SEARCH

SEARCH BY CITATION

Keywords:

  • E44;
  • G21;
  • G38;
  • K00

Abstract

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

This article investigates the impact of institutions on bank efficiency and technology, using a stochastic frontier analysis of a data set of 7,959 banks across 136 countries over 10 years. The results confirm the importance of well-developed institutions for the efficient operation of commercial banks. Furthermore, the insights reveal the impact of institutional reforms in improving bank efficiency. The results are robust to adjustments in country-specific effects, achieved by including country dummies, as well as across different risk profiles. Moreover, they provide empirical evidence in support of the public view of the banking sector.


I. Introduction

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

Growing evidence points to the importance of institutions for the efficient operations of the financial system. In particular, it appears that the economic, legal and social environments in which financial institutions and markets operate determine economic growth (e.g. La Porta et al., 1998, 2000). Better institutions can enhance competition throughout the economy and positively affect bank cost efficiency. However, the impact of these better institutions on bank efficiency is not unambiguous. For example, a political economy view of financial development argues that weak institutions can increase the cost efficiency of banks through regulatory capture, such that financial institutions use their power to reshape the regulatory agenda, and regulatory agencies in turn provide benefits, such as government subsidies, to certain financial institutions. These conflicting notions highlight the need for empirical studies that clarify the effect of institutions on bank efficiency.

Various macro-economic empirical studies have examined the effect of institutions on financial development, including the relevance of well-developed institutions for economic growth and financial development (e.g. Levine, 1998, 1999; Levine, Loayza and Beck, 2000). Yet studies fail to reveal how institutions influence the efficiency of individual banks, with one exception: Demirgüç-Kunt, Laeven and Levine (2004) examine the impact of bank regulations, market structures and institutions on bank efficiency and find some evidence of a positive effect of better institutions on bank efficiency.

To offer new empirical evidence about the relationship between institutions and bank efficiency, we use a stochastic frontier analysis (SFA) to determine the effects of institutions on bank efficiency, unlike Demirgüç-Kunt et al. (2004), who measure bank efficiency with net interest margins. Because institutions may influence the inefficient use of technologies by banks, as well as the technology available for banks to adopt, we seek in particular to evaluate the role of institutions, in terms of both adopting appropriate technology and determining the efficiency of technology that is already in place.

Moreover, this study provides indirect evidence of the empirical relevance of two contrasting views regarding the impact of institutions on bank efficiency. The public interest view argues that weak institutions affect bank efficiency negatively, because restrictions on banks prevent them from attracting funds in the cheapest way or allocating them to the most profitable investment projects. The political economy view of financial development instead indicates that weak institutions can increase the cost efficiency of banks through a regulatory capture effect. We find a robust positive relationship between better institutions and higher cost efficiency, in empirical support of the public interest view. Although we cannot rule out the possibility that the political economy view holds in some cases, our results do not offer support for a negative relationship between bank efficiency and institutional quality.

Our analysis relies on an unbalanced set of nearly 8,000 banks that function in 136 countries over a span of 10 years. With SFA, we can estimate a best practice frontier for all banks, as well as the distance to the frontier for each individual bank. We analyze whether differences in institutions explain individual distances from this frontier (i.e. extent to which institutions can explain inefficiency), as well as whether institutions determine shifts in the frontier (i.e. extent to which institutions explain an inability to adopt the most appropriate technologies). The results provide evidence for both possible effects of institutions on bank performance. Overall, our results reinforce the importance of well-developed institutions for the efficient operations of commercial banks. The insights also confirm the importance of institutional reforms as means to improve bank efficiency.

The remainder of this article is organized as follows: in section II, we provide a survey of prior literature related to the relationship between institutions and the development of financial systems. This section also explains the channels through which institutions influence bank efficiency. Section III contains our methodology, and in section IV we describe the data set. After presenting the regression results in section V, we conclude in section VI.

II. Institutions and the development of the financial system

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

For more than a decade, academics have noted the importance of institutional quality. Following North (1990, p. 107), many of them have argued that institutions are the main determinants of welfare levels and ‘the long-run performance of economies’. For example, Acemoglu, Johnson and Robinson (2004) suggest that economic institutions not only determine the incentives of and constraints on economic actors but also shape economic outcomes. They regard institutions as social decisions, chosen for their intended consequences. Because various groups and individuals benefit from different economic institutions, there tends to be conflict over social choices, and the winners of the conflicts generally are groups with greater political power. According to Acemoglu et al. (2004), the distribution of political power in society thus depends on political institutions and the distribution of resources. Economic institutions that encourage growth emerge when political institutions allocate power to groups with interests in broad-based property rights enforcement, create effective constraints on power holders and allow relatively few rents to be captured by those power holders.

Empirical support for the importance of well-developed institutions also appears in various studies, such as those by Knack and Keefer (1995), Kaufmann, Kraay and Zoido-Lobatón (1999), Acemoglu, Johnson and Robinson (2001), Easterly and Levine (2003) and Rodrik, Subramanian and Trebbi (2004). In a growth accounting framework, institutional differences across countries help explain differences in total factor productivity (TFP): some countries have lower TFP because they cannot adopt the most productive technologies, due to institutional differences. A lower TFP also might result from the inefficient use of technologies resulting from rent seeking, corruption, monopoly power or immobility in the factors of production (Jerzmanowski, 2007). The institutional environment helps determine the appropriate set of technologies in a country and the degree to which existing technology gets used efficiently (Olson, 1982).

Growing literature also points to the importance of institutions for the efficient operations of financial systems. Properly functioning institutions can help financial markets and institutions channel funds to ultimate investors efficiently, which then has a positive impact on economic growth (e.g. Beck, Demirgüç-Kunt and Levine, 2003; Beck and Levine, 2003). Such research has identified three main ‘institutions’ that drive financial development: legal, cultural and political (Haber and Perotti, 2008).

The legal view (e.g. La Porta et al., 1998) implies that the distinction between a bank and a market-based financial system is irrelevant; it is more important to establish an institutional environment in which financial systems can operate efficiently. Only that part of financial development related to the institutional environment is important for fostering economic growth. Levine (1998, 1999) and Levine et al. (2000) also undertake macro-economic empirical analyses in which they connect the legal origins of countries with their financial development and economic growth. Legal origins thus appear associated with cross-country differences in the development of banking sectors and stock markets. Morck, Strangeland and Yeung (1998), Morck, Yeung and Yu (2000), Durnev et al. (2003, 2004) and Glaeser et al. (2004) also note the impact of weak institutions on equity markets. Property rights, shareholder rights, stock market transparency and capital account openness all contribute to efficient capital allocation and economic growth.

Regarding the importance of cultural differences for financial development, Stulz and Williamson (2003) control for them across countries when examining the impact of legal origins on financial development. They show that persistent cultural values, such as religion and trust, are important for financial development (see also Guiso, Sapienza and Zingales, 2004).

Finally, North and Weingast (1989) emphasize the importance of political institutions, arguing that financial accumulation requires limited government (see also Barth, Caprio and Levine, 2006; Haber, North and Weingast, 2007). Rajan and Zingales (2003) also argue that political forces, as manifested in laws that shape the financial sector and the business environment, play important roles. Beck et al. (2003) review the relative contributions of the different types of institutions, controlling for natural resources, religion and differences in political systems to assess the relationship between law and finance (see also Acemoglu and Johnson, 2005). Ultimately, Beck, Demigüç-Kunt and Levine (2006) determine that a supervisory strategy that focuses on empowering the private monitoring of banks, by forcing banks to disclose accurate information to the private sector, lowers the degree to which corruption by bank officials hinders firms from raising external finance. They therefore argue that regulations that empower private monitoring exert a beneficial effect on the integrity of bank lending in countries with sound legal institutions.

This ample evidence strongly indicates that supporting institutions are necessary to improve financial development. Yet the impact of institutions on bank efficiency is not totally undisputed. Better institutions may enhance bank efficiency (e.g. Demirgüç-Kunt et al., 2004) by increasing competition throughout the economy, which increases bank cost efficiency and improves the allocation of credit. This public interest view of banking also holds that, subject to technological and institutional constraints, bankers allocate credit efficiently. Yet Japelli, Pagano and Bianco (2005) reveal that better institutions actually exert an ambiguous effect on bank efficiency, in that they enhance bank efficiency by decreasing the cost of financial intermediation but also improve the access of low-grade borrowers to the credit market, which raises the average rate of default, with a resultant negative effect on bank efficiency. The political economy view of financial development therefore argues that weak political institutions, by inducing capture, improve banks’ cost efficiency. Unaccountable institutions benefit their connected interests, resulting in regulatory capture, which can favour some financial institutions.

