SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

We examine the effect of competition on banking stability using a new measure of competition based on the reallocation of profits from inefficient banks to efficient ones. In a sample of European Banks, we find that this measure does capture competition, that competition is stability-enhancing, and that the stability-enhancing effect of competition is greater for healthy banks than for fragile ones. Our results suggest that efficiency is the conduit through which competition contributes to stability and that regulators must condition policy on the health of existing banks.

The question of whether more competition is good or bad for banking stability has been intensely debated in academic, as well as policymaking and regulatory circles (Keeley, 1990; Allen and Gale, 2004; Berger et al., 2004; Berger, Klapper, and Turk-Ariss, 2004; Beck et al., 2010; Martinez-Miera and Repullo, 2010; Wagner, 2010). Competition among banks has been argued to be a contributor to the instabilities that triggered banking problems in many countries (OECD, 2010). A competing argument is that the instability reflected regulatory failures or weak market discipline, and that more competition is necessary to ensure that banks are stronger. This belief is reflected in a statement expressed a few years prior to the crisis by Padoa-Schioppa (2001, p. 16) who claimed “if banks were strengthened by the gymnastics of competition, the banking system would be stronger and more resilient to shocks.”

The academic literature has yet to reach a consensus. While recent theory and evidence point toward the positive effect of competition on stability (Boyd and De Nicolo, 2005; Schaeck, Cihak, and Wolfe, 2009; Allen, Carletti, and Marquez, 2011), the traditional literature conjectures that increasing competition erodes charter values resulting in a negative trade-off between competition and stability (Keeley, 1990). In an attempt to reconcile these conflicting predictions, Martinez-Miera and Repullo (2010) consider a risk-shifting effect that increases stability via a reduction in borrower default rates as a result of lower loan rates, and a countervailing margin effect that reduces banks’ buffers against loan losses and decreases stability. As a result of these two competing forces, they predict a U-shaped relationship between competition and stability.

In this paper, we revisit the debate empirically and examine what mechanism compels competition to contribute to bank stability as suggested in several recent studies. Despite its relevance for policy and regulation, the transmission mechanism has remained an underexplored area. Our research aims to fill this gap.

We note that based on ideas put forth in the industrial organization literature, we expect efficiency to be the conduit between competition and bank stability. To examine this conjecture, we exploit the unique features of a new competition indicator, known as the Boone indicator in the literature (Boone, 2008). Using a representative panel data set of banks that encompasses European banking systems during the precrisis era to avoid government interventions such as blanket guarantees and other bailouts that undermine market discipline and, as such, blunt the hypothesized effects arising from competition, we offer several innovations in the debate on competition and stability.

First, to investigate the mechanism by which competition contributes to stability, we use the Boone (2008) indicator to capture the impact of competition on the performance of efficient banks. The use of the indicator is consistent with the industrial organization literature demonstrating that competition reallocates profits from inefficient to efficient firms (Olley and Pakes, 1996; Stiroh, 2000), and with the banking literature that offers ample evidence that more efficient banks are more stable (Berger and DeYoung, 1997). In the first step of our analysis, we also demonstrate that the Boone (2008) indicator captures a variety of competitive characteristics of banking systems in Europe.

Additionally, we analyze the link between competition and stability, exploiting the unique properties of the Boone (2008) indicator. The analysis robustly supports the stability-enhancing effect of competition via the efficiency channel. Our findings remain unaffected once we account for survivorship issues, consider subsamples of different bank types (commercial, savings, and cooperative banks), and alternative measures of stability.

Moreover, we examine heterogeneous responses to competition using quantile regressions. Here, we determine that fragile institutions benefit less from competition than stable banks. Our research carries important implications for policy and regulation in banking. The findings raise a cautionary flag against far reaching restrictions regarding the scope and scale of bank activities as proposed in recent years following stress in numerous banking systems, including in the United States (Volcker and Frenkel, 2009; Blinder, 2010). Instead, when supported by efficient supervision, policies promoting competition may have positive effects on bank efficiency and stability. In addition, the results obtained with quantile regressions highlight another previously neglected phenomenon. Policymakers need to consider that the effect of competition on stability is conditional upon the stability of the banks operating in the relevant market.

The paper is structured as follows. We develop our hypotheses in Section 'Hypothesis Development'. Section 'Data' provides an overview of the data set, while Section 'Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator' presents the intuition behind the new measure of competition. We present the results for the effect of competition on stability in Section 'The Nexus between Competition, Efficiency, and Bank Stability' and report our conclusions in Section 'Conclusion'.

I. Hypothesis Development

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

In banking, information asymmetries have substantial ramifications for competition and efficiency of lending decisions as banks obtain proprietary information through lending activities that provide information advantages over other, less informed lenders (Bouckaert and Degryse, 2006). Additionally, it is not only lending activities give banks a competitive advantage. Mester, Nakamura, and Renault (2007) find that other dimensions of relationship banking, such as transactions accounts, also substantially reduce information asymmetries and give rise to a competitive advantage.

Alternatively, competition itself is considered to improve firm efficiency in the industrial organization literature (Tirole, 1989; Hay and Liu, 1997). In addition, there is also evidence in banking that efficient institutions maintain better screening and monitoring procedures, resulting in better asset quality (Wheelock and Wilson, 1995). Consequently, an analysis that focuses on the transmission mechanism by which competition can influence bank stability needs to reflect on efficiency considerations.

We derive an empirically testable hypothesis to investigate the competition-efficiency-stability nexus. Specifically, we suggest that efficiency is the conduit through which competition enhances stability. Following the industrial organization literature, we expect competitive environments to result in greater efficiency. Ultimately, efficiency improvements will also enhance stability as inefficiencies in banking are primarily due to poor lending decisions that arise from resource intensive monitoring of delinquent borrowers, analyzing workout arrangements, and seizing and disposing of collateral (Berger and DeYoung, 1997).

For our analysis, we use a new competition indicator developed in the industrial organization literature by Boone, Griffith, and Harrison (2005), and Boone (2008). The intuition behind the indicator emanates from the efficiency hypothesis, which stresses that industry performance is an endogenous function of the growth of efficient firms (Demsetz, 1973). Put simply, the indicator gauges the strength of the correlation between efficiency (measured in terms of average costs) and performance (measured in terms of profitability).

A. The Link between Competition and Efficiency

The link between competition and efficiency has become rather well established in the industrial organization literature. The consensus is that competition tends to trigger reallocations of profits toward more efficient firms (Olley and Pakes, 1996). More efficient firms outperform their less efficient counterparts in terms of profits, and this fosters industry-wide efficiency.

The underlying theoretical arguments from the broader industrial organization literature also apply to the specific case of the banking sector. Empirical research regarding the link between competition and efficiency in banks by Berger (1995) indicates that market shares correlate positively with profitability when industry concentration is accounted for. This work suggests that the relative market power of banks matters when explaining profitability. More importantly, these findings also reject the view that scale economies are positively related to profits and market concentration. More directly related to the ideas laid out in Boone (2008) is the evidence in Stiroh (2000), who determines that a reallocation of assets from weak to well-performing banks maintains profits on the industry level. Stiroh and Strahan (2003) find that competition reallocates profits from weak to “well run” banks, while Berger and Hannan (1998) indicate that banks operating in uncompetitive markets are less efficient. Studies by Jayaratne and Strahan (1998), DeYoung, Hasan, and Kirchhoff (1998), and Koetter, Kolari, and Spierdijk (2012) confirm that competition enhances efficiency.

B. The Link between Efficiency and Stability

We propose that greater efficiency will translate into enhanced stability, reflected in the reduced likelihood of bank default and better asset quality. The intuition is as follows. Chen (2007) finds that competitive pressure improves the efficiency of screening and monitoring of borrowers, which, in turn, improves borrower performance. Allen et al. (2011) confirm that more efficient screening and monitoring by banks also attracts better credit risks. Empirical work by Wheelock and Wilson (1995) and Berger and DeYoung (1997) supports these theoretical predictions.

Based on these considerations, we formulate the Transmission Mechanism hypothesis: Competition, measured by the Boone (2008) indicator, enhances bank stability.

