“To what extent is there a relationship between internal corporate governance mechanisms and risk taking in the insurance industry?”
Abstract: This thesis researches the relation between several corporate governance variables (board size, CEO compensation and blockholdership) and risk taking in the insurance industry. Using a sample of 31 listed insurance companies from Europe over 2010 – 2014 (139 observations), we find that blockholdership is positively related to risk taking of an insurer (with the volatility of the stock as proxy). This result was consistent after robustness tests.
Andrej Jelenic
Student number: 10481923
Amsterdam Business School (UvA)
Thesis Executive Master of Finance and Control
Date: 23-10-2016
Table of Contents
1. Introduction ... 1
2. Literature review and hypotheses ... 3
2.1 Agency theory and the conflict of interest between agent and principal ... 3
2.2 Corporate governance in the insurance sector ... 5
2.3 Corporate governance variables ... 5
2.3.1 Board size ... 6
2.3.2 CEO Compensation ... 8
2.3.3 Blockholders ... 10
2.4 Hypotheses ... 12
3. Legal framework of Solvency I ... 13
3.1 Regulation in the insurance industry ... 13
3.2 The Solvency I framework ... 14
4. Data and Methodology ... 15
4.1 Sample description ... 15
4.5 Research design ... 16
4.2 Dependent variables: risk measures ... 17
4.3 Corporate governance variables ... 18
4.3.1 Board size ... 18 4.3.2 CEO Compensation ... 19 4.3.3 Blockholdership ... 19 4.4 Control variables ... 19 5. Results ... 21 5.1 Descriptive statistics ... 21 5.2 Results 2SLS regressions ... 24 5.2.1 Board size ... 26 5.2.2 CEO compensation ... 27 5.2.3 Blockholdership ... 28 5.2.4 Control variables ... 29
5.3 Implications of the results ... 30
5.4 Robustness test ... 30
6. Summary and conclusions ... 33
6.1 Summary ... 33
6.3 Limitations research ... 34
6.4 Possibilities for further research ... 34
References ... 36
1
1. Introduction
The financial crisis resulted in a catastrophe for financial institutions. Insurance companies made sizable losses due to their investments in equity markets, credit default swaps (selling protection) and mortgage backed securities. The insurance industry lost worldwide $ 225 billion dollars during the financial crisis (Impavido and Tower, 2009). As a result, a significant number of financial institutions were bailed out by the Government. For example in The Netherlands, Nationale Nederlanden (as part of ING), AEGON, VIVAT (as part of SNS) and ASR (as part of Fortis) were all
supported by a capital injection or completely taken over by the government. A debate has started on what caused the financial crisis. Some scholars argue that excessive risk taking by managers led to this disaster (Harrington, 2009; Dionne, 2009). In the time before the financial crisis, insurance companies were looking for higher yields. They invested in financial instruments without knowing the risks and were relying on rating agencies who did not have the right incentive of objective monitoring these assets (Dionne, 2009).
For different stakeholders (shareholders, regulators, and employees) it is interesting to find out how a new financial crisis can be avoided in the future. Shareholders wonder which measures will help them in the future to avoid any excessive risk taking from management and avoid losing their investments? This principal-agent problem (Jensen and Meckling, 1976) has existed for decades but is still an active subject in this field and is more important than ever after the crisis. Corporate governance is known as a mechanism to cope with agency problems between shareholders and managers. A wide range of different sort of mechanisms are investigated in the literature (Boubakri, 2011). The most known external mechanisms are takeovers, analysts, the legal environment, regulators and rating companies (Agrawal and Mandelker, 1990; Shleifer and Vishny, 1997). Internal corporate governance
mechanisms are e.g. board size, number of outside directors, board independence, managerial compensation, inside ownership, blockholders and reducing free cash flow by debt and dividend policy (Cheng, 2008; Coles et al., 2003; Laeven and Levine, 2009; Shleifer and Vishny, 1997).
2
Our main research question is to find out if there is a relationship between internal corporate governance mechanisms and risk taking in the insurance industry. In this study, we investigate the relationship between three corporate governance
mechanisms (board size, blockholders and CEO compensation) and the risk taking of an insurance company with the volatility of stock and the solvency risk as proxies. If these factors are related, can the supervisor in the insurance industry then use this information for its regulation and prevent another crisis?
Researching this topic is interesting for several reasons. While there have been a large number of studies investigating the effect of corporate governance in public companies in general (McColgan, 2001), studies on the effect of corporate
governance on companies within the insurance industry remains scarce. In addition, most of the papers which do investigate this relation mostly look into the relationship with corporate performance and not on risk taking. This paper will therefore be an addition to the existing literature. A significant amount of regulation has arisen in the insurance industry after the financial crisis. In Europe, insurance companies are investing massive resources to be Solvency II compliant at the implementation date of the 1st January 2016. Solvency II is a new regulation for insurance companies in
Europe to reduce the risk of insolvency. This paper contributes to the discussion which factors play a role in the risk taking of an insurance company. This is not only important information for the regulator drafting the regulation, but also for the
controller who plays a part in advising on the risk management framework and the set of internal controls within the company.
Section 2 discusses the agency problem theory, several corporate governance mechanisms and the resulting hypotheses. Section 3 gives an overview of the legal framework of Solvency and section 4 gives information on the data collection, the definition of the variables and the regression methods used. Section 5 presents the results whiles section 6 provides a conclusion and a brief summary of the study.
3
2. Literature review and hypotheses
In this section we first discuss the agency theory, corporate governance and its mechanisms. Then we elaborate on several corporate governance mechanisms we use in our research with previous literature and its theory. In the last paragraph of this section, we outline our main hypotheses. Appendix I give a summarized overview of the most relevant literature concerning this study on the relationship between
corporate governance and risk.
2.1 Agency theory and the conflict of interest between agent and principal
Berle and Means (1932) were the first scholars to describe the potential problems which arise when separating the ownership of a company from its control. When companies became larger and larger in the beginning of the 19th century, it was
difficult to finance these companies by a single owner. Additional capital had to be provided by external sources and ownership therefore became dispersed. The dispersed ownership resulted in that control of the company was given to the managers of the company and a separation of ownership and control was realized. This separation between the owner and the manager can create a suboptimal result for the owners.