The political economy literature also notes that some banks enjoy close ties with governments and that government policies strongly influence their appearance (for an overview, see Haber and Perotti, 2008). Some banks then may be more powerful than others and translate this power into political capital. The political economy view therefore focuses on interactions among banks, governments and politics and defines government-supported banks as rent-seeking monopolists. Some big banks reap the benefits from the economies of scale and scope to influence policymakers; Krueger (1974) argues that incumbent businesses also can exercise rent seeking through their power, in support of the rent-seeking view of big banks. Other banks in turn might use their power (i.e. political influence) to sustain institutional arrangements that benefit them but damage the rest of the economy. This view implies that regulators are routinely captured by large businesses (Stigler, 1971). Perotti and Vorage (2009) also show, in their investigation of politicians’ choices of state or private control, that bank ownership is determined endogenously by politicians’ choice. When political accountability is low, politicians prefer state ownership of banks, because it creates the greatest rents for those politicians. Some banks help governments realise political and economic aims; for example, the government may employ banks as instruments to foster state control and further industrial development. Most developing economies also confront the problem of capital scarcity, so with limited financial resources it becomes a reasonable solution to pool capital in several big businesses. In other words, governments may want to promote a few big banks to speed up economic growth.

This political economy perspective certainly is helpful for explaining political–economic factors, such as the role and functions of the state and governments. It also may be particularly relevant in emerging economies, where government authorities play a greater role, and the scope for rent-seeking activities is broad. However, the political economy perspective implies that increased cost efficiency for banks does not necessarily ensure value creation, because with government support, banks’ performance actually is a function of their rent-seeking ability and opportunities. Greater cost efficiency thus could result from the capture of political rents, which does not benefit the economy as a whole.

III. Methodology

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

As we have shown, the impact of institutions of bank efficiency remains unclear. According to the public interest view of banking, which assumes that banks allocate credit efficiently, better institutions improve both bank efficiency and credit allocation. However, the political economy view maintains that weak institutions can facilitate banks’ ability to capture political rents, which worsens the allocation of credit but improves cost efficiency. More empirical research thus is needed to determine the likely effects of institutions on bank efficiency.

We accordingly adopt a SFA model to analyze a bank's cost efficiency, in its intermediation role. The task of a bank in an intermediation model is to attract money from savers and provide it to investors. Inefficiency can affect this role in several ways. First, in terms of technical efficiency, the bank might use its inputs inefficiently, such that the amount of deposits the bank collects does not generate the amount of loans that would be possible, were the bank completely efficient. Second, the bank can choose its input mix such that its costs are higher than those it would have to pay with another mix of inputs. This allocative efficiency question directly affects the amount of loans the bank can provide to investors, because the costs of these loans are higher; potentially profitable projects might not be financed. Third, a bank can choose the wrong output mix and provide loans to projects that are not beneficial or fail to provide loans to beneficial projects. This type of efficiency is called allocative efficiency, but it is allocative particularly in its outputs.

To examine the interactive effect of technical and allocative efficiency, we estimate a cost function in which we assume that the output mix depends on the optimal contracts between borrowers and the bank. At these outputs, the bank must choose inputs to ‘produce’ the desired output mix. Moreover, given input prices, the bank must choose an input mix to minimise costs. We note a caveat: the bank must have some labour before it can negotiate contracts with borrowers, so some simultaneity exists.

Although examining cost efficiency is common in banking literature (Berger and Humphrey, 1997), we also consider if banks are wasting valuable resources or making credit allocations unnecessarily expensive, so we provide some insight into whether the political view of banking holds. That is, if governments that operate in countries with low-quality institutions favour some banks, those banks should enjoy lower average costs and appear more cost efficient. If institutional quality increases though, their cost efficiency would decrease. Furthermore, cost efficiency is only one part of the story. A highly cost efficient bank can suffer low profit efficiency if it reallocates credit poorly, though a bank that is very cost inefficient cannot ever become highly profit efficient. Therefore being cost efficient is a necessary but not sufficient condition for being profit efficient.

In line with Berger and Mester (2001), we measure cost efficiency as the proximity of a bank's cost to a best practice bank's cost for producing the same output bundle in the same conditions. Cost functions are not directly observable; we measure inefficiencies relative to an efficient cost frontier. Rather than data envelopment analysis, we use a SFA to control for measurement error and other random effects.1 When it was initially developed in the mid-1970s (Aigner, Lovell and Schmidt, 1977; Meeusen and Van den Broeck, 1977), the SFA technique estimated production functions that allowed firms to be inefficient; since then, it also has been used to evaluate banking efficiency by Berg et al. (1993, Nordic), Griffell-Tatje and Lovell (1996, Spain), Mendes and Rebelo (1999, Portugal), Berger and Mester (2001, US), and Kumbhakar and Sarkar (2003, India), Kumbhakar et al. (2001, Spain), among others. As this list indicates, efficiency studies often focus on a single country, though a few make cross-country comparisons (e.g. Pastor, Pérez and Quesanda, 1997). For our investigation, we follow Battese and Coelli's (1995) approach, which we refer to as the BC model, because it offers several advantages over the standard two-step SFA approach. For example, the BC model can estimate the cost frontier and coefficients of the efficiency variables simultaneously.2Wang and Schmidt (2002) show that the two-step approach suffers from the assumption that the efficiency term is independently identically half-normally distributed in the first step, and then in the second step, the efficiency terms are assumed to be normally distributed and dependent on the explanatory variables. This method inherently renders biased coefficients. Furthermore, the BC model supports estimations with an unbalanced panel data set, which can increase the number of observations considerably.

To examine the effects of institutions on the financial system, we estimate a cost function and examine whether institutions affect it. We then seek to evaluate the extent to which cost inefficiencies (i.e. distance from the cost frontier) can be explained by differences in institutions. We adopt the cost function for several reasons: It can be derived from a product function and input prices (e.g. Shephard, 1970), and it facilitates the estimation of a model with multiple outputs, whereas a production function approach instead would assume a single output in a SFA context. Moreover, the cost function assumes that banks minimise costs, whereas the production function approach assumes that they maximize output. Finally, the cost function approach is more appropriate in a competitive environment where input prices are given and demand determines output.

The precise specification of a bank's cost function is debatable, and several models appear in prior literature (e.g. Benston, 1965; Sealey and Lindley, 1977). We follow Sealey and Lindley (1977), who characterize a bank as an intermediary between actors with funds and those that want to borrow funds. We also apply a transcendental logarithmic (translog) form developed by Christensen, Jorgenson and Lau (1973), which offers better fit than the more common Cobb–Douglas form (Kumbhakar and Lovell, 2000).

The general BC model thus specifies a stochastic cost frontier:

  • image(1a)

where Ci,t is the total cost that bank i faces at time t, and inline image is the cost frontier. The model measures bank efficiency relative to a global best practice frontier, which implicitly assumes that banks in different countries have equal access to the same banking technology. Alternatively, we might allow for different frontiers for each country in the sample, though that would prevent us from comparing estimated efficiencies across countries. To control for the possibility that banks lack equal access to the same technology in all countries, we must allow for cross-country differences in the frontier. As we detail subsequently, we control for these cross-country differences by including proxies for the institutional environment and country dummies.

Within the cost frontier, yi,t represents the logarithm of the output of bank i at time t, wi,t is a vector of the logarithm of input prices of bank i at time t, q are country-specific variables, and β is a vector of all parameters to be estimated. The term ui,t captures cost inefficiency and has a truncated normal distribution,3 whereas vi,t captures measurement error and random effects (e.g. good and bad luck), distributed as a standard normal variable. Both ui,t and vi,t are time and bank specific.

The precise specification of the cost function we use is in equation (1b). The dependent variable is TC, or total costs. Outputs include the total customer loans (TCL) a bank issues and the amount of securities and other earning assets it holds (TSOA). The input prices are the price of funds (PF), defined as interest expenses over total deposits and total other funding, and the price of labour (PL), defined by personnel expenses over total assets.4 We also include the year (T), year squared, and year × output and input prices to detect trends.5 Moreover, among bank characteristics, we note loan loss reserves over gross loans (LLR) to control for risk taking. Other operating income over total assets (OOIOTA) provides a proxy for quality differences with regard to bank services.6 Finally, the model features some country-specific variables. The log of gross domestic product (GDP) per capita (GDPPC), GDP growth (GDPGR) and the real interest rate (RIR) of a country make the cost frontier more flexible across countries. Thus, the ultimate model is

  • image(1b)

The cost function in equation (1b) could be estimated with ordinary least squares (OLS), which would imply that we assume every bank in our data set produces optimally and with the right input output mix – which is never true in practice.7 Moreover, the total cost can be endogenous with the input price and output parameters, because firms choose their inputs with knowledge of their level of productivity. These elements reaffirm our decision to estimate the model with SFA. Whereas with OLS, the error term ɛ should be normally distributed with mean 0 and variance σ2, the SFA method separates ɛ into two terms: an error component with mean 0 and a standard deviation of σ2, and an inefficiency portion of a certain distribution that is always positive. By making an explicit assumption about the inefficiency distribution, SFA deals with possible endogeneity as well (Van Biesebroeck, 2008). The SFA model developed by Battese and Coelli (1995) is well suited for estimating an unbalanced panel data set and has the ability to model determinants for inefficiency as well. The BC model is as follows:

  • image(2)
  • image(3)
  • image(4)

In equation (2), ɛ is separated into a random error component v and an inefficiency component u. The properties of both components are in equation (3). The random error term is similar to the random error term for OLS, with mean 0 and variance inline image. The inefficiency component u reveals a normal distribution, truncated at 0 with a first parameter m and second parameter inline image. The first parameter depends on a constant and the variables z, as in equation (4). To test for whether institutional differences influence efficiency, we include a proxy for institutions that explains inefficiency (equation 5): equity over total assets (EQOTA) thus controls for scale inefficiency effects. Moreover, we include return on average assets (ROAA) to control for management effects. A time trend (T) and time trend squared in the efficiency component allow for differences in efficiency over time. Although GDPPC and GDPGR appear in the cost frontier, we also add them to the efficiency component, because efficiency levels may differ across countries:

  • image(5)

The institutions variable does not differ between banks in the same country and year, so the standard errors may be biased. In response, we calculated clustered, bootstrapped standard errors over countries and time. However, some models include country-specific effects, which already control to some extent for the biased standard errors. For these models the general standard errors are given, because a clustered bootstrap is infeasible with country-specific effects (i.e. singular matrices generated for some bootstrap samples).