II. Data

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

We assemble a panel data set from BankScope for European banks from 1995 to 2005. The sample covers Austria, Belgium, Denmark, France, Italy, Germany, Luxembourg, Netherlands, Switzerland, and the UK and consists of 17,965 bank year observations for 3,325 banks. Of the 17,965 observations, 5,705 are for savings banks, 9,297 for cooperatives, and 2,963 for commercial banks. This data set is representative for the major European banking systems and not affected by selection problems.1

Exploiting the time-series variation, this sample allows for tracking of competitive dynamics over time. Europe provides a fertile ground for analyzing the effects of competition since these banking systems experienced changes in regulation aimed at creating a level playing field for competition in the 1990s.

Our data set ends prior to the recent crisis. The rationale for restricting the sample to the precrisis period is that government intervention, liquidity support, guarantees, and public bailouts during the crisis distort competition and risk-taking incentives as demonstrated in recent papers. Hakenes and Schnabel (2010) illustrate that bailout expectations encourage protected banks to expand and intensify deposit market competition, resulting in compressed margins at competitor banks whose risk-taking incentives increase. Empirical research by Gropp, Hakenes, and Schnabel (2011) supports these predictions. They present evidence that government guarantees increase risk-taking by banks that compete with institutions that receive public support. Table I provides summary statistics for the variables used in the empirical analyses below.

III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

To examine the link between bank competition, efficiency, and stability, we use a modified version of a competition indicator proposed by Boone et al. (2005), and further developed by Boone (2008).

A. The Theory Underlying the Boone (2008) Indicator

The indicator is based on the efficient structure hypothesis that associates performance with differences in efficiency. Under this hypothesis, more efficient banks (i.e., banks with lower marginal costs) achieve superior performance in the sense of higher profits at the expense of their less efficient counterparts. This effect is monotonically increasing in the degree of competition when banks interact more aggressively and when entry barriers decline. The hypothesis is consistent with findings by Stiroh (2000) confirming that increased competition allows the transference of considerable portions of assets from low profit to high profit banks.

As demonstrated in Boone (2008), the reallocation effect is a general feature of intensifying competition, so that the indicator may be seen as a robust measure of competition.2 While different forces may cause increases in competition (e.g., increases in suppliers of banking services through lower entry cost), more aggressive interaction between banks (shift from Cournot to Bertrand competition) or banks’ relative inefficiencies as long as the reallocation conditions holds, the indicator remains valid. As the industry becomes more competitive, given a certain level of efficiency of each individual bank, the profits of the more efficient ones increase relative to those of the less efficient banks.

The Boone (2008) indicator has another appealing feature. It overcomes shortcomings of traditionally used proxies for competition, such as the Herfindahl-Hirschman index and the three-bank concentration ratio, that gauge competition by examining concentration levels (Degryse, Kim, and Ongena, 2009). This is particularly important as the recent literature has arrived at the conclusion that the link between concentration and competition is very weak in banking (for a detailed overview, see Berger et al., 2004). Unlike concentration indices, the Boone (2008) indicator is able to capture interaction among banks by focusing on conduct, whereas concentration ratios only capture the outcomes of competitive conduct. For instance, if competition leads to a bank exit via failure or merger, concentration ratios increase. Thus, relying on concentration measures will yield misleading inferences as high concentration is considered to be indicative of a lack of competition.

Unsurprisingly, studies that examine competition and concentration in banking conclude that concentration is a poor proxy for competition (Claessens and Laeven, 2004). In addition, other measures of competition, such as the Panzar and Rosse (1987) H-Statistic, require restrictive assumptions about the market existing in long-run equilibrium, while the Lerner index is criticized for not capturing product substitutability (Vives, 2008). The Boone (2008) model does not assume long-run equilibrium, nor does it suffer from the problem relating to product substitutability. What matters for the Boone (2008) indicator is how aggressively the more efficient banks exploit their cost advantage to reallocate profits from the least efficient banks in the market.

Following Boone et al. (2005) and van Leuvensteijn et al. (2007), we write a banking system demand function in which bank i produces a product (or product portfolio) qi so that

  • display math(1)

whereby each bank has a constant marginal cost ci. The parameter p denotes the price, and a captures market size, while b denotes the market elasticity of demand. We use the parameter d to characterize the extent to which consumers see the different products in a market as close substitutes for each other. It is assumed that a > ci and 0 < d ≤ b. To maximize variable profits, the bank decides on the optimal output level qi so that

  • display math(2)

The first order condition for equilibrium is then given by

  • display math(3)

For a banking system with N banks that produce positive levels of output, one obtains N first-order conditions from Equation (3):

  • math image(4)

Equation (4) illustrates the relation between output and marginal cost. We can see from Equation (2) that variable profits depend upon marginal costs in a quadratic way. Note that we define profits πi as variable profits that exclude entry costs. Therefore, a bank will enter the market if, and only if, πi ≥ ε in equilibrium.

Based on these properties, competition increases for two reasons. First, competition increases when close substitutes exist for bank products and when banks interact more aggressively (i.e., d increases assuming that d < b). Additionally, competition increases if entry costs ε decline. Boone (2008) confirms that performance of more efficient firms improves under both these regimes.3

Assuming that the relation between profits inline image and marginal costs ci is downward sloping, it follows that higher marginal costs imply lower margins per output unit for a given price. Moreover, if higher marginal costs lead to higher prices, output is reduced.

B. Estimating the Boone (2008) Indicator

In line with the theoretical exposition by Boone (2008) and the existing empirical work by van Leuvensteijn et al. (2007), which both start at the microlevel, but then move to the aggregate level, we estimate the Boone (2008) indicators from microlevel data to gauge the magnitude of the reallocation effect at the aggregate banking system (i.e., country) level. For the empirical implementation, we characterize the Boone (2008) model for bank i as

  • display math(5)

where inline image measures profits of bank i at time t, β is referred to as the Boone (2008) indicator, and inline image denotes marginal costs. Since we cannot observe marginal costs directly, we follow the suggestion in Boone et al. (2005) and use average costs as a proxy.4

The model by Boone (2008) makes no reference to financial intermediation theory. Therefore, we briefly discuss the model in the context of banks’ production technology as assumptions about inputs and outputs are essential. While the early literature in banking exhibits debates about modeling input and output, it is by now common to follow the intermediation approach (Sealey and Lindley, 1977; Koetter et al., 2012).5 Under this approach, banks use input factors, such as labor, physical capital, and deposits and other borrowed funds, to produce loans and other earning assets as outputs. Thus, the Boone (2008) model can be used to examine how profits from loans and other earning assets co-vary with the average costs of deposits and other borrowed funds, labor, and fixed assets.

The rationale for using the Boone (2008) indicator to capture the link between profitability and average cost is as follows. An increase in costs reduces profits in all markets, but the same percentage increase in a more competitive market leads to a greater decline in profits as banks are punished more harshly for being inefficient.

The indicator exploits this property because it measures the extent to which differences in efficiency are reflected in performance differences. In other words, the indicator expresses the reduction of profits that arises from cost inefficiencies. Since cost inefficiencies often reflect poor lending decisions, the indicator is well suited for expressing competition as a function of efficiency in banking.

To allow for heterogeneity within in the sample, we also include bank-specific effects. For the investigation of the transmission mechanism, it is important to allow for time variation in the effect of competition on stability. We estimate the Boone (2008) model as follows:

  • display math(6)

where inline image are the profits of bank i at time t as a proportion of its total assets, T is the total number of periods (years), dkt is a time dummy where dkt = 1 if k = t and zero otherwise, inline image are average variable costs, and uit is the error term. Closely following Boone's (2008) construction of the indicator, we use the average costs of bank i, measured as a ratio to its total income. The cost components are the sum of interest and personnel expenses, administrative, and other operating expenses. Income consists of commission and trading income, interest income, fees income, and other operating income. Profits are higher for banks with lower marginal costs (β < 0). Thus, increases in competition raise profits of more efficient banks relative to less efficient ones. The stronger the effect (i.e., the larger the β in absolute value), the stronger is competition.

C. Boone (2008) Indicators: Results

We estimate the relation between bank profitability, measured by return on assets (ROA), and average costs based on Equation (6) using a generalized method of moments (GMM)-style estimator. Our choice of a GMM-style estimator is due to concerns that performance and cost are jointly determined. For instance, banks that are large relative to the system might benefit from lower costs of production due to market power. Those banks may also be able to extract higher profits by exploiting such market power. The efficiency gains of the two-step GMM estimator relative to a traditional instrumental variables estimator is derived from the use of the optimal weighting matrix and the relaxation of the i.i.d. assumption. In line with the literature on dynamic panels (Blundell and Bond, 1998) and with recent work on bank efficiency (Koetter et al., 2012), we use one-year lags of the explanatory variables as instruments. The appendix presents estimates of the Boone (2008) indicators. The coefficients for the indicator are negative and significant.

