Italy—Assessing Competition and Efficiency in the Banking System

Contributor Notes

The paper assesses the degree of banking competition and efficiency in Italy?over time as well as compared to that in other countries, such as France, Germany, Spain, the United Kingdom, and the United States. The paper finds competition in the Italian banking sector has intensified in loan and deposit markets in recent years, but banks still operate in a highcost, high-income system, particularly with respect to retail/services, and efficiency gains have yet to fully materialize. The degree of competition falls within the range of estimates for a set of comparator countries. Greater contestability should act as a powerful force to drive banks to become more competitive and efficient. Competition policy will also continue to be an important consideration, both in enforcing Italy's antitrust laws and in ensuring that the procedures for dealing with weak banks and other merger and acquisition reviews focus on stability and competition objectives.

Abstract

The paper assesses the degree of banking competition and efficiency in Italy?over time as well as compared to that in other countries, such as France, Germany, Spain, the United Kingdom, and the United States. The paper finds competition in the Italian banking sector has intensified in loan and deposit markets in recent years, but banks still operate in a highcost, high-income system, particularly with respect to retail/services, and efficiency gains have yet to fully materialize. The degree of competition falls within the range of estimates for a set of comparator countries. Greater contestability should act as a powerful force to drive banks to become more competitive and efficient. Competition policy will also continue to be an important consideration, both in enforcing Italy's antitrust laws and in ensuring that the procedures for dealing with weak banks and other merger and acquisition reviews focus on stability and competition objectives.

Italy—Assessing Competition and Efficiency in the Banking System1

A. Introduction and Key Findings

1. The Italian banking system has been subject to deep structural transformation in the last two decades. Consolidation and privatization have permitted economies of scale in the production and distribution of services and increased risk diversification. These forces have led to lower costs and, undoubtedly, higher efficiency. However, to ensure that lower costs are passed through to households and firms, greater efficiency must be accompanied by a similar strengthening in the competitive environment in the banking sector.

2. This paper assesses the degree of banking competition and efficiency in Italy─over time as well as compared to that in other countries. Given the inherent difficulty of assessing competition from a single perspective, it relies on five main approaches: (i) indicators based on market structure, such as various concentration measures (Section B); (ii) contestability and cost indicators, including foreign bank ownership, bank retail prices and switching costs (Section C); (iii) profitability indicators (Section D); (iv) empirical efficiency estimates based on a panel of individual banks (Section E); and (v) market power indicators, such as Lerner and Panzar-Rosse indices (Section F). Whenever possible, we assess competition on an individual country basis and across time.

3. The paper finds competition in the Italian banking sector has intensified in loan and deposit markets in recent years, but banks still operate in a high-cost, high-income system, particularly with respect to retail/services, and efficiency gains have yet to fully materialize. The paper also finds the degree of competition falls within the range of estimates for a set of comparator countries. Cross-country indicators─both based on profit margins as well as on revenue elasticity─suggest the existence of monopolistic competition, as in other comparator banking sectors. However, there also are indications that competition has not been fully reflected in the pricing of services provided. More specifically, Italian banks incur significantly higher expenditures than other European banks and are only marginally more effective in generating higher revenue. These findings suggest a banking system that has undergone significant restructuring in recent years, but where efficiency gains have yet to fully materialize.

4. To secure efficiency gains, it will be important to ensure that markets are fully contestable. Greater contestability should act as a powerful force to drive banks to become more competitive and efficient. Competition policy will also continue to be an important consideration, both in enforcing Italy’s antitrust laws in the banking sector and in ensuring that bank merger and acquisition reviews focus on stability and competition objectives.

B. Consolidation and Concentration

5. In the late 1990s, the banking industry underwent rapid consolidation, but it remains relatively small compared to other EU member countries (Figures 1 and 2). Between 1995 and 2004, the number of institutions declined by a third (from 854 to 572 banks). The average size of banks (net of mutuals and foreign branches) more than doubled over this period from to € 5.5 billion to € 13.5 billion for all banks. At the end of 2004, net of mutuals, sector included 135 banks (Table 2 and Figure 2).

