Between Stability and Exposure: The Dual Effects of Income Diversification on Bank Risk

Abstract

The traditional role of banks as intermediaries has been transferred to a vast array of businesses, creating many sources of income. The present study examines the impact of income diversification on bank risk. A total of 565 commercial banks from 50 countries were examined. A dynamic panel data analysis using Maximum Likelihood with Structural Equation Modelling was performed. The study found that income diversification has no significant effect on risk-weighted assets, while it reduces the insolvency risk and liquidity risk of the bank. Multiple proxies were utilized to measure bank risk to increase the robustness of the study. The study stressed the importance of income diversification and efficient capital allocation across various investment projects to survive in a highly competitive environment. Overall, this study provides new insights into the contradictory relationship between income diversification and bank risk in the global context. This would assist in developing strategies and policies to reduce risk and increase stability in the banking sector.

1. Introduction

A bank manages a range of risks, including credit risk, liquidity risk, interest rate risk, operational risk, and others. Evidence shows that a bank’s problem arises when it faces too many liabilities coming due and does not have enough cash to meet them, ultimately resulting in liquidity risk (World Bank Group, 2017). The global financial crisis database by Laeven and Valencia (2012) identifies 147 systemic banking crises from 1970 to 2011. Cihák and Schaeck (2010) suggest that rapid credit growth and high leverage predict the likelihood of a banking crisis. Hussein (2000) also attributed Egypt’s liquidity crisis to a significant increase in private credit. A bank crisis is also exacerbated by insufficient capital, inefficient credit risk management, and insufficient portfolio diversification. (Casu et al., 2012). Market integration, financial deregulation, and increased competition have changed banking systems’ scope and operational activities. Increasingly, banks are transferring the traditional role of intermediation to a broader spectrum of businesses (Githii & Mwangi, 2018; Moudud-Ul-Huq et al., 2020). Banks have created new income streams through diversification. However, empirical evidence finds an inconclusive effect of this income diversification on bank risk (Paltrinieri et al., 2021; Sharma & Anand, 2018; Vidyarthi, 2019) in different economies. It has been found that income diversification adversely impacts bank stability, whereas asset diversification improves it (Abbas et al., 2021). Additionally, empirical evidence suggests that bank size, capital, and ownership structure influence diversification decisions.

Although diversified income can cause other risks, such as market risk, credit risk, and operational risk (Abbas et al., 2021), the theory of portfolio suggests that diversified sources of income, including non-interest income, may diversify risk (Moudud-Ul-Huq et al., 2020; Sharma & Anand, 2018). However, empirical evidence presents a contradictory relationship between income diversification and bank risk (Hoang et al., 2021; Paltrinieri et al., 2021; Vidyarthi, 2019) in different economies. Studies have also found that the cost of increased volatility in income offsets the benefits of income diversification (Stiroh, 2004, 2006).

Several key factors have motivated this study on income diversification and bank risk from a global perspective. First and foremost, it emphasizes the importance of risk management and income diversification in the banking sector, drawing on the lessons learned from the global financial crisis of 2007–2008. Secondly, the study intends to understand how banks adjust their business models and income streams in response to changes in the banking environment. It includes technological advancements and regulatory reforms, such as capital adequacy requirements. Additionally, the study acknowledges the global nature of banking and seeks to provide insights into income diversification and bank risk in different countries and regions. Finally, the study is motivated by its policy relevance, contributing to the development of policies and regulations by which the banking sector can become more stable and resilient. It aligns with the broader literature on banking risk management, financial stability, and income diversification’s impact on bank profitability.

Accordingly, this study examines the effects of income diversification on bank risk globally among large commercial banks from 2011 to 2015. Income diversification and other bank-specific variables were re-examined to determine how they affect bank risk. This study empirically examines the impact of non-interest-based income on the risk of commercial banks of Basel Committee members and non-member countries during the post-financial crisis period of implementing Basel III regulations. The result of this study indicates that income diversification is not significantly related to the bank’s risk-weighted assets. Nevertheless, income diversification decreases a bank’s insolvency and liquidity risk.

This paper extends the existing studies in several ways. Firstly, existing studies on the effect of income diversification on bank risk focus on a single economy (Vidyarthi, 2019; G. J. Wang et al., 2018), cross-country (Chunhachinda & Li, 2014; Sharma & Anand, 2018), or a specific region (C. Wang & Lin, 2021). The present study was conducted among 565 commercial banks operating in 50 countries from 2011 to 2015. Therefore, it contributes to the existing literature by adding to the findings in existing studies on income diversification and bank risk with global context. It clarifies key aspects of the debate by emphasizing the importance of income diversification for reducing bank risk and improving resilience. The study emphasizes income diversification’s effectiveness as a risk management strategy. It emphasizes the importance of striking a balance between encouraging income diversification and ensuring effective risk management practices. Furthermore, the bank’s risk is quantified from various perspectives using proxy variables covering credit risk, liquidity risk, and insolvency risk. This makes the study more robust. In addition, the study utilizes the Maximum Likelihood (ML) approach with the Structural Equation Modelling (SEM) technique to resolve the issues of endogeneity, autoregressive, and time-varying variables. In comparison with other methods of panel data analysis, such as 3SLS, Random effect, and GMM, it is more efficient. Thus, the current study also contributes to the existing literature on a methodological level.

