Chapter

4 Comparing Europe and the United States

Author(s):
Jörg Decressin, Wim Fonteyne, and Hamid Faruqee
Published Date:
September 2007
Share
  • ShareShare
Show Summary Details
Author(s)

This chapter analyzes differences in efficiency and competition between EU and U.S. banks and puts these differences in the context of historical developments and market structures.1 Financial markets in the United States offer a natural benchmark because they are the only markets that match or exceed those of the European Union in both size and level of development and because banking markets were highly fragmented in both areas until the late 1980s. By contrast, money and capital markets, as well as regulation and supervision, have always been highly integrated in the United States but not in the European Union. Specifically, the chapter examines the causes of banking market fragmentation and the role of policies in addressing it; explores the structural differences between banks and markets in the European Union and the United States; presents evidence suggesting that EU banks are less efficient in raising revenue than their U.S. counterparts and that there is less competition among banks in the European Union than in the United States; and reflects on the reasons for the lower revenue efficiency of EU banks.

One caveat is that, when focusing on the European Union as a whole, there is a risk of glossing over various complexities and cross-country differences. Therefore, whenever particularly relevant, the chapter draws on country-specific evidence.2

The History of Bank Integration

Banking markets in both Europe and the United States fragmented during the twentieth century, but only partly for the same reasons. One important common factor was the public policy response to the financial crises of the 1930s. At that time, the consensus was that restraining competition and putting restrictions on what banks could do with deposit funding would help to preserve financial stability. In some respects, the United States went much further in this than most European countries, notably by strictly separating investment and commercial banking through the 1933 Glass-Steagall Act. Restrictions on U.S. banks’ ability to act across state lines predated the 1930s, originating in the 1927 McFadden Act, but were reinforced by the 1956 Bank Holding Company Act. The latter act also curtailed bank holding companies’ ability to own stakes in nonbank firms. In European countries, the response to the 1930s’ crises was somewhat more muted in terms of restricting banks’ activities. Universal banking continued to exist, but banks’ ability to act as holding companies was restricted in most countries.3

On both sides of the Atlantic, monetary policy also played a role in restricting banks’ freedom, as central banks started to control the allocation of credit and credit ceilings became a key instrument of monetary policy. In Europe, geographical banking market segmentation was always less artificial than in the United States, being the natural result of fundamentally different legal, prudential, licensing, and consumer protection systems across sovereign nations. However, artificial fragmentation did exist within countries due to restrictions on the activities and geographical reach of different types of banks (especially public, savings, and cooperative banks).

Evolving attitudes and market forces have begun to erode this fragmentation. Attitudes toward competition have evolved in the last few decades, partly because of developments that were driven by efforts to maximize profits and by technological progress. Policymakers and banking supervisors now emphasize the benefits of competition and market discipline, when these are supported by prudential regulation and supervision; and the academic literature supports the view that there is no simple trade-off between competition and financial stability.4

Banking market integration in the United States was a gradual bottom-up process, taking place against the backdrop of a single currency and integrated money and capital markets. The federal law that prohibited commercial banks from operating across state lines, the McFadden Act, did permit cross-border banking through multibank holding companies, with state approval. In 1978, Maine took the first step and allowed entry of bank holding companies from other states, provided these states reciprocated. By 1992, virtually all states had passed reciprocal entry laws of some sort.5 The federal Riegle-Neal Interstate Banking and Branching Efficiency Act of 1994 capped this trend, allowing national bank branches across state lines from June 1, 1997, onward. As a result, between the mid-1970s and the mid-1990s, the ratio of a typical state bank’s assets held by an out-of-state bank holding company climbed from 10 percent to over 60 percent.6 In addition, the Gramm-Leach-Bliley Act of 1999 ended the separation of commercial and investment banking activities. Importantly, U.S. money and capital markets had been highly integrated all along, which spurred integration of the banking market. This is different in contemporary Europe.

Structural Characteristics

Banks and Capital Markets

Banks and capital markets offer different advantages, and an extensive literature discusses the relative merits of each system without reaching definitive conclusions.7 One key argument in favor of banks is that their long-term relationships with firms help overcome the inefficiencies related to adverse selection and moral hazard (Stiglitz and Weiss, 1981). Hence banks may be able to intermediate under more difficult information and legal environments and may be better able to smooth intertemporal risk. Financial markets, by contrast, are considered to provide better cross-sectional risk sharing. Also, markets generate more information and thus are seen by some to better fit advanced economies, as these economies exploit new production possibilities rather than catch up with existing frontiers (see, for example, Boot and Thakor, 1997). However, markets are more volatile and work well only under a complex legal and regulatory infrastructure.