Jensen and Meckling (1976) built further on the statements of Berle and Means (1932) and introduced the agency theory. In this theory there arises a conflict of interest during the agency relationship. A principal (shareholder) hires an agent (manager) to make decisions in the company on his behalf. Both seek to maximize utility and the manager can make decisions which are in his own best interest and not in the best interest of the shareholder. This can occur when their interests are not aligned and there is asymmetric information. The shareholder is less informed and therefore it is difficult for him to assess whether the manager is acting against his interest. The manager can act against the interest of the shareholder by indulging into legal and/or illegal activities. A few examples of these activities are theft, fraud, insider trading, bribes, empire building, corporate perks, excessive pay and
entrenchment. These examples are costly for the shareholder because less is left for him. The costs for hiring an agent by the principal are called agency costs. Jensen and Meckling (1976) divide agency cost in three types of cost. The first type is monitoring costs. This does not only imply monitoring the manager, but also
4
implementing rules, restrictions and compensation policies. The second type is bonding cost where the agent incurs cost to convince the principal he is working in the interest of the shareholder. The third type of agency cost is residual loss which is the loss in shareholder’s value when the principal is acting against the interest of the shareholder.
Another problem between the shareholder and the manager is that the manager could have a different perspective and preference on risk taking. There are several views on this subject (Boubakri, 2011). The first view is that shareholders can diversify their investments and get rid of the unsystematic risk tied to the company. Most managers receive their salary from a single company and are limited in
diversifying their income stream which indicates that they could be more risk averse than the shareholders who are relatively less exposed to the company. Another view is that managers take more risk because the length of their career at the company is shorter than the time frame in which a company can produce cash flows. The
managers will therefore be focused on a shorter horizon when making decisions then optimally could be for the shareholder.
There are several definitions on corporate governance (Fadun, 2013). Most of these definitions are focused on the point of view of the investor (Shleifer and Vishny, 1997; Mayer, 1997; Metrick and Ishii, 2002). For example, that corporate governance deals with getting a return on the investors capital invested and that the companies are managed in a way that are beneficial for these investors. Other scholars have a broader view of corporate governance, they include the society as a whole (Deakin and Hughes, 1997; Cadbury, 2000). This includes stakeholders such as customers (in this study policyholders), creditors, employees, the community and the
environment.
There are several mechanisms for reducing agency problems which can be divided in to internal and external mechanisms. For example, internal governance mechanisms are the board of directors, remuneration, monitoring by large shareholders. External governance mechanisms are regulation, analysts, rating agencies, takeovers and vote service providers. Vote service providers are companies that give advice on how to vote on a shareholders meeting. In this study, we focus on the corporate
5
investigate whether they are related to the degree of risk taking within an company. The choice for these three mechanisms is mainly due to data availability.
2.2 Corporate governance in the insurance sector
The link between corporate governance and the insurance sector became more interesting after the financial crisis, when insurance companies around the world struggled to keep their head above water due to their investments into mortgage-backed securities (Baranoff and Sager, 2009). Most of the studies performed to date on corporate governance mechanisms in the insurance industry are on the
differences of governance characteristics between mutual insurance funds and stock insurance companies (Lamm-Tennant and Starks, 1993; Lee et al., 1997; Mayers and Smith 1986; Mayers and Smith, 1992; Cummins et al., 2007; He and Sommer, 2011). Mutual fund insurance companies are insurance companies which are owned by the policyholders and stock insurance companies are insurance companies which are owned by investors. These studies are not in the scope of our research as we only use stock insurance companies due to data availability.
Research within the insurance industry is always a specialty. Most of the studies on corporate governance mechanisms exclude financial companies as the
characteristics of the companies are different from other companies (e.g. the structure of the balance sheet and income statement). Due to limited availability of studies on the relationship between corporate governance mechanisms and risk taking, we will mostly rely on research that cover other industries and mostly exclude financial companies. One of the factors that sets the insurance industry apart from other industries is the regulation in the insurance industry. This regulation has increased after the financial crisis and in the beginning of this year (2016). Solvency II has made its entry in the sector. One of the most profound changes in this sector is the increase in disclosure reporting the insurance industry has gone through. This increase in transparency can attribute to reducing agency problems by lowering information asymmetries between the agents and the stakeholders (Eling and Schmeiser, 2007).
2.3 Corporate governance variables
In this study we test the relationship of several corporate governance variables in relation to risk. Most of the available literature around corporate governance is the
6
relationship between corporate governance variables and corporate performance. Only in the recent period scholars have shift their interest to the relationship between corporate governance and risk taking within an organisation. A choice had been made on if we research one corporate governance variable in depth or more
corporate governance variables on a more broader base. We will research multiple governance variables as it is a complex subject. There could be a number of
corporate governance variables which have an impact on risk taking. In the following subparagraphs we will discuss the existing literature on board size, CEO
compensation and blockholdership.
2.3.1 Board size
One of the corporate governance mechanisms to align the interest between the principal and the management (the agent) is the board of directors by monitoring the lower management. In Europe, there a two type of board structures. In a one-tier board, common in the U.K., executive and non-executive directors form one board. In a two-tier board, common in Germany and The Netherlands, there are two boards. One board is the executive committee and the other the supervisory board. The task of the board of directors is to monitor the management whether they act in the interest of the shareholder thus being part of the solution to agency problems between agent and principal.
From a social psychological point of view, it is interesting to look at how the process of group decision making occurs in general concerning riskiness and extremity. There have been several studies on this topic (Adams et al., 2005; Kogan and Wallach, 1966). The main conclusion is that the final decision made by a group of people (after debating), is always the result (or the consensus) of the individual positions of those group members. We can apply the same type of reasoning to the size of boards. According to Lipton and Lorsch (1992) and Jensen (1993), larger boards are not effective because then it becomes more difficult for board members to give their opinions and discuss topics in a limited amount of time. The larger the board is, the more difficult it is to reach consensus and therefore the less likely risky proposals will be supported due to a compromise of individual opinions. This consequence could be opposite of the risk appetite of shareholders with a small holding in the company who are already diversified compared to the large blockholders. In addition, larger boards may increase agency problems. The larger the board is, the easier it becomes for
7
board directors to free-ride. For example, not being an active part of the decision making process while receiving salary (Hermalin and Weisbah, 2001). An opposite view on the relationship between board size and risk taking within a company is stated by Klein (2002). He argues that the larger the board, the more monitoring can take place as the board committees can be distributed over more directors.
