The effects of business model on bank's stability

Nguyen Thuy Thu; Hai Hong Ho; Duy Van Nguyen; Anh Cam Pham; Trang Thu Nguyen Article The effects of business model on bank's stability International Journal of Financial Studies Provided in Cooperation with: MDPI – Multidisciplinary Digital Publishing Institute, Basel Suggested Citation: Nguyen Thuy Thu; Hai Hong Ho; Duy Van Nguyen; Anh Cam Pham; Trang Thu Nguyen (2021) : The effects of business model on bank's stability, International Journal of Financial Studies, ISSN 2227-7072, MDPI, Basel, Vol. 9, Iss. 3, pp. 1-12, https://doi.org/10.3390/ijfs9030046 This Version is available at: https://hdl.handle.net/10419/257791 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ International Journal of Financial Studies Article The Effects of Business Model on Bank’s Stability Thuy Thu Nguyen 1, Hai Hong Ho 2,* , Duy Van Nguyen 3, Anh Cam Pham 4and Trang Thu Nguyen 5   Citation: Nguyen, Thuy Thu, Hai Hong Ho, Duy Van Nguyen, Anh Cam Pham, and Trang Thu Nguyen. 2021. The Effects of Business Model on Bank’s Stability. International Journal of Financial Studies 9: 46. https://doi.org/10.3390/ijfs9030046 Academic Editors: Sanjeev Acharya and Vijay Kumar Received: 23 June 2021 Accepted: 18 August 2021 Published: 26 August 2021 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). 1 Faculty of Banking and Finance, Foreign Trade University, Hanoi 10000, Vietnam; [email protected] 2Faculty of Business and Economics, Phenikaa University, Hanoi 10000, Vietnam 3QA Global Co., Hanoi 10000, Vietnam; [email protected] 4School of Economics and International Business, Foreign Trade University, Hanoi 10000, Vietnam; [email protected] 5School of Business, Law and Entrepreneurship, Swinburne University of Technology, Melbourne 3122, Australia; [email protected] *Correspondence: [email protected] Abstract: The literature shows little evidence of the effects of business models upon the volatility of banks in developing and fast-growing economies. Hence, this study examines the effects of business model choice on the stability of banks in ASEAN countries. Using GMM and other robust econometric methods on the sample of 99 joint stock commercial banks, we find significant and negative impacts of a diversification model in which banks shift toward non-interest and fees-based activities. We also find that the impacts are different between two groups of countries. For Vietnam, Indonesia and the Philippines, the diversification entails negative impacts on stability while demonstrating positive impacts for Thailand and Malaysia. Based on these findings, we draw policy implications for more sustainable development in the ASEAN banking business. Keywords: bank business model; bank stability; bank risk; ASEAN; Vietnam 1. Introduction Over the last few decades, the banking system has embarked upon drastic change on a global scale. Financial institutions like banks have adopted a diversification approach to move away from a traditional deposit-taking and lending model toward more technologicallydriven fees-based and commission-based services, such as e-banking, consumer credits, securities trading, insurance and investment brokerage. This trend appears inevitable since increasing domestic and foreign rivalry in the banking business puts pressure on traditional interest margins. In addition, the quick pace of technological progress also facilitates new services, helps cut transaction costs and saves time for clients. However, the genuine benefits of the new and non-traditional fees-based and commission-based model are still under scrutiny. While recent literature, such as Chiorazzo et al. (2008), Hamdi et al. (2017), Meslier et al. (2014) and Trivedi (2015), supports the positive impacts of non-interest income on banks’ profitability, some question whether the new sources of income may entail a higher level of risk and instability (DeYoung and Roland 2001;Stiroh 2004;Williams and Prather 2010). Upon the expansion of fees-based services in banking industry, a new line of research has been pursued and light has been shed on the causality between fees-based products and the performance of the bank. Abuzayed et al. (2018) argued that diversification of income