2.1. Effect of the institutional environment on banking performance

Institutional environment is relevant to the development of countries because it establishes the conditions for society; its scope is not only social, but also economic. An extensive literature has analyzed the role of the institutional environment on the macroeconomic system of countries, mainly through its impact on financial development. ^{Fernández and Tamayo (2017)} point out that institutional environment improvements facilitate financial development and attract direct foreign investment. This attractiveness is sustained by reductions in information asymmetries and optimal risk allocation for investors (^{Law and Azman-Saini, 2012}). In any case, the effects of the institutional environment on banking performance have been poorly analyzed.

The public-political environment is an important aspect of the institutional framework of countries. Several authors have concluded that a better public and political institutional environment promotes financial development, mostly in the banking sector (^{Gropper, Jahera and Park, 2015}; ^{Yahya, Akhtar and Tabash, 2017}). Political stability and democratic development of the economic system would facilitate banking activities and generate investor confidence, both in public and banking institutions (^{Huang, 2010}). In any case, few studies have investigated the effects of public-political institutions on bank performance. ^{Ben (2013)}, in a study conducted on banks in France, Germany, the United Kingdom and Greece for the period 2005-2011, concluded that a stable and democratic government reduces bank performance. The author argues that this scenario strengthens the State’s supervising role over banking activities and mitigates the adverse selection problem. ^{Yahya, Akhtar and Tabash (2017)} agree with previous studies and add that bad political practices increase the risk of banking activities, allowing banks to achieve higher performance. Even, ^{Gropper, Jahera and Park (2015)} argue that bank performance benefits from political connections. Corruption is another key factor in the public-political environment that impoverishes countries’ institutional quality. This fact promotes inefficiencies and bad practices in public institutions and reduces firm’s investment (^{Dike, 2005}). These bad practices increase the uncertainty surrounding banking activities (^{Méon and Sekkat, 2005}). For these arguments, some studies have found that increased corruption increases bank performance (^{Anaere, 2014}). This can be seen in the fact that banking performance has a systematic component associated with institutional deficiencies in the public-political sphere that turns risk into a greater return.

The private-regulatory environment is another relevant factor of countries’ institutional quality. This aspect is closely related to the State’s responsibility for the design and implementation of effective regulations for private activities. ^{Herger, Hodler and Lobsiger (2008)} state that regulations should promote investor confidence in State’s activities and guarantee investors’ rights protection. When the regulatory system meets these conditions, it facilitates foreign direct investment and the development of a country’s banking sector (^{Marcelin and Mathur, 2014}). However, a higher development of private-regulatory institutional environment would have negative effects on banking performance. ^{Shen and Chang (2006)}, in an empirical study for 8,113 banks in 46 countries, showed that more rigorous regulation reduces bank performance. This finding was also supported by ^{Ben (2013)}. ^{Eisenbach et al. (2017)} argue that the objective of regulation is to anticipate any event that weakens the health of the banking system. From this perspective, lower bank performance would be a consequence of more conservative risk management by banks (^{Eisenbach et al., 2017}), or lower credit growth (^{Barth et al., 2013}). Even regulation may force banks to adopt more conservative accounting criteria that mitigate the practice of manipulating financial statements (^{Kanagaretnam, Lim and Lobo, 2014}). However, ^{Arias, Maquieira and Jara (2020)} in an analysis at the bank level for 52 countries, open a space for empirical debate by finding evidence of a positive relationship between regulatory development and bank performance.