Several scholars studied the relationship between board size and risk taking within a company. Pathan (2009) investigates whether strong bank boards (characterised in this study by a small board size, more independent directors etc.) are related to bank risk taking measured by the volatility of stock returns. The sample of his study
consists of U.S. large bank holdings and the estimated method used is the
generalized least square (GLS) with random effects. Pathan (2009) finds a positive relationship between strong bank boards (small board) and bank risk. In other words, he finds a negative relationship between the size of boards and the volatility of stock returns. Cheng (2008) performed a study on the relationship between board size and the variability of corporate performance. The sample consists of 1,252 non-financial US companies over the period 1996-2004. He uses in his study a range of risk variables by measuring the volatility of corporate performance indicators such as the volatility of stock returns, the level of R&D expenditures, the frequency of acquisition and restructuring activities and accounting accruals. He finds a negative relationship between the variability of the performance measures (among which the volatility of stock returns) and board size. Nakano and Nguyen (2012) investigate 1,324 non-financial companies in Japan performing both OLS and 2sls regression. They find that larger boards are related to a lower volatility of the performance and lower bankruptcy risk. Their outcome is consistent with the results of Cheng (2008) and Pathan (2009), although the effect is not as significant as those studies which only have US companies in their sample. Huang and Wang (2015) research the
relationship between board size and corporate risk taking choices for a group of Chinese listed (non-financial) companies between 2003 and 2011. The find strong evidence that board size is negatively related to riskier company policy choices. Wang (2012) uses cash flow and stock return volatility as proxy for risk taking within a company and finds these negatively related to board size. The data sample
8
To summarize, the findings are in line with each other. We therefore expect that board size is negatively related to risk.
2.3.2 CEO Compensation
Compensation incentives are one of the mechanisms addressed by Jensen and Meckling (1976) to align the interest of the principal and the agent when contracts are not complete. From a perspective of diversification, managers are generally more risk averse than the risk appetite of the shareholder. Laeven and Levin (2009) even find in the regulated bank industry that when shareholders have more power, banks take more risk which is in line with the research of Saunders et al. (1990). For
shareholders it is possible to diversify their investments in companies by holding a diversified portfolio while most managers cannot diversify their income sources. The chances are high that the majority of income for a manager comes from the company they are working at. Furthermore, when a company faces bankruptcy the manager will face reputation damage which can have a large influence on his future career possibilities. This creates a moral hazard problem which can be mitigated by
introducing the right compensation scheme by linking the compensation of the CEO to the performance of the company (Mehran, 1995). Other research suggest that compensation with a variable component (equity- or option based) will give the right incentive (Wen and Chen, 2008; Coles et al., 2003). The compensation package of a CEO can include a fixed component (base salary, pension benefits and additional benefits such as a company car) and a variable component dependent on the performance of the company (stock, options and variable pay based on certain key performance indicators). Off course there is also an opposing view in which option compensation may lead to perverse incentives for agents and suboptimal results for the principal (Dey-Tortell et al., 2005).
Wilson and Higgins (2001) find in their results that the compensation levels of insurance companies are similar to that of unregulated non-insurance companies. The structure of the compensation package within the financial industry however could be different according to Houston and James (1995). They find that CEOs in the regulated bank industry obtain less options and stock as percentage of their total remuneration package than CEOs in other industries. Gray and Cannella (1997) find a negative relationship between total executive compensation and company risk while they expected a positive relationship. They measure risk by a different set of
9
variables namely the betas of the stock options, the standard deviation on return on sales, R&D as percentage of sales and capital intensity (fixed assets as percentage of total assets). Gray and Canella (1997) give no explanation on why the result could contradict their hypothesis. They only give a remark that it would be difficult to retain talented employees in risky companies which pay less compensation. Eling and Marek (2013) study the impact of the level of compensation with risk taking in the U.K. and insurance markets. They also find a negative relationship. Risk is measured through an opportunity asset risk (the yearly return of an asset class versus the portfolio mix of an insurance company), product risk (loss and benefits ratios) and financial risk (total investments to equity). They use several control variables such as size, a country dummy, line of business and accounting standard. Eling and Marek (2013) expected a positive relationship. They argue that the opposite result could be attributed to the corporate performance which could affect the variable part of the compensation and leads to total lower compensation. Podder and Skully (2013) investigate US listed insurance companies and they find a positive relationship between total CEO compensation and risk taking. As risk measures they used the standard deviation of the daily stock returns for each year and the underwriting risk of an insurer (the percentage change in loss reserves). We can conclude that the
existing literature on this relation is inconclusive. When we follow the theory on using compensation to align the interest between the risk neutral shareholders and risk averse agent, we expect that total CEO compensation is positively related to the insurance companies risk taking.
Other research has focused more proportion of the variable pay in total
compensation. Most results find a positive significant result that the structure of the executive compensation induces risk-taking. Most of these studies research use as variable component equity-options (Chen et al., 2001; Guay, 1999; Wen and Chen, 2008; Cohen et al., 2000), others equity (Low, 2009; Coles et al., 2004) or the proportion of the variable component versus the total compensation (Miller et al., 2002). The Sarbanes-Oxley Act (SOX) which was introduced in 2002 implements several requirements in the U.S. on the level of corporate governance. Because of this implementation, the proportion of options of the total compensation has
decreases significantly in the U.S (Carter et al., 2007). Even though we will use an European sample in our study, we will state our hypothesis on a broader base. We
10
therefore expect that the proportion of the variable incentive compensation of the total compensation of the CEO is positively related to risk taking of their company.
2.3.3 Blockholders
Another corporate governance mechanism which could align the interest between the principal and the agent is by monitoring of blockholders (large shareholders). Time, money and other resources are needed by the shareholder to monitor the managers. It depends on the size of the shareholder’s holdings whether the cost for monitoring outweighs the rise in shareholder’s value. We assume that monitoring costs are fixed but the benefits from monitoring are not (Pagano and Roell, 1998). So the more of monitoring there is, the less the agent can divert from the principal’s value
maximization. This implies that investors with a small stake in the company do not have an incentive to monitor (or can influence the outcome; Zeckhauser and Pond, 1990) but investors with larger holdings do (Shleifer and Vishny, 1986). The size of the shareholdings are therefore an important factor in this discussion. Monitoring does not only imply observing management, but it also indicates other activities the shareholders can engage in. They can vote against or in favour of proposals or mergers, support board changes, bring in their own proposal at a shareholders meeting, advise management or sell their shares (Holderness and Sheehan, 1988; Agrawal and Mandalker, 1990). Several theories concerning blockholders in relation to risk taking can be found in the literature.
One theory (the large shareholder hypothesis) is that investors can have the
incentive to induce higher risk taking in companies at the expense of the debtholder (Jensen and Meckling, 1976). By increasing the riskiness of the assets, the investor can create a semi call option for himself where the benefits are high, but the
downside is only his investment. According to Saunders et al. (1990), debtholders can only monitor and control the action of the investor imperfectly. The stockholder can induce the management of companies to take higher risks to attain higher profits. Blockholders have the largest incentive for this because they will also benefit the most. This is in line with the results of Laeven and Levine (2009) who find that diversified shareholders within the banking industry have a higher risk appetite than the debtholders and the managers. Paligorova (2010) mentions that this excessive risk taking of a company may manifest itself in the gathering of assets with a high
11
volatility. This would not be applicable in this study as the asset mix of an insurance company is strongly regulated.