The European Union’s financial system is often thought of as being bank based and the U.S. system as being market based, but this characterization is not particularly relevant. (In this chapter, EU banks can stand for either euro-area banks or banks of the 15 old EU member states.) Relative to their respective economies, EU capital markets are much smaller and bank balance sheets much larger than those in the United States (Table 4.1). However, in a typical year, EU banks lend less to the nonfinancial private sector than U.S. banks and lend only slightly more to firms. The reason U.S. banks hold smaller balance sheets is that they make extensive use of securitization, which took off during the 1980s with the development of collateralized mortgage obligations (CMOs). U.S. banks now securitize most of their standardized loans, in particular household loans and certain types of business lending. As a result, bank loans to the nonfinancial private sector in the United States do not necessarily show up on banks’ balance sheets. Because the development of capital markets has also facilitated disintermediation in corporate funding, what is left on the balance sheets of U.S. banks tends to be loans that are less well suited to standardization, notably business loans collateralized by real estate.8 In general, the trend in both the United States and the European Union is for banks to place risk in financial markets, keeping on their balance sheets only those risks for which they enjoy a particular comparative advantage. However, U.S. banks are much further along this process than their EU counterparts.

Table 4.1.Euro Area and United States: Banks and Markets, 2004(In percent of nominal GDP)
Euro AreaUnited States
Bonds123149
Government5837
Banks1,2486
Nonbank financial institutions1027
Nonfinancial corporations825
Agencies053
Other00
Equity53147
Bank assets120892
Bank loans to nonfinancial private sector92118
Firms4230
Households5087
Bank loans to general government110
Bank loans to nonbank financial corporates8
Memorandum item:
Asset-backed securities (stock)59
Asset-backed securities (new issuance)4
Sources: U.S. Federal Reserve; European Central Bank; and European Securitization Forum.

For euro area, including Eurosystem. For United States, commercial banks only.

From consolidated balance sheet of euro area Monetary Financial Institutions.

Sources: U.S. Federal Reserve; European Central Bank; and European Securitization Forum.

For euro area, including Eurosystem. For United States, commercial banks only.

From consolidated balance sheet of euro area Monetary Financial Institutions.

Data on employment and lending paint a similar picture with respect to the relative roles of banks in the United States and the European Union.9 Notwithstanding the large difference in balance sheet size, EU and U.S. per capita banking sector employment is similar (Table 4.2). Where the EU banks lead is in the accumulation of bricks and mortar (branches), and where they lag is in equity markets. The European Union is often considered to be “overbanked” and in need of consolidation—not least owing to the large number of credit institutions in Germany, France, and Italy—but the per capita number of banks and banking sector employment in the European Union is similar or lower than in the United States. Moreover, trends have diverged in recent years: the number of banks has fallen significantly in both the United States and the European Union, but more so in the latter; U.S. banks have added branches, while EU banks have closed thousands of them; and EU banks have not followed the upward trend in employment observed among their U.S. peers. Notably, over the past three decades, the number of banks was virtually halved in the United States but the number of branches doubled.10

Table 4.2.EU-15 and U.S. Banks: Structural Indicators, 1997–2003
1997199819992000200120022003
Number of banksU.S.10,92310,46410,2239,9049,6159,3549,182
EU-159,6249,3378,8728,4338,0847,7517,444
Per million peopleU.S.41393736343232
EU-1526252422212019
Number of bank branchesU.S.73,36674,79876,57676,56778,12378,50379,756
EU-15202,092200,319198,973197,513194,349190,279186,009
Per million peopleU.S.273276280278274272274
EU-15539534529523512500487
Number of bank employeesU.S.11,784,4821,863,9101,901,5961,914,5171,967,6222,017,6452,045,976
EU-152,820,5632,821,8452,862,3382,823,7842,774,175
Per million peopleU.S.6,6516,8836,9606,9466,8896,9947,023
EU-157,4937,4677,5437,4177,264
Banking system assets2U.S.73757476788082
(In percent of GDP)EU-15184185195203213211214
Sources: European Central Bank; Federal Deposit Insurance Corporation; World Economic Outlook database; and Pilloff (2004).

In full-time equivalents; raw numbers are likely to be higher.

The data here have a slightly different coverage from those in Table 4.1.

Sources: European Central Bank; Federal Deposit Insurance Corporation; World Economic Outlook database; and Pilloff (2004).