IV. Data

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

Our unbalanced data set features 7,959 banks over a 10-year period (1996–2005), operating in 136 countries, for a total of 33,790 observations. Bank-specific data come from the Bankscope database. We downloaded data for all commercial banks for which the unconsolidated information was available, and all the variables were inflation adjusted. Macro data came from the World Development Indicators, provided by the World Bank. In Appendix A, we provide the descriptive statistics of the data.

We use the set of aggregate governance indicators developed by Kaufmann et al. (2006) to proxy for institutional differences. These indicators measure alternative aspects of governance and cover 215 countries and territories for the years 1996, 1998, 2000, 2002, 2003, 2004 and 2005. A broad range of individual variables offer measures of the perceptions of governance from 31 separate data sources. From 2002 onward, yearly data are available, whereas before 2002, the set of indicators was updated only every other year. For missing years, we interpolated data from the relevant data from the years before and after.8 The perceptions of governance measure include:

  • (i)
     the process by which governments are selected, monitored and replaced;
  • (ii)
     the capacity of government to formulate and implement sound policies effectively; and
  • (iii)
     the respect of citizens and the state for the institutions that govern their economic and social interactions.

Kaufmann, Kraay and Mastruzzie (2006) suggest six indicators of the regulatory environment or ‘governance’ of a country. As we detail in Table 1, the GEF (government effectiveness) indicator refers to the government's ability to formulate and implement sound policies. A COR (control of corruption) indicator measures perceptions of corruption, interpreted as the exercise of public power for private gain, whereas LAW (rule of law) is an indicator of the extent to which agents have confidence in and abide by the rules of society. For the political instability and violence (PIV) index, they combine perceptions of the likelihood that the government in power will be destabilized or overthrown by possibly unconstitutional and/or violent means. The REG (regulatory quality) indicator summarizes the government's ability to formulate and implement sound policies. Finally, VAC (voice and accountability) uses an index of indicators of the extent to which citizens may participate in the selection of governments. These indicators use a scale of approximately −2.5 to 2.5, and higher values correspond to a ‘better’ regulatory environment. Because the different indicators for institutional differences are highly correlated, we applied a principal component analysis and included one factor that explained more than 80% of the variance of all the six indicators (Appendix A, Table A3). The weight of each variable for this score is about 0.4, and we include this factor in our regressions.9

Table 1. Governance indicators defined
Institutional indicator Definition
  1. Source: Kaufman et al. (2006).

VAC: Voice and accountabilityThe various aspects of the political process, civil liberties, political rights, and media independence
PIV: Political stability and violencePerceptions of the likelihood that the government in power will be destabilized or overthrown by unconstitutional or violent means, including domestic violence and terrorism.
GEF: Government effectivenessThe quality of public service provision and bureaucracy, the competence of civil servants, the independence of civil service from political pressures, and the credibility of the government's commitment to policies.
REG: Regulatory qualityIncidence of market-unfriendly policies such as price controls or inadequate bank supervision, as well as perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development.
LAW: Rule of lawExtent to which agents have confidence in and abide by the rules of society, such as perceptions of the incidence of crime, effectiveness and predictability of the judiciary, and enforceability of contracts.
COP: Control of corruptionExercise of public power for private gain, including both petty and grand corruption and state capture.

Institutions affect bank efficiency through various channels. Regarding the relevance of voice and accountability, we expect a higher level of media independence to increase the quality of information about local developments, which may reduce bank costs. Higher values of the political stability and violence indicator should increase bank efficiency, especially if the bank has relatively high loan loss provisions due to its risk averseness. Higher political stability and less violence also could lower banks’ costs, assuming banks run a risk of being victims of violence. Better government effectiveness reduces costs if banks face difficulty dealing with bureaucracy. Higher independence of the civil service from political pressure further lowers banks’ costs in countries where political pressure against the (entry of) banks is prevalent. Improvements in the regulatory quality help banks, if they are accompanied by more adequate banking supervision. The quality of the rule of law affects cost efficiency through the influence of the effectiveness and predictability of the judiciary. When going to court is time consuming, bank costs increase. Finally, enhanced control of corruption influences cost efficiency by lowering the costs associated with bribery. Overall, we predict that better institutions, indicated by higher values of the six governance indicators, reduce a bank's cost inefficiency. We also hypothesize that better institutions shift the cost function inside, such that better institutions enable banks to adopt more productive technologies.

V. Estimation results

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

We present three sets of estimates. In the first we test whether institutions affect efficiency, and then in the second set, we examine whether institutions affect the cost frontier. Stipulating that they do, institutions should affect technology adoption. Finally in the third set of estimates, we allow institutions to affect both the cost frontier and efficiency. The idea of including a particular variable in both the frontier and the efficiency specifications is not new (e.g. Battese and Broca, 1997; Bos et al., 2009; Lundvall and Battese, 2000), and models can be identified according to their specific distributional assumptions. We provide estimation results for the three sets in Tables 2–4. For each set, we test five models: Models 1 and 2 use the specification in equation (1b), whereas in Models 3 and 4, we exclude the two control variables for specific bank characteristics, because including them drastically reduces the number of observations. We also include country dummies to control for country characteristics other than institutions, as we detail in Models 2 and 4 for each set. Model 5 includes the country dummies in the efficiency term, not the frontier.

Table 2. Institutions in the inefficiency term;
Model [1] [2] [3] [4] [5]
  1. Notes: Clustered bootstrapped standard errors with clustering over countries and years are in square brackets; standard errors are in parentheses. TC indicates total costs. Outputs are total customer loans (TCL) and total securities and other earning assets (TSOA); the input prices are the price of labour (PL) and price of funds (PF). T denotes a time trend. GDPPC and GDPGR denote GDP per capita and GDP growth, respectively. RIR indicates real interest rate. Loan loss reserves over gross loans (LLR) is a proxy for risk taking; other operating income over total assets (OOIOTA) measures differences in services. EQ/TA stands for equity over total assets, and ROAA stands for return on average assets. T and T2 denote the time trend and the time trend squared. The variable institutions is our measure of institutional quality, obtained by applying principal component analysis on the six Kaufmann indicators. Sigma2 denotes the total amount of variance in the model. Gamma gives the ratio of variance of the inefficiency term over the total amount of variance.

  2. *Significant at 10%; **significant at 5%; ***significant at 1%.