The study is organized as follows: Section 2 presents theoretical and empirical evidence in the Literature Review, Section 3 provides the Methodology of the study, Section 4 presents the Results of the study, and finally, Section 5 and Section 6 present the Discussion and Conclusions.

2. Literature Review

2.1. Income Diversification and Bank Practices

Diversification is defined as the allocation of resources in a way that will reduce the risk exposure from any specific source, asset, or investment. The doctrine of diversification is to spread the investment across different assets in such a way that it will reduce some risk (Ross et al., 2008). The unsystematic risk can be reduced through a portfolio of different assets or incomes. Banks also reduce their risk by diversifying income or assets in different investments, firms, or geographic locations (Hsieh et al., 2013). Income diversification refers to the increase in non-interest or non-traditional income of the bank.

In the Gulf Cooperation Council (GCC) market, Maghyereh and Yamani (2022) find that banks can also decrease systematic risk through income diversification. An increase in the share of non-interest or non-traditional income indicates a bank’s increased concentration on other operating activities, i.e., to lessen the dependency solely on lending activities. It increases a bank’s income without increasing its related risk-weighted assets, i.e., credit risk. Therefore, income diversification increases the risk-adjusted rate of return of the bank (Chiorazzo et al., 2008). However, the empirical evidence presents a conflicting result for income diversification with bank risk.

Since the bank acts as an intermediary, it collects deposits, disburses loans, and earns from the spread of interest rates between lending and deposits. Increased competition in the banking industry causes a decline in the traditional source of margin income (B. Williams & Prather, 2010). Therefore, the increase in non-interest or non-traditional income will diversify the earnings of the bank by increasing fee-based income or service charges, trading activities, and underwriting insurance activities (Stiroh, 2006). Empirical studies by Park (2013) in South Korea and B. Williams and Prather (2010) in Australia found that comparative revenue from interest income has decreased over time, which implies that that the diversification of income is increasing.

2.2. Income Diversification, Performance, and Bank Risk

Winton’s (1999) model of “not putting all your eggs in one basket” proposes that diversification across loans can reduce failure probability and increase monitoring incentives. In a model for banks’ credit diversification, he demonstrated that diversification across sectors benefits the bank when the monitoring of loans adds value and bank failure is costly. Winton’s model also analyzed the effect of competitive position on the decision to diversify. When the environment is highly competitive, diversification becomes costlier. In that case, specialization in any sector brings more benefits than diversification. Moreover, in line with capital adequacy regulation, Winton observed that diversification in new areas requires more capital than concentrating on a preferred sector.

Income diversification provides the advantage of reducing the risks associated with solely relying on one source of income. The diversification theory suggests that income diversification reduces bank risk by reducing exposure to a specific sector or asset, but the existing literature finds inconsistent effects on the bank’s risk (Brahmana et al., 2018; Hsieh et al., 2013; Hoang et al., 2021; C. Wang & Lin, 2021). Banks with less diversified income are riskier, and income diversity enhances the stability of the banks. Ammar and Boughrara (2019), Brahmana et al. (2018), Hsieh et al. (2013), and C. Wang and Lin (2021) explained that a wide range of activities helps the banks to expand their source of return and at the same time reduces the risk. Hsieh et al. (2013) also point out that asset and income diversity improve the bank’s stability. Lin et al. (2012) find that, through income source diversification, some Asian countries reduced their shocks during the 1997–2005 period. Moreover, other studies (Hsieh et al., 2013; Nguyen et al., 2012) have found that bank income diversification can enhance the organization’s stability.

Furthermore, in a study on the selected commercial banks of OIC, Paltrinieri et al. (2021) found that income diversification positively contributes to the profitability of commercial banks. However, they have not found any relationship between income diversification and stability for both conventional and Islamic Shariah-following banks. In addition, Li et al. (2021) observed asymmetrical results in a study on the effect of income diversification on bank performance and risk during the COVID-19 pandemic. The study found that income diversification is positively linked to the performance of the bank and inversely related to the risk of the bank, commensurate with the findings of Luu et al. (2019), Nepali (2018), and Uniamikogbo et al. (2021).

However, Boyd et al. (1998) observed that managers receive more incentives to take excessive risks due to moral hazard problems when the banks are engaged in a wide range of activities. In this case, these banks become more complex to monitor and control, which decreases competition and efficiency. Similarly, another study on the banking industry of the U.S. by Stiroh (2006) found that though non-interest income increases the revenue of the bank, it increases the volatility of revenue and the risk of the bank. Several studies observed outweigh the diversification advantage due to the cost of increased volatility in activities like securities trading (Abbas et al., 2021; Le, 2021; B. Williams & Prather, 2010).