In full-time equivalents; raw numbers are likely to be higher.

The data here have a slightly different coverage from those in Table 4.1.

Bank Performance

On various measures of profitability and efficiency, EU banks under-perform their U.S. counterparts. Data for 2003 (Table 4.3) suggest that the pretax return on assets (ROA) of EU banks was only one-third that of U.S. banks. Reflecting higher leverage at EU banks, the gap with respect to return on equity (ROE) is considerably smaller but still quite significant. While the cost ratios of EU banks are lower, the revenue ratios fall short of those of U.S. banks by an even wider margin, with interest and other revenues contributing in similar proportions. Cross-sectional data on the top 100 EU and U.S. banks paint a similar picture (Table 4.4): not only does the median EU bank appear less profitable and less well capitalized, the same holds for the weakest decile of banks in the sample. Also, these data suggest that the relation between size and performance is unclear, as evidenced by a comparison of larger (top 50) and smaller (bottom 50) banks in the sample.11 The performance gap between EU and U.S. banks opened in the 1990s and appears to reflect a trend rather than any cyclical developments (Figure 4.1).

Table 4.3.EU-15 and U.S. Banks: Indicators of Profitability and Efficiency, 2003(In percent of assets, unless otherwise noted)
EU-15U.S.
Net interest income1.43.3
Net noninterest income1.02.5
Total income2.45.8
Staff costs0.91.4
Other costs0.61.9
Total costs1.43.3
Operating profits0.92.5
Profits, before tax0.62.1
Profits, after tax0.41.4
Return on equity (In percent of Tier 1)9.915.3
Equity-to-asset ratio4.29.2
Risk-based Tier 1 ratio (In percent)8.810.1
Overall solvency ratio (In percent)12.412.7
Table 4.4.EU-15 and U.S. Banking Sector Indicators, 1997–2003(In percent, unless otherwise noted)
EU BanksU.S. Banks
AllSmallLargeAllSmallLarge
Return on (average) assets (ROA)
10th percentile0.080.050.100.400.440.37
Median0.460.420.471.221.261.18
Return on (average) equity (ROE)
10th percentile2.641.943.747.107.866.82
Median11.4211.1011.6114.8414.1815.43
Tier-1 ratio
10th percentile5.605.685.527.307.706.90
Median7.407.507.409.459.808.50
Equity ratio
10th percentile2.011.792.414.735.784.09
Median3.874.103.808.148.517.66
Revenue1,2
10th percentile4.194.074.335.155.384.84
Median5.925.676.088.078.177.92
Cost1,3
90th percentile7.577.347.6510.4010.499.94
Median5.305.195.526.376.426.28
Operating profit1
10th percentile0.090.070.110.220.220.22
Median0.560.510.601.251.461.06
Operating expenditure1
90th percentile3.033.342.865.305.824.41
Median1.691.581.752.672.732.48
Personnel expenditure1
90th percentile1.401.461.342.162.052.18
Median0.780.690.841.311.311.32
Other revenue1
Median1.040.991.111.591.421.80
Source: BankScope database; and IMF staff calculations.

In percent of assets.

Operating revenue excluding interest expenses.

Operating costs plus interest expenses.

Sum of average return on assets and Tier-1 ratio divided by variance of average return on assets.

Source: BankScope database; and IMF staff calculations.

In percent of assets.

Operating revenue excluding interest expenses.

Operating costs plus interest expenses.

Sum of average return on assets and Tier-1 ratio divided by variance of average return on assets.

Figure 4.1.Bank Profitability and the Economic Cycle

(In percent, unless otherwise noted)

Sources: European Central Bank; OECD, Bank Profitability (2002); and Federal Deposit Insurance Corporation.

Note: Data for the European Union from 1988 to 2001 refer only to Germany, France, Italy, United Kingdom, and Spain. For 2002 and 2003, data refer to the EU-15.

EU banks appear to engage in less risky activities than U.S. banks, allowing them to be more highly leveraged. This can be gleaned from the relation between the (unweighted) equity-to-asset ratio and the risk-weighted regulatory solvency ratio. The former is more than twice as high for U.S. banks, but the latter is about the same (Table 4.3). In other words, for regulatory purposes, the assets (and off-balance-sheet activities) of EU banks carry lower risk weights, suggesting that EU banks engage in less risky activities—for example, they engage less in subprime lending. This lower riskiness explains in part the higher leverage of EU banks and their relatively lower ROA compared with U.S. banks.