To clarify the validity of the instruments, our appendix presents two diagnostics. First, we provide Anderson correlation test statistics for the null that the equation is underidentified and the instruments have low relevance. Additionally, we report Shea's (1997) Partial R2 that provides information about the correlation between the excluded instruments and the endogenous variables. We reject underidentification at conventional significance levels. Our instruments are strong in the sense that we observe, for most countries, a considerable correlation between the instruments and the endogenous regressor. Since our equations are exactly identified, we cannot perform overidentification tests.

Figure 1 illustrates how competition evolved in Europe. The Dutch banking system is the most competitive one, followed by the UK and Switzerland. In terms of the rankings of competition, our results are in line with Carbo et al. (2009). The greater variation in the Boone (2008) indicator for the Netherlands reflects a process of reorganization in the late 1990s (van Leuvensteijn et al., 2007). Germany exhibits a low degree of competition. This finding is due to the fact that major portions of the market are shielded from competition as savings banks and cooperatives only operate in local markets.

image

Figure 1. Boone Indicators

Download figure to PowerPoint

Table II, Panel A summarizes the country-by-country results for the Boone (2008) indicator and juxtaposes them with other key characteristics and measures of competition. To approximate banks’ potential to compete in different lines of business and the decline of entry costs, Table II presents an index of activity restrictions and data for the number of applications for bank licenses.6 The activity restrictions index provides insight into the scope of different activities, and one could expect competition to be higher with less activity restrictions.7 The table also reports the number of entry applications offering some information about entry costs. Given that lower entry costs tend to motivate more entry and more competition, we would expect a negative correlation between the Boone (2008) indicator and the number of entry applications.

Table I. Summary Statistics
VariableObs.MeanStd. Dev.Min.Max.
Dependent Variables     
Z-score17,96529.59622.8595.201257.506
Z-score (log)17,9653.1940.5901.6485.551
Nonperforming loans/Total loans11,4500.0460.0160.0020.078
ROA17,9650.0080.007−0.0350.112
Explanatory Variables     
Boone indicator17,965−0.0300.015−0.123−0.012
Total assets (TEUR)17,9651,374,203.003,636,810.0011,131.00113,000,000.00
Total assets (log)17,96513.1561.3089.31718.542
Asset growth17,9650.0690.221−0.8684.974
Diversification index17,9650.5880.1130.0680.999
Loan loss provisions/Total assets17,9650.0040.005−0.0580.438
Commercial bank dummy17,9650.1640.3710.0001.000
Savings bank dummy17,9650.3180.4660.0001.000
Cooperative bank dummy17,9650.5180.4500.0001.000
Herfindahl-Hirschman index17,9650.0060.0190.0000.206
Total banking system assets (log)17,96521.5491.17417.36323.199
GDP per capita17,96523,247.614,697.6817,564.9648,837.73
Unemployment17,9650.0820.0240.0020.123
Instruments     
Market share17,9650.0010.0070.0000.120
Financial Freedom17,96560.37613.7235090
Loan growth17,9650.0840.307−0.93419.714
Table II. Boone Indicators and Other Characteristics of Competition
Panel A. Levels
CountryNBooneActivityEntryStockInsuranceH-FinancialGovernmentHerfindahl
  IndicatorRestrictionsApplicationsMarketPremiums/StatisticsFreedomOwnershipHirschman
     Value/GDPGDP Index Index
Austria136−0.0311.58350.0700.0570.52270.800.000.110
Belgium29−0.0552.083240.1720.0850.58570.000.000.127
Denmark53−0.0312.25080.3610.0690.31077.900.000.127
France162−0.0211.75000.5370.0900.43050.000.000.022
Germany1,388−0.0221.583250.4600.0660.45652.3042.200.002
Italy410−0.0422.66680.3940.0580.39070.0010.000.039
Luxembourg67−0.0321.750210.0260.3100.85579.305.050.046
Netherlands11−0.0731.500121.2070.0950.97090.003.900.291
Switzerland215−0.0621.666361.8950.1190.59190.0014.120.047
UK43−0.0761.16601.0820.1370.59090.000.000.140
Panel B. Correlations and R2
   ActivityEntryStockInsuranceH-FinancialGovernmentHerfindahl
   RestrictionsApplicationsMarketPremiums/StatisticsFreedomOwnershipHirschman
     Value/GDPGDP Index Index
Correlation (of the Boone indicator0.335−0.091−0.666−0.050−0.521−0.810.270−0.666
 and the column variable)        
R2 (from OLS regression of the Boone0.1120.0080.4430.0030.2720.660.0780.443
 indicator on the column variable)        

To capture information regarding substitutes for traditional bank products, the table also provides the ratios for total stock market value traded to gross domestic product (GDP) and insurance premiums to GDP. This is to capture the point that while many banks in Europe are universal financial institutions offering brokerage and insurance services, nonbank providers, such as brokerage houses, mutual funds, and insurance corporations, compete with banks offering other, potentially substitutable financial services. The ratios of stock market value traded to GDP and insurance premiums to GDP are proxies for the extent of the financial services provided by nonbanks.

Panel B of Table II presents correlations between the Boone (2008) indicator and the other key market characteristics and competition features. Panel B also reports the coefficient of determination R2 (obtained from ordinary least square (OLS) regressions of the Boone (2008) indicator on the corresponding column variable).

Since lower (more negative) values of the Boone (2008) indicator signify more competition, the findings in Panel B support the hypothesized correlations between the indicator and the index for activity restrictions and the number of entry applications. These findings suggest less competition when regulators impose restrictions and more entry applications when the Boone (2008) indicator is lower. The corresponding coefficient of determination R2 indicates that more than one tenth of the variation in the Boone (2008) indicator can be explained by activity restrictions. The correlations between the Boone (2008) indicator and the proxies for nonbank financial services are negative supporting the notion that competition intensifies with the greater presence of substitutes. More than 44% of the variation in the Boone (2008) indicator is explained by stock market total value traded to GDP. These empirical regularities are consistent with the theoretical considerations underlying the Boone (2008) indicator.

Analyzing correlations between the Boone (2008) indicator and other competition characteristics, such as the Panzar and Rosse (1987) H-Statistic, government ownership of banks, the Financial Freedom Index, and the Herfindahl Hirschman index that captures deposit market concentration, also suggest that the indicator is linked intuitively with competition.8 The correlation between the Financial Freedom Index and the Boone (2008) indicator is negative indicating that competition is higher in systems with more freedom. Moreover, the R2 from the regression of the Boone (2008) indicator on the Financial Freedom Index highlights that 66% of the information contained in the Boone (2008) indicator is also reflected in the Financial Freedom Index. In line with intuition, government ownership is positively correlated with the indicator, whereas market concentration is inversely related to the indicator.

Of particular interest is the H-Statistic, a popular measure used in the literature to discriminate among competitive, monopolistically competitive, and monopolistic markets. It is calculated by estimating the sum of the elasticities of reduced form revenue equations with respect to factor input prices. The H-Statistic ranges between –∞ and one where higher values indicate greater competition (Claessens and Laeven, 2004). The correlation between these two competition measures is negative, as one would expect. When the Boone (2008) indicator signals more competition (is lower), the H-Statistics also tend to show greater competition (are higher). The empirical correlation is far from perfectly negative, which is not surprising. The two measures approximate the same general concept in different ways, so it is possible for them to produce somewhat different results. From a conceptual perspective, the Boone (2008) indicator is a better (more precise) measure for competition, as explained in the theoretical discussion. It is encouraging that these empirical tests underscore that other measures of competition are broadly consistent with this more rigorous measure.

IV. The Nexus between Competition, Efficiency, and Bank Stability

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

This brings us to the core of our analysis, which is the examination of the nexus between competition, efficiency, and stability. We estimate a general class of panel data models of the form:

  • display math(7)

where Zijt is a measure of bank stability for bank i in country j at time t, Bjt is the Boone (2008) indicator in country j at time t, and X and C are vectors of bank- and country-specific variables.9 All explanatory variables are lagged by one period unless stated otherwise. If we find a negative sign for the Boone (2008) indicator β, we can interpret this as evidence that the reallocation effect of profits from inefficient banks to efficient ones enhances stability.