Figure 1.
Figure 1.

Banking Sector Size and per capita GDP, end-2003

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

Source: World Development Indicators (World Bank): per capita GDP for 2003 (at constant 2000 US$).
Figure 2.
Figure 2.

Banking Sector: Total Loans and Stock Market Capitalization, end-2004

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

Table 1.

Evolution in Average Scale of Italian Banks on Consolidated Basis 1/

(In millions of euros)

article image
Source: Bank of Italy

Excludes branches of foreign banks and banks in special administration or compulsory liquidation.

Table 2.

Selected Countries: Market Concentration Indicators

article image
Source: Bankscope.

Herfinddahl-Hirschman Index by total assets.

3-firm concentration ratio is computed as the share of total assets of three largest banks.

5-firm concentration ratio is computed as the share of total assets of five largest banks.

10-firm concentration ratio is computed as the share of total assets of ten largest banks.

6. The consolidation process led to an increase in concentration, but one that was more moderate than experienced elsewhere. Market structure indicators, such as the Herfindhal-Hirschman Index (HHI)2 or the share of total Bassets held by the three, five, and ten largest institutions suggests a degree of concentration that is larger in Italy than in Germany, and the UK, but lower than in France (Table 2 and Figure 3).3 Concentration at the national level has increased (the largest five bank groups accounted for 46 percent of total assets at end-2004, compared with 37 percent a decade earlier). But after rising by more than 80 percent since 1990, the ratio of the number of branches to total population is now close to the EU average. According to the Bank of Italy (BI), this development has contributed to greater competition in provincial and regional markets, as evidenced by the rise in the average annual shift in deposit and lending market shares. The average number of banks in provincial markets is estimated to have increased from 27 to 30 in the last decade, and reached 35 at the end of 2004.4 The HHI for the provincial deposit market declined by around 12 percent from the peak it reached in 1999, falling back to the levels recorded in the mid-1990s. In regional lending markets, the index declined by 20 percent between the end of the 1990s and 2004.

Figure 3
Figure 3

Selected Countries: Concentration Indices, 1998 and 2004

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

7. The economic impact of greater concentration depends on many factors. To shed light on this issue, a number of recent papers have estimated the price effect of mergers and acquisitions in Italy in the 1980s and the 1990s. For example, Focarelli et al. (2002) account for the pricing policies of merged banks, and provide some evidence that bank mergers can allow for better risk pricing through informational benefits (i.e., closer correspondence between the price of loans and the default risk of each firm). Sapienza (2002) explores the trade-off between efficiency gains and greater market power associated with mergers and finds that in-market mergers generate higher efficiency gains than do out-of-market mergers.

Focarelli et al. (2002) find the performance of banks is affected by whether consolidation takes place through mergers or acquisitions. They provide some evidence that mergers tend to increase profitability, including through a more efficient use of capital. Acquisitions also tend to improve profitability, generally by raising the quality of the acquired bank’s loan portfolio. While this literature has helped shed light on the price impact of bank mergers, it does not aim at providing an assessment of bank consolidation on the degree of competition in the Italian banking system.

C. Contestability and Costs Indicators

Foreign ownership

8. Italy illustrates how fragmented─along national lines─the EU banking market still is. In line with some other large countries, the presence of foreign banks is concentrated primarily in investment banking and remains very limited in retail banking. So far, foreign take-overs have proven difficult to carry out, prompting scrutiny by the European competition and single market authorities. At end-2004, 7 percent of total bank assets were owned by foreigners, similar to the share in other large western European countries (Table 3), except that in Italy no major bank is majority foreign-owned.5 At end-2004, foreigners were majority owners in two medium-size banks (with total assets below € 20 billion) and 13 smaller banks (with total assets below € 7 billion), accounting in total for only 2.5 percent of total bank assets.

Table 3.

Foreign-Ownership in Banking Sector

article image
Source: Bank of Italy.

Bank with total assets above 45 EUR millions.