Productive Efficiency

The lower profitability of EU banks could be related to lower efficiency or to other factors. Specifically, EU banks could be less profitable because (for various reasons) they face less pressure to use their inputs efficiently, that is, they can tolerate lower X-efficiency. However, simple comparisons of profitability, revenue, and cost indicators fail to provide enough information to judge the operational efficiency of EU banks relative to U.S. banks. For example, EU banks may have faced higher labor costs and a less favorable yield curve than their U.S. counterparts, causing lower profits notwithstanding an efficient use of inputs. In other words, to judge efficiency, it is important to hold constant various input costs, which requires estimating revenue and cost functions. Furthermore, differences in business models and risks must be considered as well. To better gauge the relative efficiency of EU and U.S. banks, this section follows an approach that has been widely used in the literature to estimate X-efficiency.12

Gauging productive efficiency requires that an assumption be made about banks’ activities. According to the “intermediation approach,” which is followed here, banks intermediate financial services using labor and capital as inputs, with the values of loans and investments used as output measures. Given that labor and capital are the inputs, operating costs plus interest expenses are the relevant cost measures. The relevant revenue measure is operating revenue, excluding interest expenses.

A stochastic “best practices” frontier is a useful tool to gauge banks’ efficiency. This approach specifies the functional form of the efficient frontier as a translog cost or revenue function to investigate, respectively, the efficiency of cost control and revenue generation:13

In the equation above, lowercase letters denote natural logarithms. Each bank i produces two outputs y (loans and other earning assets) and relies on three inputs with prices p (labor, interest expenses, and other operating costs). In addition, the equation includes a set of exogenous variables Z, two time dummies D (which proxy for changes in the macroeconomic environment), and a constant α. The dependent variable x denotes either operating revenue, excluding interest expenses, or operating cost plus interest expenses for bank i in year t; and the dummy βEU for European banks measures their relative management efficiency on both accounts.

For a fair comparison of efficiency, an effort needs to be made to hold constant differences in banks’ business models. To do so, a number of exogenous variables Z are added to the regression equation, comprising the loan-to-asset (L/A), deposit-to-liability (D/L), asset-to-employee (A/E), and equity-to-asset (C/A) ratios, depending on the specific regression. These variables attempt to capture a number of aspects of a bank’s business model: (1) the riskiness of a bank’s assets is reflected in its capital requirements, and therefore its equity-to-asset ratio; (2) differences in sources of funding are reflected in the deposit-to-liability ratio; (3) a focus on classic intermediation versus fee-earning and financial market activities is reflected in the loan-to-asset ratio; and (4) the relative importance of on-balance-sheet assets versus activities that rely on rapid asset turnover and off-balance-sheet items is reflected in the employee-to-assets ratio. The regressions also distinguish between the 50 largest and the 50 smallest banks, as size may be a proxy for business model features not captured by the Z variables. Nonetheless, the exogenous Z variables and the splitting of the sample clearly cannot perfectly capture the differences in business models of banks, including often unobservable risks.14

The EU sample of banks is more homogenous than the U.S. sample, and the EU banks tend to hold more assets. The BankScope-sourced data sample comprises the 100 largest banks in the European Union and the United States, respectively, for 1997, 2000, and 2003. It is difficult to put an exact number on the market share of these 100 banks in each area but it exceeds 50 percent, probably by a substantial margin. The combined assets of the top 50 banks (“large” banks) are about four (EU) and seven (U.S.) times as large as those of the bottom 50 banks (“small” banks). While the ratio of EU to U.S. median assets equals 3.2, the same ratio for median employment only reaches 1.5, pointing to a more intensive use of assets and more important off-balance-sheet activities among U.S. banks.

Ordinary least squares (OLS) estimates of the efficient frontier suggest that EU banks are less efficient in generating revenue, while costs appear to be well behaved.15 On average, EU banks exhibit costs that are 5 percent lower than their U.S. counterparts, regardless of the regression specification (Table 4.5). However, they also generate up to about 18 percent less revenue, with the gap falling to about 12 percent when the equity-to-assets ratio is included in the regression as a proxy for the riskiness of assets (Table 4.6).16 Further differentiating between the large and small banks cuts the revenue efficiency gap of EU banks to some 7 percent and cuts the cost advantage to about 3 percent.

