Bank Opacity and Bank Stability in the Top UKBs of the Philippines

empirical study of bank opacity vs bank stability in the top UKBs in the Philippines

1. AN ANALYSIS OF THE RELATIONSHIP BETWEEN BANK OPACITY, RISK-TAKING BEHAVIOR, AND SOLVENCY OF THE TOP 20 UNIVERSAL AND COMMERCIAL BANKS IN THE PHILIPPINES Dr. Jovi C. Dacanay Gene Matthew E. Chua 8 November 2024

2. OVERVIEW INTRODUCTION LITERATURE METHODOLOGY RESULTS AND ANALYSIS CONCLUSIONS AND RECOMMENDATIONS Background of the Study, Problem Statement, Research Objectives, Significance of the Study, Scope and Limitations Theoretical Framework, Conceptual Framework, Empirical Methodology Dumitrescu-Hurlin Panel Causality Test Panel Generalized Method of Moments (GMM) Baron-Kenny Mediation Analysis using Regression Summary of Results, Conclusion, Recommendations, Areas for Future Studies Bank Opacity and Risk-taking, Bank Risk-taking on Solvency, Bank Opacity and Risk-taking on Solvency, Agency Theory, Mediation Analysis, Regulatory Environment and Governance, Research Gap

3. BACKGROUND OF THE STUDY THE BANKING SECTOR... Is the backbone of a country’s economy - offers financial intermediation services Has opportunities to foster economic growth, development, and stability Accelerates economic development

4. The Philippine government has been implementing significant financial system reforms to boost efficiency and stability. 1980s 1997–1998 Asian Crisis : The Philippines was not as severely affected as other Asian nations due to a stronger financial system following the reforms 2005 The Philippines’ financial sector has seen major advances as seen in quadrupled loans and deposits, and increase in number of ATMs and bank branches Present Larger and more dominant banks emerged BACKGROUND OF THE STUDY

5. BANKS MARKET SHARE ( TOTAL ASSETS) 18.14 % 13.85 % 12.95 % 12.08 % 6.52 % 5.46 % 5.12 % 4.24 % 4.19 % 4.17 % 2023 TOP 10 UKBS (TOTAL ASSETS MARKET SHARE) Source: Bangko Sentral ng Pilipinas (BSP) BACKGROUND OF THE STUDY

6. The inability of outsiders to assess risk in a bank’s assets due to opaque activities or limited account information 2007 to 2009 Global Financial Crisis (GFC): The lack of clarity and transparency had profound implications for the stability of the overall financial system BANK BEHAVIOR IN FINANCIAL SYSTEM BANK OPACITY The pursuit of strategies that increase the probability of high returns while also carrying higher degrees of risk Incentives to engage in risky operations to improve competitive positions and maximize profits Exposes banks to increased vulnerabilities, potentially resulting in financial instability, systemic risk, and negative outcomes BANK RISK-TAKING BEHAVIOR BACKGROUND OF THE STUDY

7. PROBLEM STATEMENT “Does bank opacity and heightened risk-taking behaviors lead to decreased solvency among the top 20 universal and commercial banks (UKBs) in the Philippines?” BACKGROUND OF THE STUDY

8. RESEARCH OBJECTIVES To determine the impact of bank opacity and risk-taking behaviors on the solvency of the top 20 universal and commercial banks (UKBs) in the Philippines. To empirically verify that heightened risk-taking behavior is caused by the opacity of the top 20 UKBs in the Philippines OBJECTIVE 1 To empirically verify the impact of heightened risk-taking behavior on the solvency of the top 20 UKBs in the Philippines OBJECTIVE 2 To determine the mediating effect of bank opacity on the relationship between macroeconomic variables and solvency of banks under the Top 20 Philippine UKBs OBJECTIVE 3 BACKGROUND OF THE STUDY

9. SIGNIFICANCE OF THE STUDY Address a significant gap in the banking literature on the relationship between bank opacity and risk-taking behaviors of the top UKBs in the Philippines FILLING THE RESEARCH GAP Allow banks to refine risk management practices and enhance their competitive position INFORMED DECISION-MAKING AND REGULATORY POLICIES Allow regulatory authorities to develop policies to promote transparency, stability, and efficiency in the banking sector Help stakeholders make informed decisions , contributing to a more efficient allocation of resources within the banking sector ENHANCING MARKET EFFICIENCY AND FINANCIAL STABILITY Help implement measures to mitigate systemic vulnerabilities and enhance the resilience of the banking sector to economic shocks BACKGROUND OF THE STUDY