Panel A
Frontier     
Number of observations15,05215,05233,79033,79033,790
Dependent variable     
Ln(TC)     
Intercept2.017***Country 2.318***Country 2.064***
(0.248)effects[0.219]effects[0.076]
ln(TCL)0.597*** 0.670*** 0.537*** 0.562*** 0.529***
(0.034)[0.018][0.033][0.011][0.011]
ln(TSOA)0.261*** 0.266*** 0.385*** 0.387*** 0.409***
(0.038)[0.017][0.029][0.010][0.011]
ln(PL)1.197*** 0.788*** 1.497*** 1.229*** 1.381***
(0.078)[0.040][0.072][0.025][0.026]
ln(PF)0.068 0.156***−0.0020.025−0.018
(0.057)[0.020][0.053][0.015][0.016]
ln(TCL)20.075*** 0.070*** 0.083*** 0.079*** 0.081***
(0.003)[0.001][0.002][0.001][0.001]
ln(TSOA)20.060*** 0.059*** 0.071*** 0.069*** 0.071***
(0.002)[0.001][0.002][0.001][0.001]
ln(PL)20.039*** 0.008* 0.070*** 0.041*** 0.054***
(0.007)[0.004][0.008][0.003][0.003]
ln(PF)20−0.008***0.007−0.002−0.004***
(0.004)[0.002][0.005][0.002][0.001]
ln(TCL)  ln (TSOA)−0.126***−0.121***−0.146***−0.142***−0.145***
(0.004)[0.002][0.004][0.001][0.001]
ln(TCL)  ln (PL) 0.014** 0.021***−0.0020.003−0.002
(0.006)[0.003][0.006][0.002][0.002]
ln(TCL)  ln (PF)−0.01−0.016***−0.01−0.007***−0.002
(0.010)[0.002][0.008][0.002][0.002]
ln(TSOA)  ln (PL)−0.024***−0.018***0.001−0.0010.001
(0.005)[0.003][0.005][0.002][0.002]
ln(TSOA)  ln (PF)0.004 0.006***0.001−0.003*−0.011***
(0.010)[0.002][0.008][0.002][0.002]
ln(PL)  ln (PF)0.003 0.010**−0.009−0.012***−0.023***
(0.009)[0.005][0.008][0.003][0.004]
T 0.001 0.025**0.037 0.046*** 0.034***
(0.043)[0.011][0.032][0.007][0.007]
T2 0.0030.001−0.001−0.003***−0.002***
(0.003)[0.001][0.002][0.000][0.000]
ln(TCL)T−0.006**−0.005***−0.006***−0.004***−0.004***
(0.003)[0.001][0.002][0.001][0.001]
ln(TSOA)T0.002 0.003***0.001 0.002***0
(0.003)[0.001][0.002][0.001][0.001]
ln(PL)T0.002 0.007***0−0.001−0.003**
(0.005)[0.002][0.004][0.001][0.001]
ln(PF)T0.002−0.0010.006 0.003*** 0.005***
(0.004)[0.001][0.004][0.001][0.001]
ln(GDPPC)−0.046*** 0.048***−0.027***−0.416***−0.028***
(0.011)[0.011][0.010][0.058][0.003]
GDPGR−0.01−0.006***−0.004−0.008***−0.011***
(0.007)[0.002][0.006][0.001][0.001]
RIR0−0.001*−0.001 0.003*** 0.001***
(0.002)[0.001][0.002][0.001][0.000]
LLR 0.007*** 0.006***   
(0.002)[0.000]   
OOIOTA 3.212*** 3.691***   
(0.575)[0.133]   
Panel B
Efficiency     
Number of observations15,05215,05233,79033,79033,790
Constant0.028−0.308−0.029−3.922***Country
(0.553)[0.393](0.786)[0.960]Effects
EQ/TA 0.009** 0.015*** 0.023*** 0.043*** 0.009***
(0.004)[0.002](0.007)[0.006][0.001]
ROAA−0.087***−0.115***−0.102***−0.208***−0.068***
(0.018)[0.007](0.026)[0.020][0.003]
T −0.004−0.07−0.287*−0.747***−0.118***
(0.122)[0.059](0.168)[0.139][0.021]
T2 −0.0040.0020.017 0.057*** 0.007***
(0.011)[0.005](0.015)[0.012][0.002]
ln(GDPPC)−0.029−0.051−0.0430.121−0.889***
(0.041)[0.043](0.084)[0.085][0.208]
GDPGR0.019−0.0110.027 0.084*** 0.010**
(0.022)[0.009](0.030)[0.021][0.004]
RIR−0.002−0.0010.005−0.005−0.003**
(0.004)[0.002](0.006)[0.005][0.001]
Institutions−0.184***−0.168***−0.434***−0.818***−0.061
(0.037)[0.033](0.083)[0.114][0.039]
Sigma20.3977230.5477150.723251.6520220.320459
Gamma0.9095370.9487780.9649740.98640.925876
Table 3. (a) Institutions in the frontier;
Model [1] [2] [3] [4] [5]
  1. Notes: Clustered bootstrapped standard errors with clustering over countries and years are in square brackets; standard errors are in parentheses. TC indicates total costs. Outputs are total customer loans (TCL) and total securities and other earning assets (TSOA); the input prices are the price of labour (PL) and price of funds (PF). T denotes a time trend. GDPPC and GDPGR denote GDP per capita and GDP growth, respectively. RIR indicates real interest rate. Loan loss reserves over gross loans (LLR) is a proxy for risk taking; other operating income over total assets (OOIOTA) measures differences in services. EQ/TA stands for equity over total assets, and ROAA stands for return on average assets. T and T2 denote the time trend and the time trend squared. The variable institutions is our measure of institutional quality, obtained by applying principal component analysis on the six Kaufmann indicators. Sigma2 denotes the total amount of variance in the model. Gamma gives the ratio of variance of the inefficiency term over the total amount of variance.

  2. *significant at 10%. **significant at 5%. ***significant at 1%.

Panel A
Frontier     
Number of observations15,05215,05233,79033,79033,790
Dependent variable     
Ln(TC)     
Intercept 1.815***Country 2.137***Country 2.095***
(0.249)Effects(0.269)Effects[0.080]
ln(TCL) 0.599*** 0.665*** 0.550*** 0.568*** 0.528***
(0.032)[0.018](0.035)[0.011][0.011]
ln(TSOA) 0.259*** 0.264*** 0.373*** 0.379*** 0.410***
(0.032)[0.017](0.035)[0.010][0.011]
ln(PL) 1.203*** 0.826*** 1.498*** 1.215*** 1.381***
(0.084)[0.039](0.080)[0.026][0.026]
ln(PF)0.077 0.146***0.03 0.038**−0.02
(0.057)[0.020](0.049)[0.015][0.016]
ln(TCL)2 0.075*** 0.070*** 0.081*** 0.079*** 0.081***
(0.002)[0.001](0.002)[0.001][0.001]
ln(TSOA)2 0.060*** 0.058*** 0.071*** 0.069*** 0.071***
(0.002)[0.001](0.002)[0.001][0.001]
ln(PL)2 0.040*** 0.011*** 0.071*** 0.039*** 0.054***
(0.008)[0.004](0.008)[0.003][0.003]
ln(PF)20.001−0.007*** 0.008*−0.003**−0.003***
(0.004)[0.002](0.004)[0.002][0.001]
ln(TCL) ln(TSOA)−0.125***−0.120***−0.144***−0.141***−0.145***
(0.004)[0.002](0.005)[0.001][0.001]
ln(TCL) ln(PL) 0.014** 0.021***−0.001 0.004**−0.002
(0.006)[0.003](0.006)[0.002][0.002]
ln(TCL) ln(PF)−0.01−0.016***−0.011−0.008***−0.002
(0.009)[0.002](0.008)[0.002][0.002]
ln(TSOA) ln(PL)−0.024***−0.018***0−0.0020.001
(0.005)[0.003](0.006)[0.002][0.002]
ln(TSOA) ln(PF)0.005 0.007***0.003−0.002−0.011***
(0.009)[0.002](0.008)[0.002][0.002]
ln(PL) ln(PF)0.006 0.008*−0.003−0.011***−0.023***
(0.010)[0.005](0.008)[0.003][0.004]
T −0.0050.0110.033 0.044*** 0.035***
(0.035)[0.011](0.030)[0.007][0.007]
T2 0.003 0.001*−0.001−0.003***−0.002***
(0.003)[0.001](0.002)[0.000][0.000]
ln(TCL)T−0.006**−0.005***−0.006***−0.004***−0.004***
(0.003)[0.001](0.002)[0.001][0.001]
ln(TSOA)T0.002 0.003***0.001 0.002***0
(0.002)[0.001](0.002)[0.001][0.001]
ln(PL)T0.002 0.005***−0.001−0.001−0.003**
(0.005)[0.002](0.004)[0.001][0.001]
ln(PF)T0.002−0.0020.006 0.002** 0.006***
(0.004)[0.001](0.004)[0.001][0.001]
ln(GDPPC)−0.015 0.091***−0.001−0.425***−0.033***
(0.016)[0.012](0.018)[0.062][0.005]
GDPGR−0.007−0.004***−0.004−0.006***−0.011***
(0.006)[0.001](0.006)[0.001][0.001]
RIR0−0.001**−0.001 0.002*** 0.001***
(0.001)[0.001](0.001)[0.001][0.000]
LLR 0.007*** 0.006***   
(0.002)[0.000]   
OOIOTA 3.172*** 3.635***   
(0.558)[0.134]   
Institutions−0.031***−0.070***−0.027**−0.0140.004
(0.010)[0.011](0.011)[0.009][0.003]
Panel B
Efficiency     
Number of observations15,05215,05233,79033,79033,790
Constant 1.440*** 1.001*** 3.969*** 3.579***Country
(0.440)[0.270](0.930)[0.624]Effects
EQ/TA 0.012*** 0.017*** 0.035*** 0.059*** 0.009***
(0.004)[0.002](0.008)[0.008][0.001]
ROAA−0.099***−0.123***−0.136***−0.252***−0.068***
(0.022)[0.008](0.029)[0.026][0.003]
T 0.03−0.052−0.34−0.761***−0.112***
(0.134)[0.064](0.235)[0.158][0.020]
T2 −0.0070.0010.026 0.066*** 0.007***
(0.012)[0.006](0.021)[0.014][0.002]
ln(GDPPC)−0.251***−0.251***−0.671***−1.014***−1.060***
(0.047)[0.028](0.110)[0.137][0.177]
GDPGR0.007−0.025***0.02 0.057** 0.010***
(0.023)[0.009](0.044)[0.024][0.004]
RIR−0.00100.007−0.006−0.003**
(0.005)[0.003](0.008)[0.007][0.001]
Sigma20.448880.5862550.9296011.8776110.317906
Gamma0.9178130.9507810.9726150.9879310.925256
Table 4. Institutions in the frontier and inefficiency term
Model [1] [2] [3] [4] [5]
  1. Notes: Clustered bootstrapped standard errors with clustering over countries and years are in square brackets; standard errors are in parentheses. TC indicates total costs. Outputs are total customer loans (TCL) and total securities and other earning assets (TSOA); the input prices are the price of labour (PL) and price of funds (PF). T denotes a time trend. GDPPC and GDPGR denote GDP per capita and GDP growth, respectively. RIR indicates real interest rate. Loan loss reserves over gross loans (LLR) is a proxy for risk taking; other operating income over total assets (OOIOTA) measures differences in services. EQ/TA stands for equity over total assets, and ROAA stands for return on average assets. T and T2 denote the time trend and the time trend squared. The variable institutions is our measure of institutional quality, obtained by applying principal component analysis on the six Kaufmann indicators. Sigma2 denotes the total amount of variance in the model. Gamma gives the ratio of variance of the inefficiency term over the total amount of variance.

