The Relation of CEO Compensation on the Past,
Contemporaneous and Future Performance of Dutch Hospitals
Master Thesis
Name: M.M.E. Coopmans Student number: 6032710 Date: July 2014
MSc Economics and Business Specialization: Finance Supervisor: Dr. J.E. Ligterink
Abstract
This study focuses on the relation between CEO compensation and firm performance in a semi-public sector. The study continues on the model of Banker, Darrough, Huang and Plehn (2012) about the construction of compensation contracts using past and future performance and fixed and variable components. First, compensation is explained through past performance measures, which reflect the CEO’s ability and can give indications on what the level of compensation should be for contemporaneous performance. Second, future performance is explained through fixed and variable components of compensation. The results show a positive association between fixed pay and contemporaneous performance, but also between variable pay and contemporaneous performance, indicating that there is a positive pay for performance sensitivity regarding contemporaneous performance. Fixed pay, in addition, has a positive association with future performance. The research tests theoretical predictions using CEO compensation data from 2007-2012 for a sample of 82 Dutch hospitals.
Table of contents Introduction 4 I. Literature review 6 A. Corporate Governance 6 B. Executive Compensation 7 B.1. Contracting Views 7 B.2. Model 8 C. Previous Research 9 C.1. General 9
C.2. The Nonprofit Sector 10
C.3. Hospitals in the Netherlands 11
C.4. CEO compensation and performance in time 12
II. Hypotheses and Methodology 14
A. Hypotheses 14
B. Methodology 15
B.1 Variables for Hospital Performance 15
B.2 Variables for CEO compensation 17
III. Data and Descriptive Statistics 19
A. Data Collection 19 B. Descriptive Statistics 19 IV. Results 22 A. Model 1 22 B. Model 2 26 V. Robustness Checks 31 VI. Conclusions 34 IV. References 36 Appendix 39
Introduction
Berle and Means (1932) were the first to discuss the separation between ownership and control. They described the agency problem, a problem about information asymmetry between principal and agent. The conflict arises when an agent needs funds from a principal to put to productive use and the principal needs to incentivize the agent to create an optimal return for the principal (Shleifer and Vishny, 1997). One way to deal with this problem is to set up a complete contract between both parties. In a perfect world, a contract would be set up in a way that the contract specifies in every situation how the manager should handle that particular situation (Shleifer and Vishny, 1997). Since this is not possible because future contingencies are hard to foreseee, the principal can try to set up a contract that incentivizes the manager through compensation. The difficult part is compensation must be set above the level of personal benefits to the manager, so that the manager will not use the funds to his own discretion but rather make a profit for the financiers. However, managers might have power over their own payment packages, due to poor corporate governance, which further complicates the relationship between principal and agent (Bertrand and Mullainathan, 2003).
In 2013, the law in the Netherlands has changed towards CEO compensation in public and semi-public companies. First, there was the law WOPT, which stated that the companies in the public and semi-public sector had to comply with the guideline or explain the difference. Since 2013, the law limited compensation in the public and semi-public sector. There had been a lot of media attention on the excessive salaries of these CEOs, which they received even though the country was in a financial crisis. The national newspaper Trouw (2010) reported that CEOs of hospitals had been focused on status and were earning above the limit that they agreed upon and even set themselves. Posing limits on CEO compensation in the public and semi-public sector can decrease competitiveness in comparison to the private sector. When bonuses are prohibited, it might create difficulties for principals to incentivize agents. Moreover, when these salaries are significantly lower than salaries in the private sector, the agent (best fitted for the job) might choose to switch to the private sector.
In order to investigate this, this research quantifies the relation between performance and managerial compensation in hospitals. This research is therefore concentrated on whether the variable and fixed components of CEO compensation are related to hospital performance. Past performance is studied since past performance measures can be used by principals to set a contract according to the agent’s ability (Banker, Darrough, Huang and Plehn-Dujowich (2012)). Additionally, current contracts should contain information about the agent’s future ability and thus be predictive of future performance (Banker et al., 2012). The aim of this research is to comprehend the relation between CEO remuneration and the performance of a hospital. This leads to the following research question:
What is the relation between CEO compensation and the past, contemporaneous and future performance of Dutch hospitals?
The scope of this paper is hospitals in the Netherlands. Most literature is based on CEO compensation in the US and the results of these researches are being used with a widespread range.
Research to more specific regions with different institutionalized settings can give new insights. The Netherlands can represent continental Europe since they have similar legal settings (Van Ees, Postma and Sterken, 2003). Additionally, most of the research in this field has been studying how much of CEO compensation is explained by contemporaneous performance and has not investigated thoroughly the relation between current CEO compensation and past and future performance (Banker et al., 2012). Moreover, several researches have been performed on the relation between total CEO compensation and firm performance, while the fixed and variable components of a compensation contract have different intentions. Bonus is put into a contract to induce more effort from the CEO, while salary is put in for the reservation utility (Banker et al., 2012). Without a solid base salary no agent would sign the contract, because of risk aversion. Thus, when performance has different relations with bonus and salary, the result could be that these two move in opposite directions and total compensation comes out to be not significant at all.