Likewise, empirical studies found an inverse relationship between income diversification and bank performance. A study on European banks from 2002 to 2012 observed that an increase in non-traditional income affects the bank’s profitability inversely during the global financial crisis (Maudos, 2017). It was also found in a panel data study of Ghanaian commercial banks from 2000 to 2015 that income diversification negatively affected profitability and stability (Duho et al., 2020). Furthermore, Abuzayed et al. (2018) found that income diversity is not associated with bank stability in the Gulf Cooperation Council (GCC). Macroeconomic conditions and institutional quality play a significant role in determining bank stability. An important factor in income diversification is the size of the bank. In several studies, the size of the bank has been associated with the diversification strategy. The economies of scale enable large banks to benefit from diversification benefits and gain a positive correlation between diversification and risk-adjusted returns (Duho & Onumah, 2019; Meng et al., 2018) compared to smaller institutions. Due to their greater strategic experience and better equipment, larger banks can invest more in non-traditional income sources and manage the associated risks (Brighi & Venturelli, 2014). The study also observed that the bank’s risk reduces as capital increases, while non-traditional activities and profitability increase. Moreover, some studies observed a U-shaped relationship between income diversification and bank performance and stability (Alhassan, 2015; Duho et al., 2020).

The current empirical evidence does not agree on the relationship between income diversification and bank risk. However, this study examines the impact of revenue diversification on commercial bank risk based on Winton’s (1999) theory. Specifically, the paper explains how increasing non-interest income to total operating income influences bank risk. Because of the above, the study expects an inverse relationship between income diversification and bank risk. Therefore, the following hypothesis is developed:

Hypothesis: 

Income diversification has a negative impact on the risk of banks.

In addition to investigating the impact of income diversification on bank risk, the study also attempts to clarify the influence of bank-level control variables, such as bank size, loan-to-total assets ratio, non-performing loan ratio, capital ratio, and macroeconomic variables, such as GDP growth and domestic credit to the private sector, on bank risk.

3. Materials and Methods

3.1. Sample Size and Data

It focuses on commercial banks following conventional systems implementing the Basel II or the Basel III regulations. The present study uses data from 2011 to 2015, which were deliberately chosen to capture the post-implementation effects of Basel II and the early adoption phase of Basel III across member and non-member countries. This timeframe reflects a transitional phase in global banking regulation: Basel II was being implemented in non-member countries around 2010–2011, and Basel III began taking effect among member countries from 2013 onward. By selecting this timeframe, the study aims to observe regulatory impacts during relative stability and comparability. From 2011 to 2015, the total number of active commercial banks from those 59 countries was 9412. The research employs the availability of data to finalize the sample size. As the study utilizes balanced panel data, only banks that are available throughout the entire study period (2011–2015) and have complete variable information are considered. The study begins with a total of 9412 commercial banks. The total number of commercial banks for the member and non-member countries reaches 565. The sample size was reduced because data were not available for the entire study period from 2011 to 2015. Banks that began operations after 2010, as well as those that merged or closed during this timeframe, were excluded from the study. Additionally, some banks were excluded due to the absence of specific information in the Bankscope database. Appendix A presents the list of sample countries and the number of banks observed in the present study. Banks in countries with missing data on bank-specific or macroeconomic variables were excluded. Therefore, the dataset is strongly balanced with panel data. Since one of the variables to measure the bank’s risk is calculated by the changes over the years, the study period becomes 4 (5 − 1) and 2260 (4 × 565) Bank year observations. The data is obtained from the Bankscope database, World Bank database, IMF database, Bank for International Settlements (BIS), annual reports, and websites of the sample banks. The present study includes only commercial banks. Due to the unique characteristics of Islamic banks following Shariah law, they are not included in the study. Appendix B presents the descriptive statistics of the study.

3.2. Variables

In addition to the difficulty of defining the bank’s risk, the measures used to quantify the risk are also limited (Beck et al., 2009). As a result, different proxies are used to measure risk. The present study evaluates bank risk using three proxy variables. The variables used in this study are changes in the total risk-weighted asset ratio (RWAR), liquidity ratio (LR), and Z-score. These variables represent the bank’s credit, liquidity, and default risks. Since there is no single technique to define risk, the present study used these three proxies as the dependent variable.

3.2.1. Change in Risk-Weighted Asset Ratio (RWAR)

The risk-weighted asset ratio represents the bank’s share of risky assets. It is a widely used proxy variable that measures the bank’s total risk exposure regarding assets (Shehzad et al., 2010). Risk-weighted assets are the sum of risk-weighted assets associated with credit, market, and operational risks (BCBS, 2004). Accordingly, the risk-weighted asset ratio is the ratio of the total risk-weighted assets to the total assets. This study uses the risk-weighted asset ratio (RWAR) to assess the relationship between risk change and other independent factors.