sources into non-correlated activities could decentralize risk and decrease the propensity of financial distress. Moreover, diversification might enhance the intermediation role of banks and motivate managerial efficiency (Drucker and Puri 2009;Hamdi et al. 2017). In contrast, previous literature has challenged the link between diversification and banks, stability, providing evidence that bank size, ownership structure and model are the significant determinants of stability rather than diversification (Chiorazzo et al. 2008; Kohler 2015;Lee et al. 2014). Furthermore, Lee et al. (2014) found no evidence supporting the expected benefits of diversification, perhaps because diversified banks tend to take Int. J. Financial Stud. 2021,9, 46. https://doi.org/10.3390/ijfs9030046 https://www.mdpi.com/journal/ijfs Int. J. Financial Stud. 2021,9, 46 2 of 12 more risks and operate with greater financial leverage than non-diversified counterparts (Cebenoyan and Strahan 2004). The new evidence in this recent line of research motivates this research to examine whether the diversification model has positive or negative impacts on the stability of banks. This research extends the literature in the context of ASEAN countries where little evidence is documented, and even if it is, the impacts vary from one country to another (Lee et al. 2014;Nguyen and Pham 2020). Additionally, since King and Levine (1993) find that financial diversification and bank stability are robustly associated with economic growth, we aim to shed more light on the impacts of economic growth on such a relationship in ASEAN markets. The impacts of diversification on listed banks’ stability are examined with a modelling approach described in Kohler (2015), in which Z score and risk-adjusted profitability measures are the alternative proxies for the stability. The robustness of our findings is controlled with a variety of diversification indicators, non-linear effects, and heterogeneity and endogeneity considerations. This study aims to determine the extent of the influence of business models on banks’ stability. Our contributions are the empirical evidence along with policy implications regarding business models of banks in ASEAN countries where little is known about the effects of income diversification. We find that diversification of business models significantly reduces the level of stability at banks in the ASEAN region, except for Malaysia and Thailand where diversification effects are statistically positive. Such findings demonstrate significant different effects of diversification between ASEAN and developed countries. The paper is organized as follows: Section 2reviews the literature on the relationship between bank diversification and stability before the development of hypotheses. Section 3 outlines the data, variables and methodology, while Section 4presents the empirical results. Section 5concludes along with several practical implications. 2. Literature Review Bank stability has been revisited, especially after the failure of banking system during the financial crisis in 2008–2009. Among a variety of factors, diversification emerged and was debated upon in the previous literature. In most research, bank diversification involves two main aspects: income diversification and funding sources diversification. In our paper, we focus on the impacts of income diversification, which are the increase in share of fee, net trading profit and other non-interest income (Mahdaleta et al. 2016). Until recently, the literature has suggested a mixed picture of such impacts. A variety of studies reveal positive effects of income diversification on bank stability. For instance, Froot et al. (1993) and Froot and Stein (1998) emphasize the importance of how income and asset diversification can reduce the probability of bank distress since revenues from different activities are not perfectly correlated. In a study of European banks, the researchers report that the emergence of non-interest income activities contributes to the stability of retailoriented banks by strengthening the intermediation function and reducing information asymmetries (Baele et al. 2007;Kohler 2015). Acharya et al. (2006) and Lepetit et al. (2008) add