10. SCOPE AND LIMITATIONS Focuses on the Top 20 UKBs in the Philippines (96% market share) Data collected from BSP and respective bank websites: 2005 Q1 to 2023 Q4 SCOPE The findings in this study will not immediately be relevant to smaller banks and other industries such as rural and thrift banks Data used in this study will be gathered from publicly available sites LIMITATIONS CHUA THESIS PROPOSAL BACKGROUND OF THE STUDY

11. REVIEW OF RELATED LITERATURE

12. Banks are known to inherently be highly opaque compared to other industries due to the inherent uncertainty of their financial disclosure On the Relationship between Bank Opacity, Risk-taking, and Solvency REVIEW OF RELATED LITERATURE Excessive Risk-taking erodes bank solvency and leads to financial distress (depending on the banks’ capitalization and size) Opacity incentivizes banks to increase risk-taking due to higher funding costs More opaque banks are associated with higher risk-taking behavior and higher risks of insolvency

13. Mitnick (1975) Both parties have their own self-interests and agendas which leads to agents engaging in riskier and more opaque activities to gain these incentives. Leads to problems in their relationship as the information asymmetry between the two causes a moral hazard and makes it harder for the principals to monitor and control their agents Zhang and Wang (2023) Aligns closely with the agency theory, focusing on the concept of asymmetric information Underscores the importance of transparency in maintaining financial stability, especially during periods of economic uncertainty. AGENCY THEORY REVIEW OF RELATED LITERATURE

14. Baron and Kenny (1986) Established one of the fundamental approaches to mediation analysis Highlighted a series of regressions to determine whether a mediator variable transfers the impact of an independent variable to a dependent variable. Zhang and Wang (2023) Employed the Baron-Kenny Mediation analysis in their panel data study among banks in China to determine the mediating effect of bank opacity on the relationship between economic policy uncertainty and bank stability. They found that the volatility of economic policy, influence banks to reduce information disclosure to avoid oversight and minimize dissatisfaction among their investors, which negatively affected their financial stability. Highlights the potential role of bank opacity as a mediator in complex relationships within the banking sector BANK OPACITY AS A MEDIATOR: BARON-KENNY MEDIATION ANALYSIS REVIEW OF RELATED LITERATURE

15. BSP (2023) Significant impact on bank conduct: opacity and risks. Implemented numerous regulatory reforms aimed at improving banks’ transparency, accountability, and risk management practices ADB (2009) Weaknesses in regulatory monitoring, capacity restrictions, and gaps in enforcement procedures that require further research REGULATORY ENVIRONMENT AND GOVERNANCE IN THE PHILIPPINES REVIEW OF RELATED LITERATURE

16. RESEARCH GAP Relationship between bank opacity, risk-taking behavior, and solvency of the top UKBs in the Philippines has received relatively limited attention Previous research suggested studying the relationship in the context of developing countries using various methodologies REVIEW OF RELATED LITERATURE

17. FRAMEWORK AND METHODOLOGY

18. THEORETICAL FRAMEWORK Source: Mehrabian and Russell (1974). Modified by the author for this study. FRAMEWORK AND METHODOLOGY

19. CONCEPTUAL FRAMEWORK Source: Author FRAMEWORK AND METHODOLOGY

20. DATA COLLECTION The study focuses on the top 20 UKBs in the Philippines from 2005 Q1 - 2023 Q4 Dependent and Independent variables: sourced from their financial statements on the BSP website and their respective websites Macroeconomic variables: sourced from the World Bank Data and PSA website FRAMEWORK AND METHODOLOGY

21. Variables Definition Bank Solvency The sum of shareholder’s equity and net income divided by shareholder’s equity all over the standard deviation of the ROE of the sample Bank Risk-taking The bank’s non-performing loans over total loans Bank Opacity The banks’ available-for-sale assets over total assets Bank Size, log The natural logarithm of the bank’s total assets Bank Debt The bank’s total loans over total assets GDP per capita, log The natural logarithm of GDP per capita of the Philippines Unemployment Rate The unemployment rate of the Philippines Interest Rate The interest rate of the Philippines DEFINITION OF VARIABLES

22. “H1: Increased bank opacity causes heightened risk-taking behaviors among the top 20 UKBs in the Philippines.” EMPIRICAL MODEL DUMITRESCU-HURLIN CAUSALITY OBJECTIVE: Determine whether one time series is a factor and will offer useful information in forecasting another time series ASSUMPTION: X and Y are stationary time-series Verified through the Augmented Dickey-Fuller (ADF) Test If any of the time series is not stationary, it must be made stationary through differencing and other transformations. VARIABLES: Dependent Variable : Risk-taking behavior of UKBs measured using NPL Ratio (NPL/Total Loans) Independent Variable : Bank opacity of UKBs measured using AFS Ratio (AFS/Total Assets) Number of observations : 1,520 (20 banks) Time period: 2005 Q1 - 2023 Q4 FRAMEWORK AND METHODOLOGY