Change in RWAR = (RWAR_i,t − RWAR_i,t−1)/RWAR_i,t−1 × 100.

Here, RWAR_i,t = Risk-weighted asset ratio of the bank i for year t.

RWAR_i,t−1 = Risk-weighted asset of the bank i for the previous year (t − 1).

3.2.2. Z-Score

Z-score is the relative state of capital in terms of return volatility. It measures the ability of the capital to absorb adverse shocks due to variable returns (Haque, 2019) and is used as a proxy for estimating the probability of a bank’s default (Ashraf et al., 2016; Lepetit & Strobel, 2015; Laeven & Levine, 2009). In other words, it combines accounting measures such as profitability, leverage, and volatility of returns. Since the Z-score indicates the distance from insolvency, a bank with a higher Z-score is more stable. In addition, banks with low Z-scores are more likely to default. The Z-score calculation is as follows (Haque, 2019; Lepetit & Strobel, 2015):

Z-score = (Capital Ratio + Return on Asset)/Standard deviation of Return on Asset

Here, Capital Ratio = Total Capital/Total Assets

3.2.3. Liquidity Ratio

It represents the bank’s liquidity risk, defined as the inability to meet any cash or collateral obligation without suffering unacceptable losses (BCBS, 2008). The liquidity ratio also indicates the potential risk to earnings and market value of equity arising from the bank’s inability to meet payment obligations cost-efficiently and in due time (MacDonald & Koch, 2015). A lower liquidity ratio indicates liquidity or insolvency risk. The liquidity ratio is calculated as follows.

LR = [Total liquid assets/(total deposits + total borrowings)] × 100

In the present study, the independent variables used to explain the bank risk are capital ratio (CR), income diversification (IND), bank size (SIZE), loans to total asset ratio (LTTA), non-performing loan ratio (NPL), growth or GDP (GGDP), and domestic credit to the private sector (DCPS). Table 1 presents the list of dependent and independent variables, measurements, and data sources used in the present study.

Table 1. Variable measurement and source of information.

Variables	Definition	Measurement	Source
Dependent Variables
RWAR	Risk-weighted asset ratio	Risk-weighted asset ratio = (Risk-weighted asset/Total asset)	Bankscope database and authors’ calculation
Z-score	Z-score	Z-score = (ROA+ capital ratio)/standard deviation of ROA	Bankscope database and authors’ calculation
LR	Liquidity ratio	Liquidity Ratio = 100 × [Total liquid assets/(total deposits + total borrowings)]	Bankscope Database
Independent Variables
IND	Income diversification	Non-interest income/total operating income	Bankscope Database
SIZE/LN TA	Bank size	Natural logarithm of total bank assets	Bankscope Database
LTTA	Loans to total asset ratio	Total loans/total assets	Bankscope Database
NPL	Non-performing loan ratio	Total non-performing loans/total loans	Bankscope Database
CR	Capital ratio	Total capital/total risk-weighted assets of the bank	Bankscope Database
GGDP	Growth of GDP	Rate of change in GDP from one year to another of a country.	World Bank database
DCPS	Domestic credit to private sector	Financial resources provided to the private sector by the financial organization, measured as a percentage of the GDP	World Bank database

3.3. Model

The trade-off between risk and return is an integral part of banking transactions (Ahmed, 2015). Performance and risk are closely related in the banking industry. Bank performance is influenced by the risk-taking practices of the bank (Moudud-Ul-Huq et al., 2020), while the risk-taking attitudes of the bank can be influenced by the performance of the bank. According to the Moral Hazard hypothesis, poorly performing banks are more likely to engage in risk-intensive activities compared with highly performing banks (Moudud-Ul-Huq et al., 2020).

Therefore, to find out the effect of a bank’s income diversification on risk, the present study follows the model developed by Shrieves and Dahl (1992). It is modified to incorporate the associated variables analyzed in the present literature. According to Shrieves and Dahl (1992), rather than analyzing the relationship between risk and income diversification, we examine the relationship between changes in risk and the share of non-interest income to total operating revenue as an indicator of income diversification changes.

∆RISK_i,t = α0 + α1 ∆IND_i,t + α2SIZE_i,t + α3LTTA_i,t + α4 NPL_i,t + α5CR_i,t + α6GGDP_i,t + α7DCPS_i,t − α8RISK_i,t−1 + ε_i,t

(1)

Another two risk equations are formulated to quantify risk by the proxy variables of the Z-score and the liquidity ratio. Z-score and liquidity ratio are widely used to measure the bank risk in terms of default risk and liquidity risk, respectively (Ashraf et al., 2016; Laeven & Levine, 2009). Incorporating bank-specific and macroeconomic variables, the present study formulates the following equations:

Z_i,t = γ0 + γ1IND_i,t + γ2SIZE_i,t + γ3LTTA_i,t + γ4NPL_i,t + γ5GGDP_i,t + γ6DCPS_i,t + ν_i,t

(2)

LR_i,t = δ0 + δ1IND_i,t + δ2SIZE_i,t + δ3LTTA_i,t + δ4NPL_i,t + δ5CR_i,t + δ6GGDP_i,t + δ7DCPS_i,t + μ_i,t

(3)

The risk Equations (1)–(3) express the effect of different explanatory variables of bank i during the period of t. In risk Equation (1), RISK_i,t−1 is the risk-weighted asset (RWA) of bank i in the previous year (t − 1). In risk Equation (2), the explanatory variable Capital ratio (CR) is not included, as the calculation of the dependent variable Z-score also employs the capital ratio. Since the Z-score exhibits an inverse relationship with risk, the sign of the coefficients with the Z-score to explain risk (Model 2) should also be interpreted inversely. Similarly, the relationship between the independent variables and the liquidity ratio (Model 3) is interpreted inversely in terms of risk. This is because a lower liquidity ratio indicates a bank’s higher liquidity risk, while higher liquidity ratios indicate less risky positions.

Many studies use Shrieves and Dahl’s (1992) capital and risk model. However, the current model differs significantly from the Shreve model because it attempts to explain risk using different explanatory variables. This study aims to assess the impact of income diversification, bank size, and macroeconomic variables on earnings.

3.4. Method of Analysis

Considering the panel nature of this study, linear dynamic panel data model estimation using the Maximum Likelihood (ML) and Structural Equation Modelling (SEM) is followed (Allison et al., 2017). The dataset is cross-sectional, has a medium time dimension, and is balanced in nature. Because of potential bank-specific effects and possible endogeneity problems (Law et al., 2013), standard panel models, i.e., the Pooled OLS regression model, random effect model, and Fixed effect model, are not suitable. Moreover, estimation by OLS on dynamic panel data can also be biased due to the correlation between lagged dependent variables and the error term. The present study employs the Maximum Likelihood approach to resolve endogeneity, autoregressive, and efficiency. In dynamic panel data, the endogeneity issues can be addressed by the Maximum Likelihood (ML) estimation of Structural Equation Modelling (SEM) (Allison et al., 2017).

Moreover, the ML-SEM approach is more efficient than the GMM method in the event of the absence of finite sample bias. Furthermore, the SEM approach is more advantageous than GMM because of the absence of “incidental parameters” problems. It can also handle the missing data problem (Allison et al., 2017). Therefore, the study chooses to employ the ML-SEM approach. Additionally, the ML approach with SEM is most effective when the data are small in time and the panel is large (R. Williams et al., 2015). A series of alternative tests i.e., Likelihood ratio, Akaike’s information Criteria (AIC), Bayesian Information Criterion (BIC), RMSEA, Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Standardized Root Mean Square Residual (SRMR), and Coefficient of Determination (CD) was performed to determine the model fitness, based on ML estimation. The statistical software used to perform the analysis is Stata 14.0.

4. Results

4.1. Findings

The collinearity among the independent variables was measured before the estimation of the model. Table 2 presents the correlation among the explanatory variables in the regression model. A weak correlation between the variables predicts the absence of multicollinearity in the model. If the correlation between the variables exceeds 0.7, the multicollinearity issue becomes a problem (Kennedy, 2008). Hence, the weak collinearity implies that the model is not biased by collinearity among the explanatory variables.

Table 2. Pearson correlation matrix among explanatory variables.

	RWA	IND	LN TA	LTTA	NPL	CR	GGDP	DCPS
RWA	1
IND	−1.37 **	1
LN TA	−2.12 **	0.113 **	1
LTTA	0.454 **	0.324 **	−0.108 **	1
NPL	−0.76 **	0.094 **	−0.087 **	−0.86 **	1
CR	−1.77 *	0.133	−0.149 **	−0.25 **	0.11 **	1
GGDP	0.100 **	0.027 *	0.143 **	−0.07 **	−0.12 **	−0.02 *	1
DCPS	0.164 **	−0.244 **	−0.046 **	0.17 **	−0.26 **	−0.14 **	−0.47 **	1

* Correlation is significant at 0.05 level (two-tailed). ** Correlation is significant at 0.01 level (two-tailed).

Table 3 presents the Variance Inflator Factor (VIF), used to measure the linear dependence among three or more variables, unlike the correlation between two variables. Therefore, it helps to detect multicollinearity more precisely. It represents the value of 1/(1 − R²). The rule of thumb is that VIF exceeding the value of 5 obliges further review and VIF exceeding 10 indicates high multicollinearity and direct corrections (Kennedy, 2008). In the present study, the maximum value of VIF 2.02 indicates the absence of multicollinearity in the model. Therefore, all of the variables can be used to estimate the model.

Table 3. VIF Matrix among the explanatory variables.