that non-interest income could boost competition and financial innovation. The positive evidence of diversification in EU and US banks is also documented in the studies relating to Asia and other emerging countries (e.g., Sanya and Wolfe 2011;Nguyen et al. 2012;Amidu and Wolfe 2013;Nguyen and Pham 2020). Other findings supported the positive impacts by providing further evidence on positive influence of non-credit income on profitability and risk reduction. Boyd and Prescott (1986) and Drucker and Puri (2009) show that expanding operations across different products and services as well as geographically reduces risk concentration, thereby decreasing the likelihood of financial distress. Such desirable effects are achieved through lower monitoring costs, greater efficiencies, and scale economy of managerial skills. Moreover, evidence from the US, Pakistan, India and the Philippines advocates the positive impacts Int. J. Financial Stud. 2021,9, 46 3 of 12 of income diversification on profitability (Ismail et al. 2015;Trivedi 2015;Li et al. 2021). Meslier et al. (2014), Saunders and Walter (2014) and Sissy et al. (2017) highlight that bank performance is boosted via cross-border diversification. In contrast, some opposing evidence indicates negative or ambiguous causality from diversification to bank stability. With the development of non-interest activities, banks in the US and EU encounter greater levels of risk, depending on bank size, ownership and the type of noninterest income diversification (DeYoung and Roland 2001;DeYoung and Torna 2013;Lepetit et al. 2008;Stiroh and Rumble 2006;Yang et al. 2020). Williams (2016) found similar results for Australian banks. These studies also specify several reasons for the negative impacts, such as the reduction in bank’ incentives (Acharya et al. 2006) or the growing involvement into derivatives markets or larger loan portfolios undermine the buffer of capital tiers (Demsetz and Strahan 1997;DeYoung and Roland 2001). Meanwhile, a paper on banks in emerging countries finds that, despite lowering banks’ risk, income diversification is unable to enhance bank stability (Abuzayed et al. 2018). Moreover, the literature shows mixed results on the effects of diversification upon risk and return. While Craigwell and Maxwell (2006) and Stiroh and Rumble (2006) find that fees-based activities actually darkened the risk profile, they still impose positive effects on profitability. Sianipar (2015) claims that although income diversification can lower idiosyncratic and total risk, it does not raise the market value of the bank significantly. Others find that income diversification neither increases the return nor reduces the banks’ risk (Acharya et al. 2006;Hayden et al. 2007). Under some circumstances, the relationship varies across different types of risks (e.g., Abedifar et al. 2018;Akhigbe and Stevenson 2010;Carlson 2004;De Vries 2005;Elyasiani and Wang 2012;Banwo et al. 2019). Hence, no apparent consensus has been reached on the diversification effects. The lack of consensus also extends geographically since studies in different areas, such as the US, EU (DeYoung and Roland 2001;DeYoung and Torna 2013;Lepetit et al. 2008;Stiroh and Rumble 2006), emerging countries (Sanya and Wolfe 2011;Wu et al. 2020), Africa (Sissy et al. 2017;Adesina 2021) and Asia (Meslier et al. 2014;Nguyen et al. 2012;Nguyen and Pham 2020), yield inconsistent results. Such inconsistency suggests that the effects of income diversification on banks’ stability is country-specific. As a result, the link between income diversification and banks’ stability in ASEAN might be distinctive. Little evidence is found in existing papers involving ASEAN banking systems, and almost none is found on the direct causality of income diversification on banks’ stability. A rare paper (Meslier et al. 2014) provides evidence of positive relationship between income diversification and risk-adjusted performance of banks in Philippines; however, the relationship is insignificant for small banks. Nguyen and Pham (2020) examined listed and unlisted banks in Vietnam and demonstrated evidence of diversification impacts on risk mitigation; and such impacts are stronger for unlisted banks and for banks undergoing restructuring. Given the heterogeneity and divergence in size, dynamics, efficiency and technology adopted among ASEAN banking system, income diversification is likely to have dissimilar effects on banks’ stability in each country. Our paper, therefore, seeks to shed more light on that matter. 