and The Netherlands than in Spain. On the other hand, there is no systematic evidence that banks' behaviour in Germany, Denmark and the UK are any less

5Since the explanatory variables are identical for the two equations, OLS estimates are efficient and identical to seemingly unrelated methods.

13 collusive than that observed in Spain. These findings can be related to the characteristics of the national regulatory framework described above; Spain had by far the most protective system of regulation and it also turns out to be the most collusive country. In addition, Germany, Denmark and the UK all had a system of specialized institutions for the provison of mortgages. These countries, which have relatively light regulations in general, still turn to be undistinguishable from Spain. This may suggests that indeed the separation of banking activities into specialized institutions favors the coordination of behaviour. For the corporate market, all countries but The Netherlands and the UK are significantly more competitive than Spain.

The most competitive corporate sector appears to be Germany (see

Table 4), where regulation is indeed light. It still comes as a surpise that The Netherlands and the UK cannot be distinguished from Spain, given that these countries are also lightly regulated. Other factors, besides regulation, must be at work in those countries to account for such high level of coordination. Finally, we checked whether behaviour had changed over time by allowing time varying parameters in the fixed country effects6. We find no evidence that the degree of coordination had been reduced over time, despite the widespread deregulation described above. In the case of France, we even find that the degree of coordination may have increased over time in both corporate and household segments.

V. Concluding Remarks and Policy Implications This paper develops an aggregate, multiproduct model for the European Banking industry and estimates a structural model with data on demand and costs from 7 countries between 19811989. In general we reject non-coorporative Nash behavior in favour of more collusive cartel­ like conduct. W e report some mixed evidence regarding the impact of regulation on the coordination of behaviour among banks. On the one hand, we observe that the most protective regulation is always associated with the highest degreee of coordination and that the mortgage segment is particularly collusive when it is supplied by specialized institution. On the other hand, we observe that the degree of collusion in the corporate segment is high in relatively lightly regulated countries, suggesting that other factors may be at work. We cannot trace out either the effect of the widespread deregulation that has occured in the eighties. Of course, it may be that our sample is to short to pick up the effect of deregulation, given that it has taken place mostly in the mid to late eighties.

6The results are not reported here but can be obtained upon request from the authors.

14 Our analysis also sheds some light on the structure of the banking industry itself, irrespective of the regulation. First, we find that the elasticity of demand for household and corporate loans is relatively high, suggesting that much of the market power observed in the industry should indeed be associated to the coordination of behaviour. We also find that there are some economies of scale in the provison of loans.

Finally, we observe that bank's behaviour

is more collusive in the household segment than in the corporate one. Accordingly, there seems to be an implicit cross-subsidization between the two segments in most European countries in general, and particularily so in Germany.

15

Appendix - The Data and Descriptive Statistics

The data only refers to the banking sector, stricto sensu, and hence excludes savings banks (and building societies). The loans to the household sectors refer to the outstanding stock of mortgage loans at the end of each period (year). Loans to the corporate sector refer the outstanding stock of short and long term bank loans (including trade credit) at the end of the period. The data has been gathered from national sources; published sources are the monthly review of The National Bank of Belgium, the Quarterly Reviews of the Bundesbank, the Banque de France, the Bank of Italy and the Statistical Abstract of the UK. W e could not get data for mortgages in Italy. For Spain and the Netherlands, the data is not published and was provided to us by the respective central bank. All data were converted in real 1981 Ecus, using Exchange rates and inflation rates from the IMF, International Financial Statistics. The interest rates (interbank rate, corporate loans and deposits and household mortgages and deposits) were obtained from the OECD, financial statistics Monthly (various issues). For deposits (household and corporate) and corporate loans, we use the interest rate that applies to instruments of average maturity (given that our data on loans and deposits includes all maturities). All nominal rates were converted into real rates using inflation rates from the IMF, International Financial Statistics. The demand data used in (1*) are obtained from ECONOMIE EUROPEENNE (1993). CONS is private consumption, INVPER is annual capital investment in percentage change, PIMPPER is price index of import goods and services in percentage change, WAGESHR is the share of the total wage in GNP. The index of regulatory framework and regulatory changes across countries were constructed from various sources including the articles by Bruni; Caminal, Gual and Vives; de Boissieu; Rudolph; and Mayer in Dermine (eds), European Banking in the 90s, the Financial Times Business Information (1988 and 1991), OECD publications (Competition in Banking, 1989 and Financial Markets Trens, various issues) and Melitz (1990). A detailed description of regulatory changes is also presented in Gual and Neven (1992).

16 Data on total wage bills and number of employees were obtained from the OECD, Rentabilite des Banques, 1988 and 1992. GDP data were obtained from EUROSTAT (Economie Europeenne, Supplement Statistique). The market shares of the four largest banks (C4 concentration ratios) in the banking sector were obtained from Gual and Neven (1992). Missing years were extrapolated using a linear trend fitted to the available data. From the C4 indices, we then derive an Herfindhal index, using the correspondence between the two measures derived by Sleeuwagen and Dehandschutter (1986). These authors derive upper and lower bounds to the Herfindhal index which may be associated with any given value of the C4 concentration ratio. We take the average Of these bounds. These indices involve firms with different markets shares. Our model does however assume that firms are symmetric and the estimation involves the number of firms. Accordingly, instead of using the Herfindhal indices in the regressions, we compute and use the number of symmetric firms which would give rise to equivalent indices.