To answer the research question a sample of 82 hospitals from the Netherlands is formed. Data is publicly available on a governmental website, where information with respect to the WOPT law is published. Additionally, the annual reports of the hospitals are published on this website. Another government website, the BIG-register, displays information about doctors, thus making it possible to check if the CEO has a medical background. The methodology is built on a two-period agent-principal model by Banker et al. (2012) with a clear distinction between bonus and salary. This distinction between bonus and salary is because of different underlying ideas. Bonus is focused on inducing more effort, therefore deals with a moral hazard problem of the unobservable effort the agent can put in. Salary deals with adverse selection, which starts in the succeeding period when the information asymmetry is reduced and the principal knows about the agent’s ability. Two models are formed, the first to analyze fixed and variable components of compensation on past and contemporaneous performance. The second model gives insight on fixed and variable components with respect to future performance. Data on the performance of the hospital and of the CEO has been gathered for the period 2007-2012. To gain insights in to the data, panel data methods are used. Specifically, the research uses the random effects model generalized least squares in order to estimate the parameters.
In the next section the related existing literature is summarized and explained, an overview of empirical evidence in the field is given and studies so far are examined for the creation of the hypotheses in the third section. In the third section the hypotheses are given, followed by the methodology. The fourth section elaborates on the data and shows some descriptive statistics. The fifth section shows the results of the research, whereas the sixth section presents some robustness checks. The last section contains the conclusion.
I. Literature review
This section discusses the agency problem, the optimal contracting and the managerial power view, the institutional background, nonprofit firms, Dutch hospitals and reviews previous research in the field.
A. Corporate Governance
Corporate governance is about how suppliers of finance to corporations can assure themselves of getting a return on their investment (Shleifer and Vishny, 1997). It is a system between control and supervision, concerning healthy relations between managers, supervisors and shareholders (Rijksoverheid, 2014).
The Dutch system of corporate governance represents a unique combination of characteristics of the market-based Anglo-Saxon regime and the bank-based Continental-European regime (Van Ees, Postma and Sterken, 2003). In the Netherlands the Corporate Governance Code, also known as Tabaksblat, is used for good corporate governance in listed firms. The code enforces to comply with the general rules or announce publicly the reasons of deviating. For the health care sector there is a more specific set of guidelines, the Zorgbrede Governance Code. The Zorgbrede Governance Code operates with the comply- or explain-principle as well. In 2009 close to 80% of the health care organizations were complying with the code (Adamini, Nelissen and Oortwijn, 2010). The government interferes occasionally in order to meet the policy goals of accessible and affordable health care of high quality (Kocsis, Koning, and Mot, 2011). Government officials reacted to the news about the excessive levels of compensation by enforcing a new law, WOPT. This law (Wet Openbaarmaking uit Publieke middelen gefinancierde Topinkomens), enforced all remunerations above 130% of the salary of the minister that come from public funds to be reported. Since January 1st of 2013, the government has enforced a newer law, namely WNT (Wet Normering bezoldiging Topfunctionarissen publieke en semipublieke sector). The WNT law states that all remunerations of members of the boards in public and semi-public corporations need to be limited to 130 per cent of the salary of a minister. Newly employed board members will directly follow this law, already employed directors or employees with a salary above this norm will go into a process of four years in order to decrease their salary until the norm.
The agency theory is part of the contractual view of the firm developed by Coase (1937), Jensen and Meckling (1976) and Fama and Jensen (1983). It explains a conflict in the separation between ownership and control (Shleifer and Vishny, 1997). Berle and Means (1932) were one of the first to distinguish that the difficulty lies in making management interested in profit maximization for financiers. The conflict arises between a principal and an agent, because of asymmetric information between the two parties. The financiers of a corporation, the principal in this case, need to incentivize the manager, the agent in this case, to maximize profits. In a perfect world the financiers would take
all the control and sign a contract on what the manager should exactly do in every state of the world (Shleifer and Vishny, 1997). Since future contingencies are hard to describe and foresee, making such a contract is technologically infeasible (Shleifer and Vishny, 1997). It is therefore necessary to optimally use the contract and allocate the residual control rights to either manager or financiers (Grossman and Hart, (1986), Hart and Moore, (1990), Shleifer and Vishny (1997)). The residual control rights allocated to the manager can lead to moral hazard, where the manager uses his discretion to allocate the funds in a way that is more optimal for himself than for the financiers. Managerial expropriation of funds can take elaborate forms, such as just taking cash out through transfer pricing, consumption of prerequisites, expanding the firm greater than is rational or entrenching themselves (Shleifer and Vishny, 1997). Additionally asymmetric information can lead to adverse selection, meaning it is impossible for the principal to confirm on the agent’s ability (Canton, Adamini, Canoy, Lubberman, Molenaar, and Volkerink, (2010), Banker et al., (2012)). The financiers can use incentive compensation packages to try to ascertain behavior that is beneficial to the principal. In some cases, the financiers can give an implicit threat or promise to impose that they will take action when the manager is not acting on their behalf (Shleifer and Vishny, 1997). Incentive contracts can consist of many things, such as share ownership, stock options, or they can be a threat of dismissal when income is low (Jensen and Meckling (1976), Fama (1980), Shleifer and Vishny (1998)).