Bank with total assets between 20 and 45 EUR millions

Bank with total assets between 7 and 20 EUR millions

Bank with total assets below 7 EUR millions.

Includes banks for which shareholding is less than 15 per cent.

Includes non controlled banks for which shareholding is greater than 15 per cent.

Costs of banking services

9. The pricing data suggest relatively high costs of banking in Italy. According to one international survey, the average price of basic banking services (adjusted for local consumption patterns) appears to be among the highest in Europe (Table 4). This survey, however, does not provide a comprehensive cost estimate for basic banking services and should be interpreted with caution. Adjusting for joint-ownership of current accounts and the higher implied average balances, as well as the remuneration of accounts, another study found that the average price of holding a current account in Italy is still some 23 percent higher than the average for the EU countries surveyed.6 The high cost of services does not seem to be associated with delivery of high quality services: a survey on the quality of financial services in Europe─measured by consumers’ assessments of aspects such as the quality of information provided by banks, the ease of settling disputes with banks, the extent to which they trust banks’ advice, and use of internet for banking services─suggests dissatisfaction with the quality of services in Italy.7 These findings suggest a low level of competition in the services provided by Italian banks in the retail sector.

Table 4.

Selected Countries: Cost of Banking Services

article image
Source: CapGemini et al. World Retail Banking Report (2005).

Price of banking services are adjusted for local consumption patterns.

Switching costs

10. Switching costs can provide additional insight into the degree of banking competition. In this area, the Italian authorities are concerned that high switching costs hamper competition. The BI, jointly with the Competition Authority, has initiated an investigation regarding banking costs and depositors’ mobility in local markets. The BI surveyed a representative sample of 300 branches on the costs actually incurred by customers who closed their current accounts. Preliminary results suggest an average cost of closing a current account of €34, with wide variation among banks (from €0 to €100). This suggests that for some banks, high switching costs can hamper customers’ mobility or help keep customers captive, to the detriment of a more competitive environment. Cross-country comparisons on switching costs, however, are not available.

11. Persistently high operating profits, coupled with high revenues and/or high costs, are frequently associated with non competitive behavior. Relative to banks in other large industrial countries, Italian banks could fit this profile. For example, focusing on the top 50 banks, Italian banks enjoy relatively high operating income, surpassed only by US banks (Figure 4).8 However, because of high operating expenses, the net operating profit of Italian banks is only slightly higher than that of UK and Spanish banks (Figures 5 and 6). In this exercise we focus as much as possible our figures on the 50 largest banks in the countries under examination, as they are most likely to drive (or hamper) competition in their domestic markets. The overall trends, however, may be slightly different when looking at the entire banking sector.9

Figure 4
Figure 4

Selected Countries: Top 50 Banks, 2004 Operating Revenues

(In percent of total assets)

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

Figure 5.
Figure 5.

Top 50 Banks: Profitability Indicators, end-2004

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

Source: Bankscope.
Figure 6.
Figure 6.

Top 50 Banks: Indicators of Efficiency, 2004

Citation: IMF Working Papers 2007, 026; 10.5089/9781451865905.001.A001

Source: Bankscope.

12. The profitability of banks in Italy underwent two very distinct phases in the 1980s and 1990s, which has been interpreted as evidence of intensified competition in the banking industry in recent years (Ciocca, 2005).

  • In the 1980s, the industry was highly inefficient but profitablE, suggesting low levels of competition. Despite rising labor costs (in real terms) and low (albeit rising) productivity (assets per employee), profit rates, remained high (double digit) until the end of the decade.

  • In the 1990s, the degree of inefficiency was greatly reduced, and productivity increased steadily and rapidly, by just under 4 percent annually. In the meantime, the growth of labor costs moderated sharply. However, the rate of profits declined steadily, to close to zero by the mid-1990s. Only later in the decade, driven by banks’ continued efficiency gains, did profit rates recover.

13. Return on equity components can help identify determinants of banks’ performance. To help shed light on how much of the change in profitability was due to improved efficiency, risk exposure, or other factors, it is possible to decompose the return on equity (ROE) for banks as follows (Table 5):

Table 5.