Table 4.5.Measures of Cost Efficiency: EU Banks Compared with U.S. Banks, 1997–2003
EquationEfficiency CoefficientAdj. R2Degrees of Freedom
All banks
No exogenous variable–4.81***0.99423
L/A, D/L, A/E ratios added–4.66***0.99414
L/A, D/L, A/E, C/A ratios added–5.21***0.99413
Specification 4–5.70***0.99411
Large banks
No exogenous variable–4.01***0.99182
L/A, D/L, A/E ratios added–4.37***0.99174
L/A, D/L, A/E, C/A ratios added–3.99**0.99173
Specification 4–4.37**0.99171
Small banks
No exogenous variable–2.130.99223
L/A, D/L, A/E ratios added–2.610.99219
L/A, D/L, A/E, C/A ratios added–2.640.99218
Specification 4–3.110.99217
Notes: Dependent variable is the log of operating expense plus interest expense. Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Notes: Dependent variable is the log of operating expense plus interest expense. Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Table 4.6.Measures of Revenue Efficiency: EU Banks Compared with U.S. Banks, 1997–2003
EquationEfficiency CoefficientAdj. R2Degrees of Freedom
All banks
No exogenous variable–18.29***0.99417
L/A, D/L, A/E ratios added–18.02***0.99408
L/A, D/L, A/E, C/A ratios added–12.39***0.99407
Large banks
No exogenous variable–15.13***0.99176
L/A, D/L, A/E ratios added–16.36***0.99168
L/A, D/L, A/E, C/A ratios added–7.70***0.99167
Small banks
No exogenous variable–11.42***0.98217
L/A, D/L, A/E ratios added–11.19***0.98213
L/A, D/L, A/E, C/A ratios added–6.43*0.98212
Sources: Fitch IBCA database; and IMF staff calculations.Notes: Dependent variable is the log of operating income plus interest expense. Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Sources: Fitch IBCA database; and IMF staff calculations.Notes: Dependent variable is the log of operating income plus interest expense. Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Overall, the evidence suggests that differences in business models explain a substantial part of the relative revenue gap between EU and U.S. banks. The remaining gap could reflect lower X-efficiency. Alternatively, the gap may be due to factors not considered by the explanatory variables, such as missing capital markets that hinder more efficient intermediation, differences in risk that are not captured by differences in equity ratios, or greater competition in Europe, a factor that is explored in the next section.

Competition

Measures of concentration can be misleading when gauging competition. Concentration and competition tend to be inversely related, although this finding does not receive unambiguous support in the literature, which stresses the importance of contestability.17 One simple concentration measure, the per capita number of banks, would suggest that there is more competition in the U.S. market than in the EU (see Table 4.2). However, the number of banks in itself does not provide information about the effective degree of competition among the banks. In Europe as well as the United States, markets are segmented by geography and other factors. Many banks also belong to larger groups (such as a bank holding company) that should be considered monolithic when analyzing competition. In many EU countries, a majority of licensed banks are small cooperative and savings banks that belong to networks within which there is no competition. More sophisticated concentration indicators cannot fully overcome these fundamental challenges, nor can they gauge the impact of potential competition in contestable markets.

The relationship between a bank’s costs and its revenues provides better information on competition than standard indicators of concentration. The underlying idea is that the reaction of a bank to an increase in its cost base depends on the degree of competition it faces, and therefore that the competitiveness of a market can be derived indirectly from this reaction. The advantage of this indirect approach is twofold: it does not require microdata on the prices of comparable services, and there is no need to specify a geographic market. The Panzar and Rosse (1987) H-statistic is designed to capture this relation between costs and revenues, and is given by

where R = R (d, c, w) denotes a bank’s revenue as function of a vector of input prices w, as well as exogenous variables that shift demand d or cost c. A number of standard assumptions need to be satisfied for the H-statistic to be useful, including (1) profit maximization; (2) homo-thetic production functions; (3) exogenous factor prices; (4) an elasticity of demand that rises with the number of rivals in the market; and (5) a market that is in long-run equilibrium. Notice that these conditions can potentially cause problems, notwithstanding the widespread assumption in the literature that they are satisfied. Since the analysis here focuses on the top 100 banks in each area, homotheticity should not be a major issue because these banks are fairly large and returns to scale are seen as running out at smaller levels.18 Long-run equilibrium might be a different matter, however, given the rapid pace of change in the financial services industry. Notice that under all five conditions,

  • H ≤ 0 for a monopoly market. Any increase in cost prompts the monopolist to cut back output, which leads to a loss in revenue—the relation between cost and revenue is negative.19

  • 0 < H < 1 for a market characterized by monopolistic competition. An increase in a bank’s costs prompts an increase in prices but revenues do not rise one for one, as the bank’s demand curve slopes downward. Notice that a larger H-statistic implies a more elastic demand curve and thus less market power (Vesala, 1995).

