CEO power and bank risk in the UAE

The lessons from the 2008 global financial crisis show that excessive risk taking and governance failures contribute to the failure of several banks. As a result, the relationship between corporate governance mechanisms and risk taking has been the subject of many studies. However, extant studies report inconclusive results. Therefore, this study aims to investigate the relationship between CEO power and bank risk in the UAE using data over the period of 2015–2018 and a sample of 19 UAE banks. The study uses a Pearson pairwise correlation to analyze the relationship between CEO power and bank risk. In addition, a two-tailed t-test is used to examine the differences between conventional and Islamic banks in terms of CEO power and risk-taking. The results of the study show that CEO power measured using CEO duality and CEO tenure reduces risk. Furthermore, the paper indicates that larger boards and higher CEO ownership tend to increase risk. The study also reports that conventional banks have higher return variability, larger boards and powerful CEOs than Islamic banks. However, Islamic banks tend to have higher non-performing finances than conventional banks. The study provides important insights on the relationship between CEO power and bank risk and concurs with earlier studies. The findings can be of interest to policy makers and can be used as input data for the development of corporate governance mechanisms. Shareholders can also use the survey results as input when appointing a CEO for their banks.


INTRODUCTION
The banking sector is of paramount importance to the country's economic development and growth. Specifically, in countries like the United Arab Emirates, where the financial system dominates the banking sector, banks play a significant role in the economy. Naturally, banks are likely to take more risk to generate more cash flows. However, taking excessive risk may lead to a bank failure. Most of the banking transactions involve contracts between two parties (e.g., lender and borrower), where adverse selection and moral hazard is a big concern, and these information asymmetry problems may lead to a failure. A failure of a single bank can have a domino effect that can easily affect other banks in an economy. Moreover, unlike other industries, banks are highly interlinked with the global financial market, and the failure of one global bank may have a spillover effect on banks in other countries (Gebba & Aboelmaged, 2016). A good example is the financial crisis in 2008, which has started in the USA and then spread over to other parts of the globe, including the Middle East. The 2008 financial crisis led to a failure of several banks and recorded significant losses on stock exchanges. After the financial crisis, several studies have focused on the causes of the financial crisis in banks. These studies indicate that excessive risk and inappropriate governance practices lead to the failure of several banks. Since then, policy makers stressed more emphasis on the stability of the banking industry. In an effort to maintain the stability of the banking sector, policy makers constantly update banking regulations and legislations. The central point of these regulations and legislations is to control excessive risk taking, ensure financial stability, and protect the interest of different stakeholders such as shareholders, depositors, borrowers, etc. (Gebba & Aboelmaged, 2016). Accordingly, enhancing the corporate governance of banks is also one of the priorities of central banks. The abovementioned factors motivate this study to examine the relationship between CEO power and bank risk.
CEO power indicates the extent of managerial entrenchment the CEO of the bank exercises. Several studies argue that if the manager is more powerful, the board of directors will be less effective in monitoring and controlling the actions of the manager. Besides, a powerful manager has a significant influence on the strategic decisions of the bank. Higher autonomy in managerial decision making can increase agency conflicts and agency costs and, as a consequence, lead to lower firm value, lower cash flows, lower credit rating and higher debt costs (Bebchuk, Cremers, & Peyer, 2011). The impact of CEO power on the firm's performance is still an emerging area of research, and results from existing studies are inconclusive. Research examining the impact of managerial power and bank risk is also scarce. 1 Islamic banks' corporate governance is complex compared to conventional banks, including several stakeholders and an additional supervisory board, which is called a Shariah board. The Shariah board is entitled to review the activities of the board of directors and the management team (Gebba, & Aboelmaged, 2016).