ROE Decomposition, 1994–2004 1/

article image
Source: Bank of Italy

“Major” banks are those with total “medium banks” are those with total balance-sheet items in excess of 45 billion euros, “large” banks are those with total balance-sheet items between 20 and 45 billion euros, balance-sheet items between 7 and 20 billion euros, “small” banks have total balance-sheet items amounting to less than 7 billion euros.

The efficiency ratio is given by overall administrative costs divided by gross income.

Not including branches of foreign banks. Data refer only to banks that have submitted income statement reports and provided information about the number of banking staff.

14. The steady improvements in efficiency in the second half of the 1990s suggest that improved competition was likely at play. Between 1994 and 2000, the efficiency ratio improved by some 12 percent. The ROE decomposition (Table 5) suggests that in the same period major and large banks had the ability to generate more value added per unit of assets adjusted for the risk assumed. All banks shifted toward more risky activities.

15. Over the last five years, however, banks’ efficiency gains stagnated, and, except for major banks, profitability continued to improve, suggesting that competitive pressures may have receded, at least in some segments of the banking sector. An important factor behind banks’ rising ROE was the improved quality of their loan portfolios, reflected in a higher net profit ratio. This reduction in NPLs, however, was largely driven by temporary tax incentives, and remained short-lived. In the case of major banks, despite higher income ratios, profits declined as a result of higher administrative costs (efficiency losses). This suggests that lower competitive forces among major banks may have allowed these banks to generate higher income without creating corresponding efficiency gains.

16. However, simple comparisons of profitability, revenue, and cost indicators do not provide sufficient information to assess the operational effectiveness of Italian banks relative to other banks. For a fair comparison of banks’ effectiveness, size, regulatory environments, input costs, and business models need to be held constant. The next section complements the analysis above by controlling for the impact of such exogenous effects on various profitability indicators. A description of the data used for this analysis is provided in Box 1.

Data Sources

The data used to compute the analysis come from Bankscope, a comprehensive database containing harmonized detailed balance sheets and income statements of individual banks across countries. This database allows a reasonably consistent crosscountry comparison of banking systems. To avoid double-counting of banks within the country selected, our data is based on consolidated statements, when available.

The data set covers six large countries over a seven-year period from 1998 to 2004. In particular, it includes a total of over 3,500 large banks (i.e., banks with total assets greater than US$1 billion) in France, Germany, Italy, Spain, the United Kingdom, and the United States. Table 6 lists the number of banks represented each year for each selected country. Sample statistics are presented in Table 7, at the end of the paper.

Table 6.

Selected Countries: Banks Coverage, 1998–2004

article image
Table 7.

Selected Countries: Sample Statistics, 2004

(In million euros, unless specified otherwise)

article image
article image
article image
Source: Bankscope

Interest income minus interest expense over total assets.

Operating income minus operating expenses.

D. Efficiency Estimates

Cross-country profitability differences

17. We start by examining whether Italian banks earn profits that are statistically different from those of banks in other countries. As a first approximation, we concentrate on differences between countries rather than variations over time. We estimate a pooled weighted least squares regression using the between-effect estimator, controlling for bank and country specific variables and including country dummies to capture cross-country differences.10 In particular, we run the following regression:

Dijt=μ+β1 (BkSijt¯) +β2(HHI_TAjt)+β3(Macrojt¯)+β4(FOjt)+β5(CDj¯)+eit(1)

where the subscripts represent, respectively, individual bank i, country j, and year t. The dependent variable (Dijt) represents profits, which are measured as the net interest margin (interest income minus interest expense over total assets)11 and as operating profits to total assets (operating income minus operating expenses). μt represents the time fixed effects. The vector of bank-specific variables (BkSijt) is expressed as a share of total assets and includes: gross income, total equity, total loans, loan loss provisions, total banks deposits, customer deposits plus money market funds, and, in some regressions, operating expenses. To control for the level of asset concentration in the banking sector, we include the Herfindahl-Hirschman Index (HHI_TAjt), which is computed as the sum of the squares of the shares of total assets (expressed in percentage) held by each bank in the respective countries.12 Macrojt is a vector of macroeconomic variables and includes per capita GDP, inflation, Treasury-bill rates, and GDP growth. We also include a dummy equal to one if a bank is majority foreign-owned (FOijt) and zero otherwise and a vector of country dummies (CDj). The results are presented in Tables 8 and 9.