  • H = 1 for perfect competition. If the market is perfectly competitive then there must be free entry and exit, which sets the price equal to minimum average cost; thus, any increase in cost must be matched one-for-one by revenue.

Implementing the Panzar-Rosse method also requires that an assumption be made about banks’ activities. As in the previous section on productive efficiency, the intermediation approach is followed here. Accordingly, the following regression is run:

where the subscripts i and t denote bank i at time t; rev denotes operating revenue, excluding interest expenses; pers_exp denotes personnel expenditure divided by employment; int_exp denotes interest expenditure divided by liabilities; and oth_exp is other expenditure divided by assets. These variables, including total assets, are in natural logarithms. The H-statistic is given by: H = β1 + β2 + β3. The exogenous variables Z are the same as those used previously. The time dummies D proxy for changes in the macroeconomic environment.

The key findings are that large banks are equally competitive in the United States and the European Union, but small EU banks behave less competitively relative to both large EU banks and to small U.S. banks. The estimate for the H-statistic for the full sample of EU banks is about 0.5, while that for U.S. banks is about 0.7 (Table 4.7). The confidence intervals permit the rejection of the hypotheses of pure monopoly or perfect competition in both cases, suggesting that monopolistic competition prevails.20 Standard test statistics21 point to similar competition among large EU and U.S. banks (Table 4.8). However, the smaller EU banks appear to behave significantly more monopolistically than their U.S. counterparts. Furthermore, at the 10 percent significance level, large EU banks exhibit a higher H-statistic (implying a more competitive environment) than their smaller counterparts (Table 4.9), whereas the reverse appears to be the case in the United States.

Table 4.7.Measures of Competition for EU and U.S. Banks (a), 1997–2003
Confidence Region
EquationH-StatisticLower boundUpper bound
All banksEUU.S.EUU.S.EUU.S.
No exogenous variable0.500.730.430.640.580.82
L/A, D/L, A/E ratios added0.530.710.460.600.600.82
L/A, D/L, A/E, C/A ratios added0.540.680.480.580.610.79
Large banks
No exogenous variable0.640.710.520.580.760.85
L/A, D/L, A/E ratios added0.650.600.520.430.770.77
L/A, D/L, A/E, C/A ratios added0.680.500.570.320.800.68
Small banks
No exogenous variable0.540.840.420.690.670.98
L/A, D/L, A/E ratios added0.530.830.410.680.650.99
L/A, D/L, A/E, C/A ratios added0.630.830.510.680.740.98
Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.
Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.
Table 4.8.Measures of Competition for EU and U.S. Banks (b), 1997–2003
EquationEU Banks’ H-Statistic

Relative to U.S. Banks
Confidence Region
Lower boundUpper bound
All banks
No exogenous variable–0.23***–0.34–0.13
L/A, D/L, A/E ratios added–0.18***–0.31–0.05
L/A, D/L, A/E, C/A ratios added–0.14**–0.26–0.02
Large banks
No exogenous variable–0.08–0.270.10
L/A, D/L, A/E ratios added0.04–0.170.25
L/A, D/L, A/E, C/A ratios added0.18*–0.030.39
Small banks
No exogenous variable–0.34***–0.53–0.16
L/A, D/L, A/E ratios added–0.34***–0.53–0.15
L/A, D/L, A/E, C/A ratios added–0.23***–0.42–0.04
Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly different from zero at the 1 percent level.

denotes that the coefficient is significantly different from zero at the 5 percent level.

denotes that the coefficient is significantly different from zero at the 10 percent level.

Table 4.9.Measures of Competition for Small EU Banks Compared with Large EU Banks, 1997–2003
EquationSmall Banks’

H-Statistic

Relative to U.S. Banks
Confidence Region
Lower boundUpper bound
No exogenous variable–0.12*–0.290.05
L/A, D/L, A/E ratios added–0.13*–0.300.04
L/A, D/L, A/E, C/A ratios added–0.07–0.230.09
Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly negative at the 10 percent level.

Sources: BankScope database; and IMF staff calculations.Notes: Confidence region is two standard errors wide on each side of point estimate. L/A is loan-to-asset ratio; D/L is deposit-to-liability ratio; A/E is asset-to-employee ratio; and C/A is equity-to-asset ratio.

denotes that the coefficient is significantly negative at the 10 percent level.