Theoretical framework
The study follows the literature on corporate governance and employs agency theory to explain the relationship between CEO power and bank risk taking. Jensen and Meckling (1976) indicate that agency relationships exist in all forms of organizations where an agent (manager) is expected to maximize the wealth of a principal (shareholder), but sometimes the manager's interests may differ from those of shareholders, which can lead to a conflict of interest (agency costs). Effective corporate governance mechanism is essential to reduce agency costs arising due to the differences between shareholders' and managers' interests. Such differences can be regulated through constant monitoring and control, as well as managerial compensation (e.g., Jensen & Meckling, 1976). Boards of directors are delegated by shareholders to control and monitor managerial decision-makings and are an effective tool for maintaining the separation between shareholders and management (Fama & Jensen, 1983). An effective board of directors helps to monitor and control self-interest managerial behavior and, as a consequence, mitigate managerial opportunism (Boyd, Haynes, & Zona, 2010). However, these monitoring and incentive mechanisms sometimes may fall short in motivating the firm manager to work in the best interests of shareholders. As a result, the CEO of a firm will tend to have more influence on the decision-making of the firm, and will be more likely to pursue actions and make financial decisions that maximize their financial interests (Pathan, 2009;Sheikh, 2019).
Shareholders can have a diversified portfolio and neutralize their risk by investing in multiple investment securities. In contrast, managers are less diversified, and their risk profile is tied up to the human and financial capital they manage in the company (Sheikh, 2019). Therefore, managers have an incentive to divert company's resources for their personal financial interests and avoid risky projects, leading to a conflict of interest with shareholders. The study argues that this incentive tends to be stronger for powerful CEOs. Therefore, the paper investigates whether the CEO power affects firm risk and addresses the following re-  3) An expert power is measured using a CEO tenure, which equals the number of years the manager of a bank served as a CEO since his first appointment (Pathan, 2009). 4) Prestige power refers to the reputation of a manager, educational background from elite schools and the network of the manager with other partners (Diga & Kelleher, 2009;Finkelstein, 1992). Prestige power can also be measured using the number of board memberships a CEO belongs to (Finkelstein, 1992 2 It is believed that the combination of these four power indicators contributes to the overall managerial power in a company. However, due to data limitations, this study considers structural, ownership and expert power indicators. The study uses the size of the board and CEO duality to measure structural power, CEO tenure to measure expert power and the percentage of share ownership the CEO owns to measure ownership power.

Hypotheses
The extant literature on the relationship between board size and risk taking behavior is mixed. For example, Adams and Ferreira (2009) and Haider and Fang (2016) report a negative association be-   (Jensen, 1993). Therefore, based on the above literature, the study puts forward the following hypothesis: H1: Larger board size reduces bank risk.
CEO duality refers to a situation where a person possesses both chief executive officer and board chairperson positions (Hermalin & Weisbach, 1998). In such a dual leadership structure, the CEO will obtain an opportunity to dominate the board of directors and diminish their effectiveness in controlling and monitoring the management (Daily & Johnson, 1997). With little power of the board, the CEO tends to be more influential, thereby increasing managerial entrenchment (Peng, Zhang, & Li, 2007). Agency theory argues that duality increases CEO power and information asymmetry and weakens the independence of the board of directors (Fama & Jensen, 1983). As a result, the CEO may choose risky projects, which results in lower firm value (Jensen, 1993). Thus, the study argues that CEO duality tends to increase risk-taking behavior and suggests the following hypothesis: H2: CEO duality increases bank risk.
CEO tenure is defined as the number of years the manager serves as the company's CEO (Pathan,   3 As of July 2019, there were 22 national banks in the UAE. However, the study excludes three banks from the sample due to lack of adequate data, thus 19 banks are included.

2009). Sheikh (2019) indicates that CEOs with
higher tenures tend to develop influential power on their boards and, as a result, diminish the controlling power of the board of directors, increase managerial autonomy (Hermalin & Weisbach, 1998) and lower CEOs risk taking behavior to protect their personal financial interests. In contrast, Chen and Zheng (2014) show that CEOs with long tenure are overconfident in their skills and experience to manage uncertainties and take higher risk compared to short-tenure managers. This indicates that CEO's tenure is positively associated with firm risk. Although, the results on the relationship between CEO tenure and bank risk are mixed. Following Sheikh (2019), the study proposes the following hypothesis: H3: CEO tenure is negatively related to bank risk.
One of the ways a CEO maintains power in a firm is by owning some shares of the company. A CEO who owns shares in the company is expected to converge his interests with shareholders' interests (Fama & Jensen, 1983). As a result, the CEO exhibits an incentive to maximize firm value. Allen (1981) also indicates that CEOs with higher share ownership in a company tend to have greater influence on the strategic financial decision making of the company. A recent study by Sheikh (2019) shows that a higher share ownership makes the CEO of the company more concerned about losing firm value due to his less diversified portfolio, and this provides an incentive to lower company risk. Accordingly, the following hypothesis is proposed: H4: CEO ownership is negatively related to bank risk.

