Bank Market Structure, Competition, and SME Financing Relationships in European Regions

Abstract

How do concentration and competition in the European banking sector affect lending relationships between small and medium sized enterprises (SMEs) and their banks? Recent empirical evidence suggests that concentration and competition capture different characteristics of banking systems. Using a unique dataset on SMEs for selected European regions, we empirically investigate the impact of increasing concentration and competition on the number of lending relationships maintained by SMEs. We find that competition has a positive effect on the number of lending relationships, weak evidence that concentration reduces the number of banking relationships and weak persistent evidence that they tend to offset each other.

Appendices

Appendix I: Definitions of explanatory variables

Variable	Description	Source
Multi-banking relationships	Whether SME has 1,2, 3, or 4 bank relationship	Cambridge SME Survey
Firm Type	Whether the SME is public or private company (1 = private, 0 otherwise).	Cambridge SME Survey
Size (employees)	Firm size measured as an index variable that is increasing in the number of employees. The index takes the value of 1 if there are 1–9 employees; 2 if there are 10–19 employees; 3 if there are 20–49 employees; 4 if there are 50–99 employees and 5 if there are more than 100 employees.	Cambridge SME Survey
Amount of bank finance used	This is an index variable that is increasing in the volume of bank financing. It takes on the value 1 if the firm used up to GBP99,999 (DEM337,999); 2 if GBP 100,000–249,999 (DEM338,000–843,999); 3 if GBP250,000–499,999 (844,0000–1,687,999); 4 if GBP 500,000–999,999 (DEM1,688,000–3,374,999); and 5 if GBP 1 m and over (DEM 3,375,000 and over).	Cambridge SME Survey
Easiness to obtain finance	This index is increasing in the degree of difficulty to attract bank financing. It takes on the value 1 if the firm considers obtaining financing ‘very easy’; 2 if the firm considers it ‘fairly easy’; 3 if the firm considers it ‘fairly difficult’; or 4 if the firm considers obtaining financing ‘very difficult’.	Cambridge SME Survey
Distance	Distance of bank from firm. Variable takes value of 1 if <10 miles; value of 2 if between 10–49 miles; value of 3 if >= 50 miles.	Cambridge SME Survey
Firm Growth	Firm growth in terms of changes in the number of employees over 5 years.	Cambridge SME Survey
Attitude	Dummy variable that takes on the value 1 if the firm considers the main bank’s attitude as ‘supportive’ or ‘helpful’ and zero otherwise.	Cambridge SME Survey
Herfindahl-Hirschman index	Sum of the squared market shares in terms of total assets	BankScope and authors’ calculations
H-Statistic (1)	Measure of the degree of competition, dependent variable is interest revenue	BankScope and authors’ calculations
H-Statistic (2)	Measure of the degree of competition, dependent variable is total revenue	BankScope and authors’ calculations

Appendix II: Computation of the H-Statistic and of the HHI

We present in this appendix a brief overview of the Panzar and Rosse (1987) H-Statistic that we utilise to gauge competition. This statistic is widely used to test for banking competition (e.g., Molyneux et al. 1994; Claessens and Laeven 2004; Schaeck et al. 2009). Subsequently, we also briefly discuss the way we compute the HHI.

1.1 Computation of the H-Statistic

The H-Statistic is derived from reduced-form revenue equations and measures market power by the extent to which changes in factor input prices are reflected in revenue. Assuming long-run equilibrium, a proportional increase in factor prices will be mirrored by an equiproportional increase in revenue under perfect competition. Under monopolistic competition, however, revenues increase less than proportionally to changes in input prices. In the monopoly case, increases in factor input prices will be either not reflected in revenue, or will tend to decrease revenue.¹¹ The magnitude of H can be interpreted in the following way:

H ≤ 0 :

indicates monopoly equilibrium

0 < H  < 1 :

indicates monopolistic competition

H = 1 :

indicates perfect competition

To calculate H-Statistics for local banking markets in which the SMEs in the survey operate in, we proceed as follows. In the first step, we obtain additional information from the Survey of the Financing of Small and Medium-sized Enterprises in Western Europe about the first three digits of the postcodes where the SME’s are located.¹² This allows us to ascertain exactly in which regions of Bavaria and the South-East of England the firms operate. Next, we obtain bank data from BankScope and download all savings, co-operative, and commercial banks operating in these regions that we identify with the postcodes. This exercise allows us to manually match firms with banks operating in these regions. We retrieve BankScope data for the years 1998–2001 to be able to capture competitive dynamics when calculating H-Statistics. We impose the criterion that we need at least 15 bank-year observations in each region to obtain reasonable estimates of the H-Statistic.