Variables	VIF	1/VIF
Income diversification	1.29	0.775
LN total asset	1.52	0.658
Loans to total assets	1.80	0.556
NPL ratio	1.26	0.794
Capital ratio	1.27	0.787
Growth of GDP	2.02	0.495
Domestic credit to private sector	2.00	0.500

Table 4 presents the regression results for the models explaining the bank’s risk from different perspectives: credit risk measured by risk-weighted assets, default risk, and liquidity risk.

Table 4. Dynamic panel data estimation with ML and SEM for models (1), (2) and (3).

Variables	Model 1 (RWAR) *	Model 2 (Z-Score) *	Model 3 (LR) *
Lag ∆RISKi,t−1	0.0002 (0.49)	-	-
Lag Z-score	-	0.47 (11.87 *)	-
Lag liquidity ratio	-	-	0.29 (7.85 *)
∆Income diversification	−0.0007 (−0.25)	0.04 (4.23 *)	0.08 (2.24 **)
LN total asset	−0.0248 (−0.819)	0.0553 (3.36)	0.13 (1.04)
Loans to total asset ratio	0.0115 (1.68 *)	0.0580 (1.80 *)	−0.466 (−11.66 *)
NPL ratio	0.0323 (6.26 ***)	0.0724 (6.22 *)	0.080 (3.24 *)
Capital ratio	−0.015 (−2.47 **)	0.27 (32.56 *)	−0.04 (−2.15 **)
Growth of GDP	0.0439 (1.90 **)	−0.0031 (−0.35)	0.01 (0.56)
Domestic share of credit to private sector	0.4030 (1.46)	−0.0191 (−0.50)	−0.11 (−1.2)
Lag RISKi,t−1	−0.0617 (25.99 *)	-	-

Note: The z values are shown in parentheses; *, **, *** represent statistically significant at 1%, 5% and 10%, respectively, in a two-tailed t-test. * Dependent variables RWAR, Z-score, and LR.

Model (1) expresses bank risk in terms of the bank’s risk-weighted asset ratio and the coefficient of the lagged dependent variable (∆RISK_i,t−1), which is positive and statistically insignificant. In models (2) and (3), the lagged Z-score and liquidity ratio coefficients are both positive and statistically significant at 1%.

In model (1), the coefficient for the change in income diversification is negative and matches the prediction, but is statistically insignificant. However, in models (2) and (3), the coefficient of change in income diversification for both models is positive and statistically significant, respectively, at 1% and 5% levels, which supports the study’s hypothesis. Since the Z-score and the liquidity ratio both indicate an inverse relation with risk, the sign of the coefficient with the Z-score and the LR in explaining risk should be interpreted inversely. This implies that the increase in income diversification will increase the bank’s solvency, reducing the bank’s insolvency or default risk. Similarly, the increase in non-interest income increases the performance as well as the liquidity position of the bank.

The coefficient for the logarithm of total assets representing bank size in model (1) is negative and statistically insignificant. Equations (2) and (3) both present a positive association of bank size with the Z-score and the liquidity ratio, implying an inverse relation between bank size and insolvency risk and liquidity risk. However, this relation is not statistically significant for all models.

In model (1), the coefficient for loans to total assets is positive and statistically significant at the 1% level. This indicates that the increase in the loans to total asset ratio, i.e., an increase in loans or traditional sources of income, will increase the bank’s risk, which matches the expectation of the study. In model (2), the coefficient for loans to total assets is positive with the Z-score and statistically significant at the 1% level. However, the loans to total asset ratio in model (3) presents a negative and significant relationship with the liquidity ratio, i.e., a positive association between loans to total assets and liquidity risk. This justifies the idea that due to the increase in loans, the liquid assets of the bank decrease, implying an increase in liquidity risk.

The coefficient for the NPL ratio is positive and statistically significant at the 10% level in model (1), with risk measured in terms of change in risk-weighted assets. However, in model (2) and model (3) the coefficient for NPL is also positive and significant, implying the negative association of non-performing loans and insolvency risk and the liquidity risk of the bank.

Furthermore, in model (1) the capital ratio is negatively associated with the RWAR and is significant at the 10% level. This states that the risk of the bank will decrease due to the increase in capital ratio. Similarly, in model (2), the positive association of capital ratio with the Z-score is significant at the 1% level, which implies that the increase in capital will enhance the solvency of the bank. However, in model (3)’s equation for liquidity risk, it is observed that the capital ratio is significant and negatively associated with the liquidity position of the bank. It states that with the capital increase, the liquidity of the bank decreases, indicating an increase in liquidity risk.

Finally, macroeconomic variables like GDP growth and domestic credit to the public sector were included to determine the effect on banks’ credit and stability risk. In model (1), the GGDP is found to be positive and significantly associated with the risk-weighted RWAR or the total risk of the bank. In models (2) and (3), the coefficients for GGDP are not statistically significant. Similarly, the result for the relation of domestic credit to the private sector and bank risk is not statistically significant in any of the risk models.