Table 8.

Selected Countries: Panel Regression Results on Net Interest Margin Between Estimator with Weighted Least Squares

(Dependent variable: net interest margin)

article image
Source. Bankscope.
Table 9.

Selected Countries: Panel Regression Results on Operating Profits to Total Assets Between Estimator with Weighted Least Squares

(Dependent variable: operational profits to total assets)

article image
Source: Bankscope.

18 In this simple framework, our results suggest that the net interest margins operating profits of Italian banks do not appear to be, on average, statistically different from those of the other banks included in the sample. This is shown in Column 1, where the coefficient on the dummy variable for Italian banks (itad) is not statistically significant. However, the data suggest some interesting cross-country differences. In particular, we find that Italian banks exhibit significantly higher net interest margins than German banks (Column 3). Italian banks, however, do not seem to be able to maintain this advantage as they generate an overall lower level of operational profits than their German counterparts. Thus, while Italian banks make good returns on their lending business, they lag behind German banks in generating net revenues from non interest based activities. The picture is reversed in the case of Spanish banks, which earn higher net interest margins than Italian banks but lower overall operating profits (Column 5). Moreover, our results suggest that the average Italian bank earns a level of profits that is broadly in line with its French counterpart (Column 2).

Effects of bank characteristics on revenue generation and cost control

19 Next, we compare the ability of Italian banks to control costs and generate revenues relative to banks in other large countries In this framework, we allow variables to change across banks and across time, while controlling for bank characteristics13:

Xit=μi+μt+β1lnTA+β2(lnTA)2+β3(II/TA)+β4(NCFTR/TA)+β5(PE/TA)+β6(TL/TA)+β7(LLP/TA)+β8(TCD/TA)+β9(TEQ/TA)+(2)β10(FO)+β11(HHI_TA)+β12(CDi¯)+eit

where the dependent variable Xit is, respectively, operating expenses to total assets (OpE/TA), operating income to total assets (OpI/TA), and net operating profits to total assets (Opp/TA). μi captures the individual fixed effects, while μi represents the time fixed effects. Explanatory variables include interest income to operating income (II/TA), net commission and fee income to total assets (NCR/TA), personnel expense to total assets (PE/TA), total loans to total assets (TLN/TA), total customer deposits to total assets (TCD/TA), and total equity to total assets (TEQ/TA). We also include a dummy equal to one if the bank is majority foreign-owned and zero otherwise (FO), the country-specific HHI (as a share of total assets) as a concentration proxy, and a vector of country dummy variables (CD). The results are presented in Tables 10, 11, and 12.

Table 10.

Cross-Sectional Time-Series FGLS Regression Allowing for Heteroskedastic Panels and Common AR(1) Coefficient for All Panels

(Dependent variable: operational income to total assets)

article image
Source: Bankscope.
Table 11.

Cross-Sectional Time-Series FGLS Regression Allowing for Heteroskedastic Panels and Common AR(1) Coefficient for All Panels

(Dependent variable: operational expenses to total assets)

article image
Source: Bankscope.
Table 12.

Cross-Sectional Time-Series FGLS regression Allowing for Heteroskedastic Panels and Common AR(1) Coefficient for All Panels

(Dependent variable: operational profits to total assets)

article image
Source: Bankscope.