For the German data, we are then able to establish four regional banking markets in Bavaria based on the survey data. All these markets are located in the south of Bavaria and closely match administrative districts, the so-called “Landkreise” to which the operations of small and regional banks are typically constrained. For the UK, we identify six regional banking markets based on postcodes, these markets overlap with “counties”.

Using a fixed effects panel data estimator we estimate the following reduced-form revenue equation for each one of the local banking markets as follows

$$ \begin{array}{*{20}c} {\ln \left( {P_{it} } \right) = \alpha + \beta_1 \ln \left( {W_{1,it} } \right) + \beta_2 \ln \left( {W_{2,it} } \right) + \beta_3 \ln \left( {W_{3,it} } \right)} \\ {\quad \;\;\, + \gamma_1 \ln \left( {Y_{1,it} } \right) + \gamma_2 \ln \left( {Y_{2,it} } \right) + \gamma_3 \ln \left( {Y_{3,it} } \right) + \varepsilon_{it} } \\ \end{array} $$

(A.1)

where P _it is the ratio of interest revenue to total assets (proxy for output price), W _1,it is the ratio of interest expenses to total deposits and money market funding (to proxy for the input price of deposits), W ₂ , _it is the ratio of personnel expense to total assets (proxy for the price of labour) and W _3,it is the ratio of other operating and administrative expenses to total assets (proxy for price of fixed capital), with i denoting bank i and t denoting time t. Y _1,it is a control variable for the ratio of equity to total assets, Y _2,it controls for the ratio of net loans to total assets and Y _3,it is the log of total assets to capture size effects. All variables enter the equation in logs. H is calculated as β ₁ + β ₂ + β ₃. For a subsequent sensitivity test reported in Section 4 we re-run Eq. (A.1) with the ratio of total revenue to total assets as dependent variable.

Since the H-Statistic assumes long-run equilibrium, we perform the following equilibrium test and estimate Eq. (A.1) with the pre-tax return on assets as dependent variable.

$$ \begin{array}{*{20}c} {\ln \left( {ROA_{it} } \right) = \alpha + \beta_1 \ln \left( {W_{1,it} } \right) + \beta_2 \ln \left( {W_{2,it} } \right) + \beta_3 \ln \left( {W_{3,it} } \right)} \\ {\; + \gamma_1 \ln \left( {Y_{1,it} } \right) + \gamma_2 \ln \left( {Y_{2,it} } \right) + \gamma_3 \ln \left( {Y_{3,it} } \right) + \varepsilon_{it} } \\ \end{array} $$

(A.2)

The modified H-Statistic is the equilibrium statistic and it is again calculated as β ₁ + β ₂ + β ₃. We test if the equilibrium statistic E = 0, using an F-test. This test aims to establish whether input prices are uncorrelated with industry returns since a competitive system will equalise risk-adjusted rates of return across banks in equilibrium. If this hypothesis is rejected, the market is assumed to be in disequilibrium. The results from our equilibrium test indicate that the markets under consideration are in long run equilibrium.

1.2 Computation of the Herfindahl-Hirschman index

Whereas the panel data estimator yields exactly one observation for each one of the ten different local markets, we proceed as follows to compute HHIs that can be used for the estimation in our cross-sectional regression models. First, we calculate annual HHIs for the local markets for each year. Second, we then take the average of the annual HHIs for the local markets to use them in our subsequent estimations.

We present an overview of the two different H-Statistics and the HHIs for the local banking markets in Table 5. To explore the interrelationships, we have also explored further the relation between HHI and the H-Statistic. Our visual inspection already suggests a positive relationship between the two variables. We also examine the correlation and find that they are significantly positively correlated (correlation coefficient 0.64***). This finding is in line with the results presented in Claessens and Laeven (2004), and further challenges the predominant view that concentration and competition are inversely related. Fig. 1

Fig. 1

Scatter plot HHI and H-Statistic

Table 5 Summary of market structure variables by region