2.2. Effect of market structure and income diversification on banking performance

The banking industry structure and income diversification strategies are factors that many researchers have catalogued as relevant on bank performance. Diverse studies have discussed the effects of market concentration on bank performance. Seminal works established the theoretical bases that associate higher bank performance with more concentrated market structures (^{Klein, 1971}). But the empirical evidence has analyzed this positive relationship in different ways. Some studies have argued that, in concentrated market structures, banks achieve higher performance because they would have greater control over industry prices. This vision is known as the structure-conduct-performance hypothesis (SCP) and several researchers have corroborated it for developed and emerging markets (^{Garza-García, 2012}). Other studies have indicated that banks can reduce their production costs through greater activity volume and generate higher profitability (^{Shepherd, 1983}). These facts are common in the more concentrated banking industry, where large banks take advantage of the scale factor associated to size (^{Berger, 1995}). This perspective is known as the relative-market-power hypothesis (RMP). ^{Chortareas, Garza-García and Girardone (2011)} corroborate this hypothesis for Latin American and Organization of the Petroleum Exporting Countries (OPEC) banks, respectively. They argue that market concentration is associated with a broader financial product portfolio that allows banks to reduce the cost of operational activities.

Although both hypotheses analyze the effect of market concentration on banking performance through different views, their results have shown that banks achieve greater benefits from monopolistic behaviors (^{Jara, Arias and Rodriguez, 2014}). This would generate a smaller number of banks in the industry, where large banks absorb the largest profits in the sector due to a broader portfolio of financial products (^{Maudos and Solís, 2009}). It is important to note that other studies support that an excessive market concentration could have side effects on bank profitability. A higher market concentration could lead to greater banking income instability, incidence of riskier borrowers and lower efforts to maximize profits. These facts could reduce bank performance (^{Mirzaei, Moore and Liu, 2013}).

Financial deregulation also has favored the development of non-traditional activities by banks. Distributing banking costs into various financial products with similar production technologies allows banks to achieve greater benefits in economies of scope than economies of scale (^{Gregoire and Mendoza, 1990}). For this reason, income diversification would increase bank profitability. ^{Lee, Hsieh and Yang (2014)}, in a study for 2,372 banks of 29 Asia-Pacific countries, corroborate that income diversification increases banking performance. ^{Elsas, Hackethal and Holzhäuser (2010)} add that income diversification increases banks’ market value through a study for 380 banks from 9 developed countries. ^{Adem (2022)} demonstrates the positive impact of diversification on the banking performance and stability of African countries. Even more, all these findings are consistent with the RMP hypothesis. In a more concentrated banking industry, larger banks would have higher profitability because they would develop a higher production volume of diversified financial services (^{Deng and Elyasiani, 2008}). Many researches have corroborated these findings for both developed and emerging countries (^{Sanya and Wolfe, 2011}; ^{Jara, Arias and Rodriguez, 2014}; ^{Ali et al., 2022}).

In spite of the above, there is a lack of consensus regarding the effects of income diversification on banking performance. ^{Chiorazzo, Milani and Salvini (2008)} analyzed Italian banks and showed that income diversification strategies reduced bank profitability. They add that income diversification generates a trade-off between risk and return on bank performance. ^{Berger, Hasan and Zhou (2010)} corroborate these findings for Chinese banks and add that income diversification reduces bank performance because it increases their costs. Other researchers support that income diversification should increase bank risk and performance instability because it deepens the adverse selection problem on financial service portfolio (^{Stiroh and Rumble, 2006}). Even more recent studies support that the negative impact of the COVID-19 pandemic on banking performance could have been accelerated by the high degree of diversification as well as capital requirements (^{Susanti, Putra and Bahtiar, 2023}).

4.1. Data

The data were extracted from Global Financial Development (GFD), World Developing Indicators (WDI) and Worldwide Governance Indicators (WGI), all of them from World Bank databases. The data cover annual periods between 1994 and 2019 and were organized into a panel data for 168 countries.

Table 1 shows the geographical composition of the sample. Countries’ distribution according to their income level reveals that 38.69% correspond to high-income countries. Of those countries, more than 50% are located in Europe and Central Asia, and to a lesser degree, in Asia-Pacific and Latin America. Also, 26.79% of countries are of upper-middle income, located mainly in Latin America, Europe and Central Asia, and 23.81% of countries have the lower-middle income classification, represented mostly by countries of Asia-Pacific, Europe and Central Asia, and Middle East and North Africa. Finally, 10.71% are low-income countries, which are almost entirely Sub-Saharan African nations.