  2. *significant at 10%; ** significant at 5%; ***significant at 1%.

Panel A
Frontier     
Number of observations15,05215,05233,79033,79033,790
Dependent variable     
Ln(TC)     
Intercept 2.308***Country 2.388***Country 2.107***
(0.330)Effects(0.268)Effects[0.080]
ln(TCL) 0.596*** 0.665*** 0.534*** 0.562*** 0.527***
(0.030)[0.018](0.032)[0.011][0.011]
ln(TSOA) 0.269*** 0.265*** 0.388*** 0.387*** 0.410***
(0.033)[0.017](0.031)[0.010][0.011]
ln(PL) 1.174*** 0.823*** 1.495*** 1.230*** 1.382***
(0.096)[0.039](0.074)[0.025][0.026]
ln(PF)0.052 0.147***−0.0060.024−0.02
(0.053)[0.020](0.046)[0.015][0.016]
ln(TCL)2 0.076*** 0.070*** 0.083*** 0.079*** 0.081***
(0.002)[0.001](0.002)[0.001][0.001]
ln(TSOA)2 0.060*** 0.059*** 0.071*** 0.069*** 0.071***
(0.002)[0.001](0.002)[0.001][0.001]
ln(PL)2 0.038*** 0.011** 0.070*** 0.041*** 0.054***
(0.009)[0.004](0.008)[0.003][0.003]
ln(PF)20.001−0.007*** 0.008**−0.002−0.003**
(0.004)[0.002](0.004)[0.002][0.001]
ln(TCL) ln(TSOA)−0.127***−0.121***−0.146***−0.142***−0.145***
(0.004)[0.002](0.004)[0.001][0.001]
ln(TCL) ln(PL) 0.014** 0.021***−0.0020.003−0.002
(0.006)[0.003](0.006)[0.002][0.002]
ln(TCL) ln(PF)−0.011−0.016***−0.01−0.007***−0.002
(0.010)[0.002](0.008)[0.002][0.002]
ln(TSOA) ln(PL)−0.023***−0.018***0.001−0.0010.001
(0.005)[0.003](0.005)[0.002][0.002]
ln(TSOA) ln(PF)0.004 0.006***0.001−0.003*−0.011***
(0.010)[0.002](0.008)[0.002][0.002]
ln(PL) ln(PF)0.002 0.008*−0.009−0.012***−0.023***
(0.010)[0.005](0.008)[0.003][0.004]
T 0.0080.0140.039 0.048*** 0.036***
(0.038)[0.011](0.031)[0.007][0.007]
T2 0.002 0.001*−0.001−0.003***−0.002***
(0.003)[0.001](0.002)[0.000][0.000]
ln(TCL)T−0.007***−0.005***−0.006***−0.004***−0.004***
(0.002)[0.001](0.002)[0.001][0.001]
ln(TSOA)T0.002 0.003***0.001 0.002***0
(0.002)[0.001](0.002)[0.001][0.001]
ln(PL)T0.002 0.006***0−0.001−0.003**
(0.005)[0.002](0.003)[0.001][0.001]
ln(PF)T0.004−0.002 0.007* 0.003*** 0.005***
(0.004)[0.001](0.004)[0.001][0.001]
ln(GDPPC)−0.104** 0.086***−0.038*−0.443***−0.034***
(0.048)[0.013](0.020)[0.063][0.005]
GDPGR−0.014*−0.005***−0.004−0.007***−0.011***
(0.008)[0.001](0.005)[0.001][0.001]
RIR0.001−0.001**−0.002 0.003*** 0.001***
(0.002)[0.001](0.002)[0.001][0.000]
LLR 0.007*** 0.005***   
(0.002)[0.000]   
OOIOTA 3.362*** 3.654***   
(0.605)[0.134]   
Institutions0.06−0.059***0.0110.011 0.006*
(0.053)[0.011](0.014)[0.010][0.003]
Panel B
Efficiency     
Number of observations15,05215,05233,79033,79033,790
Constant−0.579−0.093−0.31−3.927***Country
(0.627)[0.411](0.805)[0.950]Effects
EQ/TA 0.006** 0.016*** 0.021*** 0.042*** 0.008***
(0.003)[0.002](0.006)[0.006][0.001]
ROAA−0.069***−0.119***−0.095***−0.206***−0.067***
(0.019)[0.008](0.023)[0.020][0.003]
T −0.036−0.068−0.268*−0.741***−0.120***
(0.096)[0.063](0.155)[0.137][0.021]
T2 −0.0010.0020.016 0.056*** 0.007***
(0.008)[0.006](0.014)[0.012][0.002]
ln(GDPPC)0.11−0.095**0.0150.13−0.850***
(0.100)[0.046](0.091)[0.085][0.206]
GDPGR0.021−0.0140.025 0.082*** 0.010***
(0.018)[0.009](0.030)[0.020][0.004]
RIR−0.00300.006−0.005−0.003**
(0.004)[0.002](0.005)[0.005][0.001]
Institutions−0.278***−0.130***−0.450***−0.820***−0.075*
(0.073)[0.035](0.071)[0.114][0.039]
Sigma20.2893840.5746470.6583621.6339490.316004
Gamma0.8823630.950640.9616540.9862520.92484

We initially considered the impact of institutions on efficiency (see Table 2). The input prices and outputs, as well as their squared counterparts, were positive and significant, with the exception of ln(PF)2, at the usual significance levels.10 In line with our expectations, an increase in input prices or outputs increases costs. Furthermore, the impact of the included trend variables was not stable, and total customer loans had negative impacts on total costs over time. Our proxy for risk, LLR was positive and highly significant, indicating that banks that confront more risk suffer extra costs. The proxy that measures differences in the quality of services was positive and significant at the 1% level; therefore, banks that offer more services confront higher costs. Perhaps most important, institutional variables in the inefficiency term (cf. the model in which country effects are in the efficiency component) were negative and highly significant: banks are more efficient if the institutional environment is better. The insignificant effect of institutions on efficiency in the model with country efficiency effects might be explained by the persistence of institutions over time, such that the effects of institutions occur on the country level. However, this result also might reflect the incidental parameter problem, because there are many parameters to be estimated in the nonlinear part of the model (cf. Greene, 2005).11 We test the impact of efficiency with a likelihood ratio test, as we detail in Appendix B, together with its interpretation. For all models, the impact of efficiency is high.

These findings mirror those of Demirgüç-Kunt et al. (2004), who use a different concept for efficiency and reject the political economy view that weaker institutions increase bank efficiency. However, the findings cannot confirm if the greater cost efficiency results from technical efficiency and using inputs more effectively to generate outputs or else from allocative efficiency, in which case the input mix is optimal, which then reduces costs.

We therefore turn to the impact of institutions on the frontier (see Table 3). The impact of the control variables was qualitatively the same as in Table 2. However, our institutions variable was not significant for the models with the complete data set with country effects. Therefore, institutions are less important for technology adoption than for efficiency.

Finally, in Table 4 we present the estimates when we allow institutions to affect both the frontier and the efficiency term.12 The results for the cost function and bank-specific variables remained qualitatively the same, and the findings for our institution variable also were in line with our preceding findings: the variable was always negative and significant in the efficiency part, though one significance reached only 10%, so institutions are of great importance, even when controlling for the effect on technology. The influence of institutions on technology adoption also was negatively significant at a 1% level. Institutions can foster technology adoption, but this result is not very robust. Because institutions included in the frontier only affect technology, and technical efficiency and technology adaption are closely related, this result offers a limited indication that allocative efficiency might be a main effect.