23. EMPIRICAL MODEL DUMITRESCU-HURLIN CAUSALITY MODEL Expected Relationship: The level of bank opacity in the top 20 UKBs homogeneously causes the risk-taking behavior of these banks

24. “H2: Heightened risk-taking behaviors among the top 20 UKBs in the Philippines are negatively associated with solvency.” EMPIRICAL MODEL PANEL GENERALIZED METHOD OF MOMENTS MODEL OBJECTIVE: Observe the relationship between variables while accounting for endogeneity, heteroscedasticity, and cross-section correlation of residuals DIAGNOSTIC TEST: Unit Root and Variance-Ratio Tests for each variables F-Tests for relevance and validity of instruments J-Test, Arellano-Bond Test, and Durbin-Wu-Hausman specification test for endogeneity, serial correlation, and autocorrelation. Weighting procedure of regressors and estimation covariance matrix for correction of cross-section dependence. VARIABLES: Dependent Variable : Solvency of UKBs measured using their Z-score (ROE) Independent Variable: Risk-taking behavior of UKBs measured using NPL Ratio (NPL/Total Loans) Macroeconomic Variables: GDP per capita, Unemployment Rate, Interest Rate Number of observations: 1,520 (20 banks) Time period: 2005Q1 - 2023Q4 FRAMEWORK AND METHODOLOGY

25. Variables Expected Signs Sources Dependent Variable Bank Solvency Cao & Juelsrud (2020), Demirguc-Kunt et al. (2010), Fosu et al. (2017), Martynova et al. (2015), Oino (2021), Ratnovski (2007) Independent Variable Bank Risk-taking - Bekir (2023), Tran et al. (2022), Cao & Juelsrud (2020), Fosu et al. (2017), Behr (2012) Macroeconomic Variables GDP per capita +/- Akawee (2023), Gizycki (2001), Trenca et al. (2015) Macroeconomic Variables Unemployment Rate - Moinescu (2008), Trenca et al. (2015), Macroeconomic Variables Interest Rate -/+ Akawee (2023), Gizycki (2001) EMPIRICAL MODEL PANEL GENERALIZED METHOD OF MOMENTS MODEL

26. “H3: Bank opacity plays an intermediary role between macroeconomic changes and solvency among the top 20 UKBs in the Philippines.” EMPIRICAL MODEL BARON-KENNY MEDIATION ANALYSIS MODEL OBJECTIVE: Identify the mediating effect of the variable in question on the relationship between the dependent and independent variables. ASSUMPTIONS: The variables need to be using a continuous scale. 2. The variables of interest should have a linear relationship 3. The data must not show multicollinearity 4. There should be no spurious outliers, and the distribution of the variables should be approximately normal. VARIABLES: Dependent Variable/s: Solvency of UKBs measured using their Z-score I ndependent Variables: GDP per Capita, Unemployment Rate, Interest Rate Mediator Variable: Bank opacity of UKBs measured using AFS Ratio (AFS/Total Assets) Number of observations: 1,520 (20 banks) Time period: 2005 Q1 - 2023 Q4 FRAMEWORK AND METHODOLOGY

27. Variables Expected Sign Sources Dependent Variable Bank Solvency Cao & Juelsrud (2020), Demirguc-Kunt et al. (2010), Fosu et al. (2017), Martynova et al. (2015), Oino (2021), Ratnovski (2007) Independent Variables GDP per capita, log +/- Akawee (2023), Gizycki (2001), Trenca et al. (2015) Independent Variables Unemployment Rate - Moinescu (2008), Trenca et al. (2015), Independent Variables Interest Rate -/+ Akawee (2023), Gizycki (2001) MEDIATION ANALYSIS MODEL: OUTCOME REGRESSION FRAMEWORK AND METHODOLOGY EMPIRICAL MODEL

28. Variables Expected Sign Sources Dependent Variable Bank Opacity Bekir (2023), Tran et al. (2022), Cao & Juelsrud (2020), Fosu et al. (2017), Behr (2012) Independent Variables GDP per capita, log +/- Akawee (2023), Gizycki (2001), Trenca et al. (2015) Independent Variables Unemployment Rate + Moinescu (2008), Trenca et al. (2015), Independent Variables Interest Rate -/+ Akawee (2023), Gizycki (2001) MEDIATION ANALYSIS MODEL: MEDIATOR REGRESSION FRAMEWORK AND METHODOLOGY EMPIRICAL MODEL