Table 2 presents the variables used in this research. The dependent variable was banking performance (BPER) measured by return on the asset (ROA). According to several empirical studies, this ratio compares the net banking income to total assets. This measure is a global performance indicator for banks (^{Mercieca, Schaeck and Wolfe, 2007}; ^{Yahya, Akhtar and Tabash, 2017}). We also used net interest margin (NIM) as an alternative bank performance measure.

The institutional environment (IE) was measured through six proxy variables. Corruption control (CORR), government effectiveness (GOVEF), and political stability (PST) variables measure the country’s public-political institutional quality; regulatory quality (REQL), rule of law (RLAW) and accountability (ACCOUNT) variables represent private-regulatory institutional quality. Each indicator normally fluctuates between -2.5 and +2.5, where positive (negative) values indicate a high (low) institutional quality. Some empirical studies have used this type of measurement to evaluate institutional environment quality, although not necessarily in relation to bank performance (^{Ben, 2013}; ^{Anaere, 2014}).

The effects of market concentration and income diversification on banking performance are variables of interest for our research. The Lerner Index (LER) measures market concentration¹. This indicator oscillates between 0 and 1, values that indicate a competitive and a monopolistic market structure, respectively. Any value within this range represents a noncompetitive market structure for a country’s banking industry (^{Klein, 1971}). We also measured banking concentration through the asset concentration of the five largest banks (BCON). On the other hand, bank income diversification (DIV) was measured by the non-operating income to total income ratio, as has been suggested by different studies (^{Stiroh and Rumble, 2006}; ^{Chiorazzo, Milani and Salvini, 2008}; ^{Lee, Hsieh and Yang, 2014}).

We also incorporate other control variables. At the industry-level, we use the bank capital ratio (CAPASS), bank deposits to gdp ratio (DEP), non-performing loans to total bank loans ratio (NPL), bank credits to gdp ratio (FD), gross margin ratio (EFIC), and the Z-Score indicator as a proxy of banking stability (FST) as previous studies indicate (^{Maudos and Solís, 2009}). At the macroeconomic level, we use the economic growth (GROWTH), annual inflation (INF) and the CRISIS dummy variable (^{Yahya, Akhtar and Tabash, 2017}; ^{Saleh and Abu Afifa, 2020}; ^{Alam et al., 2021}).

4.2. Econometric methodology

To quantify the effects of market concentration and income diversification on banking performance, conditioned to institutional environment quality, we use the Generalized Method of Moments (GMM) estimator for dynamic panel regression proposed by ^{Arellano and Bond (1991)}:

Where BPER _it is the dependent variable that measures bank performance for country i in period t. The BPER _it-1 is the lag in t-1 period for bank performance. Banking performance is controlled over institutional environment quality (IE _it ), where β₂ measures the direct channel of institutional environment on banking performance. LER _it is the Lerner index that measures banking market concentration and DIV _it measures income diversification. The variables (LER _it × HIQ _it ) and (DIV _it × HIQ _it ) measure the indirect channel of institutional environment quality, through market concentration and income diversification, on banking performance. The indirect channel describes the way in which institutional quality conditions the effect of these variables on banking performance. We define HIQ _it as a dummy variable that adopts the value 1 when a country has a high level of institutional quality and 0 otherwise. To distinguish countries’ institutional quality, the value 1 was assigned when the value of each institutional quality indicator was greater than 0. We use this dummy variable to attenuate the multicollinearity problem associated to low variance of institutional development variables.

