Determinants of Capital Adequacy Ratio of Banks in Botswana

Capital Adequacy Ratio (CAR) plays a very important role in the financial success of banks and acts as a buffer to prevent and absorb any unexpected losses. This study examines explanatory variables that influence CAR for nine banks in Botswana. Multiple linear regression was used for analysis, with CAR as the dependent variable and thirteen financial ratios as the independent variables. The study period is 2015-2019. Based on the data for this period, it was established that out of the thirteen financial ratios utilised, only four were found to have significant impact on the CAR of the nine banks under study, which are: Asset to Equity Ratio (A E), Return on Equity (ROE), Non-Performing Loans Ratio (NPL RATIO) and the Cost-to-Income Ratio (C I). The A E Ratio was found to be the most influential driver of the CAR and the NPL Ratio was found to be the least influential driver of the CAR for the banks under study.


Introduction
Capital adequacy ratio is an important measure of "Safety and Soundness", hence sufficient capital is required by banks as it acts as a buffer to prevent and absorb any unexpected losses. Mishkin (2013) states that banks have to make decisions about the amount of capital they need to hold for the following three reasons: • Bank capital helps to prevent bank failure, • The amount of capital affects returns for the owners (equity holders) of the bank, and • A minimum amount of bank capital (bank capital requirements) is required by bank authorities.
Capital adequacy ratio refers to the amount of equity capital and other securities which a bank holds as reserves against risky assets. According to Basel II, CAR is measured using Tier 1 (Core Capital) and Tier 2 (Supplementary Capital). The sum of the Tier 1 Capital and Tier 2 Capital is divided by the Risk Weighted Assets (RWA) as shown in the equation below: Capital Adequacy Ratio (CAR) = Tier 1 Capital + Tier 2 Capital Risk Weighted Assets (1) where, • Tier 1 Capital is the core capital and it can absorb a reasonable amount of loss without forcing the bank to stop its trading. It includes shareholders' equity and retained earnings.
• Tier 2 Capital is the supplementary capital and it can sustain a loss if there is a liquidation. It includes subordinated loans, revaluation reserves, undisclosed reserves and general provisions.
• Risk Weighted Assets (RWA) are a key measure in risk management, and they are used to determine the minimum amount of capital that must be held by banks in order to reduce the risk of insolvency. Banks can increase their capital adequacy ratios by either increasing the amount of regulatory capital held, or by decreasing their RWA (Das & Amadou, 2012). RWA include credit risk, operational risk and market risk.

Literature Review
El-Ansary and Hafez (2015) carried out a study to investigate the determinants of capital adequacy ratio on commercial banks in Egypt for the period 2003-2013 using multiple linear regression analysis. Some of their major findings were that, profitability had no impact on CAR, however, return on assets was found to be significantly positively correlated to CAR. Liquidity and management quality were found to be positively correlated to CAR while size of bank and risk was found to be negatively correlated to CAR. Asset quality was not correlated to CAR. Al-Tamini and Obeidat (2013) examined the factors that determine capital adequacy of commercial banks in Jordan for the period 2000-2008. Some of the major findings were that there was a statistically significant positive correlation between capital adequacy of commercial banks and liquidity risk, as well as rate of return on assets. Moreover, in their study they found out that there is an inverse relationship with statistical significance between capital adequacy and rate of return on equity and interest rate risk. In a study by Kartal (2019), the factors that influence CAR in the Turkish Banking Sector (TBS) for the period 2006/Q1 -2019/Q1 were studied using Multivariate Adaptive Regression Splines (MARS) on 14 explanatory variables. It was found out that credits/total asset ratio, legal equities, risk weighted assets, non performing loans (NPL), NPL/total credits ratio and credit/deposits ratio were influential factors on CAR in Turkey.
Using panel data regression, Raharjo, Hakim, Manurung, and Maulana (2014) studied the determinants of capital ratio of the state-owned banks in Indonesia and it was established that the capital ratio was affected by the size of the bank, the bank's leverage, the quality of management, and the interest rate risk. Bhattarai (2020) carried out a study to determine the CAR of 11 commercial banks in Nepal for the period 2013/14 to 2017/18. The Pooled OLS, Fixed Effects Model and Random Effects Model were used for data analysis and it was established that liquidity, size of bank and inflation were the major determinants of CAR in Nepal with liquidity having a positive and statistically significant effect, whereas the size of the bank and inflation rate had negative and statistically significant results. El-Ansary, El-Masry, and Yousry (2019) carried out a comparative analysis of CAR determinants between 38 Islamic Banks (IBs) and 75 Conventional Banks (CBs) in MENA countires using GMM. CAR was used as the dependent variable and profitability, liquidity risk, credit risk, bank size, deposits to assets, operational efficiency, portfolio risk and two macro-economic variables as independent variables. In their study, it was found out that both Islamic banks and Conventional Banks have a significant association between CAR and bank size, operational efficiency and DP growth rate. A significant association between CAR and Deposits to Assets ratio was seen for IBs, while for CBs it was found out that there is an association between CAR and profitability, credit risk and portfolio risk. Alajmi and Alqasem (2015) studied the effects of seven internal factors on capital adequacy ratios of five conventional banks in Kuwaiti for the period 2005 to 2013 using fixed effect model and random effect model. When using the fixed effect model, it was found out that size and ROA have a significant and negative relationship with CAR whereas the other five factors did not have any impact on CAR. When using the random effect model, it was found out that CAR was adversely affected by bank size and ROA was found to have a significant and positive relationship with CAR, whereas, LDR showed a significant and positive relationship with CAR. The other four factors did not have a significant effect on CAR.
Dreca (2013) Thoa and Anh (2017) examined the relationship between six internal banking factors (independent variables) and CAR (dependent variable) in the Vietnamese banking system. The study was carried out using linear regression analysis and the study period was 2011-2015. Some of the findings in the study were that, bank size and leverage have no impact on the CAR, while the net interest margin and liquidity have a siginificant positive impact on CAR.