3. Data and Methodology 3.1. Data The sample of listed banks in ASEAN is arranged in panel form. The dependent and independent variables data are extracted from Refinitiv Eikon and cross-checked with the official periodical reports of listed banks in ASEAN stock exchanges in 2011–2019 period. Bank data has tremendous advantages over corporate data with respect to accuracy, especially for listed banks, since such data are input by better-trained staff and checked by internal audit, creditworthy auditors and the central bank. Using equity screening application on Refinitiv Eikon, we collected the initial sample consisting of 236 banks and 2585 observations from 2010 to 2019 of five ASEAN countries (Vietnam, Thailand, Indonesia, Malaysia and Philippines). In all countries, except for Viet- Int. J. Financial Stud. 2021,9, 46 4 of 12 nam, under the name of one bank there were numerous types of equity, including ordinary shares, American Depositary Receipts (ADRs), non-voting depositary receipts, and rights and preference shares. However, we just keep the ordinary shares to avoid duplication, resulting in a drop of 124 datapoints and 693 observations. We further filtered out missing dependent variables for two consecutive years to keep the series in each datapoint sufficiently long to be consistent with a more robust econometric technique (i.e., the DGMM). The final sample comprised 99 banks and 987 observations. All of the observations were free from outlier problem since they were winsorized at the 5% and 95% levels. In the sample, we employed several alternative proxies for bank’s stability, including Z-score, risk-adjusted return on assets (RAROA) and risk-adjusted capital adequacy ratio. Z-score was calculated by the sum of ROA and CAR, divided by SDROA, where ROA was return on assets, CAR was the fraction of total equity divided by total assets and SDROA was the standard deviation of ROA. To enhance the robustness, RAROA and RACAR were also considered as varieties of risk-adjusted measures for banks’ stability. According to Kohler (2015), the higher Z-score or RAROA or RACAR entails lower risk and higher stability. Regarding independent groups of variables, net interest income (NII) is the one that best captures the traditional interest-based business model. The spread between lending and deposit rates has been a major income source of banks and hence is supposed to exert positive effects on banks’ stability (Kohler 2015). Alternatively, net non-interest income (NNII) and trading are proxies for non-traditional and fees-based sources of income. NNII and trading activities help banks diversify their revenue streams to spread the risks inherently concentrated on conventional interest-based products and services. NNII and trading could enhance banks’ sustainability for commercially-oriented banks, while they may undermine stability since excessive engagement in fee-generating activities such as currency trading or off-balance sheet securitization tends to damage one’s risk profile (Kohler 2015;Altunbas et al. 2011;DeYoung and Roland 2001). In principle, most of the variables in the final sample are kept in the fraction form, except for total assets, which is converted to log form to reduce the skewness and enhance variance stability (Lütkepohl and Xu 2010). Detailed formulae and description for each variable are presented in Table 1. Table 1. The description of variables. Variable Description Expected Alternative Dependent variable Z-score Fraction of (ROA+CAR)/SDROA RAROA Faraction of ROA/SDROA RACAR Fraction of CAR/SDCAR Independent variables NII Fraction of abs(net interest income)/abs(total income) (+) NNII Fraction of net non-interest income/net operating income (+) TRADE Fraction of trading income/net operating income (−) Control variables CAR Fraction of equity/total assets SIZE Ln(Total assets) (−) LOANS Fraction of total loans/total assets (+) NIM Fraction of net interest revenue/average earning assets (+) Source: Authors’ synthesis. 