29. Variables Expected Sign Sources Dependent Variable Bank Solvency Cao & Juelsrud (2020), Demirguc-Kunt et al. (2010), Fosu et al. (2017), Martynova et al. (2015), Oino (2021), Ratnovski (2007) Independent Variables GDP per capita, log +/- Akawee (2023), Gizycki (2001), Trenca et al. (2015) Independent Variables Unemployment Rate - Moinescu (2008), Trenca et al. (2015), Independent Variables Interest Rate -/+ Akawee (2023), Gizycki (2001) Mediator Variable Bank Opacity +/- Bekir (2023), Tran et al. (2022), Cao & Juelsrud (2020), Fosu et al. (2017), Behr (2012) MEDIATION ANALYSIS MODEL: MEDIATED REGRESSION FRAMEWORK AND METHODOLOGY EMPIRICAL MODEL

30. SUMMARY OF CONCEPTUAL MAP FRAMEWORK AND METHODOLOGY

31. PRESENTATION, INTERPRETATION, AND ANALYSIS OF DATA

32. BANK OPACITY AND RISK-TAKING RESULTS AND ANALYSIS Both ratios are generally decreasing from 2005 Q1 to 2023 Q4 AFS Ratio is more volatile - sharp decline in 2008 (GFC) and increase in 2020-2021 (COVID-19) NPL Ratio has minor fluctuations with a significant increase in 2020-2021 (COVID-19) Simultaneous increase/decrease suggests the possible interrelationship between the two Mean AFS Ratio vs. Mean NPL Ratio Correlation Coefficient: 0.18 “H1: Increased bank opacity causes heightened risk-taking behaviors among the top 20 UKBs in the Philippines.”

33. Variable IPS ADF-Fisher Variable Level Level Z-index (Return on Equity) -9.54808*** (0.0000) [6] 182.546*** (0.0000) [6] Available-For-Sale Securities Ratio -3.02232*** (0.0013) [2] 66.2002*** (0.0031) [2] Non-Performing Loans Ratio -9.71877*** (0.0000) [4] 149.841*** (0.0000) [4] BANK OPACITY AND RISK-TAKING RESULTS AND ANALYSIS Results: Bank solvency, Bank Opacity and Bank Risk-taking behavior had stationary properties at level. These variables can be applied to the Dumitrescu-Hurlin Panel Causality Test PANEL UNIT ROOT TEST RESULTS

34. Null Hypothesis: p-value Interpretation NPLR does not homogeneously cause AFSR AFSR does not homogeneously cause NPLR 0.0018*** 0.0156*** Bilateral causality: Bank Opacity (AFSR) homogeneously causes bank risk-taking (NPLR), and vice versa. AFSR does not homogeneously cause ZROE ZROE does not homogeneously cause AFSR 0.0023*** 0.1085 Bank Opacity (AFSR) homogeneously causes ZROE, but not the other way around NPLR does not homogeneously cause ZROE ZROE does not homogeneously cause NPLR 0.9209 0.3152 No causality between Bank Risk-taking (NPLR) and Bank Solvency (ZROE) BANK OPACITY AND RISK-TAKING RESULTS AND ANALYSIS DUMITRESCU-HURLIN PANEL CAUSALITY RESULTS

35. BANK RISK-TAKING ON SOLVENCY RESULTS AND ANALYSIS Scatterplot exhibits an inverted U-shaped curve ZROE initially increases (3.00) when NPLR increases, indicating the initial returns of increased risk-taking Further increases in NPLR lead to decreased ZROE (stabilizing at a lower level) Correlation coefficient = -0.16, signifying a weak negative correlation “H2: Heightened risk-taking behaviors among the top 20 UKBs in the Philippines are negatively associated with solvency.” Correlation Coefficient: -0.16 NPLR vs. ZROE across Top 20 UKBs

36. BANK RISK-TAKING ON SOLVENCY Bank Size and Bank Debt: Insignificant Coefficients Appeared to have introduced multicollinearity in the model -> VIF Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 26 Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Regressors and Expected Signs Regressors and Expected Signs Coefficient Std. Error T-Statistic p-value Constant Constant 4.9016*** 1.4299 3.4278 0.0006 Bank Solvency (-1) (+) 0.7415*** 0.0342 21.675 0.0000 Bank Risk-taking (-) -1.7409* 0.8975 -1.9396 0.0526 GDP per capita (-/+) -0.4168** 0.1842 -2.2624 0.0238 Unemployment Rate (-) -3.6861** 1.8064 -2.0405 0.0415 Interest Rate (-/+) 4.7274** 2.1135 2.2368 0.0255 Bank Size (+) 0.0127 0.0782 0.1617 0.8715 Bank Debt (-) 0.0487 0.2540 0.1919 0.8479 R-squared R-squared 0.8150 J-Stat (Sargan Statistic) J-Stat (Sargan Statistic) 3.02E-16 Adjusted R-squared Adjusted R-squared 0.8117 Arellano-Bond Serial Correlation Test (m-statistic/p-value): Arellano-Bond Serial Correlation Test (m-statistic/p-value): Arellano-Bond Serial Correlation Test (m-statistic/p-value): Durbin-Watson Stat Durbin-Watson Stat 2.3840 AR(1) AR(2) -6.21 (0.0000)*** 0.92 (0.3516) -6.21 (0.0000)*** 0.92 (0.3516)