The model incorporates other k control variables, both at banking industry and macroeconomic levels, grouped in the X _kit matrix (see Table 2). The model also includes temporal effects (ƞ _t ) related to year t and individual effects (ƞ_i) related to country i. It also includes dummy variables associated to crises periods, geographic zones and income levels in order to control the sample heterogeneity. The crisis dummy variable adopts the value 1 in the periods of Asian (1997-1998), Subprime (2008-2009) and European (2011-2012) crises. Finally, ε_it is the remaining random disturbance. Specifically, we use the GMM estimator proposed by ^{Arellano and Bond (1991)} because contemporaneous banking performance can be affected by the persistence of shocks that have occurred in past banking performance (^{Arias, Maquieira and Jara, 2020}). This fact justifies the inclusion of the lagged dependent variable at t-1, which is treated as an endogenous regressor variable. To control the endogeneity problem, we implement the one-step GMM estimator, and we use the lags at t-2 and t-3 of the dependent variable as instruments since they are autocorrelated with the same dependent variable, but not with the error term. However, for the model to be correctly specified, ^{Arellano and Bond (1991)} also pointed out that these estimators must be consistent, and the model must be instrumentally overidentified. To guarantee GMM estimator consistency, the presence of first-order autocorrelation is necessary, but not a higher-order of autocorrelation. The instrumental overidentification of the model was verified through the Sargan Test. All these models were estimated through robust variance.

5.1. Descriptive analysis

Table 3 shows the statistical description, correlations and unit root tests. Banking industry performance at the international level shows an average return on assets of 1.46%. Within this figure, it is worth noting that there is lower bank performance in high-income countries (0.88%) and a higher return for banks in low-income countries (2.04%). In fact, the banking sector of high-income countries is the only one that shows a performance below the world average. Net interest margin (NIM) follows a similar pattern.

Institutional environment quality shows different results according to countries’ income level. Although the public-political and private-regulatory institutional indicators have a value close to 0, the high institutional quality of high-income countries contrasts with the poor institutional environment of low-income countries. Even these indicators correlate negatively and significantly with bank performance measures such as ROA, and NIM. This fact suggests that higher banks returns are also a consequence of countries’ institutional weaknesses.

Preliminarily, market concentration and income diversification are positively correlated with banking performance. According to the Lerner Index (LER) and the five largest banks assets concentration (BCON), the banking market structure is clearly not competitive. Although this quality is very similar among countries, banks in high-income countries exhibit a less concentrated structure than the world average, which could explain the lower industry performance. On the other hand, income diversification indicates that 37.97% of bank income comes from non-operational activities (non-traditional), a figure that rises to 43.25% in low-income countries.

The banking industry-level variables also show interesting results. The capital requirements, which on average oscillate around 10.21%, is positively correlated with bank profitability. This positive correlation is also observed with bank liquidity, operational efficiency and financial stability. These results preliminarily suggest that banks achieve higher returns in banking industries with greater financing independence, liquidity, cost control efficiency, and financial stability. Other banking system characteristics, such as its non-performing loans and banking development are negatively correlated with bank performance. It should be noted that the banks of high-income countries show the lowest capital requirements and non-performing loans, as well as higher levels of liquidity and banking penetration when compared to banks of lower-income countries.

Regarding the macroeconomic characteristics, we observed an annual growth of 4.12%. At this point, there is dissociation between the low growth of high-income countries, with a figure of around 2.98%, and the higher economic activity of low-income countries, whose growth is above the average global growth. Finally, unit root test results indicate that all variables are stationary processes.

5.2. Effect of institutional environment, income diversification and market concentration

Table 4 shows the results of model (1). This model satisfies the necessary specification conditions indicated by ^{Arellano and Bond (1991)}. The GMM estimators are consistent because z-test AR1 reveals the presence of first-order autocorrelation, while the AR2 test discards second order autocorrelation. The Sargan test indicates that the model is instrumentally overidentified. Moreover, the unit root test indicates that ROA is a stationary process. This result guarantees that its lags are not weak instruments. Finally, the Wald test is a chi-square test that, for all cases, supports the model’s global significance.