3.2. Methodology The potential causality between income diversification and banks’ stability is examined with a standard linear function. To ensure the robustness of findings, the usual suspects of endogeneity, spatial dependence and country heterogeneity are controlled by Driscoll–Kraay’s robust error and Difference Generalized Method of Moments (DGMM). Int. J. Financial Stud. 2021,9, 46 5 of 12 We adopt the functional model from Kohler (2015) as follows: Bank0stability =αi+β1×Business modelit +β2×Control variablesit +εit (1) With a cross-country panel data, one of the potential selection biases lies in the timeinvariant country confounder in that the hypothetical change in the independent or control variables within the same country could inflate the impacts on the dependent variables in an OLS regression model (Mummolo and Peterson 2018). This paper limits selection bias caused by country heterogeneity with country fixed-effects model. We further control for spatial dependence, heteroscedasticity and serial correlation problems with Driscoll– Kraay’s robust standard errors. Finally, one of the key econometric issues in panel data analysis is potential endogeneity bias since it can yield misleading conclusions or even incorrect signs of the coefficients (Ullah et al. 2018). Indeed, Wooldridge (2010) proved that the omitted and observed/unobserved variables unincorporated in the model are the causes of such endogeneity. Accordingly, we control for the omitted variables as well as potential pairwise correlation between independent variables and the error term with the endogenous treatment in the Difference Generalized Method of Moments (DGMM) model. Specifically, the initial equation in DGMM model is as follows: Yit =(β0+υi)+β1Yit−1+β2Xit +εit (2) Equation (2) is transformed into first-difference form to suppress potential fixed effects assumed in panel data. ∆Yit =β1∆Yit−1+β2∆Xit +εit where : υit =νi+εit ∆υit =(νi+εit)−(νi+εit−1)=∆εit (3) 4. Results 4.1. Descriptive Statistics The descriptive results show that for all countries, the average Z-score is 39.08. The maximum value of Z-score is 267.249, and the minimum is 0.052. They also indicate that the average value of RAROA is 2.947 and that of RACAR is 36.113. For business model metrics, NII averages around 2.321 and NNII 1.017. The remaining variables are summarized in Table 2. It is noteworthy that Vietnamese banks’ Z-scores, RACAR, and RAROA appear to be relatively smaller than the counterparts of other ASEAN countries (Table 3). Table 2. Descriptive statistics for all countries. Variable Name Mean SD Min Max Z-score 39.080 35.975 0.052 267.249 RAROA 2.947 2.662 −2.786 21.254 RACAR 36.113 33.766 0.000 248.724 NII 2.321 10.322 0.005 188.302 NNII 1.017 2.643 −1.089 69.551 TRADE −0.519 20.988 −635.648 105.737 ROA 1.503 1.489 −10.700 10.210 CAR 17.146 19.353 0.000 489.580 SDROA 0.827 0.903 0.002 4.720 SDROE 0.064 0.045 0.022 0.230 LOANS 0.604 0.129 0.020 0.920 NIM 4.149 4.163 −0.009 45.030 Note: The mean Z-score is remarkably higher than the safety level for non-manufacturers (2.90). Some banks are still subject to bankruptcy since the minimum level of Z-score is below the critical level (0.15). The sample contains 987 observations (N) for 99 banks in ASEAN banks. Source: Authors’ synthesis. Int. J. Financial Stud. 2021,9, 46 6 of 12 Table 3. Descriptive statistics for Vietnam and ASEAN countries. Variable Name Mean SD Min Max Vietnam Z-score 25.670 18.007 0.052 99.718 RAROA 2.061 1.584 −2.602 6.613 RACAR 23.556 17.269 0.000 96.548 NII 0.236 0.718 0.005 9.650 NNII 0.222 0.296 −1.089 3.003 NIM 0.032 0.014 −0.009 0.094 LOANS 0.556 0.120 0.145 0.744 TRADE −3.264 52.344 −635.647 105.737 ROA 0.009 0.008 −0.060 0.029 CAR 0.108 0.064 0.000 0.402 SDROA 0.006 0.005 0.002 0.023 ASEAN countries Z-score 41.765 38.370 0.864 267.249 RA_ROA 3.138 2.790 −2.786 21.254 RA_CAR 38.627 36.037 2.507 248.724 NII 2.749 11.045 0.014 188.301 NNII 1.191 2.788 0.011 69.551 NIM 5.074 3.943 0.260 45.030 LOANS 0.000 0.000 0.000 0.000 TRADE 0.009 0.022 0.000 0.077 ROA 1.828 1.504 −10.700 10.210 CAR 20.404 18.705 6.150 489.580 SDROA 1.011 0.903 0.070 4.720 Note: The mean