37. Variables VIF (with Bank Size and Debt) VIF Bank Solvency (-1) 1.1511 1.1342 Bank Risk-taking 1.3919 1.2548 GDP per capita 4.8597 2.2647 Unemployment Rate 1.3087 1.2845 Interest Rate 1.6409 1.6088 Bank Size 3.4587 NA Bank Debt 1.1532 NA BANK RISK-TAKING ON SOLVENCY RESULTS AND ANALYSIS With Bank Size and Debt, VIF values for GDP per capita and Bank Size were relatively high but below a VIF value of 10 (moderate multicollinearity) When Bank Size and Debt were removed , VIF value of GDP per capita significantly decreased and other VIF values decreased Variance Inflation Factors (VIF) Results

38. Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2005Q3 2023Q4; Total panel (balanced) observation: 1406 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Regressors and Expected Signs Regressors and Expected Signs Coefficient Std. Error T-Statistic p-value Constant Constant 4.8555*** 1.3358 3.6349 0.0003 Bank Solvency (-1) (+) 0.7417*** 0.0339 21.855 0.0000 Bank Risk-taking (-) -1.7805** 0.8514 -2.0912 0.0367 GDP per capita, log (-/+) -0.3952*** 0.1247 -3.1688 0.0016 Unemployment Rate (-) -3.6458** 1.7727 -2.0566 0.0399 Interest Rate (-/+) 4.7840** 2.0751 2.3054 0.0213 R-squared R-squared 0.8150 J-Stat (Sargan statistic) J-Stat (Sargan statistic) 3.45E-17 Adjusted R-squared Adjusted R-squared 0.8119 Arellano-Bond Serial Correlation Test (m-statistic/p-value): Arellano-Bond Serial Correlation Test (m-statistic/p-value): Arellano-Bond Serial Correlation Test (m-statistic/p-value): Durbin-Watson stat Durbin-Watson stat 2.3836 AR(1) AR(2) -6.21 (0.0000)*** 0.95 (0.3423) -6.21 (0.0000)*** 0.95 (0.3423) BANK RISK-TAKING ON SOLVENCY Results: All variables were significant at the 1% and 5% level Bank Solvency: strong persistence Bank Risk-taking: negative coefficient at -1.78 Macroeconomic Variables: GDP per capita: negative coefficient at -0.395 Unemployment Rate: negative coefficient at -3.65 Interest Rate: positive coefficient at 4.78 Model passed all diagnostic tests

39. “H3: Bank opacity plays an intermediary role between macroeconomic changes and solvency among the top 20 UKBs in the Philippines.” BANK OPACITY AS A MEDIATOR RESULTS AND ANALYSIS Scatterplot exhibits an inverted U-shaped curve ZROE initially increases (3.00) when AFSR increases, indicating the initial returns of increased opacity Further increases in AFSR lead to decreased ZROE Correlation coefficient = 0.13, signifying a weak positive correlation Correlation Coefficient: 0.13 AFSR vs. ZROE across Top 20 UKBs

40. Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Regressors and Expected Signs Regressors and Expected Signs Coefficient Std. Error T-Statistic p-value p-value Constant Constant 4.3586** 1.7258 2.5255 0.0117 0.0117 Bank Solvency (-1) (+) 0.7468*** 0.0374 19.968 0.0000 0.0000 GDP per capita (-/+) -0.3514** 0.1649 -2.1308 0.0333 0.0333 Unemployment Rate (-) -3.5273* 1.8357 -1.9214 0.0549 0.0549 Interest Rate (-/+) 3.9956* 2.4162 1.6536 0.0985 0.0985 R-squared R-squared 0.8259 J-Stat (Sargan statistic) J-Stat (Sargan statistic) 3.75E-16 3.75E-16 Adjusted R-squared Adjusted R-squared 0.8227 Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Durbin-Watson stat Durbin-Watson stat 2.3103 AR(1) AR(2) -4.93 (0.0000)*** -0.065(0.9590) -4.93 (0.0000)*** -0.065(0.9590) -4.93 (0.0000)*** -0.065(0.9590) BANK OPACITY AS A MEDIATOR Results: All variables exhibited significant coefficients at different levels Bank Solvency: strong persistence Macroeconomic Variables: GDP per capita: negative coefficient at -0.35 Unemployment Rate: negative coefficient at -3.53 Interest Rate: positive coefficient at 4.00 Model passed all diagnostic tests Baron-Kenny Mediation Analysis - Step 1: Outcome Regression

41. Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 23 Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Dependent Variable: Bank Opacity (AFSR) Regressors and Expected Signs Regressors and Expected Signs Coefficient Std. Error Std. Error T-Statistic p-value Constant Constant 0.0738 0.0829 0.0829 0.8901 0.3736 Bank Opacity (-1) (+) 0.8791*** 0.0180 0.0180 48.906 0.0000 GDP per capita (-/+) -0.0071 0.0080 0.0080 -0.8933 0.3719 Unemployment Rate (+) 0.1411 0.0893 0.0893 1.5797 0.1144 Interest Rate (-/+) 0.0936 0.1174 0.1174 0.7969 0.4257 R-squared R-squared 0.8745 J-Stat (Sargan statistic) J-Stat (Sargan statistic) J-Stat (Sargan statistic) 1.54E-15 Adjusted R-squared Adjusted R-squared 0.8721 Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Durbin-Watson stat Durbin-Watson stat 1.967 AR(1) AR(2) AR(1) AR(2) -1.11 (0.2676) -0.95 (0.3403) -1.11 (0.2676) -0.95 (0.3403) Results: Macroeconomic variables were statistically insignificant Bank Opacity: strong persistence Macroeconomic Variables: GDP per capita: negative coefficient at -0.007 Unemployment Rate: positive coefficient at 0.14 Interest Rate: positive coefficient at 0.09 Model passed all diagnostic tests Macroeconomic variables weakly explains Bank Opacity BANK OPACITY AS A MEDIATOR Baron-Kenny Mediation Analysis - Step 2: Mediator Regression

42. Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Method: Panel GMM EGLS (Period Random Effects) Sample (adjusted): 2008Q2 2023Q4; Total panel (balanced) observation: 1197 White cross-section instrument weighting matrix; Swamy and Arora estimator of component variances; Cross-section weights (PCSE) standard errors & covariance (d.f. corrected); Instrument Rank = 24 Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Dependent Variable: Bank Solvency (ZROE) Regressors and Expected Signs Regressors and Expected Signs Coefficient Std. Error Std. Error T-Statistic p-value Constant Constant 4.0865** 1.6757 1.6757 2.4387 0.0149 Bank Solvency (-1) (+) 0.7414*** 0.0375 0.0375 19.7876 0.0000 Bank Opacity (+/-) 0.5527** 0.2724 0.2724 2.0290 0.0427 GDP per capita (-/+) -0.3290** 0.1600 0.1600 -2.0558 0.0400 Unemployment Rate (-) -3.6554** 1.7846 1.7846 -2.0483 0.0407 Interest Rate (-/+) 3.9673* 2.3456 2.3456 1.6914 0.0910 R-squared R-squared 0.8265 J-Statistic (Sargan statistic) J-Statistic (Sargan statistic) J-Statistic (Sargan statistic) 3.83E-16 Adjusted R-squared Adjusted R-squared 0.8231 Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Arellano-Bond Serial Correlation Test (m-Statistic/ p-value): Durbin-Watson stat Durbin-Watson stat 2.3052 AR(1) AR(2) -4.95 (0.0000)*** 0.20 (0.8435) -4.95 (0.0000)*** 0.20 (0.8435) -4.95 (0.0000)*** 0.20 (0.8435) Results: All variables exhibited significant coefficients at different levels Bank Solvency: strong persistence Macroeconomic Variables: GDP per capita: negative coefficient at -0.33 Unemployment Rate: negative coefficient at -3.66 Interest Rate: positive coefficient at 3.97 Bank opacity: significant positive coefficient at 0.55 Model passed all diagnostic tests BANK OPACITY AS A MEDIATOR Baron-Kenny Mediation Analysis - Step 3: Mediated Regression

43. Decreasing ZROE Decreasing ZROE Constant ZROE Constant ZROE Constant ZROE Increasing ZROE Increasing ZROE Bank Trend Bank Bank Trend Bank Trend Chinabank 4.0 BDO BDO 4.0 AUB > 4.0 DBP 3.0 BPI BPI 4.0 East West > 5.0 PNB < 2.0 Citibank Citibank 2.0     HSBC < 1.0 Bank of Commerce Bank of Commerce < 2.0         LandBank LandBank 5.0         Maybank Maybank 1.5         Metrobank Metrobank 2.0         MUFG MUFG 1.5         PBCom PBCom < 1.0         PhilTrust PhilTrust 1.0         RCBC RCBC 4.0         Security Bank Security Bank 3.0         UnionBank UnionBank 4.0     Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Note: The ideal level for ZROE is at least 2.0 Several banks face significant risks as they increase their AFS securities. This underscores the critical need for effective management of opacity to mitigate the risk of insolvency for these banks. BANK OPACITY AS A MEDIATOR Trend of ZROE vis-à-vis AFSR Based on Individual Bank Scattergram (ZROE Stabilizes to a certain level when AFSR increases) Though some banks effectively manage the risks associated with bank opacity, this analysis reveals a concerning trend:

44. Step 1: Outcome Regression Step 2: Mediator Regression Step 3: Mediated Regression Dependent Variable: ZROE Dependent Variable: AFSR Dependent Variable: ZROE Constant 4.3586** (0.0117) 0.0738 (0.3736) 4.0864** (0.0149) ZCAR(-1)1,3/ AFSR(-1)2 0.7468*** (0.0000) 0.8791*** (0.0000) 0.7414*** (0.0000) GDP per capita, log (LNGDPPC) -0.3513** (0.0333) -0.0071 (0.3719) 0.3290** (0.0400) Unemployment Rate (UNEMP) -3.5272* (0.0549) 0.1411 (0.1144) -3.6554** (0.0407) Interest Rate (INTR) 3.9956* (0.0985) 0.0935 (0.4257) 3.9673* (0.0910) Bank Opacity (AFSR) N/A N/A 0.5527** (0.0427) Outcome Regression: Macroeconomic variables (GDP per capita, Unemployment rate, Interest rate) significantly explain Bank Solvency Mediator Regression: Pathway from the macroeconomic variables to the mediator variable (Bank Opacity) is weak Mediated Regression: Mediator variable (Bank Opacity) significantly explains Bank Solvency Macroeconomic variables are still significant (with minor changes) BANK OPACITY AS A MEDIATOR Summarized Results of Baron-Kenny Mediation Analysis

45. Summary of Conceptual Map with Results FRAMEWORK AND METHODOLOGY 0.5527** -1.7805** Bilateral Causality

46. SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

47. Methods and Indicators Results Panel Unit Root Test and Dumitrescu-Hurlin Panel Causality Test Bank Risk-taking Behavior (NPL Ratio) Bank Opacity (AFS Ratio) Bank Solvency (ROE-based Z-index) There is a bilateral causality between Bank Opacity and Bank Risk-taking Behavior Bank opacity homogeneously causes bank risk-taking behavior of the top 20 Philippine UKBs, and vice versa. Bank Opacity homogeneously causes Bank Solvency, but not vice versa. Significant role of bank opacity in influencing bank solvency, strengthening its potential mediating role . These causal relationships lay the groundwork for understanding how risk-taking behavior, influenced by bank opacity, translates into solvency outcomes, impacting the overall stability of the banking sector. SUMMARY OF RESULTS Hypothesis 1: Increased bank opacity causes heightened risk-taking behaviors among the top 20 UKBs in the Philippines. Objective 1: To empirically verify that heightened risk-taking behavior is caused by the opacity of the top 20 UKBs in the Philippines Objective 1: To empirically verify that heightened risk-taking behavior is caused by the opacity of the top 20 UKBs in the Philippines Objective 1: To empirically verify that heightened risk-taking behavior is caused by the opacity of the top 20 UKBs in the Philippines Objective 1: To empirically verify that heightened risk-taking behavior is caused by the opacity of the top 20 UKBs in the Philippines

48. Methods and Indicators Results Panel Generalized Method of Moments (GMM) Dependent Variable: Bank Solvency (ROE-based Z-index) Independent Variable: Bank Risk-taking behavior (NPL Ratio) Macroeconomic Variables: GDP per capita, Unemployment rate, Interest rate There is a significant negative relationship between bank risk-taking behavior and solvency of the top 20 UKBs in the Philippines. Banks with higher values of NPLs, indicating more risk-taking, had lower solvency levels. This highlights the importance of risk management in maintaining bank solvency, as increased risk-taking is detrimental to financial stability. SUMMARY OF RESULTS Hypothesis 2: Heightened risk-taking behaviors among the top 20 UKBs in the Philippines are negatively associated with solvency Objective 2: To empirically verify the impact of heightened risk-taking behavior on the solvency of the top 20 UKBs in the Philippines Objective 2: To empirically verify the impact of heightened risk-taking behavior on the solvency of the top 20 UKBs in the Philippines Objective 2: To empirically verify the impact of heightened risk-taking behavior on the solvency of the top 20 UKBs in the Philippines Objective 2: To empirically verify the impact of heightened risk-taking behavior on the solvency of the top 20 UKBs in the Philippines