The preceding analysis strongly suggests that a healthy institutional environment provides banks with abilities to adapt to new technologies and use existing technologies more efficiently. Yet our estimation strategy is open to criticism, in that institutions slowly change over relatively long periods, so it would be difficult to identify the impact of institutions on banks in a panel framework. We did not encounter identification problems; as a robustness check though, we present an additional set of estimates in Table 5 in which we suppressed the time-series variance by performing a SFA on the means of all banks in the sample. By taking the unit means, we treat all changing variables as time invariant, which improves identification. This analysis complements our prior regressions: a better institutional environment again enhances bank efficiency and enables banks to adopt better technologies. Our analysis thus strongly supports the view that the institutional environment is extremely important for banks’ behaviour.

Table 5. Estimates of unit means
Model [1] [2] [3] [4] [5] [6]
  1. Notes: Standard errors are given in square brackets. TC indicates total costs. Outputs are total customer loans (TCL) and total securities and other earning assets (TSOA); the input prices are the price of labour (PL) and price of funds (PF). Loan loss reserves over gross loans (LLR) is a proxy for risk taking; other operating income over total assets (OOIOTA) measures differences in services. EQ/TA stands for equity over total assets, and ROAA stands for return on average assets. The variable institutions is our measure of institutional quality, obtained by applying principal component analysis on the six Kaufmann indicators. Sigma2 denotes the total amount of variance in the model. Gamma gives the ratio of variance of the inefficiency term over the total amount of variance.

  2. *significant at 10%; **significant at 5%; ***significant at 1%.

Panel A  
Frontier      
Number of observations8,2648,2648,2643,8363,8363,836
Dependent variable      
Ln(TC)      
Intercept 1.893*** 1.942*** 1.949*** 1.475*** 1.559*** 1.540***
[0.105][0.109][0.105][0.176][0.173][0.173]
Ln(TCL) 0.473*** 0.514*** 0.483*** 0.541*** 0.566*** 0.560***
[0.018][0.019][0.018][0.030][0.031][0.030]
Ln(TSOA) 0.466*** 0.411*** 0.450*** 0.283*** 0.269*** 0.268***
[0.017][0.018][0.018][0.029][0.030][0.029]
Ln(PL) 1.404*** 1.327*** 1.388*** 1.103*** 1.073*** 1.095***
[0.042][0.043][0.041][0.069][0.068][0.068]
Ln(PF)0.0330.0310.025 0.121*** 0.087** 0.110***
[0.026][0.027][0.026][0.038][0.039][0.039]
Ln(TCL)2 0.082*** 0.081*** 0.082*** 0.078*** 0.078*** 0.078***
[0.001][0.001][0.001][0.002][0.002][0.002]
Ln(TSOA)2 0.075*** 0.075*** 0.075*** 0.065*** 0.065*** 0.065***
[0.001][0.001][0.001][0.002][0.002][0.002]
Ln(PL)2 0.063*** 0.056*** 0.061*** 0.035*** 0.033*** 0.034***
[0.005][0.005][0.005][0.007][0.007][0.007]
Ln(PF)2 0.005***0.0020.003−0.004−0.007**−0.006**
[0.002][0.002][0.002][0.003][0.003][0.003]
Ln(TCL) ln(TSOA)−0.151***−0.149***−0.151***−0.131***−0.133***−0.132***
[0.002][0.002][0.002][0.003][0.004][0.003]
Ln(TCL) ln(PL)−0.014***−0.008**−0.013*** 0.012** 0.016*** 0.014**
[0.003][0.003][0.003][0.006][0.006][0.006]
Ln(TCL) ln(PF)−0.003−0.001−0.002−0.012***−0.006−0.011**
[0.003][0.003][0.003][0.004][0.004][0.004]
ln(TSOA) ln(PL) 0.017*** 0.009*** 0.015***−0.015***−0.019***−0.018***
[0.003][0.003][0.003][0.006][0.006][0.006]
ln(TSOA) ln(PF)−0.001−0.002−0.001 0.008*0.005 0.008*
[0.003][0.003][0.003][0.005][0.005][0.005]
Ln(PL) ln(PF)0−0.002−0.0020.0120.0080.011
[0.006][0.006][0.006][0.008][0.009][0.009]
LLR    0.004*** 0.005*** 0.004***
   [0.001][0.001][0.001]
OOIOTA    3.376*** 3.801*** 3.524***
   [0.268][0.302][0.290]
Institutions −0.064***−0.022*** −0.061***−0.027***
 [0.002][0.004] [0.003][0.006]
Panel B  
Efficiency      
Constant−0.151*−6.821***−0.797*** 0.151**−1.989***−0.382
[0.081][1.872][0.238][0.064][0.616][0.234]
EQ/TA 0.012*** 0.074*** 0.017*** 0.005*** 0.028*** 0.010***
[0.002][0.016][0.003][0.001][0.006][0.003]
ROAA−0.082***−0.291***−0.109***−0.064***−0.150***−0.090***
[0.006][0.059][0.012][0.006][0.025][0.013]
Institutions−0.326*** −0.382***−0.182*** −0.193***
[0.022] [0.041][0.016] [0.028]
Sigma20.3841.6620.5560.2260.5500.336
Gamma0.9450.9840.9600.8520.9150.888

VI. Conclusions

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

This article examines whether institutional differences between countries improve the efficient operation of commercial banks. We have tested whether well-developed intuitions affect the adoption of technology by banks, as well as their efficient uses of technology already in place. By applying a SFA to a panel of approximately 8,000 banks in 136 countries over 10 years, we show that banks operating in countries with better institutions apply more cost-reducing technologies and can use existing technologies more efficiently. Overall, these results confirm the importance of well-developed institutions for the efficient operations of commercial banks. The related insights also reveal the scope for institutional reforms to improve bank efficiency. Finally, our results offer empirical evidence in support of the public interest view of banking and challenge the political view.

However, our database does not contain full information about all banks. Although the database tries to be as complete as possible, it likely suffers from a selection bias in favour of larger banks. These large banks are more likely to possess sufficient power and connections to reshape the regulatory agenda to their benefit, so this limitation cannot explain our finding of an opposite result.

Furthermore, this study cannot reveal if the main impact of institutions on cost efficiency functions through improved technical efficiency or allocative efficiency. It is not possible with our existing knowledge to use an SFA model to disentangle these two effects. To determine allocative efficiency properly, we would need detailed information about input prices. Thus far only aggregated interest expenses are available in the database, though it would be preferable to split these expenses into different deposit types. If such data were to become available, it would be helpful to research the impact of allocative efficiency more thoroughly.

Finally, this study focuses mainly on the supply side of banks’ intermediation role. We do not examine profit efficiency, or the multiplication of technical efficiency and allocative efficiency on the supply side, such as by determining whether a bank selects the correct output mix. Only loan loss reserves appear in this analysis, which could be a cost of choosing the wrong output mix. The analysis of institutions on the demand side of the intermediation role of banks remains for further research.

Footnotes
  • 1

    Non-parametric techniques, like data envelopment analysis, do not allow for measurement error and luck factors, so any deviation from the best practice bank gets attributed to technical inefficiency. For an extensive review of non-parametric and parametric approaches, see Matousek and Taci (2005).

  • 2

    Kumbhakar and Lovell (2000) discuss other SFA models that solve for exogenous influences on efficiency simultaneously.

  • 3

    The total costs a bank faces are never lower than the costs of the frontier. In a graphical representation, Berger, Hancock and Humphrey (1993) show that inefficiency is determined by both technical and allocative inefficiency.

  • 4

    Personnel expenses over total number of employees would be a better proxy, but the number of employees is not available in our data set.

  • 5

    The concept of multiplying the trend variable by the input prices and output is common for a translog production function (e.g. Coelli, Rao and Battese 1998). Although it is possible to include time dummies, our approach allows the parameters of input prices and outputs to change over time.

  • 6

    Because LLR and OOIOTA are ratios, they are not log transformed.

  • 7

    An overview of 130 studies dealing with banks’ ability to produce optimally (Berger and Humphrey, 1997) indicates that banks are not able to do so.

  • 8

    Excluding these years from the analysis, instead of interpolating the data, does not change the main results.

  • 9

    We also included the six variables separately in our regressions; the results did not qualitatively differ. These results are available on request.

  • 10

    Based on a clustered bootstrap, with clustering over countries and time and 200 replications.

  • 11

    Because of the incidental parameter problem, we only included country efficiency effects for the large set with 33,790 observations.

  • 12

    The frontier and efficiency coefficients cannot be compared directly, because the efficiency effect is nonlinear which makes the interpretation of the size of the effect cumbersome. However, the effect is monotonic so the sign of the effect can easily be interpret.