The other, most straightforward theory is that blockholders have such a concentrated stake in a specific company that their risk is not diversified. They are therefore more risk averse than the other more diversified shareholders which leads to a reduced risk taking by managers through monitoring (Cheng et al., 2011). They then push for conservative investment to protect their cash flow. This theory is called the
suboptimal diversification theory and was first discussed by Fama and Jensen (1983). Another line of research on the topic of large shareholders is the
expropriation by minority shareholders (Demsetz and Lehn, 1985; Stulz, 1988). The interest of the blockholder could differ with the interest of the other shareholders (or stakeholders) which can result in taking private benefits by the large shareholder when having enough control. This could be at the expense of the minority
shareholders.
Cole et al. (2011) study the role of several shareholders in the monitoring process in relation to risk. Not only have they used the variance in return on assets (ROA) as an risk measure, but they have also used the Best’s capital adequacy ratio which
combines multiple variables of the insurer’s risk including asset, credit, underwriting, off-balance sheet, and interest rate risks. They find that a higher concentration of ownership is related to lower risk. Cheng et al. (2011) investigate the relationship between ownership of institutional investors and risk taking within insurance
companies and find that institutional ownership stability (ownership during a longer period) is negatively related to total risk where total risk is derived from the volatility in stock returns. They control for size, trading volume and P/E ratio. Eling and Marek (2013) study the relationship between blockholdership and risk taking in the U.K. and Germany for insurance companies. They measure risk taking by several measures such as asset (standard deviation of a theoretical asset return), product (loss and benefit ratios) and financial risk (logarithm of the ratio of total investments to total shareholder equity) and find that the presence of blockholders has a negative impact on risk taking within insurance companies.
12
To summarize, most studies find that the presence of blockholders has a negative impact on risk taking within a company. Therefore, we expect that the presence of a blockholder is negatively related to risk taking.
2.4 Hypotheses
Related to our considerations in the previous paragraphs, we can state the following hypotheses:
H1: Board size is negatively related to risk taking in insurance companies.
H2a: Total CEO compensation is positively related to the insurance companies risk taking.
H2b: The proportion of the variable incentive compensation of the CEO is positively related to risk taking of their company.
13
3. Legal framework of Solvency I
This section first explains why there is regulation in the insurance industry on the basis of the public interest theory. Then, we focus on the Solvency I framework and describe this framework.
3.1 Regulation in the insurance industry
The public interest theory argues why the insurance industry is regulated. The theory states that regulation of insurance companies exists due to inefficiencies created by costly information and agency problems (Doff, 2008). Because especially the life insurance industry deals to a large extent with long-term obligations to their
policyholders, it is of vital importance for these policyholders, that these companies can survive volatility in the markets and do not become insolvent. As a policyholder, it is difficult to assess how safe every insurer is when becoming a policyholder, and how safe the insurer will be in the future. The insurer can also change its risk appetite after the choice of the consumer. In addition, the owners of insurance companies might have an opposite incentive to the solvency/safety of these companies as only their investment is at risk (Klein, 1995). The regulators therefore play an important role in protecting the interests of the policyholders.
One of many regulations the regulators introduced in the insurance industry is
regulation on the solvency of the insurance companies. The EU implemented the Life and General Insurance directives in 2002. The Solvency I ratio calculates whether the required capital of an insurance company is covered by its assets and future profits (Schwarz et al., 2011). It gives an insight whether the company can meet its future obligations. The next paragraph will explain the Solvency I framework further. In January 2016, the Solvency II directive came into force EU-wide and replaced the earlier Solvency I framework. In this framework, insurance companies have to
reserve capital for several risks (such as market risk, credit risk, liquidity risk and operational risk) and not just insurance risks as was required during Solvency I. Dependencies between risks are also integrated in the framework. The Solvency II framework is implemented and executed since 2016 by the EU insurance companies. The Solvency II framework is not part of this thesis due to lack of data on the
Solvency II ratio which is required for insurance companies to report since only this year. The Solvency II framework is not part of this thesis due to lack of data on the
14
Solvency II ratio which is required for insurance companies to report since only this year.
3.2 The Solvency I framework
The basis of the solvency I framework originates from the 1970s when the European Union (EU) initiated several directives to centralise and unify the regulation on
solvency across the EU member states (life and non-life insurance EU directives). This solvency regulation was focused on the size of the insurance liabilities and its simplified insurance risk which does not give an insurer the right incentives because the insurance risk is not the only risk it faces. The Solvency I framework is backward-looking and deals with a fixed discount rate. The required solvency ratio for life insurance companies was calculated by the available shareholder’s funds on the balance sheet divided by the highest figure of the insurance liabilities x 4% or 0.3% x the capital sum at risk. The higher the insurance liabilities an insurer had, the higher solvency was required. For general insurance companies, the highest figure of 18% x gross premium or 26% x claims expense was applicable (Wiener, 2007). Next to these requirements, there is regulation on the investments of the insurance company. In the beginning of the 21st century, several measures were introduced to increase
the capital requirements on categories of insurance business with a higher volatility (Linder and Ronkainen, 2004). In this Solvency I framework, there are clearly no qualitative requirements on the subject of internal governance mechanism.
15
4. Data and Methodology
In the next paragraphs, we give a description of the sample and justify our
methodology and the two types of regression we use in this study. In addition we describe all the dependent variables, explanatory variables and control variables which are part of the regression.
4.1 Sample description
The sample consists of 31 listed EU+2 insurance companies between 2010 and 2014. EU+2 is defined as all the countries in the EU with Norway and Switzerland included. The geographical condition of EU+2 has been chosen because of extensive reporting requirements in these countries and/or the obligation of complying with the Solvency I framework which was in effect during the time of the sample period. It also broadens the potential size of the final sample.
Most of the dataset is collected manually from annual reports, while the stock returns are extracted from Datastream. Only companies which report under IFRS are
included in the sample (for reason of comparison) with a minimum annual operating turnover of $10 million. The data in the sample covers a period of 2010-2014. The decision for this time period has two reasons. First, in these most recent years, more and more information is included in the annual reports concerning corporate
governance which contributes to the data availability. Secondly, a range of four years is used because of the time consuming activity of collecting the data. We had no access to samples in databases where the combination of corporate governance variables for European companies in the insurance sector is available.