20 To account for the identified heteroscedastic error structure and a first-order autocorrelation process, we fit our panel data using a generalized least squares (FGLS) regression. Our data failed to reject the null hypothesis of no first-order autocorrelation based on the Wooldridge panel test. After finding that a random effects model was not an appropriate model structure for our data (based on both the Breusch and Pagan Lagrangian multiplier test for random effects and the Hausman specification test), we switched to a panel FGLS model. This model structure allows us to control for cross-country differences, something that would not have been possible under a simple fixed effects model, while correcting for first-order autocorrelation and testing for the presence of heteroscedasticity. Iterated GLS produce maximum likelihood estimates, which make it possible to use a likelihood ratio test to test the null hypothesis of a homoscedastic error structure. Our results supported the presence of heteroskedasticity, a result that was confirmed by the modified Wald test for groupwise heteroskedasticity conducted in our fixed effect regression models.

21. Our results suggest that Italian banks incur significantly higher expenditures than other European banks and generate lower revenues than other banks, except for French banks, which earn even lower revenues than Italian and German banks. When we allow for variables to change across banks and across time, and after controlling for characteristics that affect banks’ ability to generate revenue and control costs, Italian banks have a statistically higher level of operating expenses to total assets and lower operating income to total assets than the other banks included in the sample (Column 3). Columns 4 to 6 compare the costs and revenues of Italian banks to, respectively, those of German, French, and US banks. Overall, the combination of higher expenses and lower revenues translates into lower net operating profits for Italian banks, after controlling for banks’ characteristics. This situation could indicate a low-competition environment, where banks are not pressured to reduce their costs to compensate for low income margins.

22. In Italy, foreign banks are relatively more profitable that domestic banks owing to their ability to better manage their cost structure (Column 1). This contrast with foreign banks in Germany and France, where foreign banks’ higher income capacity is fully offset by their higher operating expenditures (Column 4 and 5). These differences can be indicative of lower competition in Italy than in other markets. Moreover, while the degree of banks’ asset concentration in the sector, as measured by HHI, appears to be statistically significant, its impact is negligible for the countries under study.

Productive efficiency

23. It may not be sufficient to control for a bank’s profile (i.e., in terms of balance sheet structure and profit and loss accounts) to assess its productive efficiency. For example, a bank may have relatively higher personnel costs than other banks and yet be more efficient, if these costs help provide high-value-added services that require a highly qualified staff. Alternatively, a bank’s profitability may be lower because it faces less pressure to use its inputs efficiently. Thus, it is important to control for endogenous factors that affect banks’ ability or motivation to generate higher revenues and/or manage costs more effectively.

24. A stochastic “best practices” frontier approach is a useful tool to assess banks’ efficiency. This approach estimates indirect levels of revenues and costs for a given level of output and for given input prices, while allowing a number of other factors to affect total factor productivity.14 One must specify a functional form for the efficiency frontier. A common approach in the literature, and the one adopted here, is to use the translog specification:

Xit=μi+μt+j=14ϕjZijt+j=12αjlog yijt+j=13βjlog pijt+12j=12k=12ηjklog yijtlog yikt+12j=13k=13ϕjklog pijtlog pikt+j=12k=13ρjklogyijtlogpikt+eit(3)

where Xit is either revenues (operating income to total assets) or costs (operating expenses to total assets) for bank/in year t; the individual fixed effects (μi) capture relative measures of management effectiveness across banks groups, while the time fixed effects (μt) represent technological progress and aggregate shock. Zit is a vector of exogenous variables (total customer deposits to total liabilities, total bank deposits to total assets, loan loss provisions to total assets, equity to total assets, total assets to number of employees, a dummy equal to one for majority foreign-owned banks, and the country-specific HHI) that affects efficiency but not the estimate frontier. These variables are intended to proxy for bank differences in the business model and in the regulatory environment. yijt and yikt are bank output as a share of total assets (loan and other earning assets); and pijt and pikt are a bank input price as a share of total funding (personnel expense, interest expense, and total operating expense net of personnel expense); eit is an error term. The results are presented in Tables 13 and 14. For clarity, we do not provide the full set of results and focus on country-specific effects captured by the country dummies.

Table 13.

Cross-Sectional Time-Series FGLS Regression Allowing for Heteroskedastic Panels and Common AR(1) Coefficient for All Panels

(Dependent variable: operational expenses to total assets)

article image
Source: Bankscope.