To measure bank stability, we use the Z-score, calculated as

  • display math(8)

where ROA is return on assets, E/A denotes the equity to asset ratio, and inline image is the standard deviation of return on assets. We use a three-year rolling time window for inline image to allow for variation in the denominator of the Z-score. This approach avoids that the Z-scores are exclusively driven by variation in the levels of capital and profitability.

The Z-score is a widely used measure of bank stability (Laeven and Levine, 2009; Demirguc-Kunt and Huizinga, 2010; Houston et al., 2010). It combines banks’ buffers (capital and profits) with the risks they face (measured by the standard deviation of returns). The Z-score measures the number of standard deviations a return realization has to fall in order to deplete equity. A higher Z-score implies a lower probability of insolvency, providing a direct measure of stability that is superior to analyzing leverage. We use a log-transformation of the Z-score because it is skewed. For brevity, we label the variable Z-score when we refer to the transformed version of the Z-score in the remainder of the paper.

We use total assets (log) to control for size as larger banks are frequently subject to too big to fail policies. Asset growth is included to account for differences in risk preferences. To consider the fact that better diversified banks are assumed to be less risky, we control for diversification, measured by a diversification index (Laeven and Levine, 2007).10 We use the ratio of loan loss provisions to total assets as a measure of asset quality.

On the country level, we use a Herfindahl-Hirschman index of concentration to control for the effect of market structure. Work by Claessens and Laeven (2004) and Schaeck et al. (2009) has demonstrated that concentration is not a good proxy for competition. Rather, concentration has independent effects on bank performance. Thus, while the Boone (2008) indicator captures competition, we control for market structure with the Herfindahl-Hirschman index. Since we compare Herfindahl indices across different markets, we also include total banking system assets (log) to account for the size of the systems (Bresnahan, 1989). GDP per capita (log) and unemployment adjust our regressions for the macroeconomic environment. Finally, we include a time trend to capture the gradual nature of changes in the regulatory environment. The time trend is calculated as the current year minus the start date of the sample period.

A. Main Results

We use different estimation techniques, including two-stage least squares estimators, to adjust for endogeneity between measures of competition and stability and cluster standard errors at the bank level. Panel data models with bank-fixed effects are reported in Columns (1) and (2) in Table III. The negative sign at the 1% level for the Boone (2008) indicator supports the positive link between competition and stability. Since the indicator captures competition via a reallocation effect in more efficient banks, we interpret this result as evidence that the mechanism is running through the efficiency channel in line with the Transmission Mechanism hypothesis.

Table III. The Effect of Competition on Bank Stability
EstimatorFixed EffectsTwo-Stage Least SquaresTwo-Stage Least Squares
 (1)(2)(3)(4)(5)(6)(7)
ModelBankBank andBankBank andZ-score Components
SetupLevelMacroLevelMacro 
 VariablesVariablesVariablesVariables 
Dependent variableZ-scoreZ-scoreZ-scoreZ-scoreCapital RatioROAS.D. ROA
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Bank-Specific Variables       
Total assets (log)−0.2044***−0.2010***−0.2198***−0.2060***−0.0213***−0.0016***−0.0004***
 (−15.26)(−14.40)(−14.86)(−14.70)(−10.80)(−4.99)(−2.89)
Asset growth−0.0003−0.01080.0211*−0.00660.0005−0.00030.0002**
 (−0.04)(−1.04)(1.79)(−0.61)(0.48)(−1.32)(1.91)
Diversification index−0.5871***−0.5970***−0.5846−0.5970***−0.0494***−0.0066***−0.0016***
 (−14.21)(−14.60)(−13.79)(−14.60)(−9.35)(−6.16)(−2.86)
Loan loss provisions/Total assets−4.2818***−4.3660***−3.6263***−4.2080***−0.0308−0.2560***0.0216***
 (−8.83)(−8.82)(−7.87)(−8.59)(−0.48)(−10.54)(3.08)
Country-Specific Variables       
Herfindahl Hirschman index−0.2900***−0.2900***−0.1086−0.2380***−0.0464***−0.0067*−0.0038*
 (−3.63)(−3.55)(−1.16)(−2.73)(−3.93)(−1.89)(−1.95)
Banking system assets (log)0.0079**0.0092***0.00210.0071**0.0011***−6.49e-053.16e-05
 (2.28)(2.60)(0.57)(2.03)(2.85)(−0.62)(0.57)
GDP per capita (log, t−2) 0.4760** 0.5210**0.0421***−0.00060.0022
  (2.33) (2.46)(3.12)(−0.16)(1.28)
Unemployment (t−2) 0.0003 −0.00090.0002−0.0003***0.0001***
  (0.12) (−0.35)(1.02)(−6.22)(4.14)
(Continued )
Competition Indicator       
Boone indicator−5.0054***−5.2630***−9.1704***−6.5400***−0.2260***−0.2510***−0.0043
 (−16.64)(−16.50)(−8.30)(−10.50)(−3.74)(−14.65)(−0.45)
Time Effect       
Time trend0.0331***0.0254***0.0406***0.0269***0.0017***0.0002***−5.45e−05
 (28.96)(7.60)(16.01)(8.40)(6.65)(3.79)(−1.62)
Observations17,96517,96517,56817,56817,56817,56817,568
Number of banks3,3253,3252,9282,9282,9282,9282,928
R20.2570.2610.2300.2580.1900.1470.016
Durbin-Wu-Hausman testn/an/a98.99***10.28***1.1124.74***6.31**
Anderson testn/an/a512.801***2,843.872***803.831***803.831***803.831***
Shea's partial R2n/an/a0.0640.1760.1760.1760.176
Weak identification F-testn/an/a401.025***1,644.563***1,644.563***1,644.563***1,644.563***
Hansen J-testn/an/a2.0911.7930.4544.640**1.894

Among the control variables, we observe lower Z-scores in concentrated banking systems. This finding captures a pure effect arising from the market structure in regressions that explicitly consider competition. We believe that the inverse relationship between the Herfindahl-Hirschman index and the dependent variable suggests that banks in concentrated systems are more likely to be considered too big to fail. We also find that size and loan loss provisions are negatively related to Z-scores. The diversification index also enters negatively.

We are concerned that the Boone (2008) indicator is endogenous because more fragile institutions may “gamble for resurrection” by increasing risk via the origination of risky loans that, by itself, can be interpreted as a sign of increased competition. To address these concerns, we use a two-stage estimator and exploit the Financial Freedom Index and an interaction term of market share and loan growth as instruments for the Boone (2008) indicator.

Recall that the Financial Freedom Index provides information concerning how independent a banking system is from government control and state interference. We argue that the Financial Freedom Index is suitable to instrument the Boone (2008) indicator as less state ownership and interference directly affect competition. This conjecture is also supported in Table II that presents a strong correlation between the index and the Boone (2008) indicator. We use an interaction term of the bank's market share with loan growth since it increases whenever market share or loan growth or both increase. Such increases signal aggressively competing institutions and the rapid growth of banks, and consequently affect the competitive nature of the market since these factors impact upon the operating costs of institutions.

The indicator remains significantly negative and increases in magnitude in Columns (3) and (4) in Table III indicating a bias in our previous estimates. To verify that the two-step estimator is warranted, we perform a Durbin-Wu-Hausman test. This test rejects the exogeneity of competition, confirming the choice of a two-stage estimator. The specification tests for under-identification and over-identification imply that the model is identified and that the instruments are exogenous. We also present a weak identification test for the null hypothesis that the instruments are weak (Stock and Yogo, 2005). These tests support the idea that our instruments are strong.

B. Extensions

To better understand the driving forces behind the hypothesized mechanism from competition via efficiency to stability, we perform three additional tests. First, we focus on the components of the Z-score to establish whether the beneficial effect of competition on stability is attributable to the effects of competition on capitalization, profitability, or on the volatility of profits. Columns (5)-(7) report an inverse correlation of the indicator with all components, yet only the coefficients in the regressions with the capital ratio and ROA assume significance. Taken together, these findings suggest that competition principally drives Z-scores higher via incentives to hold higher capital ratios and through the reallocation effect for profits.