The final sample consists of 139 observations. A significant amount of companies is eliminated from the sample due to data availability. The insurance companies in the final sample are from thirteen countries in Europe with the most observations from the U.K. and Germany. All the foreign currency figures are converted into Euro’s with the corresponding exchange rate at the end of the observation year. The sample includes insurance companies with only, or a combination of life insurance, general insurance and reinsurance activities. Because of the selection of only publicly traded insurance companies and the limited availability of data on corporate governance variables in the annual reports for smaller companies, only 31 companies are left in
16
the sample. This is not unusual for research of financial reporting of insurance companies (see Chen et al, 2001; Borde et al., 2004; Staking and Babbel, 1995). The sample list of companies is derived from Orbis:
Table 1: Number of companies in the sample size
Company selection # of companies
EU+21 publicly listed insurance companies 333
And comply with accounting practice IFRS 266
With a minimum $10m operating turnover2 72
Gross number of companies in sample 72
Incomplete data availability -41
Net number of companies in sample 31 1. Countries in the European Union with Norway and Switzerland included.
2. Last year available.
4.5 Research design
An OLS regression is employed to test the several corporate governance variables and dependent risk measures.
The following regression equation is formulated:
, = + (),+ (),+ (),+ ( ),+ , (1)
Where is the error term, denotes the several insurance companies in the sample and the years within the chosen time period. till are the parameters to be
estimated; board size, CEO compensation and the existence of blockholdership (shareholders with at least a 5% stake). The control variables are size, the natural logarithm of the yearly average daily trading volume divided by the number of shares outstanding, year dummy variables, one-tier versus two-tier dummy and a dummy variable for line of business (industry).
One of the assumptions of using an OLS regression is that the explanatory variables should be exogenous. Concerning the field of corporate governance in relation to corporate risk taking, there is evidence that the corporate governance variables are actually endogenous (Cheng, 2008; Cheng et al., 2011; Eling and Marek, 2013). Because of these endogeneity problems of the corporate governance variables we will perform an additional regression using the two-stage least square method to
17
control for the results of the OLS regression. The following regression equations are therefore formulated: , = + (),+ ( ),+ , (2) , = + (),+ ( ),+ , (2) , = + (),+ ( ),+ , (3) , = + (),+ ( ),+ , (3) , = + (),+ ( ),+ , (4) , = + (),+ ( ),+ , (4)
Where is the error term, denotes the insurance companies in the sample and the
years within the chosen time period. till are the parameters to be estimated;
board size, CEO compensation and the existence of blockholdership (shareholders with at least a 5% stake). The control variables are size, the natural logarithm of the yearly average daily trading volume divided by the number of shares outstanding, year dummy variables, one-tier versus two-tier dummy and a dummy variable for line of business (industry).
4.2 Dependent variables: risk measures
In this study, total risk will be measured by the total stock return volatility and the insolvency risk measured by the natural logarithm of the Solvency I ratio.
The first risk measure is stock return volatility. This is calculated as the standard deviation of the daily stock returns. This proxy for risk is the most used in studies which involve risk taking (Cheng et al., 2011; Lai and Lin, 2008; Pathan, 2009; Nakano and Nguyen, 2012; Podder and Skully, 2013). The higher the volatility of the stock the higher risk the company represents. The stock returns have been adjusted for dividends and stock splits.
The Solvency I ratio measures to what extent the minimum requirement of the insurance companies capital set by the regulator is covered by the financial position
18
of the company. The lower the Solvency I ratio for the company, the higher the insolvency risk. Before the financial crisis, the Solvency I ratio was just a measure to report to the regulator. This changed significantly after the financial crisis. Investors and analysts started to follow the Solvency I ratio and this ratio became a priority for the management of insurance companies. The Solvency I ratio has to my knowledge not been used in any previous research on the topic of corporate governance in relation to risk.
4.3 Corporate governance variables
The three corporate governance variables we use in our research are board size, CEO compensation and blockholdership. These variables are chosen mainly because of data availability. It becomes more and more difficult to gather data on corporate governance subjects when the size of the insurance company becomes smaller. The available data in the annual reports becomes also less detailed in that case.
4.3.1 Board size
As we are facing different board structures in our sample, so we have to make a decision on how to define the board size. There is only one study which sample consists of one-tier and two-tier board structures (Eling and Marek, 2013). They investigate the role of the number of board meetings but fail to specify how they cope with counting the board meetings in their sample of companies with a two-tier
structure. As stated in our theoretical framework, the larger the board, the more difficult it is to reach consensus and therefore the less likely risky proposals will be supported due to a compromise of individual opinions. As most of the decision-making process takes places in the executive board of companies with a two-tier structure, we will therefore only count the number of board members in the executive board. These are all the directors in the board for companies with a two-tier board structure and all the board directors for companies with an one-tier board structure. A retirement of one of the board directors followed by a succession by a new board member during the year is counted as 1 board member. A control variable will be added in the regression which separates the two board structures with a dummy (0 for a two-tier board structure and 1 for an one-tier board structure).
19 4.3.2 CEO Compensation
CEO compensation is the logarithm of the total compensation of the CEO during the year. This includes a fixed component (base salary, pension costs and other benefits) and a variable component. Only direct compensation is included in this variable, compensation dependent of the future is not included. When a CEO steps down during the year (mostly because of retirement), the compensation of the stepped down CEO and the compensation of the new CEO will count as the total
compensation of the CEO during the year. As we also want to check the relation between the proportion of the variable incentive compensation of the CEO and risk taking of their company, we also run a regression with the variable part of the CEO compensation package (variable compensation / total compensation) in line with Miller et al. (2002). The variable part of the compensation can consist of cash and options or stock granted. The performance targets are mostly is based on key performance indicators, both financially and non-financially and reflect the strategy and long-term objective of the insurance companies.
4.3.3 Blockholdership
The third variable is blockholdership. Kole and Lehn (1999) investigate the number of outside blockholdings of the U.S. airline industry and find that these outside
blockholdings increased after deregulation of the industry. This implies that from the perspective of agency problems, there is less need for large shareholders who need to monitor when the industry becomes more regulated. The size of the insurance companies are on average higher than in other industries. Investors who have a 5% stake in an insurance company, have invested therefore on average in absolute term more than the investor with a 5% stake in a different industry. This could have an impact on the incentive of the investor to monitor. We use in this study a minimum stake of 5% as the threshold for a blockholder. This percentage is in line with other studies such as Eling and Marek (2013), Core et al. (1999) and Andersen and Fraser (2000). The ownership stakes were measured at the end of the year. Changes during the year are not taken into account.