The results suggest that the smaller, less internationally oriented banks operate under more sheltered conditions in the European Union. The literature provides further evidence on this. Bikker and Haaf (2002), for example, use data on 23 countries and find that competition is weaker among small banks, which operate in local markets, than among large banks, which operate predominantly in international markets, while medium-size banks take an intermediate position. Furthermore, Guevara, Maudos, and Perez (2005) show that the problems related to a lack of competition in European banking markets are more pronounced in the retail sector, suggesting that there continue to be barriers to entry.

Losses related to market power have been large in the EU banking system. Guevara and Maudos (2004, 2005), for example, estimate the welfare losses in the European banking system related to excessive market power to be the equivalent of between 1½ and 2½ percent of GDP. Evidence in the literature suggests that subjecting banks to more competition may have beneficial effects for household and firms.22

Efficiency, Competition, and Financial Market Structure

The empirical evidence presented here is not consistent with the hypothesis that competition is the culprit for lower bank revenues in Europe. Perfect competition among banks does not appear to be the rule in either the European Union or the United States. However, judging by per capita employment in the banking sector, the number of banks per capita, competition from other sources of funds, and the results of a more sophisticated technique to gauge competition (the H-statistic), this analysis indicates that EU banks—particularly the relatively smaller, more domestically oriented ones—operate in a more sheltered environment.

Various other factors could explain the lower revenue efficiency of EU banks. These factors include

  • Less financial innovation. Lower competitive pressure may reduce the incentives for banks to introduce new financial services and for markets to develop new financial instruments.

  • Less bank turnover, particularly less market exit. Lower market exit not only causes excess capacity, as argued by many, but also leads to an inefficient use of existing capacity. Controlled market exit of weaker players goes hand in hand with more competition in fostering efficient bank business.23

  • More stringent laws and regulations. In many countries, legal or regulatory obstacles hinder the development and delivery of a broad range of mortgage products. Similarly, usury laws might inhibit the emergence of a broader market for consumer credit, and tax laws might discourage securitization, as was the case until recently in Germany.

Box 4.1.Securitization and the Integration of Local Housing Markets

The integration of banking markets and securitization both played a role in reducing the divergence of house price increases across the nine Office of Federal Housing Enterprise Oversight (OFHEO) regions in the United States.

The integration of state banking markets is evidenced by the increase in the weighted average of interstate asset ratios (the percent of bank assets held by out-of-state bank holding companies). Morgan, Rime, and Strahan (2004) show that interstate banking has made state business cycles smaller.

Concomitantly, securitization of mortgage loans rose rapidly and the share of deposits that fund mortgages fell significantly—from over 70 percent to less than 40 percent—as shown by Schnure (2005). He establishes a significant negative relation between the share of securitized mortgages, on the one hand, and the cross-sectional standard deviation of increases in the OFHEO house price indices (for nine regions), on the other hand.

U.S. Mortgage Market Structure

(In percent of total mortgages)

Source: Schnure (2005).

Note: GSEs = Government-sponsored enterprises.

Interestingly, notwithstanding much lower interregional migration flows and resulting relative price adjustments, house prices appear to diverge more across the EU countries than across the nine OFHEO regions in the United States. Nonetheless, the BIS data show falling divergences over time.

Standard Deviation of U.S. Real Housing Price Indices

(In percent, based on four-quarter changes)

Standard Deviation of EU Real Housing Price Indices

(In percent, based on year-on-year changes)

Source: IMF staff calculations.

  • Less scope for reaching a broad customer base. Differences in the legal and regulatory environment across countries hamper both the provision of financial services across national borders and the potential for cross-border bank mergers and acquisitions. This may reduce the payoffs to innovation and thus revenues.

  • More public sector intervention. Intervention can be explicit, through ownership of credit institutions, or implicit, through influence on the decision-making processes of major banks, notably those that were formerly publicly owned. Explicit intervention is still fairly widespread in the European Union, although much less so among the sample of banks considered here.

  • Competition from non-profit-maximizing financial institutions. The European Union not only has more publicly owned banks, it also has a large number of cooperative banks, which do not necessarily have profit maximization and innovation as their primary objectives. While potentially beneficial to consumers, competition from such institutions may reduce the returns banks can achieve, in particular through retail activities.