Z-score of Vietnamese banks; minimum and maximum values are all lower than counterparts of other ASEAN averages. The mean values of RAROA and RACAR (alternative proxies for bank’s stability) in Vietnam are also lower than those of other ASEAN countries. The aggregate of statistics suggests a different pattern in the overall stability of Vietnamese banks. Source: Author’s synthesis. Figures 1and 2compare the general trends of ROA and Z-score from 2010 to 2019. In Figure 1, these two indicators seem to have a negative correlation. The bigger the Z-score (the greater the risk), the lower the level of ROA, notably in the period from 2010 to 2015. Until 2019, these two indicators tend to converge. In contrast, for Vietnamese banks the correlation between ROA and Z-score tend to be opposite as they both decrease in the period 2010–2016 and increase gradually since 2016. This partly shows the difference in the relationship between ROA and Z-score in Vietnam and other countries in the region. Int.J.FinancialStud.2021,9,xFORPEERREVIEW7of13    Figure1.Z‐scoreandROAforASEANcountries.  Figure2.Z‐scoreandROAinVietnam. Table4illustratesthedistributionoflistedbanksbycountry,including43banksin Indonesia,accountingfor43.43%,18banksinVietnam(18.18%),14banksinPhilippines (14.15%),13banksinMalaysia(13.13%)and11infromThailand(11.11%).  Table4.Thedistributionoflistedbanksbycountry. CountryNo.ofBanksPercentage Indonesia4343.43% Philippines1414.15% Vietnam1818.18% Malaysia1313.13% Thailand1111.11% Total99100% Source:Author’ssynthesis. 4.2.TheDriscoll–Kraay’sRegressionResults TheregressionresultsyieldedthroughDriscoll–Kraay’srobuststandarderrors methodillustratethatallofnetinterestincome(NII),netnon‐interestincome(NNII)and Figure 1. Z-score and ROA for ASEAN countries. Int. J. Financial Stud. 2021,9, 46 7 of 12 Int.J.FinancialStud.2021,9,xFORPEERREVIEW7of13    Figure1.Z‐scoreandROAforASEANcountries.  Figure2.Z‐scoreandROAinVietnam. Table4illustratesthedistributionoflistedbanksbycountry,including43banksin Indonesia,accountingfor43.43%,18banksinVietnam(18.18%),14banksinPhilippines (14.15%),13banksinMalaysia(13.13%)and11infromThailand(11.11%).  Table4.Thedistributionoflistedbanksbycountry. CountryNo.ofBanksPercentage Indonesia4343.43% Philippines1414.15% Vietnam1818.18% Malaysia1313.13% Thailand1111.11% Total99100% Source:Author’ssynthesis. 4.2.TheDriscoll–Kraay’sRegressionResults TheregressionresultsyieldedthroughDriscoll–Kraay’srobuststandarderrors methodillustratethatallofnetinterestincome(NII),netnon‐interestincome(NNII)and Figure 2. Z-score and ROA in Vietnam. Table 4illustrates the distribution of listed banks by country, including 43 banks in Indonesia, accounting for 43.43%, 18 banks in Vietnam (18.18%), 14 banks in Philippines (14.15%), 13 banks in Malaysia (13.13%) and 11 in from Thailand (11.11%). Table 4. The distribution of listed banks by country. Country No. of Banks Percentage Indonesia 43 43.43% Philippines 14 14.15% Vietnam 18 18.18% Malaysia 13 13.13% Thailand 11 11.11% Total 99 100% Source: Author’s synthesis. 4.2. The Driscoll–Kraay’s Regression Results The regression results yielded through Driscoll–Kraay’s robust standard errors method illustrate that all of net interest income (NII), net non-interest income (NNII) and trading income ratio (TRADE) have negative impacts on Z-score, RACAR and RAROA (Table 5) . According to this table, the increase in non-interest income tends to reduce banks’ stability. Thus, the expansion toward fees-based activities can, on one hand, cushion income volatility. On the other hand, such activities appear inefficient and undermine the overall performance, perhaps because the fees-based expansion compromises incentives for supervision of traditional loans, leading to lower loans quality and profitability (Abuzayed et al. 2018;Acharya et al. 2006). Furthermore, risk mitigation through diversification involves a trade-off with an increase in exposure to more volatile non-traditional services, which is referred to as “the dark side” of diversification (Stiroh and Rumble 2006). Overall, our findings are consistent with some evidence found in previous studies on the significant and negative impacts of fees-based expansion (Abuzayed et al. 2018; Acharya et al. 2006;Hayden et al. 2007;Lepetit et al. 2008;Williams and Prather 2010). Such findings lend support to the detrimental effects of diversification model on ASEAN banking system. In addition, other significant variables such as TRADE and NIM are also consistent with our prediction that the increase in non-traditional trading services will impair banks’ stability as it diverts managerial focus away from traditional deposit and lending activities. Int. J. Financial Stud. 2021,9, 46 8 of 12 Table 5. Regression results for ASEAN countries. Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Variables Zscore Zscore Racar Racar Raroa Raroa NII −0.0365 ** −0.0272 * −0.00935 *** (0.0149) (0.0149) (0.00266) NNII −0.126 * −0.107 −0.0192 *** (0.0709) (0.0682) (0.00436) TRADE −0.0225 *** −0.0224 *** −0.0223 *** −0.0223 *** −0.000182 −0.000184 (0.00717) (0.00719) (0.00658) (0.00661) (0.000748) (0.000746) SIZE 0.0969 0.120 0.107 0.121 −0.00983 −0.000158 (0.0811) (0.0798) (0.0747) (0.0739) (0.0148) (0.0119) LOANS −23.38 *** −23.53 *** −22.18 *** −22.32 *** −1.198 * −1.209 * (4.973) (4.964) (5.018) (5.011) (0.684) (0.675) NIM −0.179 ** −0.174 * −0.260 *** −0.257 *** 0.0819 *** 0.0826 *** (0.0896) (0.0898) (0.0958) (0.0954) (0.0180) (0.0180) CAR 0.492 *** 0.492 *** 0.490 *** 0.490 *** 0.00199 0.00185 (0.0263) (0.0265) (0.0247) (0.0249) (0.00271) (0.00264) Constant 43.58 *** 43.02 *** 39.90 *** 39.62 *** 3.687 *** 3.406 *** (2.407) (2.229) (2.402) (2.254) (0.420) (0.465) Observations 914 914 914 914 914 914 No. of groups 96 96 96 96 96 96 Note: the regression results are generated with Driscoll and Kraay estimator. Standard errors in parentheses. *** p< 0.01, ** p< 0.05 and *p< 0.1. 4.3. Robustness Check for Endogeneity To control for the endogeneity problems potentially leading to bias coefficients, and to check whether the relationship between the diversification of bank business model and stability is robust, a GMM estimator is employed. Consistently, the GMM model (Table 6) yields similar results, especially on the impacts of diversification on bank stability, in that net non-interest income (NNII) has negative and statistically significant effects on the Z-score. The Hansen test statistic is in the good range, according to (Roodman 2009), and supports the rejection of the null hypothesis of over-identifying restrictions. On country level, the estimated results (Table 7) demonstrate that the impacts of net non—interest income (NNII) are similar in Vietnam and Indonesia in the sense that increasing non-interest income statistically reduces stability. This evidence supports the negative impacts of diversification model on the efficiency of banks. By contrast, Malaysia and Thailand have more in common as NNII has positive and significant impacts on stability, which supports the positive effects of diversification model on banks’ stability. Perhaps the difference in the level of financial market development among ASEAN countries mediates the impacts of NNII and diversification. TRADE is the other proxy of income diversification in the regression model, and its negative impacts on banks’ stability apparently converge in all countries except for Philippines and Malaysia, suggesting that foreign exchange trading activities statistically undermine the stability of banks. Contrary to the theoretical suggestions, we find negative correlation between NIM and stability. In principle, the implicit assumption in banking business is that banks cope with default and various types of risks by setting the loan–deposit spread, and consequently the NIM, wide enough to buffer against adverse consequences. Thus, the increase in NIM lowers the risk of interest-based activities, and hence NIM accounts for the majority of income in a traditional bank. Against our anticipation, our significant evidence of inverse relationship between NIM and banks’ stability suggests that the increase in NIM, perhaps as the consequence of higher lending rate, comprises higher risk of non-performing loans (Brock and Rojas Suarez 2000), lowers quality of earning assets and leads to an increased level of risk in the next period. Our findings confirm those of Brock and Rojas Suarez (2000) 1 and Marinkovi´c and Radovi´c