4.2. Fit Statistics of the Model

Table 5 presents the fit statistics for the models (1), (2), and (3). For model (1), model vs. saturated chi-square is the test used for 53 over-identifying restrictions. The RMSEA value of 0.093 is close to the accepted level of 0.05 to 0.10. The standardized root mean squared residual or SRMR of 0.015 also meets the criteria of good fit (<0.08). Moreover, for model (1), the coefficient of determination, or the R2, is 0.711 or 71.1%.

Table 5. Goodness of fit indices for models (1), (2) and (3).

FIT Statistics	Model 1 (RWAR) *	Model 2 (Z-Score) *	Model 3 (LR) *
Likelihood ratio
Chi2_ms (53)	312.49 Prob > chi2 = 0.00	-	-
Chi2 _bs (93)	895.99 Prob > chi2 = 0.00	-	-
Chi2 ms (48)	-	336.735 Prob > chi2 = 0.00	-
Chi2 bs (84)	-	7184.68 Prob > chi2 = 0.00	-
Chi2_ms (53)	-	-	219.962 Prob > chi2 = 0.00
Chi2 _bs (93)	-	-	3768.302 Prob > chi2 = 0.00
Wald test
Chi2 (14)	704.21	-	-
Chi2 (13)	-	1452.29	-
Chi2 (14)	-	-	486.94
Population error
RMSEA	0.093	0.100	0.080
90% CI, lower bound	0.083	0.093	0.069
Upper bound	0.103	0.110	0.090
Pclose	0.000 Prob RMSEA ≤ 0.05	0.000 Prob RMSEA ≤ 0.05	0.000 Prob RMSEA ≤ 0.05
Information criteria
AIC	18,981.404	18,266.866	21,359.871
BIC	21,327.627	20,205.427	23,298.432
Baseline comparison
CFI	0.677	0.959	0.953
TLI	0.433	0.929	0.918
Size of Residuals
SRMR	0.015	0.003	0.005
CD	0.711	0.998	0.979

Note: LR = Liquidity ratio, RMSEA = Root mean square error of approximation, AIC = Akaike’s Information Criterion, BIC = Bayesian Information Criterion, CFI = Comparative Fit Index, TLI = Tucker–Lewis Index, SRMR = Standardized Root Mean Squared Residual, CD = Coefficient of Determination. * Dependent variables RWAR, Z-score, and LR.

For model (2), the RMSEA value of 0.10 meets the accepted level of 0.05–0.10 and is marginally better than the poor fit. The CFI value of 0.959 confirms the good fit of the model, and the Tucker–Lewis Index (TLI) of 0.929 meets the acceptable level of 0.90, near to good fit (0.95), of the model. The SRMR value also meets the good fit (<0.08). The coefficient of determination is 99.8%, which indicates that the model is a good fit. In model (3), the chi-square value suggests a good fit of the data. Moreover, other fit statistics like RMSEA, CFI, TLI, SRMR, and R2 also confirm the good fit of the model.

5. Discussion

The study demonstrates that the increase in non-interest income reduces the insolvency risk and enhances the performance and liquidity position of banks. This result matches the findings of Ammar and Boughrara (2019), Brahmana et al. (2018), and Li et al. (2021). When a bank increases its engagement in non-traditional activities like fee-based income or service charges, then it increases returns with fewer risk-weighted assets. This increases the bank’s solvency and decreases the bank’s insolvency risk in terms of Z-score. This result corresponds with the findings by Haque (2019). Similarly, due to the increase in income diversification, the liquidity position of the bank improves. The findings are consistent with Hsieh et al. (2013) and Mundi (2019).

An increase in non-interest or non-traditional income sources generates downward disbursement of loans and advances, enabling the bank to attain enough liquid assets to meet its obligations and reducing the liquidity risk. Nevertheless, Saif-Alyousfi and Saha’s (2021) research challenges this conclusion. The present study findings contradict the evidence about the banking sector of the Gulf Cooperation Council (GCC). In particular, their research revealed that non-traditional activities within highly capitalized banks increased risks and decreased stability. The absence of a statistically significant relationship between bank size and risk suggests that there is no significant association between bank size and bank risk. Nevertheless, these findings contradict Sharma and Anand (2018)’s cross-country analyses.

The statistically significant relationship between loans to total assets ratio and RWAR indicates an increase in the bank’s risk-weighted assets. Moreover, it is consistent with the agency theory that states that if monitoring is costly and ineffective, bank owners link up the personal benefit and incentives of the management and staff accordingly to the bank’s performance. This makes them more accountable to the owners’ interests (IMF, 2014). Therefore, the agency problem may be reduced by encouraging bank growth through increased loans and advances. However, it also increases the risk associated with the bank’s operations, which can appear due to increasing the portfolio size with inferior quality loans or assets that may default in the long term. The positive association between loans to total assets and Z-score indicates that the growth of loans and advances also increases the bank’s solvency, hence reducing the insolvency risk. Simultaneously, it also decreases the liquid assets, as observed by the significant positive association with liquidity risk, which supports the evidence of different studies (Khan et al., 2017; Chen et al., 2021). This result supports the evidence of the relationship between the solvency and liquidity risk of the bank (Wagner, 2007; Chen et al., 2021).