Data source
The study focuses on national banks of the UAE. Financial performance and governance data are collected from banks' annual reports and the Bloomberg database. The study combines both governance and financial performance data, and this results in a sample of 19 banks over the period of 2015-2018. 3 The study uses two measures of bank risk as dependent variables. (1) Loan portfolio risk is measured using Non-performing loans (NPL), Loan loss coverage ratio (COVERAGE), and Loans to total asset ratio (LOANS). Non-performing loan ratio (NPL) is measured using the size of loans or Islamic financing more than 90 days overdue as a proportion of the gross loans or Islamic financing. Non-performing loan ratio is the most commonly used measure of bank risk in the banking literature (e.g., Dong, Girardone, & Kuol, 2017; Pathan, 2009). Loan loss coverage ratio (COVERAGE) is measured using the amount of loan loss reserve as a proportion of non-performing loans. This ratio measures the ability of a bank to absorb the risk of loan losses. The study includes the Loans to total asset ratio (LOANS) to measure bank risk in terms of the volume of loans distributed as a proportion of total assets of a bank. The paper argues that a higher ratio of loans to assets indicates higher probability of default. (2)  Z-score is computed using a moving standard deviation of ROA over the previous two years, and these values are combined with a current return on assets and capital to asset ratios (Delis, Tran, & Tsionas, 2012). 4 Therefore, Z-score is computed as the sum of current year return on assets and the capital to asset ratio divided by the moving average standard deviation over the previous two years, using the following equation: where ROA is return on assets, Cap/TA represents the capital to total assets ratio and . STD ROA refers to the standard deviation of return on assets, estimated using a two-year moving average. 4 Z-score is interpreted as an inverse of the probability of insolvency, and a higher value shows the farthest distance to bankruptcy, indicating a lower bank risk (Laeven & Levine, 2009).
Several studies confirm that Z-score values are considerably skewed (e.g., Laeven & Levine, 2009), and it is highly recommended to transform Z-score values using the natural logarithm. Therefore, Z-score values are transformed using natural logarithms to make the data normally distributed.
The study constructs the CEO power variable using four measures, which are widely used in the banking industry: 3) CEO's share ownership is measured using the proportion of shares of the company owned by the CEO (Sheikh, 2019).
4) The size of the board is also included as a measure of CEO power. It shows that larger boards face coordination and communication challenges, resulting in higher managerial entrenchment and opportunism that increase the CEO power. Table 1 provides a description of the variables.
Following the literature, the study uses several control variables to control for bank level differences.
1) Bank size, the natural logarithm of total assets is used to measure the size of the bank.
2) Bank profitability, measured using return on assets to control for the differences in annual returns, and 3) Capital to assets ratio, measured as the capital to assets ratio to control for the capitalization differences between the banks.

Method of analysis
First, the study runs a summary of statistics to analyze the mean, median, standard deviation, minimum and maximum values for CEO power and bank risk indicators. Second, it employs a Pearson pairwise correlation to examine the relationship between CEO power and bank risk. Similar to Altunbaş, Thornton, and Uymaz (2019), this study recognizes that CEO power is multifaceted, so several indicators were used to measure it. However, due to the limited sample size in this study, it is not possible to formulate the composite CEO power variable using principal component analysis. Nevertheless, to further analyze the strength of the relationship between CEO power and bank risk, the study uses Pearson's pairwise correlation between the interaction of CEO power indicators and bank risk variables. Finally, a twotailed T-test was used to investigate if there are differences between conventional banks and Islamic banks in terms of CEO power and risk taking behavior.

RESULTS
This section reports the results of the study. First, the descriptive statistics are reported to provide an overview of the variables used in the study, followed by the Pearson correlation and paired t-test results.