49. Baron-Kenny Mediation Analysis Baron-Kenny Mediation Analysis Results Outcome Regression: Dependent Variable: Bank Solvency (ROE-based Z-index) Independent Variables: per capita GDP, Unemployment Rate, Interest Rate Bank opacity partially mediates the relationship between the macroeconomic variables (GDP per capita, unemployment rate, interest rate) and bank solvency. The inclusion of bank opacity in the model shows that it significantly affects bank solvency. The macroeconomic variables continue to significantly affect bank solvency. There is a significant positive relationship between bank opacity and bank solvency of the top 20 UKBs in the Philippines. Although some banks effectively manage the risks associated with bank opacity, further analysis reveals that several banks face significant risks as they increase their available-for-sales securities. Mediator Regression: Dependent Variable: Bank Opacity (AFS Ratio) Independent Variables: Same as for the outcome regression Bank opacity partially mediates the relationship between the macroeconomic variables (GDP per capita, unemployment rate, interest rate) and bank solvency. The inclusion of bank opacity in the model shows that it significantly affects bank solvency. The macroeconomic variables continue to significantly affect bank solvency. There is a significant positive relationship between bank opacity and bank solvency of the top 20 UKBs in the Philippines. Although some banks effectively manage the risks associated with bank opacity, further analysis reveals that several banks face significant risks as they increase their available-for-sales securities. Mediated Regression: Dependent Variable: Bank Solvency (ROE-based Z-index) Independent Variables: Same as for the outcome regression Mediator Variable: Bank Opacity (AFS Ratio) Bank opacity partially mediates the relationship between the macroeconomic variables (GDP per capita, unemployment rate, interest rate) and bank solvency. The inclusion of bank opacity in the model shows that it significantly affects bank solvency. The macroeconomic variables continue to significantly affect bank solvency. There is a significant positive relationship between bank opacity and bank solvency of the top 20 UKBs in the Philippines. Although some banks effectively manage the risks associated with bank opacity, further analysis reveals that several banks face significant risks as they increase their available-for-sales securities. SUMMARY OF RESULTS Hypothesis 3: Bank opacity plays an intermediary role between macroeconomic changes and solvency among the top 20 UKBs in the Philippines. Objective 3: To determine the mediating effect of bank opacity on the relationship between macroeconomic variables and solvency of banks under the Top 20 Philippine UKBs Objective 3: To determine the mediating effect of bank opacity on the relationship between macroeconomic variables and solvency of banks under the Top 20 Philippine UKBs Objective 3: To determine the mediating effect of bank opacity on the relationship between macroeconomic variables and solvency of banks under the Top 20 Philippine UKBs Objective 3: To determine the mediating effect of bank opacity on the relationship between macroeconomic variables and solvency of banks under the Top 20 Philippine UKBs

50. Enhancing transparency, risk management, and informed regulatory interventions ensures the robustness and sustainability of the Philippine banking sector in the long-term and in the face of economic fluctuations. Significant influence of bank opacity on risk-taking behaviors, and bank profitability and stablity CONCLUSIONS Macroeconomic variables (GDP per capita, unemployment rate, and interest rate) have significant influences on bank solvency Bank opacity and risk-taking behavior mutually influence each other Heightened risk-taking behavior significantly reduces bank solvency 2. Maintaining macroeconomic stability to ensure the financial health of its banking sector 3. Fostering an environment of transparency and accountability can help mitigate the adverse effects of opacity 1. Promoting transparency could be an effective measure for strengthening financial stability

51. RECOMMENDATIONS SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS 1 2 3 To implement stricter disclosure requirements to reduce bank opacity To mandate robust risk management frameworks that include stress-testing, comprehensive risk assessments, and the establishment of risk limits aligned with each banks’ risk appetite To implement policies aimed at maintaining economic growth, reducing unemployment, and managing interest rates effectively FOR POLICYMAKERS,

52. RECOMMENDATIONS SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS 1 2 3 To promote a culture of transparency and accountability within banks. To invest in advanced risk management technologies and data analytics to improve risk identification and mitigation. To pursue growth strategies that prioritize long-term stability over short term gains. FOR BANK MANAGERS AND OWNERS,

53. RECOMMENDATIONS SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS 1 2 3 To expand the scope of research by including a larger sample of banks and conducting longitudinal studies to examine the long-term effects of bank opacity, risk-taking behaviors and macroeconomic variables on bank solvency To investigate different measurements of bank opacity, risk-taking, and solvency that may generate more robust results To explore different methods of mediation analysis such as the bootstrapping method and the Sobel test FOR FURTHER STUDIES

54. AN ANALYSIS OF THE RELATIONSHIP BETWEEN BANK OPACITY, RISK-TAKING BEHAVIOR, AND SOLVENCY OF THE TOP 20 UNIVERSAL AND COMMERCIAL BANKS IN THE PHILIPPINES Dr. Jovi C. Dacanay Gene Matthew E. Chua 8 November 2024