Definition of Variables
When using multiple linear regression, the goal is to predict the dependent variable, which in this study is defined as: • Y= CAR =(Tier 1 Capital + Tier 2 Capital) / Risk Weighted Assets The independent variables X i are: • X 1= A E = Assets / Equity

The Modelling Approach
The study was based on 9 banks in Botswana and the data sources were their annual reports and financial statements.

Multicollinearity Test
The dependent and independent variables are tested for multicollinearity using the correlation matrix shown in Table 2 and it was established that there was no significant correlation between CAR and the 13 independent variables.

Multiple Regression Analysis
The general CAR model is: where; • i ranges from 1 to 13, • β 0 is the constant term to be determined, • β i are the coefficients to be determined, • X i is the i th driver of CAR, • µ is a random error. The CAR model is:: From the CAR model and Table 3, it is evident that A E, ROE, NPL RATIO and C I are the most influential drivers of CAR for the banks under study. It was also established that A E is the largest driver with a coefficient of -0.0115, therefore for every 1-unit increase in A E, there is a 0.0115 decrease in CAR, assuming all other variables are held constant. Therefore, A E and CAR are negatively correlated. The least influential driver with a coefficient of -0.9682 is NPL RATIO, meaning that for every 1-unit increase in NPL RATIO there is a 0.9682 decrease in CAR, assuming all other variables are held constant. Therefore, the NPL RATIO and CAR are negatively correlated. In the absence of all drivers, banks in Botswana generally have a CAR of 43.79% as derived from the constant of 0.4379 in the CAR model above.
The R 2 and adjusted R 2 values of 74.0% and 71.4%, respectively, shown in Table 4 indicate a good explanatory power for the regression model, as the R 2 value shows that 74.0% of the variation in the dependent variable (CAR) is explained by the independent variables. The same applies for the adjusted R 2 value, that is, 71.4% of the variation in the dependent variable (CAR) is explained by the independent variables.  Table 5, the model has a significance value=0.000<0.05 (chosen significance level). This implies that the model is significant at the chosen level of significance of 5%. Therefore, A E, ROE, NPL RATIO and C I have a significant impact on the CAR of the 9 banks in this study.

Back Testing
The model was back-tested using a data set compiled from 2015-2019 and the results in Table 6 were obtained. The model was able to predict the CAR for all the banks with an average error of 0.70%, 0.33%, 0.34%, -0.27% and -1.11% for 2015, 2016, 2017, 2018 and 2019, respectively. On average, for the study period, the model performed the best on 2018 data with an average error of -0.27% and performed the least on 2019 data with an average error of -1.11%. Using the 5 year average, the model was found to be most accurate when predicting the ABSA performance followed by Bank of Baroda, and the least accuracy was noted when predicting the CAR for BancABC.

Model Accuracy Test
The performance of the model in predicting the CAR of the banks under study was measured using the Mean Absolute Error (MAE), Mean Square Error (MSE) and Root Mean Square Error (RMSE). The model was found to perform well and the results are shown on Table 7.

Discussion of the Results and Conclusion
The research analyses the influence of thirteen financial ratios on the capital adequacy ratio of nine banks in Botswana. Asset to Equity Ratio (A E), Return on Equity (ROE), Non-Performing Loans Ratio (NPL RATIO) and Cost to Income Ratio (C I) were found to be the most influential drivers of the CAR in Botswana banks. The NPL ratio was found to be an influential driver of the CAR for the banks under study, which is in agreement with the study by Kartal (2019). The ROE, being one of the most influential drivers, is in agreement with Al-Tamini and Obeidat (2013), who found out that there is an inverse relationship between the capital adequacy and rate of return on equity. The A E Ratio was found to be the most influential driver of the CAR for the banks under study. For the study period, the model performed the best on 2018 data with an average error of -0.27% and performed the least on 2019 data with an average error of -1.11%. Using the 5 year average, the model was found to be most accurate when predicting the ABSA performance followed by Bank