Summary statistics
The mean (median) loan portfolio at risk ratio (PAR) for 2015-2018 reported in Table 2 is 5.9% with a standard deviation of 2.2%, indicating that 5.9% of gross loans of UAE banks are overdue for 90 days. The average loan loss coverage ratio (COVERAGE) of banks in the UAE is 108.4% with a standard deviation of 24.1%. Table 2 also indicates that the average gross loan to total asset ratio (LOANS) is 58.6% with a standard deviation of 10.5%.
The average standard deviation of return on assets (STD.ROA) and standard deviation of return on equity (STD.ROE) is 0.28% and 2.23%, respectively, and the mean (median) Z-score is 87.72 (62.4).
The mean (median) board size (BSIZE) reported in Table 2 indicates that UAE banks have eight board members on average. The amount of loans or Islamic financing more than 90 days overdue as a proportion of gross loans and advances or Islamic financing. Loan loss coverage ratio (COVERAGE) The amount of loan loss reserve as a proportion of non-performing loans. Gross Loans to total Asset ratio (LOANS) The size of loans distributed as a proportion of total assets. Standard deviation of return on assets (STD.ROA) Standard deviation of bank's ROA for the previous two years. Standard deviation of return on equity (STD.ROE) Standard deviation of bank's ROE for the previous two years. Insolvency risk (Z-score) (ROA + Capital to assets ratio)/Standard deviation of ROA.

Independent variables
Board size (BSIZE) The number of people who comprise the board.

Control variables
Bank size (SIZE) The natural logarithm of the bank's total assets. Bank profitability (ROA) The ratio of net income to the average total assets of the bank. Bank capital ratio (CAPRATIO) The ratio of owners' equity to the average total assets of the bank.
The mean (median) profitability (ROA) and size (SIZE) of a UAE bank is 1.31%( 1.50%) and 128.5 (58.2) billion Dirhams, respectively. The average capital to asset ratio (CAPRATIO) of banks in the sample is 14%.

Pearson correlation
As reported in Table 3 Both CEOtenure and CEOWN are positively and significantly associated with Z-score (ρ = 0.401 and 0.475, respectively). Table 4 reports the correlation between measures of risk and the interaction of CEO power variables.
Using an interaction helps understand the effect of a combination of several measures of CEO power on bank risk. The interaction between BSIZE and CEOdual is negatively and significantly associated with PAR (ρ = -0.471) and positively and significantly associated with COVERAGE (ρ = 0.418) and LOANS (ρ = 0.246). The interaction between BSIZE and CEOtenure is negatively and significantly associated with STD.ROA (ρ = -0.409) and STD.ROE (ρ = -0.396). In addition, the interaction between BSIZE and CEOtenure is positively and significantly associated with Z-score (ρ = 0.324).
On the other hand, the interaction between BSIZE and CEOWN is positively and significantly asso-  Table 3. Pearson correlation analysis of the relationship between dependent and independent variables  Table 5 presents a statistical comparison between two independent samples -the Islamic and conventional banks, and assesses if they have differ-  Note: Significance levels at 10% ( * ), 5% ( ** ) and 1% ( *** ). Note: Significance levels at 10% (*), 5% (**) and 1% (***). ences in their CEO power structure and risk taking behavior. Islamic banks differ from conventional banks mainly in the way they generate their revenue, which is mainly from different financial services based on the virtue of sharing costs and benefits. The study argues that these differences affect the risk taking behavior of CEOs in conventional and Islamic banks.

Paired t-test analysis
The average PAR for conventional banks is 5.21%. Islamic banks' PAR accounts for 6.6%. Column 5 of Table 5 shows that there is a significant difference between conventional and Islamic banks, indicating that Islamic banks have higher PAR than conventional banks. However, the COVERAGE ratio for Islamic banks (111.87%) is greater than for conventional banks (106.42%).  Table  5 indicates that there is no significant difference between conventional banks and Islamic banks in terms of loans to assets ratio (LOANS), standard deviation of return on equity (STD.ROE) and CEO duality (CEOdual).