The study result confirms that an increase in the NPL ratio increases the risk-weighted asset ratio of the bank, which matches the evidence found by Hsieh et al. (2013) and Shrieves and Dahl (1992). However, the result between the NPL ratio and the Z-score contradicts our assumption and violates the relationship between asset quality and the bank’s solvency risk. Since a higher Z-score can result from the increase of ROA or the increase of capital ratio, the positive association between the NPL and Z-scores presents evidence of an increase in capital ratio. This implies a positive effect of regulatory pressure which compels the banks to increase the required capital level due to an increase in NPL and risk-weighted assets.

In the same way, findings for capital ratio indicate an inverse association with the bank’s total risk and insolvency risk, as opposed to the finding by Ashraf et al. (2016). However, the positive relation between capital ratio and liquidity risk is consistent with the findings of Altunbas et al. (2007). This can be explained by the increased level of loans and advances that utilize additional capital flows to increase the rate of bank return. It contradicts the findings of Ben Bouheni (2014) and affirms the result of Maji and Kumar (2015).

The positive association between GGDP and the bank’s total risk contradicts the present evidence in different economic sectors (Ashraf et al., 2016; Hsieh et al., 2013). These studies observed that countries with higher GDP growth and economic development support the asset diversity of the banks that accelerate the profitability and bring down the risk (Abuzayed et al., 2018; Hsieh et al., 2013). However, the present study results imply that economic expansion increases risk-weighted assets within the banking industry, consequently contributing to a rise in credit risk exposure.

Income diversification is an imperative strategy for managing a company in a competitive environment, minimizing risks, and increasing profits. Accordingly, this study provides evidence that capital requirements regulations improve banks’ liquidity position and stability. Therefore, managers can take advantage of an increase in capital level by investing efficiently in projects with low-risk profiles. It is also possible for banks to reduce their risks through the efficient use of capital across different sectors.

Additionally, a positive association between GDP growth and bank risk can be an influential factor for bank managers in deciding the loan criteria and disbursement of loans and advances during the economic boom. A conservative or strict credit evaluation can help in approving the appropriate borrower and credit that will have a lower chance of default in the subsequent period. A policy framework should be developed to control the risk, keeping in mind factors like capital ratio, lending ratio, and the economic condition of the country in which the bank is operating.

However, we acknowledge several limitations. First, the dataset does not accurately reflect recent developments in the banking sector, including the impact of COVID-19 and the implementation of Basel III. Second, the analysis does not account for the widespread adoption of digital technologies and AI tools, which are known to impact non-interest income. Third, the present study has not examined any country-level effects, as the focus remains on bank-level variables. Future research could extend the dataset to include post-2015 trends, explore the role of ICT adoption in shaping bank revenues, and incorporate country-specific factors to better capture cross-country heterogeneity and policy relevance. Future research could examine how technological advancements—such as fintech integration and digital platforms—have reshaped banking performance.

This study is limited to only conventional banks. Islamic banks are not included in the present study due to the different nature of their principles. Therefore, future research could be conducted on Islamic banks from different regions to obtain a clear picture of the banking industry. Additionally, a large number of sample banks come from the USA. A different perspective may be obtained by excluding the banking industry of the United States. Consequently, future studies could be conducted on the comparative position of the banking industry in developing and developed countries.

6. Conclusions

This study examines the effect of income diversification on the risk of 565 commercial banks from 50 countries during the post-financial crisis period of 2011 to 2015. The role of non-interest income was observed through changes in risk in terms of the risk-weighted asset ratio, bank insolvency risk, and liquidity risk. It also determines the effect of other bank-specific and macroeconomic indicators on bank risk.

The study finds that income diversification is inversely related to the bank’s insolvency risk and liquidity risk. Therefore, an increase in non-interest income increases the solvency or Z-score as well as the liquidity position of the bank. It is observed that bank size alone does not necessarily determine its risk level; other factors play a more significant role in the risk profile. Although loans to total assets improve the Z-score of the bank, it is also positively related to the total risk and liquidity risk of the bank, as is to be expected because of agency problems. However, because of the increased revenue, the bank’s solvency (Z-score) is also improved. Thus, this study demonstrates the importance of diversifying income sources through non-traditional activities to increase solvency and liquidity. This study contributes to the existing literature by filling the gap with an analysis of the effect of income diversification on bank risk during the post-financial crisis in a global context. It tries to assess the effect of macroeconomic variables on the risk of commercial banks. In addition, it employs several approaches to measure the risk of the bank, as there is no single technique to define risk, which makes the study more robust.