References

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices
  • Acemoglu, D. and Johnson, S. (2005). ‘Unbundling institutions’, Journal of Political Economy, Vol. 113, pp. 949995.
  • Acemoglu, D., Johnson, S. and Robinson, J. A. (2001). ‘Colonial origins of comparative development: an empirical investigation’, American Economic Review, Vol. 91, pp. 13691401
  • Acemoglu, D., Johnson, S. and Robinson, J. A. (2004). ‘Institutions as the Fundamental Cause of Long-run Growth’, NBER Working Paper 10481.
  • Aigner, D., Lovell, C.A.K. and Schmidt, P. (1977). ‘Formulation and estimation of stochastic frontier production function models’, Journal of Econometrics, Vol. 6, pp. 2137.
  • Barth, J. R., Caprio, Jr, G. and Levine, R. (2006). Rethinking Bank Regulation: Till Angels Govern, Cambridge University Press, Cambridge, MA.
  • Battese, G. E. and Broca, S. S. (1997). ‘Functional forms of stochastic frontier production functions and models for technical inefficiency effects: a comparative study for wheat farmers in Pakistan’, Journal of Productivity Analysis, Vol. 8b, pp. 395414.
  • Battese, G. E. and Coelli, T. J. (1995). ‘A model for technical inefficiency effects in a stochastic frontier production function for panel data’, Empirical Economics, Vol. 20, pp. 325332.
  • Beck, T., Demirgüç-Kunt, A. and Levine, R. (2003). ‘Law, endowments and finance’, Journal of Financial Economics, Vol. 70, pp. 137181.
  • Beck, T., Demirgüç-Kunt, A. and Levine, R. (2006).‘Bank supervision and corruption in lending’, Journal of Monetary Economics, Vol. 53, pp. 21312163.
  • Beck, T. and Levine, R. (2003). Legal Institutions and Financial Development, World Bank Policy Research Working Paper 3136.
  • Benston, G. J. (1965). ‘Branch banking and economies of scale’, Journal of Finance, Vol. 22, pp. 312331.
    Direct Link:
  • Berg, S. A., Førsund, F. R., Hjalmarsson, L. and Suominen, M. (1993). ‘Banking efficiency in the Nordic countries’, Journal of Banking & Finance, Vol. 17, pp. 371388.
  • Berger, A. N., Hancock, D. and Humphrey, D. B. (1993).‘Bank efficiency derived from the profit function’, Journal of Banking & Finance, Vol. 17, pp. 317347.
  • Berger, A. N. and Humphrey, D. B. (1997). ‘Efficiency of financial institutions: international survey and directions for future research’, European Journal of Operational Research Vol. 98, pp. 175212.
  • Berger, A. N. and Mester, L. J. (2001). ‘Explaining the Dramatic Changes in Performance of U.S. Banks: Technological Change, Deregulation, and Dynamic Changes in Competition’, Wharton School Working Paper 01-06.
  • Bos, J. W. B., Koetter, M., Kolari, J. W. and Kool, C. J. M. (2009). ‘Effects of heterogeneity on bank efficiency scores’, European Journal of Operational Research, Vol. 195, pp. 251261.
  • Christensen, L. R., Jorgenson, D. W. and Lau, L. J. (1973). ‘Transcendental logarithmic production frontiers’, Review of Economics and Statistics, Vol. 55, pp. 2845.
  • Coelli, T. J., Rao, D. S. P. and Battese, G. E. (1998).An Introduction to Efficiency and Productivity Analysis, Kluwer Academic Publishers, Boston.
  • Demirgüç-Kunt, A., Laeven, L. and Levine, R. (2004). ‘Regulations, market structure, institutions, and the cost of financial intermediation’, Journal of Money, Credit, and Banking, Vol. 36, pp. 593622.
  • Durnev, A., Li, K., Morck, R. and Yeung, B. (2004). ‘Capital markets and capital allocation: implications for economies in transition’, Economics in Transition, Vol. 12, pp. 593634.
  • Durnev, A., Morck, R., Yeung, B. and Zarowini, P. (2003).‘Does greater firm-specific return variation mean more or less informed stock pricing?’Journal of Accounting Research, Vol. 41, pp. 797835.
  • Easterly, W. and Levine, R. (2003).‘Tropics, germs, and crops: how endowments influence economic development’, Journal of Monetary Economics, Vol. 50, pp. 339.
  • Glaeser, E.L., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A.(2004). ‘Do institutions cause growth?’Journal of Economic Growth, Vol. 9, pp. 271303.
  • Greene, W. (2005). ‘Reconsidering heterogeneity in panel data estimators of the stochastic frontier model’, Journal of Econometrics, Vol. 126, pp. 269303.
  • Grifell-Tatje, E. and Lovell, C.A.K. (1996). ‘Deregulation and productivity decline: the case of Spanish savings banks’, European Economic Review, Vol. 40, pp. 12811303.
  • Guiso, L., Sapienza, P. and Zingales, L. (2004). ‘The role of social capital in financial development’, American Economic Review, Vol. 94, pp. 526556.
  • Haber, S., North, D. C. and Weingast, B. R. (2007). Political Institutions and Financial Development, Stanford University Press, Stanford.
  • Haber, S. and Perotti, E. (2008). The Political Economy of Financial Systems, Tinbergen Institute Discussion Paper 2008-045/2.
  • Japelli, T., Pagano, M. and Bianco, M. (2005). ‘Courts and banks: effects of judicial enforcement on credit markets’, Journal of Money, Credit and Banking, Vol. 37, pp. 223244.
  • Jerzmanowski, M. (2007). ‘Total factor productivity differences: appropriate technology vs. efficiency’, European Economic Review , Vol. 51, pp. 20802110.
  • Kaufmann, D., Kraay, A. and Mastruzzie, M. (2006). ‘Governance matters V: aggregate and individual governance indicators for 1996–2005’, WGI Working Paper, Available at: http://www.govindicators.org.
  • Kaufmann, D., Kraay, A. and Zoido-Lobatón, P. (1999). Governance Matters, World Bank Policy Research Working Paper, No. 2196.
  • Knack, S. and Keefer, P. (1995).‘Institutions and economic performance: cross-country tests using alternative institutional measures’, Economics and Politics, Vol. 7, pp. 207227.
  • Kodde, D. A. and Palm, F. C. (1986).‘Wald criteria for jointly testing equality and inequality restrictions’, Econometrica, Vol. 45, pp. 12431248.
  • Krueger, A. (1974). Foreign Trade Regimes and Economic Development: Turkey, UMI, New York.
  • Kumbhakar, S. C. and Lovell, C. A. K. (2000). Stochastic Frontier Analysis, Cambridge University Press, Cambridge.
  • Kumbhakar, S. C., Lozano-Vivas, A., Lovell, C. A. K and Hasan, I. (2001). ‘The effects of deregulation on the performance of financial institutions: the case of Spanish savings banks’, Journal of Money Credit and Banking, Vol. 33, pp. 101120.
  • Kumbhakar, S. C. and Sarkar, S. (2003). ‘Deregulation, ownership, and productivity growth in the banking industry: evidence from India’, Journal of Money Credit and Banking, Vol. 35, pp. 403424.
  • La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998). ‘Law and finance’, Journal of Political Economy, Vol. 106, pp. 11131155.
  • La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (2000). ‘Investor protection and corporate governance’, Journal of Financial Economic, Vol. 58, pp. 327.
  • Levine, R. (1998). ‘The legal environment, banks, and long-run economic growth’, Journal of Money, Credit, and Banking, Vol. 30, pp. 596620.
  • Levine, R. (1999).‘Law, finance, and economic growth’, Journal of Financial Intermediation, Vol. 8, pp. 3667.
  • Levine, R., Loayza, N. and Beck, T. (2000). ‘Financial intermediation and growth: causality and causes’, Journal of Monetary Economics, Vol. 46, pp. 3177.
  • Lundvall, K. and Battese, G. E. (2000). ‘Firm size, age and efficiency: evidence from Kenyan manufacturing firms’, Journal of Development Studies, Vol. 36, pp. 146163.
  • Matousek, R. and Taci, A. (2005). ‘Efficiency in banking: empirical evidence from the Czech Republic’, Economics of Planning, Vol. 37, pp. 225244.
  • Meeusen, W. and van den Broeck, J. (1977). ‘Efficiency estimation from Cobb-Douglas production functions with composed error’, International Economic Review, Vol. 18, pp. 435444.
  • Mendes, V. and Rebelo, J. (1999). ‘Productive efficiency, technological change and productivity in Portuguese banking’, Applied Financial Economics, Vol. 9, pp. 513521.
  • Morck, R. K., Strangeland, D. A. and Yeung, B. (1998). Inherited Wealth, Corporate Control, and Economic Growth: the Canadian disease? NBER Working Paper 6814.
  • Morck, R. K., Yeung, B. and Yu, W. (2000).‘The information content of stock markets: why do emerging markets have synchronous stock price movements?’Journal of Financial Economics, Vol. 58, pp. 215260.
  • North, D. C. (1990). Institutions, Institutional Change and Economic Performance, Cambridge University Press, Cambridge.
  • North, D. C. and Weingast, B. R. (1989). ‘Constitutions and commitment: the evolution of institutions governing public choice in seventeenth-century England’, Journal of Economic History, Vol. 44, pp. 803832.
  • Olson, M. (1982).The Rise and Decline of Nations: Economic Growth, Stagflation, and Social Rigidities, Yale University Press, New Haven, CT.
  • Pastor, J. M., Pérez, F. and Quesada, J. (1997).‘Efficiency analysis for banking firms: an international comparison’, European Journal of Operational Research, Vol. 98, pp. 395407.
  • Perotti, E. and Vorage, M. (2009). Entry: Direct Control or Regulation?, Amsterdam Business School Working Paper, University of Amsterdam.
  • Rajan, R. G. and Zingales, L. (2003). ‘The great reversals: the politics of financial development in the 20th century’, Journal of Financial Economics, Vol. 69, pp. 550.
  • Rodrik, D., Subramanian, A. and Trebbi, F. (2004).‘Institutions rule: the primacy of institutions over geography and integration in economic development’, Journal of Economic Growth, Vol. 9, pp. 131165.
  • Sealey, C. W. and Lindley, J. T. (1977).‘Inputs, outputs, and a theory of production and cost at depository financial institutions’, Journal of Finance, Vol. 32, pp. 12511266.
  • Shephard, R. W. (1970). Theory of Cost and Production Functions, Princeton University Press, Princeton, NJ.
  • Stigler, G. (1971). ‘The theory of economic regulation’, Bell Journal of Economics and Management Science, Vol. 2, pp. 321.
  • Stulz, R. M. and Williamson, R. (2003).‘Culture, openness, and finance’, Journal of Financial Economics, Vol. 70, pp. 313349.
  • Van Biesebroeck, J. (2008). ‘The sensitivity of productivity estimates: revisiting three important debates’, Journal of Business & Economic Statistics, Vol. 26, pp. 311328.
  • Wang, H.-J. and Schmidt, P. (2002). ‘One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels’, Journal of Productivity Analysis, Vol. 18, pp. 129144.