4.4 Control variables
We add the following control variables in the regression: size, trading volume divided by the number of outstanding shares, net income to assets, year dummy variables,
20
board structure dummy variable and a dummy for line of business. Earlier research shows that size is negatively correlated with company risk (John et al., 2008; Cheng, 2008). The larger the company, the larger its ability of diversification which can have an effect on the total risk taking of an insurance company. Colquitt and Hoyt (1997) find that larger life insurance companies are more inclined to make use of hedging activities because they have economies of scale and the resources and knowledge required for it. In addition, other control variables are the natural logarithm of yearly average daily trading volume divided by the number of shares outstanding (Pathan, 2009; Cheng et al., 2011). This ratio is significant as a control variable in the study of Pathan (2009). Furthermore, net income to assets, year dummy variables, one-tier versus two-tier dummy and dummy variables for line of business are used as control variables. The companies are divided into life business, non-life business (general insurance or property-liability insurance) and multi-line insurance. In addition, a reinsurance dummy is added for insurance companies which have reinsurance activities. The threshold for a company to be classified as life business or non-life business is 70% of the total last reported revenue attributable to the life or non-life segment. Multi-line insurance is defined as when a company has both life and
general insurance activities but does not reaches the threshold of 70% of the revenue for one of these activities. This is in line with the definition in the study of Eling and Marek (2013).
21
5. Results
This section starts with an overview of the sample and the descriptive statistics. Then, we present the results of the 2SLS regressions. We compare the results with the literature reviewed in our theoretical framework and show the implication these results. Lastly, the results of the robustness test are discussed.
5.1 Descriptive statistics
Table 2
Descriptive statistics of the collected data
N Minimum Maximum Mean Std. Deviation
Dependent: Solvency I ratio 139 110% 582% 233% 75.90 Volatility stock 139 0.13 0.58 0.26 0.09 CG variables Boardsize 139 2 12 6 3.13 CEO pay in € th. 139 176 9,503 2,985 2,257 Control variables Total assets in € th. 139 1,767 840,069 184,040 199,453
net income to assets 139 -0.02 0.11 0.01 0.02
trading volume / shares
outstanding 139 0.004 2.033 0.620 0.505
Note: In addition to above variables, the following dummy variables are included in the regression. 1) Corporate governance variable blockholdership is a dummy. The sample
consists of 0 (no blockholdership) and 1 (blockholdership). 2) Five year dummies from 2010 till 2014 are included 3) Three control dummies for line of business (life insurance, general insurance, multi-line insurance
and reinsurance) and a tier control dummy is included for separating companies with a one-tier and one-tier board
structure (1 for one-tier board). N = company years.
The natural logarithm of some variables are used for a normal distribution of the data. The table in appendix II gives an overview. Table 2 describes the statistics of the sample before adding the natural logarithm to certain variables. The Solvency I for the insurance companies is on average 233%. This is overall a comfortable level. The board size of the companies varies between two and twelve persons with an average of six executive board members in the board. The average size of the board is lower than in the research of Cheng et al. (2008) and Pathan (2009). They have an average of respectively nine and twelve board members in their sample. The
22
the compensation of the CEOs in the sample. The minimum is €176 thousand versus the maximum of €9.5 million.
The companies in the sample size are on average large. The average size of the balance sheet is EUR 184.0 billion and varies between EUR 1.8 billion and EUR 840.1 billion. The large size of the balance sheet for insurance companies can be explained by the long-term obligations towards their policyholders on the liability side of the balance sheet, and the investments to cover these obligations on the asset side. Concerning the dummy variables, 46 of the 139 insurance companies years in the sample have a one-tier board structure. The companies with an one-tier board structure are, as expected, mostly from the U.K. A little more than a half of the insurance companies years (71) have a blockholder with at least 5% or more of shares as an investor. From the total sample, 81 insurance companies years have multi-line business, 31 only life business and 24 insurance companies years are included in the dataset as only general insurance activities.
Table 3 gives an overview of the correlation between the variables. No correlation between variables is higher than 0.75. The correlation matrix show that volatility is negatively related to the Solvency I ratio. This is plausible given that in the recent years more and more focus lies on the Solvency ratios as the insurance companies shift to the Solvency II framework1. Furthermore, as expected we see a significant
positive correlation between the size of the company (represented by total assets) and the salary of the CEO. What is more surprising is the negative significant correlation (at the 1% significant level) between the existence of a blockholder and the total CEO compensation. These correlations are in line with findings of Core et al. (1999) and Eling and Marek (2013).
1Especially during the writing of this thesis, we see large volatility on the Q3 results of the insurance
companies Delta Lloyd and Aegon where the focus of these quarterly updates lies more on news around the Solvency ratios of these companies than any other results.
23 Table 3
Correlation matrix
Vol stock LN(Solvency) Boardsize
LN(CEO compensation) Blockholder dummy One tier dummy LN(total assets) LN(net income to assets) Life dummy General insurance dummy Reinsurance dummy 2014 dummy 2013 dummy 2012 dummy 2011 dummy 2010 dummy trading volume to shares outstanding Vol stock 1 -.277** -.244** -.183* .029 .142 .130 -.324** .320** -.213* -.153 -.388** -.218** .006 .544** .086 .369** LN(Solvency) -.277** 1 .126 .064 -.018 -.019 -.205* .424** -.152 .301** -.052 .151 .039 .084 -.197* -.089 -.133 Boardsize -.244** .126 1 .490** .013 -.677** .194* .252** -.263** -.025 .377** -.065 -.015 .018 .018 .048 -.195* LN(CEO compensation) -.183* .064 .490** 1 -.229** -.097 .708** -.143 .080 -.058 .133 .036 .058 .016 -.065 -.050 .217* Blockholder dummy .029 -.018 .013 -.229** 1 -.342** -.174* .205* -.217* .129 -.001 -.026 -.008 .011 .013 .013 -.258**
One tier dummy .142 -.019 -.677** -.097 -.342** 1 -.033 -.077 .174* .245** -.254** .040 .015 -.010 -.024 -.024 .170*
LN(total assets) .130 -.205* .194* .708** -.174* -.033 1 -.569** .355** -.502** -.008 .001 -.005 -.005 .008 .001 .317**
LN(net income to assets) -.324** .424** .252** -.143 .205* -.077 -.569** 1 -.546** .604** .183* .040 .087 .043 -.131 -.046 -.343**
Life dummy .320** -.152 -.263** .080 -.217* .174* .355** -.546** 1 -.245** -.226** .013 -.020 -.011 .009 .009 .449**
General insurance dummy -.213* .301** -.025 -.058 .129 .245** -.502** .604** -.245** 1 -.140 .038 .000 .008 -.024 -.024 -.033
Reinsurance dummy -.153 -.052 .377** .133 -.001 -.254** -.008 .183* -.226** -.140 1 .023 -.019 -.012 .004 .004 -.231** 2014 dummy -.388** .151 -.065 .036 -.026 .040 .001 .040 .013 .038 .023 1 -.269** -.263** -.252** -.252** -.116 2013 dummy -.218** .039 -.015 .058 -.008 .015 -.005 .087 -.020 .000 -.019 -.269** 1 -.258** -.246** -.246** -.074 2012 dummy .006 .084 .018 .016 .011 -.010 -.005 .043 -.011 .008 -.012 -.263** -.258** 1 -.241** -.241** .057 2011 dummy .544** -.197* .018 -.065 .013 -.024 .008 -.131 .009 -.024 .004 -.252** -.246** -.241** 1 -.230** .021 2010 dummy .086 -.089 .048 -.050 .013 -.024 .001 -.046 .009 -.024 .004 -.252** -.246** -.241** -.230** 1 .119
trading volume to shares
outstanding .369
** -.133 -.195* .217* -.258** .170* .317** -.343** .449** -.033 -.231** -.116 -.074 .057 .021 .119 1
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
24
The non-life dummy refers to the insurance with at least 70% general insurance (property-liability insurance). This dummy control variable is positive significantly correlated to the Solvency I ratio. The sample period is a time in which interest rates are declining. Low interest rates for a longer period are negative for insurance
companies because they need to discount their insurance liabilities on their balance sheet with a much lower discount rate. Companies which are defined as non-life offer more short-term insurance products and are therefore less prone to declining interest rates. This has an effect on the Solvency I ratio of these companies. It therefore makes sense that we see a significant positive correlation between non-life insurance companies and the Solvency I ratio and a significant negative correlation between insurance companies with mostly ‘life’ activities and the Solvency I ratio.