17

References Aghion, P., P. Bolton and M. Dewatripont, (1988), Interbank lending and contagious bank runs, mimeo, ULB, Brussels. Bruni, 1990, F. "Banking and financial reregulation towards 1992; the Italian case" in Dermine (ed), European Banking in the 90s. Basil Blackwell, Oxford. Caminal, R., Gual, J. and X. Vives, (1990), "Competition in Spanish Banking", in Dermine (ed), European Banking in the 90s. Basil Blackwell, Oxford. de Bandt, O., and P. Jacquinot, (1992), "The Financing of Corporate Firms in France", Economic Modelling. July, pp 253-269. Bemheim, B. Douglas, and Michael D. Whinston, 1990, "Multimarket Contact and Collusive Behavior," Rand Journal of Economics. 21(1), pp. 1-26. de Boissieu, C., (1990), "The French Banking sector in the light of European financial integration," Dermine (ed). European Banking in the 90s. Basil Blackwell, Oxford. Bresnahan, Timothy F., 1989, "Empirical Studies in Industries with Market Power," in Handbook of Industrial Organization. Vol. II, 1011-1058. Diamond, D. and D. Dybvig, (1983), Bank runs, deposit insurance and liquidity, Journal of Political Economy. 91.401-419. ECONOMIE EUROPEENNE, (1993), No. 54, Rapport Economique Annuel. Financial Times Business Information, Banking in the EC. 1988 and 1991. Gual, J and D. Neven, 1993, Deregulation in the European Banking Industry, 1980-1990, European Economy. 3,151-183. Mayer, C., (1990), "The regulation of financial services: lessons from the United Kingdom in 1992", in Dermine (ed), European Banking in the 90s. Basil Blackwell, Oxford. Melitz, J., (1990), "Financial deregulation in France", European Economic Review. 34, 394-402. OECD, Competition in Banking. 1989. OECD, Financial markets trends. 1991. Rudolph, B., (1990), "Capital requirements of German banks and the European Economic Community proposals on Banking Supervision", in Dermine (ed), European Banking in the 90s. Basil Blackwell, Oxford.

18 Sleuwagen, L. and W. Dchanschutter, 1986, "The critical choice between the concentration ratio and the H-index in assessing industry performance", The Journal of Industrial Economics. XXXV (2), 193-208.

19

Table 1 EUROPEAN BANKING INDUSTRY - Summary Statistics

Mean

Country

Minimum

Maximum

4.429

1.000

8.000

85.000

81.000

89.000

Household Loans

53640.430

1151.420

153314.900

Corporate Loans

137123.090

3404.550

372263.340

Household Rate

6.557

2.250

11.100

Corporate Rate

6.214

1.625

10.300

CONS

208.072

28.900

439.550

GNP

316.350

51.500

723.430

INVPER

2.873

-19.200

17.100

PIMPPER

4.037

-16.200

29.800

71.998

65.200

78.300

5.333

0.757

10.817

19539.680

3059.000

39812.000

WAGE

23.137

16.827

32.963

EQN

11.185

2.060

26.660

Year

WAGESHR INTER NBR

20

Table 2 EUROPEAN BANKING INDUSTRY - COURNOT BEHAVIOR Nonlinear 3SLS* Parameter

t-Stat

-2.855

-5.86

INVPER

.050

1.51

PIMPPER

-.019

-.65

WALESHR

3.885

6.89

Loans

-2.619

-5.86

CONS

8.429

2.39

Household Loans

-.072

1.01

Corporate Loans

.070

1.26

WAGE

-.008

-.28

INTER

.401

6.53

NBR

.000

.09

Household x Corporate

-.004

-1.79

1.255

3.24

1.563

2.32

Demand - Corporate Loans

Demand - Household

•

Marginal Cost

Behavioral - Corporate SC = 0 C -

1 EQN

Behavioral - Household oH

jj{

= 0H .

1 EQN

Second-order conditions are imposed

21

Table 3 EUROPEAN BANKING INDUSTRY - CARTEL BEHAVIOR Nonlinear 3SLS* Parameter

t-Stat

-2.974

-6.40

INVPER

.049

1.54

PIMPPER

-.017

-.61

WALESHR

4.022

7.47

Loans

-2.534

-1.52

CONS

7.836

1.44

Household Loans

.038

.54

Corporate Loans

.030

.53

WAGE

-.024

-.88

INTER

.426

6.84

NBR

-.000

-.50

Household x Corporate

-.006

-2.85

.524

1.48

.974

.71

Demand - Corporate Loans

Demand - Household

Marginal Cost

Behavioral - Corporate

sc = ec - i Behavioral - Household sH = e H .i

jjc

Second-order conditions are imposed

22

Table 4 EUROPEAN BANKING INDUSTRY - COUNTRY EFFECTS

Household Loans Belgium

1.379

9.42

Denmark

1.797

15.03

France

1.472

12.31

Germany

1.544

12.92

NL

1.429

11.27

Spain

1.813

12.38

UK

1.493

10.20

Belgium

1.297

8.05

Denmark

1.048

7.97

France

1.174

8.93

Germany

0.895

6.80

NL

1.614

11.57

Spain

1.719

10.67

UK

1.366

8.48

Corporate Loans