The institutional environment has significant effects on banks performance. Public-political institutional environment variables, such as corruption control (CORR), government effectiveness (GOVEF), and political stability (PST) have a negative and significant effect on bank’s ROA. Public-political institutional inefficiencies raise the perception of bad practices in this sector, raising the uncertainty surrounding banking activities (^{Méon and Sekkat, 2005}; ^{Anaere, 2014}). These fragile institutional conditions favor banking performance, but would accentuate adverse selection problem (^{Dike, 2005}; ^{Ben, 2013}). Private-regulatory environment variables, such as regulatory quality (REQL), rule of law (RLAW) and accountability (ACCOUNT), also have a negative and significant impact on ROA. The regulation on private activities, such as banking, inhibits the adverse selection and risky policies of these institutions (^{Ben, 2013}). These conditions reduce the credit expansion rate or avoid accounting manipulation in bank financial statements (^{Barth et al., 2013}). Thus, bank performance would be reduced. The direct channel of the countries’ institutional quality on banking performance reveals that their profitability is supported by institutional, political and regulatory weaknesses.

Market concentration is also another key factor affecting bank performance. Our results indicate that the Lerner index (LER) has a positive and significant effect on bank’s ROA. This result is not a novelty because several past studies have supported this relationship (^{Jara, Arias and Rodríguez, 2014}). However, country institutional quality works as an indirect channel that affects banks performance through market concentration. The interactive variable (LER × HIQ _it ) has a negative and significant effect on ROA. Banking market concentration reduces banking profitability in high institutional quality countries, a result that corroborates hypothesis H1a. This result is important for financial policy design because it shows that banking competition (concentration) is compatible with higher (lower) performance, an idea that had been largely discarded by international evidence. Thus, better institutional quality, public-political and private-regulatory, is essential for banking regulation to promote sector competition and higher performance. This result also reveals that the positive effect of market concentration on bank performance is attributable to institutional fragilities.

Income diversification strategy is another relevant factor affecting bank performance. Table 4 shows that the DIV variable has a positive and significant effect on ROA. Income diversification generates a greater banking return given that economies of scope result in a greater benefit for banks compared to economies of scale. Although there is still no clear consensus on this matter, this result is in line with several international studies (^{Stiroh and Rumble, 2006}; ^{Elsas, Hackethal and Holzhäuser, 2010}; ^{Sanya and Wolfe, 2011}; ^{Lee, Hsieh and Yang, 2014}). However, the quality of the institutional environment also affects how income diversification impacts bank performance. The variable (DIV × HIQ _it ) has a negative and significant effect on ROA, which validates hypothesis H2a. Income diversification in countries with high institutional quality reduces banking profitability. So, this fact induces them to reduce their non-traditional operations or develop more specialized services to support their performance. This result suggests that income diversification is a risky source of banking profitability, and it is related to institutional fragilities.

5.3. Robustness analysis

We sensitized model (1) through the use of net interest margin (NIM) as alternative measure of bank performance. The NIM is measured by interest-minus-expense to total assets ratio. Model (1) was estimated through the one-step GMM-SYS estimator proposed by ^{Arellano and Bover (1995)}. Obviously, the GMM-SYS estimator of ^{Arellano and Bover (1995)} is a more general case than the traditional GMM estimator of ^{Arellano and Bond (1991)} because it allows controlling the endogeneity associated with the lag of the dependent variable as well as a regressor variable, in this case, the operational risk (NPL). According to ^{Stiglitz and Weiss (1981)} and ^{Saleh and Abu Afifa (2020)}, a higher bank return is endogenously related to operational risk. The authors argue that banks have incentives to take on higher risks when they seek higher returns by granting funds to riskier borrowers. However, this decision would diminish the quality of banks’ credit portfolio because it accentuates the adverse selection problem. From an econometric point of view, we also use the lags at t-2 and t-3 of the dependent variable (BPER) and the regressor variable (NPL) as instruments. Table 5 shows the results of the robustness tests using NIM as dependent variables, and the five largest banks’ assets concentration (BCON) as a measure of market concentration. These robustness tests support all previous estimates.