EU banks’ revenue efficiency may also be constrained by missing or less developed markets. As discussed above, EU banks keep financing activities on their balance sheets that U.S. banks typically refinance in or direct toward financial markets. These activities, which generally require less specialized banking knowledge and involve less risk, can be expected to generate lower income streams and require less capital. This interpretation is consistent with the results in the section here on productive efficiency, which show that the revenue gap between U.S. and EU banks shrinks considerably when holding constant for the lower capital held by the latter. An improved ability to securitize also permits a higher turnover of assets, thus allowing banks to earn more fee income for given levels of assets and capital, which may explain part of the remaining gap. Be it as it may, EU banks hold larger volumes of assets that could be sold off in markets in the future, including mortgages and large corporate loans.24 In sum, what is captured here as a lower revenue efficiency might in part be a reflection of missing complementary capital markets in Europe.

The interplay among government intervention, market forces, and the regulation of financial activity might be at the root of this relative under-development of markets.25 In general, government intervention in Europe has been relatively less market- and more bank-friendly than in the United States. Many European authorities entered credit markets directly, via ownership of banks, and therefore had an incentive to favor banks over markets. EU capital markets were fragmented due to legal and regulatory differences between countries. In the United States, by contrast, public intervention fragmented the banking system, but capital markets could serve the entire country. The U.S. public sector also played a crucial role in developing securitization, notably with the introduction of mortgage-backed securities by Ginnie Mae in the 1960s and the not-uncontroversial role more recently of Fannie Mae and Freddie Mac in the market for such instruments. There is evidence that securitization, alongside the integration of state banking markets, has played an important role in integrating regional housing markets in the United States and better distributing housing-market-related risks (Box 4.1).

Conclusions

Europe’s financial system presently appears to offer less scope for banks and markets to leverage each other’s comparative advantages. Despite major progress, especially in recent years, EU capital markets remain smaller, less developed, and more fragmented than those in the United States. Banking systems, by contrast, are larger in the European Union in terms of assets, but other indicators suggest that their economic importance is about the same as that of banks in the United States. However, EU banks have different business models, as they engage less in placing risks in financial markets. It is questionable that these differences in business models result from positive choices. In many EU countries, the types of markets that banks would need to pursue different business models, in particular markets for securitization, are only in their infancy.

This chapter is based on Decressin and Kudela (2005).

For example, see Allen and Gale (2000a), as well as the many references therein.

Germany was an exception in continuing to allow banks to hold large equity stakes in nonbank companies.

For a succinct review of changing attitudes, see Padoa-Schioppa (2001).

More recent data cannot be produced because, since the mid-1990s, holding companies have been able to consolidate their assets at their headquarters.

For a survey, see Allen and Gale (2000a). For a discussion of the macroeconomic implications of the structure of financial systems, see International Monetary Fund (2006, Chapter IV).

See Samolyk (2004) for a comprehensive review.

Data for 1948–2001 point to a broadly stable employment share of credit agencies in the United States (Samolyk, 2004).

In the empirical literature on banking, the label “small” is typically reserved for banks that hold many fewer assets than those ranked between 50 and 100.

For a survey of bank efficiency studies based on parametric and nonparametric frontier approaches, see Berger and Humphrey (1997).

More fundamentally, the stochastic frontier has further limits when applied to banks, notwithstanding its wide use in the banking literature. Specifically, it relies on a traditional production function that is, obviously, less well suited to modern financial institutions than to, say, a typical manufacturing firm.

In estimating the translog cost function, the standard restrictions are imposed (see, for example, Johnston, 1984). Notice that in theory the error term should have a skewed, non-normal distribution. But in practice studies have found that using OLS does not make much difference, partly because the skewness is limited, which is the case here too.

Opposite findings for cost and revenue efficiency are not unusual. See, for example, Maudos and others (2002).

For a literature review, see Northcott (2004).

If the cutback in output did not lead to a loss in revenue, then the monopolist would not be acting to maximize profits.

These results are in line with the findings of Brunner and others (2004) for EU countries, as well as with those of De Bandt and Davis (2000) and Bikker and Haaf (2002).

Making use of interactive slope dummies in the regressions.

During the 1980s, U.S. banks failed in numbers not seen since the Great Depression. The total number of federally insured commercial and savings banks that were closed or received assistance from the Federal Deposit Insurance Corporation (FDIC) reached 1,617 during 1980–94. These numbers do not include failed savings and loan associations (S&Ls). See Hanc (1998) for further information. Stiroh (2002) emphasizes that the dynamic reallocation effects—entry and exit—increased the U.S. banking industry’s return on equity by several percentage points in the late 1980s.

In some European countries, notably Germany, banks rely more on (on-balance-sheet) covered bonds as a source of funding.

For a review of developments in the United States, see DeYoung, Hunter, and Udell (2003).

    Other Resources Citing This Publication