Appendices

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II. Institutions and the development of the financial system
  5. III. Methodology
  6. IV. Data
  7. V. Estimation results
  8. VI. Conclusions
  9. References
  10. Appendices

Appendix A: Descriptive statistics

Table A1. Descriptive statistics: complete set
Variable Observations Mean SD Min Max
  1. Notes: lTOC is the log of total operating cost. lTCL and lTSOA are the logs of total customer loans and total securities and other earning assets, respectively. lPF and lPL are the logs of the price of funds and labor. Institutions is a principal component score of the six Kaufmann indicators. EQUOTA and ROAA refer to equity over total assets and return on average assets, respectively. LGDPPC and GDPGR indicate the log of GDP per capita and GDP growth, respectively. RIR denotes the real interest rate.

lTOC33,7903.2861.850−1.60911.234
lTCL33,7906.1581.9511.13112.670
lTSOA33,7905.9911.9921.28112.582
lPF33,790−4.3420.587−7.046−2.618
lPL33,7901.0720.972−2.4933.959
Institutions33,7902.7621.694−4.1974.861
EQOTA33,7909.0898.048−335.13096.630
ROAA33,7900.6712.086−61.93062.560
LGDPPC33,7909.6621.1184.67610.611
GDPGR33,7902.3862.349−13.12727.462
RIR33,7906.9717.688−72.55578.794
Table A2. Correlation matrix: complete set
  lTOC lTCL lTSOA lPF lPL Institutions EQOTA ROAA LGDPPC GDPGR RIR
  1. Notes: lTOC is the log of total operating cost. lTCL and lTSOA are the logs of total customer loans and total securities and other earning assets, respectively. lPF and lPL are the logs of the price of funds and labour. Institutions is a principal component score of the six Kaufmann indicators. EQUOTA and ROAA refer to equity over total assets and return on average assets, respectively. LGDPPC and GDPGR indicate the log of GDP per capita and GDP growth, respectively. RIR denotes the real interest rate.

lTOC1          
lTCL0.9201         
lTSOA0.8760.841        
lPF04−040.018−0.22−0.2311       
lPL0.001−0.10−0.0950.1811      
Institutions−0.0890.070.016−0.224−0.2361     
EQOTA−0.140−0.26−0.2010.2420.140−0.2391    
ROAA−0.037−0.05−0.0220.0480.089−0.1230.2951   
LGDPPC−0.0150.130.090−0.206−0.4000.868−0.189−0.1371  
GDPGR0.0420.030.032−0.0320.152−0.2170.0690.149−0.3161 
RIR0.026−0.06−0.0140.1810.392−0.2000.0620.050−0.266−0.0271
Table A3. Principal components analysis
Component Eigenvalue Difference Proportion Cumulative
Comp15.100554.767970.85010.8501
Comp20.3325730.0202970.05540.9055
Comp30.3122770.1481980.0520.9576
Comp40.1640790.1157840.02730.9849
Comp50.0482950.0060650.0080.993
Comp60.04223.0.0071
Table A4. Weights of the principal components analysis
Variable Institutions
COP:Control of corruption0.4203
LAW:Rule of law0.4289
REG:Regulatory quality0.41
GEF:Government effectiveness0.4269
PIV:Political stability and violence0.3789
VAC:Voice and accountability0.3813
Table A5. Descriptive statistics for institutions: complete set
  Mean SD Min. Max. Observations
Overall0.6639112.134074−4.1971914.860676 N=944
Between 2.109594−3.471264.655147 n=136
Within 0.28587−0.7595141.774501 T-bar = 6.94118
Table A6. Descriptive statistics: reduced set
Variable Observations Mean SD Min. Max.
  1. Notes: lTOC is the log of total operating cost. lTCL and lTSOA are the logs of total customer loans and total securities and other earning assets, respectively. lPF and lPL are the logs of the price of funds and labour. Institutions is a principal component score of the six Kaufmann indicators. EQUOTA and ROAA refer to equity over total assets and return on average assets, respectively. LGDPPC and GDPGR indicate the log of GDP per capita and GDP growth, respectively. RIR denotes the real interest rate.

lTOC15,0523.8651.903−1.60911.234
lTCL15,0526.6942.0771.13112.670
lTSOA15,0526.4402.1161.28112.579
lPF15,052−4.3630.636−7.046−2.621
lPL15,0520.9621.343−2.4933.959
Institutions15,0522.1521.930−4.1974.861
EQOTA15,0529.8828.629−335.13096.630
ROAA15,0520.8312.566−61.93059.890
LLR15,0524.0926.021−0.260140.940
OOIOTA15,0520.0080.022−0.9540.715
LGDPPC15,0529.4471.3974.77210.611
GDPGR15,0522.8752.934−13.12727.462
RIR15,0527.13610.689−72.55578.794
Table A7. Correlation matrix: reduced set
 # 1 2 3 4 5 6 7 8 9 10 11 12
  1. Notes: lTOC is the log of total operating cost. lTCL and lTSOA are the logs of total customer loans and total securities and other earning assets, respectively. lPF and lPL are the logs of the price of funds and labour. Institutions is a principal component score of the six Kaufmann indicators. EQUOTA and ROAA refer to equity over total assets and return on average assets, respectively. LGDPPC and GDPGR indicate the log of GDP per capita and GDP growth, respectively. RIR denotes the real interest rate.

lTOC11           
lTCL20.9171          
lTSOA30.8600.8541         
lPF4−0.014−0.277−0.3091        
lPL5−0.023−0.174−0.1830.2991       
Institutions60.1180.3100.227−0.331−0.3331      
EQOTA7−0.203−0.294−0.2720.2640.268−0.2391     
ROAA8−0.083−0.074−0.0550.0530.154−0.1020.3201    
LLR9−0.015−0.178−0.1090.2570.131−0.3550.107−0.1831   
OOIOTA100.1230.0040.0110.2190.165−0.0980.1100.3540.0381  
LGDPPC110.1690.3410.284−0.304−0.4700.895−0.242−0.138−0.300−0.0971 
GDPGR12−0.081−0.072−0.073−0.0150.173−0.1190.0550.150−0.0730.055−0.2501
RIR13−0.003−0.122−0.0540.2220.397−0.2750.1690.0860.1180.113−0.291−0.061
Table A8. Descriptive statistics for institutions
  Mean SD Min. Max. Observations
Overall0.7136992.073252−4.1971914.860676 N=809
Between 2.07479−3.471264.677545 n=123
Within 0.284869−0.7097271.824289 T-bar = 6.57724

Appendix B: Generalized likelihood ratio test

To test the impact of efficiency on our model, we perform a generalized likelihood ratio (LR) test. The model has two types of variances, variance of the error term and variance of the efficiency term, or inline image and inline image, respectively. The total variance of the model thus is:

  • image( (B1))

Including the inefficiency term adds a certain amount of variance to the model. A measure of this amount, with respect to the total variance, is:

  • image( (B2))

The other effects of including efficiency in the model can be determined by the coefficients δ0δi. The total effect of efficiency can be tested with a null hypothesis that it has no effect:

  • image( (B3))

according to the generalized LR test:

  • image( (B4))

which can be calculated by obtaining the log-likelihood of a standard OLS model (h0) and the log-likelihood of Battese and Coelli's (1995) model. This LR test follows a mixed chi-square distribution with degrees of freedom equal to the number of restrictions (Coelli et al., 1998). The significance level is determined according to Kodde and Palm's (1986) table 1.