Summary statistics
The summary statistics reported in Table 2 show that UAE banks have relatively lower non-performing loans (5.9%), and the quality of loan portfolios of UAE banks is improving from year to year, with a declining non-performing ratio of 6.2% in 2015 to 5  ROA) and standard deviation of return on equity (STD.ROE) presented in Table 2 indicate that UAE banks report low return variability. These results are supported with a lower mean (median) Z-score of 87.72 (62.4), which shows that UAE banks have low insolvency risk, suggesting a relatively low level of failure risk.
The average board size of UAE banks in the sample is eight, which is comparable to the findings of Gebba and Aboelmaged (2016). Overall, the summary statistics show that UAE banks' profitability is lower compared to their size. However, UAE banks perform more than the average return on assets for GCC countries' banks (KPMG, 2018).

Correlation analysis
The negative and significant relationship between BSIZE, CEOdual, CEOtenure and PAR reported in Table 3  As reported in Table 4, the interaction between BSIZE and CEOdual is associated with lower loan portfolio at risk (PAR) ratio and higher coverage ratio (COVERAGE), indicating that banks with larger board size and CEO duality tend to improve their asset quality by lowering non-performing loans and maintaining adequate loan loss reserves. Banks with larger boards and experienced CEOs are associated with lower return variability (STD. ROA & STD.ROE) and insolvency risk (Z-score). In addition, banks with larger boards and with CEOs owning significant number of shares of a bank tend to increase return variability and insolvency risk. While Sheikh (2019) indicates that a higher share ownership makes the CEO of a company more worried about losing firm value, the results of this study contend agency theory expectations. This may be due to the increased power of the CEO that comes from the inconclusive large board and ownership power, which forces the CEO to make extreme financial decisions. Table 4 also shows that a bank with an experienced manager (CEOtenure) who is the chairperson of a board (CEOdual) tends to reduce bank risk. In addition, banks with a larger board (BSIZE), whose CEO owns a large proportion of shares (CEOWN) in the bank and who is also a chairperson of the board (CEOdual), tend to reduce bank risk. The results of this study are consistent with previous studies such as Haider and Fang (2018), Hermalin and Weisbach (1998) and Pathan (2009), which report a negative association between CEO power and firm risk.

Paired t-test analysis
Column 5 of Table 5 shows that Islamic banks have higher loan portfolio at risk (PAR) than conventional banks. However, Islamic banks maintain more loan loss reserve (COVERAGE) compared to conventional banks. This may be due to higher non-performing loans associated with Islamic banks.
Conventional banks have higher return variability and lower insolvency risk than Islamic banks. They also have larger boards and experienced managers. In addition, CEOs of conventional banks tend to own more shares than CEOs of Islamic banks.
Overall, conventional banks have powerful CEOs and report lower non-performing loans and failure risk than Islamic banks.

CONCLUSION
This paper examines the relationship between CEO power and bank risk using the data from 19 banks in the UAE over the period of 2015-2018. Overall, UAE banks have lower bank risk, lower variability of returns and lower bankruptcy risk. The results indicate that conventional banks have more powerful CEOs that are associated with higher performance variability and lower insolvency risk compared to Islamic banks. However, Islamic banks have higher non-performing loans compared to conventional banks. The paper also provides evidence that CEO duality and CEO tenure are negatively associated with bank risk. The results provide little evidence to support the impact of board size and CEO ownership on bank risk taking.
Several stakeholders could benefit from this paper. First, the results may be of interest to policy makers and can be used as input in designing corporate governance regulations and legislations. It is important to emphasize that the structure and extent of CEO power need to be taken into account when assessing the risk taking behavior of UAE banks. Second, shareholders can also use the results of this study as input while appointing a CEO for their banks. Identifying the features of different CEO power indicators and their potential impact on risk taking behavior is essential in appointing a CEO who fits in with the interest of shareholders to maximize their wealth.
This study is limited to UAE banks, and it is recognized that the results obtained cannot be generalized. Future research is suggested to expand the findings in a broader context by conducting a multi-country analysis to examine the impact of CEO power on bank risk-taking and contribute to the ongoing debate about the impact of CEO power on bank risk.