The findings concerning the capital ratio are consistent with the literature. Allen et al. (2011) propose that competition incentivizes banks to hold more capital, a prediction empirically supported by Schaeck and Cihak (2012). The negative association of the Boone (2008) indicator with profits reflects the intuition behind the indicator and is in line with the evidence of the effect of deregulation on reallocation of bank profits by Stiroh and Strahan (2003).

Moreover, we use quantile regression to allow for heterogeneous responses to competition (Koenker and Bassett, 1978) in Table IV by conditioning on bank stability. The intuition is as follows. Particularly weak banks (i.e., banks with low Z-scores) may respond in a different way to competition than stable banks. Such varying effects indicate that more than one single slope parameter is necessary to describe the correlation between competition and stability. This situation calls for the use of quantile regression as it permits inferences about the impact of regressors conditional upon the distribution of the measure of stability. We note two important differences between quantile regression and OLS. First, quantile regression provides information about the slope at different points of the dependent variable given the set of explanatory variables, while OLS provides information about the slope at different points of the explanatory variables. Additionally, quantile regression is more robust to departures from normality as linear estimators are more likely to produce inefficient estimates.

Table IV. The Effect of Competition on Bank Stability—Quantile Regression Estimates
EstimatorTwo-Stage-Quantile Regressions
 (1)(2)(3)(4)(5)
Model Setup10th25th50th75th90th
 QuantileQuantileQuantileQuantileQuantile
Dependent VariableZ-scoreZ-scoreZ-scoreZ-scoreZ-score
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Bank-Specific Variables     
Total assets (log)−0.0102*−0.0162***−0.0341***−0.0403***−0.0304***
 (−1.67)(−3.41)(−3.97)(−7.79)(−4.13)
Asset growth−0.135***−0.0937***−0.0439−0.0339−0.0523*
 (−3.45)(−3.58)(−1.10)(−1.26)(−1.73)
Diversification index−1.034***−0.959***−0.745***−0.543***−0.187
 (−12.0)(−15.0)(−6.83)(−8.26)(−1.62)
Loan loss provisions/total assets−19.86***−20.26***−20.61***−18.11***−8.469*
 (−13.4)(−13.9)(−10.3)(−6.12)(−1.65)
Country-Specific Variables     
Herfindahl Hirschman index−2.685***−2.668***−1.564***−1.346***−1.117
 (−6.34)(−6.53)(−4.65)(−2.92)(−1.27)
Banking system assets (log)−0.0180**−0.0462***−0.0566***−0.0598***−0.0660***
 (−2.22)(−6.81)(−8.71)(−6.36)(−4.34)
GDP per capita (log, t-2)−0.377***−0.379***−0.1280.309***0.579***
 (−3.54)(−4.54)(−1.27)(4.35)(4.03)
Unemployment (t-2)0.000320−0.00309−0.007710.003290.00222
 (0.035)(−0.49)(−0.98)(0.67)(0.22)
Competition Indicator     
Boone indicator−6.462***−6.783***−10.02***−12.53***−16.33***
 (−8.36)(−11.6)(−9.39)(−15.2)(−12.4)
      
Time Effect     
Time trend0.0390***0.0385***0.0446***0.0436***0.0505***
 (12.6)(18.0)(14.1)(15.7)(12.6)
Observations17,96517,96517,96517,96517,965
Number of banks3,3253,3253,3253,3253,325
F-test for equality of quantile coefficients   58.52*** 

We remain cautious about the endogeneity of the Boone (2008) indicator and use a two-stage quantile estimator (Amemiya, 1982). First, we regress the indicator on the excluded instruments, the Financial Freedom Index, and the interaction term of the bank's market share and loan growth and the exogenous variables. In the second stage, we regress the Z-score on the predicted value for the Boone (2008) indicator and the exogenous variables. Since the standard errors from the second stage are incorrect, we use a bootstrapping procedure based on 1,000 replications to correct them.

Our conjecture that the transmission mechanism from competition via the efficiency channel depends upon the stability of the banks in question is confirmed in the quantile regressions. Table IV reports the coefficients for the 10th, the 25th, the 50th, the 75th, and the 90th quantile of the distribution of the Z-score.

To illustrate the effect of a one-unit change of the Boone (2008) indicator on stability with the other covariates held constant, in Figure 2, we plot the quantile regression estimates as a solid curve. The vertical axis indicates the effect of competition, while the horizontal line represents the quantile scale. The grey area reports a 95% confidence interval for the quantile regression and the dashed line represents the OLS estimator and the confidence interval.

image

Figure 2. Quantile Regression Estimates of Boone Indicator on Z-Score

Download figure to PowerPoint

The coefficient of competition remains negative and significant across the quantiles. The quantile regressions offer additional insights.

Figure 2 highlights departures from the previous estimates of the Boone (2008) indicator at the upper and the lower tails of the distribution of the Z-score. The inference from the visual inspection is also validated when we use an F-test to determine whether the coefficients of the Boone (2008) indicator are equal across all quantiles. Our F-test rejects the null hypothesis for the equality of the coefficients. This suggests that relying on a single measure of central tendency may be insufficient to evaluate the effect of competition.

The quantile regression result highlights the fact that policymakers should consider that any competition-increasing policy may affect stability differently in the relevant banking market depending upon the health of the banks. Moreover, the increasing magnitude of the coefficient of the Boone (2008) indicator underscores banks at the lower tail of the distribution of the Z-score benefit less from competition. This is intuitive. A fragile institution is likely to have a low capital ratio, lower and more volatile profits, and is likely to operate at higher costs. Such an institution will find it harder to survive increases in competition than more efficient banks.

Additionally, Martinez-Miera and Repullo (2010), in recent work, revisit the trade-off between competition and the likelihood of bank default and predict a U-shaped relationship that arises from two competing forces. A risk-shifting effect results in lower loan rates due to competition, which, in turn, lowers the probability of loan defaults and increases bank health, and a margin effect from lower loan revenues due to lower loan rates in competitive environments. The margin effect reduces the buffer against loan losses and, consequently, increases bank risk. The risk-shifting effect dominates in monopolistic markets, whereas the margin effect dominates in competitive markets, resulting in a U-shaped relationship.

To test for the presence of such a nonlinear link, we introduce a squared term of the Boone (2008) indicator into our regressions. Irrespective of whether we use fixed effects or a two-step estimation, the squared term remains insignificant. This result does not support the prediction by Martinez-Miera and Repullo (2010), and we decide not to report these regressions to preserve space. The results are available upon request.

C. Robustness Tests

We perform robustness checks in Table V and constrain the discussion to the key results. Our first robustness test uses the aggregate ratio of nonperforming loans to total loans on the country level as a dependent variable. The nonperforming loans ratio is another regularly used measure of stability (Jimenez and Saurina, 2006). This analysis not only allows us to determine whether measurement issues drive the significant association of the Z-score with the Boone (2008) indicator, but also sheds light on the question whether systemic risk, approximated by the aggregate level of nonperforming loans, is affected by competition in the banking system. Our analysis produces a positive correlation between the Boone (2008) indicator and the level of nonperforming loans, confirming our suspicions that competition contributes to stability, also on the systemic level.