5.2 Results 2SLS regressions
First, the regression model is performed with the first dependent variable, the
volatility of the stock. A higher volatility of an insurer’s stock represents a higher risk the company is taking. Then, the second regression model is performed with the Solvency I ratio as dependent variable. A higher Solvency I ratio represents a lower risk (insolvency risk) the company is taking. A regression is performed using the two-stage least square method to account for any possible endogeneity issues. To
perform the 2SLS regression we need to use instrumental variables. In line with previous literature (Nakano and Nguyen, 2013) we use company size (in terms of total assets) and ROE as instrument variable for the CEO compensation variable, total assets and companies age for board size and market capitalization (Chen et al., 2011) and number of analyst coverage for blockholdership (Cornett et al, 2007). The results of the 2sls regression equations are presented in table 4 and the results of both OLS regression equations are presented in appendix III.
25 Table 4
Result of estimating the 2sls regression equations
Sample: 139 observations (period of 2010 - 2014). Note: *** indicates a significant level of 1%, ** indicates a significant level of 5%
Volatility stock SI (LN) Volatility stock SI (LN) Volatility stock SI (LN)
CG variables CEO pay (LN) -0.01 0.02 0.01 0.03 Boardsize -0.02 0.14 0.04 0.26 Blockholdership 0.11** 0.24 0.05 0.23 Control variables Total assets (LN) 0.02 -0.26 -0.01 0.01 0.06 0.44 0.01 0.02
Net income to assets 0.01 0.12*** 0.02 -0.62 0.01 0.13***
0.01 0.03 0.15 1.04 0.01 0.04
Life insurance dummy 0.03** 0.06 0.01 -0.23 0,04** 0.07
0.01 0.07 0.05 0.39 0.02 0.08
General insurance dummy -0.06*** 0.03 -0.03 0.67 -0.09*** -0.01
0.02 0.08 0.11 0.82 0.03 0.11
Reinsurance dummy -0.01 -0.11 0.01 -0.01 0.00 -0.06
0.02 0.07 0.03 0.21 0.02 0.09
One tier board dummy 0.02** -0.02 0.16 -1.00 0.07*** 0.07
0.01 0.05 0.23 1.68 0.02 0.11 2011 dummy 0.09*** -0.04 0.09*** -0.18 0.09*** -0.04 0.02 0.07 0.03 0.24 0.02 0.08 2012 dummy -0.01 0.07 -0.01 0.18 -0.01 0.08 0.02 0.07 0.03 0.22 0.02 0.08 2013 dummy -0.04*** 0.03 -0.05 0.21 -0.04** 0.05 0.02 0.07 0.04 0.30 0.02 0.08 2014 dummy -0.07*** 0.10 -0.08** 0.25 -0.06*** 0.12 0.02 0.07 0.04 0.27 0.02 0.08
trading volume / shares 0.04*** -0.02 0.03 -0.03 0.06*** 0.03
26 5.2.1 Board size
Board size is not statistically significant in relation to corporate risk taking from the 2SLS regressions. In our hypothesis, we exptected that board size is negatively related to risk taking of insurance companies. Our stated hypothesis is rejected, there is no statistically significant relationship between board size and risk taking within a company. We can therefore not support the theory of Lorsch (1992) and Jensen (1993) that the larger the board is, the more difficult it is to reach consensus and therefore the less likely risky proposals will be supported due to a compromise of individual opinions (as a proxy for risk taking).
Our result is opposite to the findings of Nakano and Nguyen (2012), Pathan (2009), Cheng (2008), Huang and Wang (2015) and Wang (2012) who find a negative
statistically significant relationship between board size and the variability in corporate performance. The scholars mentioned in our literature review all adress the
endogeinity issue in a different way. Nakano and Nguyen (2012) investigate the relationship between board size and the variability of corporate performance using the 2sls regression method. Pathan (2009) performs next to GLS (Generalised Least Squares) also a three-stage least square regression and finds the same significant negative relation for both regressions. Huang and Wang (2015), Cheng (2008) and Wang (2012) adress endogeinity issues by using lag values for the possible
endogneous variables but this results in the same conlusion as from their OLS
regressions. The results of these additional regressions methods do not change their earlier findings. In all cases, it confirms their earlier findings that board size is
negatively related to corporate risk taking. Concerning our outcome, we see the same consistent finding as board size from both our regression methods shows the same result (not statistically significant).
The explanation of why our results vary compared to other scholars might be found in our sample. If we look at differences between our sample and the samples of
previouse literature, we see that Nakano and Nguyen (2012), Wang (2012) and Cheng (2008) excluded financial companies from the sample before running their regressions. The difference between results might therefore be explained by industry differences. The effect of the strict regulation in the insurance sector could
27
(2009), which investigates the period 1997-2004, included only large banks from the US. One explanation between different results with Pathan could be the different time period. The period 1997 till 2004 is characterised as a period with less regulation than the period after the financial crisis which started in 2007 (Pathan, 2009). Pathan (2009) also used more detailled variables such as the independency of the directors and the power of the CEO. In companies where the companies CEO was powerfull (defined as when the CEO is also the board chairman and if the CEO was hired internally), there was a significant relationship between CEO power and risk taking. Due to the data availability and capacity of this research, such variables could not be incorporated in this study. It could be interesting to find out in further research where the differences in these results come from.
5.2.2 CEO compensation
Looking at the results of the 2SLS regression, we do not find a statistically significant relationship between the CEO compensation and the insurance companies risk taking. As we stated in our literature review, the results so far on this topic are mixed. Gray and Canalla (1997) and Eling and Marek (2013) find a significant negative relationship. Eling and Marek (2013) state that the incentive part of the total
compensation contributes to the negative relationship. When a bigger part of the total compensation is contingent based, it can induce risk taking. We will test this with our next hypothesis. Podder and Skully (2013) find a positive relationship between total CEO compensation and risk taking. Existing literature on this subject also adresses the endogeneity issue which therefore can not be an explanation for the differences. For example, Eling and Marek (2013) account for endogeinity by using a structural equation model (SEM). We can reject our hypothesis as it stated that we expect a positive relationship between CEO compensation and risk taking which is not the case.
To find the explanation for our different result compared to previous literaturen, we also performed a regression with the ratio of variable to fixed compensation for the CEO. The results of this regression are presented in appendix IV en V. For the regression with the volatlity of the stock as dependent variable, we see no significant result. However, for the regression with the Solvency I ratio as a variable we see a significant negative relationship (on a 5% significance level) between the ratio
28
a higher percentage of variable pay for the CEO is related to higher risk taking of the company (a lower Solvency I ratio increases insolvency risk).This is in line with
existing literature on this subject (Podder and Skully, 2013; Miller et al., 2002). On the other hand, the result of the 2SLS regression (see Appendix V) show no significant result for this variable. The results are therefore not consistent across the different regression methods.
5.2.3 Blockholdership
The result from the 2SLS regression shows that the blockholder dummy has a positive significant relationship with the volatility of the stock. This indicates that the existence of a shareholder with a least 5% of the shares induces risk taking within a company. This surprisingly contradicts our hypothesis. This also contradicts the theory that investors with a blockholding have such a concentrated stake that they are therefore more risk averse which will result in inducing lower taking by the management (Cheng et al., 2011). One explanation can be that the definition of a blockholder (a 5% stake) is too small for this theory to have an affect on the volatility of the stock. Furthermore, the size of the investor also plays a role in this theory. If a large investors has a blockholder stake (> 5%) in a small company, which can be a relatively insignificant a amount of money for the investor, the theory does not hold as well. Allthough, this seems unlikely since the insurance companies in the sample are relatively large and have a substantial market capitalization.
There might be another explanation for these different results in this insurance sample compared to the existing literature which mostly covers non-financial companies. The insurance industry is heavily regulated which might reverse the incentive of blockholders to reduce the risk of the insurance companies. Possibly, the regulators already reduce risk within the insurance company and even to a lower level than the risk appetite of the large shareholder. The blockholder has therefore an incentive to induce higher risk taking within insurance companies than companies without blockholder. This is in line with research from Saunder et al. (1990) who state that in a time of deregulation of banks, shareholders have a greater incentive to increase risk taking within banks. On the other side, Eling and Marek (2013) do find a positive significant relationship between this corporate governance variable and
29
different proxies of risk but they use the number of blockholders instead of only the existence of blockholders in their research.
The results of the second 2SLS regression shows that there is no significant relationship between the existence of investors with at least a 5% stake in an
insurance company and the Solvency I ratio. We can therefore state that our overall result is not consistent on this corporate governance variable and we therefore reject our precious stated hypothesis that there is a significant positive relationship between the existence of a blockholder in relation to risk taking within an insurance company. However, the volatility of the stock of the insurer is a proxy for the overall risk of the company while the Solvency I variable only reflects the insolvency risk of the
company from a regulatory perspective.
5.2.4 Control variables
The results from the 2SLS regression show that most control variables are significant related to the volatility of stock as dependent variable. The year 2011 and 2014 are respectively positively and negatively significant related for all these regressions. In the regressions of the corporate governance variables CEO compensation and blockholdership, we see a positive significant relationship between the life dummy and volatility (on a significance level of 5%) and a negative significant relationship between the general insurer’s dummy and volatility (on a significance level of 1%). As stated before, this could be explained by that the insurance companies with general insurance activities are less influenced by the changing market interest rates which results in less volatile stocks. General insurance products don’t result in long-term insurance liabilities as the case is with life insurance companies. In addition for the before mentioned corporate governance variables (CEO compensation and
blockholdership), we see a positive significant relationship with the trading volume control variable on a significance level of 1%. This is also in line with our
expectations.
We see some other results for the regression with the SI ratio as dependent variable. For these regressions, we only see a strong significant relationship between the control variable net income to assets and the SI ratio with a significance level of 1% which is not a surprising result. This is logical since the profit and losses are directly affecting the shareholders’ funds on the balance sheet of the insurer.
30 5.3 Implications of the results
We can conclude that the CEO compensation and board size does not play a significant role in the risk the insurance companies in taking (proxy by volatility and insolvency risk) and therefore should not be incorporated in the regulation
requirements of the supervisor. However, blockholdership is positive related to the insurance companies risk taking. This finding could be interesting for the regulators which they can take in account for future adjustments in regulation of the insurance industry.
Our research also has its limitations. As we collected our data from annual reports manually, the total size of the sample is small and sensitive to errors. Especially concerning the control variables, a consideration had to be made on the amount of time to be invested in collecting the data versus the completeness of the data and control variables. In addition, there is a scarce amount of research on corporate governance in the insurance industry which makes it difficult to compare with the existing literature on this subject. Results in the insurance sector can differ greatly from research performed with a sample that excludes financial companies.
Furthermore, corporate governance is a complex subject because of the intertwinings of the corporate governance variables with each other. External regulation from a supervisor (which is strongly present in the insurance sector) can make other corporate governance variables less necessary.
We used the Solvency I ratio in our research due the lack of sufficient data on Solvency II ratios as the regulation came into force on January 2016. In future research, when more years of data is available on Solvency II figures, it would be interesting to find out what kind of results the Solvency II ratios will bring in respect to their relation with corporate governance. The Solvency II is a far more
comprehensive ratio on the actual risks the company faces which can lead to other results and new insights.
5.4 Robustness test
We also performed a robustness test to examine the quality and validity of the results. The resinsurance companies were excluded for this sample and the
31
are located in Germany, Switzerland, France and the U.K. with the almost the half of these observations from Germany. The results from the 2SLS regression are
presented in table 5, the results of the OLS regression in appendix VI. The results from the 2SLS regressions give the same outcome with respect to the corporate governance variables. There still is a positive relationship between the existence of blockholders and the volatility of stock. The other coporate governance variables do not result in a significant relation as earlier is concluded. The results are therefore not influenced by the sample of the reinsurance companies.