Table V. Robustness Tests for the Effect of Competition via Efficiency on Bank Stability
EstimatorTwo-Stage Least Squares
 (1)(2)(3)(4)(5)(6)(7)
Model SetupAggregateCommercialSavingsCooperativeSurvivingNonsurvivingAdjustment
 Nonperforming Loans/BanksBanksBanksBanksBanksfor Generated
 Total Loans     Regressor Problem
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

Bank-Specific Variables       
Total assets (log)0.2350***−0.3250***−0.1730***−0.1750***−0.1420***−0.2440***−0.2820***
 (2.67)(−8.75)(−7.76)(−12.8)(−6.59)(−13.62)(−9.49)
Asset growth0.0848−0.1110***0.01090.0331***−0.0218−0.0029−0.0489*
 (1.39)(−4.76)(0.78)(4.00)(−1.16)(−0.15)(−1.82)
Diversification index−1.5900***−0.7270***−0.5390***−0.4670***−0.6030***−0.58000***−0.6020***
 (−4.78)(−6.15)(−9.62)(−10.2)(−8.66)(−11.59)(−6.59)
Loan loss provisions/total assets6.2170**−3.0520***−7.7160***−3.9260***−4.8780***−4.0380***−3.3740***
 (2.29)(−2.59)(−9.65)(−8.82)(−7.35)(−6.05)(−3.17)
Country-Specific Variables       
Herfindahl Hirschman index−1.0701−0.5880***−0.09860.7910−0.2630**−0.2700**−0.4730***
 (−0.47)(−4.07)(−0.70)(1.41)(−1.96)(−2.36)(−5.79)
Banking system assets (log)−0.0421*0.02920***0.0646***0.00100.00210.0138***0.0418***
 (−1.80)(3.19)(5.68)(0.22)(0.36)(3.47)(7.11)
(Continued )
GDP per capita (log, t−2)0.38100.5670*−0.50701.8350***0.04380.8400***0.0055
 (0.12)(1.87)(−1.59)(3.70)(0.12)(3.32)(0.02)
Unemployment (t−2)0.2730***−0.0238***−0.0258***0.0179***−0.0082*0.00−0.0329***
 (5.55)(−3.41)(−4.79)(4.55)(−1.86)(0.57)(−6.80)
Competition Indicator       
Boone indicator78.8900***−9.6750***−5.5790***−10.7600***−4.9750***−7.2510***−2.4070*
 (2.68)(−3.22)(−5.99)(−11.3)(−5.06)(−8.44)(−1.91)
Time Effect       
Time trend−0.0898**0.0209***0.0357***0.0159**0.0297***0.0228***0.0319***
 (−2.21)(3.40)(7.98)(2.52)(5.99)(5.63)(7.41)
Observations11,4502,8555,6289,0856,94810,62017,568
Number of banks2,4625358451,5487722,1562,928
R20.3180.2300.3960.2870.2950.2430.352
Durbin-Wu-Hausman test210.05***8.93***0.6527.10***1.8318.560.18
Anderson corr.11.71***66.01***1,107.13***3,489.43***338.94***485.19***1,312.03***
Shea's partial R20.0100.0280.2060.2450.1940.1530.085
Weak identification F-test9.1127.00***985.60***1,174.00***1,077.00***714.70***339.40***
Hansen J-test1.794.16**0.360.617.81***0.700.13

Additionally, our previous regressions render the inclusion of bank type dummies infeasible as they are perfectly collinear with the bank-fixed effect. To examine whether different types of bank react similarly to competition, we separately rerun regressions for commercial, savings, and cooperative banks, but the results remain unaffected. Further, we test whether survivorship bias affects our findings as merger and acquisition activities can substantially alter the structure, efficiency, and performance of banks. Since our main regressions do not consider these effects, we now run the analyses separately for the surviving banks (i.e., banks that remain in the sample during the entire sample period) and for nonsurviving institutions, defined as banks that exit the market (typically via mergers and acquisitions) during the sample period. These tests also confirm the stability-enhancing effect of competition.

Moreover, we use an estimation procedure that assigns less weight to observations where the Boone (2008) indicator is estimated with greater variance to account for the fact that the indicator is derived from a regression. The key reason for this test is that in some instances, the indicator may be biased due to a low correlation between the excluded instruments and the endogenous variables. These tests confirm the inverse association of the Boone (2008) indicator with the Z-score.

In further regressions, we replicate the findings for the European sample for a set of small, single market banks operating in rural counties in the United States. This sample has the benefit that it allows a better demarcation of the geographic boundaries of the banking market to precisely establish the relevant competitors (Adams, Breevort, and Kiser, 2007) which is a limitation of the European data set. Unlike the European sample, however, the US sample is not representative due to its restriction on single market banks. In these tests, we obtain qualitatively identical inferences as the European sample (not reported).11

V. Conclusion

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

In this paper, we contribute to the important debate among academics and policy makers regarding the effects of competition on bank stability. We use a new competition measure, the Boone (2008) indicator, that analyzes the cost elasticity of performance by capturing the link between competition and efficiency. We use this measure to examine the mechanism through which competition affects stability. Underlying our approach is the notion that in banking, as in other industries, competition incentivizes banks to enhance cost efficiency as it reallocates profits from unsuccessful (inefficient) banks to successful (efficient) ones. In other words, banks with strong performance pass a market test and survive, while weak banks shrink, sell out, and exit the market.

Using a sample for major European economies prior to the recent crisis to avoid competitive distortions arising from government intervention such as public guarantees, we establish, in a preliminary test, that the new competition indicator is intuitively correlated with a variety of other characteristics of competition. Before we proceed to summarize the findings, some caveats are in order. Our investigation does not explicitly account for contagion among banks arising from the failure of inefficient institutions. The recent crisis has highlighted that banks are interconnected via credit derivatives in a way that is difficult to trace, and loss of confidence of counterparties may appear quickly in liquidity dry-ups that may not be captured by our measures of stability. These complex interlinkages might constitute alternative transmission channels between competition and stability (other than the “efficiency channel” examined in this paper). We do not include data from the recent crisis episode, and testing for such effects is beyond the scope of the present analysis. Thus, we cannot establish the relative size of the importance of such alternative mechanisms in comparison to the mechanism we focus on in our work. Such an investigation is an important area for future research.

These caveats notwithstanding, our analysis leads to relatively strong results. Our key finding is that competition robustly improves stability via the efficiency channel. This result is reinforced in a variety of sensitivity checks that account for, inter alia, endogeneity of competition, different bank types (commercial, savings, and cooperative banks), and survivorship bias. Moreover, we also find that the aggregate level of nonperforming loans, a possible proxy for systemic risk, is lower in competitive environments.

In additional analyses that focus on the factors behind the positive effect of competition on stability via efficiency, we observe that bank capital and profitability increase as a result of accelerating competition. Our final set of tests that analyze heterogeneous responses to competition indicates that fragile banks benefit less from competition.

These results have important implications for policymaking and regulation. Promoting competition has benefits for efficient operations and stability if it drives inefficient banks out of business in an orderly manner. In this context, our results support the findings reported in studies by Allen et al. (2011), Berger et al. (2009), and Schaeck et al. (2009). Our findings obtained form quantile regressions highlight the necessity of policymakers to consider how increasing the level of competition may affect the stability of the institutions in the relevant banking market. In light of intensive discussions about tightening regulation and restricting bank activities (for an overview, see Blinder, 2010), we believe caution should be exercised when redesigning regulatory policies that substantially limit the scope and scale of activities due to the potentially harmful effects on banks’ incentives to operate efficiently.

Appendix

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References

Boone Indicators

We report the estimates of the Boone (2008) indicator based on average costs with ROA as the dependent variable, adjusted for heteroskedasticity. The estimates are obtained using a two-step GMM panel data estimator with bank-fixed effects whereby we employ one-year lagged values of the explanatory variables as instruments. All regressions have considerable explanatory power. We present the Anderson canonical correlation coefficient and Shea's (1997) Partial R2. The Anderson correlation coefficient is a test of the null hypothesis that the equation is underidentified and the instruments have low relevance. Shea's (1997) measure captures the correlation between the excluded instruments and the endogenous regressor. Since all of the equations are exactly identified, no tests for overidentification can be performed.

YearAustriaBelgiumDenmarkFranceItalyGermanyLuxembourgNetherlandsSwitzerlandUnited Kingdom
 Boonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-valueBoonet-value
  1. ***Significant at the 0.01 level.

  2.  **Significant at the 0.05 level.

  3.   *Significant at the 0.10 level.

1996−0.059−2.431−0.048−3.015−0.051−3.643−0.014−1.228−0.079−13.793−0.037−15.287−0.037−6.288−0.130−2.550−0.060−4.282−0.100−7.420
1997−0.038−2.590−0.052−3.216−0.034−2.231−0.012−0.837−0.067−11.481−0.033−13.469−0.034−6.312−0.082−1.533−0.063−5.452−0.093−9.295
1998−0.037−2.304−0.055−3.781−0.030−2.244−0.019−1.626−0.058−13.082−0.029−10.543−0.039−5.079−0.067−1.459−0.069−6.381−0.090−8.477
1999−0.034−2.424−0.057−3.986−0.018−1.331−0.025−2.240−0.042−8.945−0.026−8.519−0.027−6.370−0.095−1.933−0.061−5.608−0.081−7.571
2000−0.036−3.123−0.066−4.131−0.035−2.565−0.024−2.470−0.045−11.366−0.021−6.940−0.045−7.671−0.123−2.749−0.076−6.648−0.081−7.508
2001−0.034−2.474−0.055−3.414−0.022−1.546−0.025−2.644−0.043−10.033−0.016−5.026−0.033−6.145−0.098−1.843−0.070−4.885−0.070−6.622
2002−0.028−2.183−0.059−3.426−0.024−1.725−0.020−2.310−0.039−8.236−0.015−5.409−0.030−4.440−0.059−1.517−0.063−4.525−0.066−6.041
2003−0.026−2.315−0.053−3.473−0.045−3.357−0.019−2.509−0.036−8.205−0.019−7.675−0.024−5.053−0.061−1.100−0.059−4.498−0.062−7.652
2004−0.029−2.646−0.047−3.567−0.035−2.635−0.024−3.210−0.037−9.171−0.017−7.727−0.026−6.196−0.030−0.716−0.053−4.160−0.082−5.221
2005−0.030−2.890−0.048−3.849−0.042−3.416−0.023−3.064−0.039−9.155−0.024−10.302−0.028−8.000−0.064−1.791−0.053−4.704−0.056−4.754
Observations1,074 282 480 1,631 3,145 12,670 625 97 1,642 346
R20.1267 0.5764 0.3833 0.1295 0.4339 0.3433 0.3997 0.4674 0.2287 0.3204
Anderson corr.22.063** 17.911* 188.527*** 79.951*** 730.112*** 3750.837*** 88.536*** 19.902** 74.773*** 76.845***
Shea's Partial R20.241 0.560 0.206 0.270 0.136 0.094 0.503 0.533 0.196 0.320
  1. 1

    Whenever possible, we use consolidated data to avoid double counting. We exclude Spain and Sweden as we do not have sufficient data to compute estimates for the Boone (2008) indicator. Nonetheless, the sample covers the majority of European banks in terms of their number and total assets.

  2. 2

    The reallocation effect can operate via different mechanisms, but the triggers are not relevant. The literature offers insight into sources of profit reallocations. Zarutskie (2009) demonstrates that banks respond to competition by specialization. They adjust their lending and focus on certain types of loans to lower the costs of processing and originating loans or they become better at screening particular groups of borrowers. Dick and Lehnert (2010) also find evidence that competition raises lending efficiency. A widely used approach to increase the efficiency of lending decisions is to use credit scoring models as shown by Roszbach (2004).

  3. 3

    A limitation with Boone's (2008) approach is that it does not consider competition based on service quality. If an incumbent is faced with threat of entry, it may opt for raising service quality via better qualified staff. While this increases the incumbent's costs, the cost increase may be compensated by profitability increases above the rise in cost if customers value service to the extent to which they are willing to pay for it. Theoretically, this would turn the prediction of Boone's (2008) model upside down. However, there is no evidence that competition on service quality raises profitability. Research regarding the lifting of branching restrictions and the impact on service quality and performance by Dick (2006) indicates that banks respond to competition by investing in branch networks and staff to improve service quality. She observes these investments raise costs and banks charge higher service fees supporting the notion that some customers are willing to pay for service. However, she finds no evidence that profitability increases.

  4. 4

    We acknowledge that using this proxy assumes that average costs are neither rising nor falling as output levels increase or decrease. This is in line with theory. The Boone (2008) model does assume constant marginal cost which is another weakness of the new competition indicator.

  5. 5

    The alternative approach is the production approach found in the earlier literature (Benston, 1965). Under the production approach, banks produce loans and deposits and input factors are labor and physical capital.

  6. 6

    The data for activity restrictions and the number of entry applications are taken from Barth, Caprio, and Levine (2001) and averaged for the three rounds of the survey. We use an index of activity restrictions that takes on values between one and four. It provides information about whether banks can engage in securities, real estate, and insurance activities, and whether banks can hold stakes in nonfinancial firms. Larger values indicate more restrictions.

  7. 7

    This expectation is based on the prevalent view of the literature (Barth et al., 2004). Nonetheless, it is important to acknowledge that it is possible that under certain conditions more lines of business can also facilitate collusion and reduce competition. This was documented, for example, in local US markets for savings and loans (Mester, 1987).

  8. 8

    Information regarding ownership is obtained from the data set in Barth et al. (2001) and averaged over the three waves of the survey. The Financial Freedom Index is obtained from the Heritage Foundation. It measures banking independence from government control and state interference into banking business (ranging from 0 = no freedom to 100 = maximum freedom).

  9. 9

    In a previous version of the paper, we exploit Granger-causality tests to examine intertemporal correlations between competition and efficiency using efficiency scores from a stochastic frontier model based on a Translog cost function. We also estimate regressions where we model these efficiency scores as a function of the Boone (2008) indicator and control variables. These tests suggest that increases in competition precede increases in efficiency, and that the Boone (2008) indicator is significantly negatively related to bank efficiency indicating that competition improves cost efficiency. Results are available from the authors upon request.

  10. 10

    We use a diversification index that is increasing in the degree of diversification. It is defined as: inline image.

  11. 11

    We focus on banks that operate in noncore-based statistical areas. This approach considers that banking markets are local in nature as customers obtain services from nearby providers of services (Adams et al., 2007). Moreover, the policy interest in competitive dynamics relating to antitrust considerations focuses on geographically delimited markets. Finally, such small banks are not subject to “too big to fail” policies that may distort competition. The results of these regressions are presented in Schaeck and Cihak (2010).

References

  1. Top of page
  2. Abstract
  3. I. Hypothesis Development
  4. II. Data
  5. III. Competition Expressed as a Function of Efficiency: The Boone (2008) Indicator
  6. IV. The Nexus between Competition, Efficiency, and Bank Stability
  7. V. Conclusion
  8. Appendix
  9. References
  • Adams, R.M., K.P. Breevort, and E.K. Kiser, 2007, “Who Competes with Whom? The Case of Depository Institutions,” Journal of Industrial Economics 55, 141-167.
  • Allen, F., E. Carletti, and R. Marquez, 2011, “Credit Market Competition and Capital Regulation,” Review of Financial Studies 24, 983-1018.
  • Allen, F. and D. Gale, 2004, “Competition and Financial Stability,” Journal of Money, Credit, and Banking 36, 453-480.
  • Amemiya, T., 1982, “Two Stage Least Absolute Deviations Estimators,” Econometrica 50, 689-711.
  • Barth, J.R., G. Caprio, Jr., and R. Levine, 2001, “The Regulation and Supervision of Banks Around the World – A New Database,” 1, World Bank Policy Research Working Paper # 2725.
  • Barth, J.R., G. Caprio, Jr., and R. Levine, 2004, “Bank Regulation and Supervision: Shat Works Best?Journal of Financial Intermediation 13, 205-248.
  • Beck, T., D. Coyle, M. Dewatripont, X. Freixas, and P. Seabright, 2010, “Bailing Out the Banks: Reconciling Stability and Competition. An Analysis of State-Supported Schemes for Financial Institutions,” CEPR Report, London, UK.
  • Beck, T., A. Demirgüç-Kunt, and R. Levine, 2000, “A New Database on Financial Development and Structure,” World Bank Economic Review 14, 597-605.
  • Benston, G.J., 1965, “Economies of Scale and Marginal Cost in Banking Operations,” National Banking Review June, 507-549.
  • Berger, A.N., 1995, “The Profit-Structure Relationship in Banking – Tests of Market-Power and Efficient-Structure Hypotheses,” Journal of Money, Credit and Banking 27, 404-431.
  • Berger, A.N., A. Demirgüc-Kunt, R. Levine, and J. Haubrich, 2004, “Bank Concentration and Competition: An Evolution in the Making,” Journal of Money, Credit, and Banking 36, 433-451.
  • Berger, A.N. and R. De Young, 1997, “Problem Loans and Cost Efficiency in Commercial Banks,” Journal of Banking and Finance 21, 849-870.
  • Berger, A.N. and T.H. Hannan, 1998, “The Efficiency Cost of Market Power in the Banking Industry: A Test of the “Quiet Life” and Related Hypotheses,” Review of Economics and Statistics 80, 454-465.
  • Berger, A.N., L. Klapper, and R. Turk-Ariss, 2009, “Banking Competition and Financial Stability,” Journal of Financial Services Research 35, 99-118.
  • Blinder, A.S., 2010, “It's Broke, Let's Fix It: Rethinking Financial Regulation,” International Journal of Central Banking 6, 277-330.
  • Blundell, R. and S. Bond, 1998, “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,” Journal of Econometrics 87, 115-143.
  • Boone, J., 2008, “A New Way to Measure Competition,” Economic Journal 118, 1245-1261.
  • Boone, J., R. Griffith, and R. Harrison, 2005, “Measuring Competition,” AIM Research Working Paper 22-August-2005.
  • Bouckaert, J. and H. Degryse, 2006, “Entry and Strategic Information Display in Credit Markets,” Economic Journal 116, 702-720.
  • Boyd, J. and G. de Nicolo, 2005, “The Theory of Bank Risk-Taking and Competition Revisited,” Journal of Finance 60, 1329-1343.
  • Bresnahan, T.F., 1989, “Empirical Studies of Industries with Market Power,” in R. Schmalensee and R. Willig, Eds. Handbook of Industrial Organization, Vol. 2, Amsterdam, Elsevier Science.
  • Carbo, S., D. Humphrey, J. Maudos, and P. Molyneux, 2009, “Cross-Country Comparisons of Competition and Pricing Power in European Banking,” Journal of International Money and Finance 28, 115-134.
  • Chen, X., 2007, “Banking Deregulation and Credit Risk: Evidence from the EU,” Journal of Financial Stability 2, 356-390.
  • Claessens, S. and L. Laeven, 2004, “What Drives Bank Competition? Some International Evidence,” Journal of Money, Credit, and Banking 36, 563-583.
  • Degryse, H., M. Kim, and S. Ongena, 2009, “Microeconometrics of Banking: Methods, Applications and Results, Oxford, UK, University Press.
  • Demirgüc-Kunt, A. and H. Huizinga, 2010, “Bank Activity and Funding Strategies: The Impact on Risk and Returns,” Journal of Financial Economics 98, 626-650.
  • Demsetz, H., 1973, “Industry Structure, Market Rivalry, and Public Policy,” Journal of Law and Economics 16, 1-9.
  • DeYoung, R., I. Hasan, and B. Kirchhoff, 1998, “The Impact of Out-of-State Entry on the Cost Efficiency of Local Commercial Banks,” Journal of Economics and Business 50, 191-203.
  • Dick, A., 2006, “Nationwide Branching and Its Impact on Market Structure, Quality, and Bank Performance,” Journal of Business 79, 567-592.
  • Dick, A. and A. Lehnert, 2010, “Personal Bankruptcy and Credit Market Competition,” Journal of Finance 65, 655-686.
  • Gropp, R., H. Hakenes, and I. Schnabel, 2011, “Competition, Risk-Shifting, and Public Bailout Policies,” Review of Financial Studies 24, 2084-2120.
  • Hakenes, H. and I. Schnabel, 2010, “Banks Without Parachutes: Competitive Effects of Government Bail-Out Policies,” Journal of Financial Stability 6, 156-168.
  • Hay, D.A. and G.S. Liu, 1997, “The Efficiency of Firms: What Difference Does Competition Make?Economic Journal 107, 597-617.
  • Houston, J.F., C. Lin, P. Lin, and Y. Ma, 2010, “Creditor Rights, Information Sharing, and Bank Risk Taking,” Journal of Financial Economics 96, 485-512.
  • Jayaratne, J. and P.E. Strahan, 1998, “Entry Restrictions, Industry Evolution, and Dynamic Efficiency: Evidence from Commercial Banking,” Journal of Law and Economics 41, 239-273.
  • Jimenez, G. and J. Saurina, 2006, “Credit Cycles, Credit Risk, and Prudential Regulation,” International Journal of Central Banking 2, 65-98.
  • Keeley, M.C., 1990, “Deposit Insurance, Risk and Market Power in Banking,” American Economic Review 80, 1183-1200.
  • Koenker, R.W. and G. Bassett, Jr., 1978, “Regression Quantiles,” Econometrica 46, 33-50.
  • Koetter, M., J.W. Kolari, and L. Spierdijk, 2012, “Enjoying the Quiet Life Under Deregulation? Evidence from Adjusted Lerner Indices for US Banks,” Review of Economics and Statistics 94, 462-480.
  • Laeven, L. and R. Levine, 2007, “Is There a Diversification Discount in Financial Conglomerates?Journal of Financial Economics 85, 331-367.
  • Laeven, L. and R. Levine, 2009, “Bank Governance, Regulation and Risk Taking,” Journal of Financial Economics 93, 259-275.
  • Martinez-Miera, D. and R. Repullo, 2010, “Does Competition Reduce the Risk of Bank Failure?Review of Financial Studies 23, 3638-3664.
  • Mester, L.J., 1987, “Multiple Market Contact Between Savings and Loans: Note,” Journal of Money, Credit and Banking 19, 538-549.
  • Mester, L.J., L.I. Nakamura, and M. Renault, 2007, “Transactions Accounts and Loan Monitoring,” Review of Financial Studies 20, 529-556.
  • OECD, 2010, Competition, Concentration and Stability in the Banking Sector, Paris, Organisation for Economic Co-operation and Development.
  • Olley, S.G. and A. Pakes, 1996, “The Dynamics of Productivity in the Telecommunications Industry,” Econometrica 64, 1263-1297.
  • Padoa-Schioppa, T., 2001, “Competition and Supervision in the Banking Industry: A Changing Paradigm?European Finance Review 5, 1-5.
  • Panzar, J.C. and J.N. Rosse, 1987, “Testing for Monopoly Equilibrium,” Journal of Industrial Economics 35, 443-456.
  • Roszbach, K., 2004, “Bank Lending Policy, Credit Scoring, and the Survival of Loans,” Review of Economics and Statistics 86, 946-958.
  • Schaeck, K. and M. Cihak, 2010, “Competition, Efficiency, and Soundness in Banking: An Industrial Organization Perspective,” European Banking Center Discussion Paper No. 2010-20S.
  • Schaeck, K. and M. Cihak, 2012, “Banking Competition and Capital Ratios,” European Financial Management 18, 836-866.
  • Schaeck, K., M. Cihak, and S. Wolfe, 2009, “Are Competitive Banking Systems More Stable?Journal of Money, Credit, and Banking 41, 711-734.
  • Sealey, C.W. and J.T. Lindley, 1977, “Inputs, Outputs and a Theory of Production and Cost at Depository Financial Institutions,” Journal of Finance 32, 1251-1265.
  • Shea, J., 1997, “Instrument Relevance in Multivariate Linear Models: A Simple Measure,” Review of Economics and Statistics 79, 348-352.
  • Stiroh, K.J., 2000, “Compositional Dynamics and the Performance of the US Banking Industry,” Federal Reserve Bank of New York Staff Report No. 98.
  • Stiroh, K.J. and P.E. Strahan, 2003, “Competitive Dynamics of Deregulation: Evidence from US Banking,” Journal of Money, Credit, and Banking 35, 801-828.
  • Stock, J.H. and M. Yogo, 2005, “Testing for Weak Instruments in Linear IV Regression,” in D.W. Andrews and J.H. Stock, Eds. Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, Cambridge, UK, Cambridge University Press, 80-108.
  • Tirole, J., 1989, Theory of Industrial Organization, Cambridge, MA, MIT Press.
  • van Leuvensteijn, M., J.A. Bikker, A.A.R.J.M. van Rixtel, and C.K. Sorensen, 2007, “A New Approach to Measuring Competition in the Loan Markets of the Euro Area,” European Central Bank Working Paper No. 768, Frankfurt.
  • Vives, X., 2008, “Innovation and Competitive Pressure,” Journal of Industrial Economics 56, 419-469.
  • Volcker, P.A. and J.A. Frenkel, 2009, “Financial Reform: A Framework for Financial Stability,” Group of Thirty (G30), Financial Reform Working Group.
  • Wagner, W., 2010, “Loan Market Competition and Bank Risk-Taking,” Journal of Financial Services Research 37, 71-81.
  • Wheelock, D.C. and P.W. Wilson, 1995, “Explaining Bank Failures: Deposit Insurance, Regulation, and Efficiency,” Review of Economics and Statistics 77, 689-700.
  • Zarutskie, R., 2009, “Competition and Specialization in Credit Markets,” paper presented at the 45th Bank Market Structure Conference, Federal Reserve Bank of Chicago.