B. Executive Compensation
Designing an optimal contract that would work anywhere at any time would not be difficult if shareholders had complete information regarding the activities of the CEO (Jensen and Murphy, 1990, Shleifer and Vishny, 1997). Unfortunately managerial actions and investment opportunities are not perfectly observable by shareholders (Jensen and Murphy, 1990). Still various different compensation policies are designed trying to ensure the managers are taking actions and choosing investments that are in line with the shareholders best interests. Regarding compensation packages there are two views, the optimal contracting and the managerial power view.
B.1. Contracting Views
The optimal contracting view is the more dominant view to studying executive compensation, according to Bebchuk and Fried (2003). The optimal contracting view states that executive compensation package should align the wealth of a manager with the wealth of the shareholders, therefore minimizing the agency problem. A contract that ties the welfare of the CEO to the welfare of the shareholder would ensure the CEO to take appropriate actions (Jensen and Murphy, 1990). In this case there is a belief that contingent, performance-based compensation provides a solution to the agency problem by aligning both interests (Abowd, 1990). The principal determines the level of performance-sensitivity in the manager’s compensation contract for controlling the tradeoff between better management and increased compensation costs (Abowd, 1990).
The managerial power view, also called skimming view (Bertrand and Mullainathan, 2001), explains that executive compensation is not only a remedy to the agency problem, but at the same time is a part of the problem (Bebchuk and Fried, 2003). According to this view the manager has sufficient power to affect the design of the executive pay, through entrenchment or the complexity of the pay process (Bebchuk and Fried, 2003). The manager can deviate the contract from its optimal form towards his own wealth.
This research follows the optimal contracting view in the sense that compensation packages provide a solution to the agency problem and principals can set the level of performance-sensitivity in the contract.
In the last decades there has been much to say about excessive compensation. When the Dutch government decided to put a cap on compensation in the public and semi-public sector, this was a tool against excessive compensation. Excessive executive compensation could lead to lower performance, since the executives can no longer objectively manage the firm (Brick, Palmon and Wald, 2006, Wright, Ferris, Sarin, and Awasthi, 1996). Managers might focus solely on the objectives, which increase their compensation according to the package. This could leave behind factors that are not included in the contract, which might not necessarily be less important.
B.2. Model
Banker, Darrough, Huang and Plehn-Dujowich (2012) continue on the optimal contracting view on executive compensation. In their research they create a model, which helps to create an optimal contract. The two-period principal-agent model incorporates moral hazard and adverse selection. Moral hazard, in this case, defines the unobservable effort the agent can make and adverse selection, in this case, defines the unobservable ability the agent has. Since the model contains two periods, the principal is able to observe the agent’s ability in the first period and set the optimal compensation in regards to the agent’s ability, in the second period. The model separates bonus and salary, as opposed to prior research, which mostly focuses solely on total compensation. Because of this separation, bonus and salary can be determined to have different purposes. In case of the model of Banker, Darrough, Huang and Plehn-Dujowich (2012), the sensitivity of bonuses to performance is used to induce effort and rewarding better ability types. Additionally, that same higher ability type will receive less of a salary. Time is in this case introduced, since the agent’s ability is assumed to be positively correlated over time (Banker, et al., 2012). This means that for a given agent, salary increases over time.
In an optimal contract the information rent, rent an agent receives from having information not accessible to the principal, would be zero for the agent with the lowest ability. The information rent increases with the ability of the agent to induce truth-telling (Banker, et al., 2012). On the other hand, a higher ability agent has a lower salary and a larger bonus component in his compensation
contract. In this case the other terms in the compensation formula decrease with ability and more than offset the greater information rent (Banker et al., 2012).
An optimum compensation structure trades off productive efficiency with information rent extraction (Banker et al., 2012). The model predicts that salary increases with past performance, since past performance can give an indication of the ability level of the agent. This is confirmed by Murphy (1985), who states that an executive’s remuneration in a given year will depend part on past performance as well as current performance. On the same line of reasoning, higher current salary predicts higher future performance. Since bonus is designed to extract more effort, bonus is expected to be positively associated with contemporaneous performance and to be negatively associated with past performance and future performance. While the predictions on salary are fairly intuitive, the predictions on bonus with respect to past and future performance are predicted by the two-period model of Banker, Darrough, Huang and Plehn-Dujowich (2012). Because the predictions differ between salary and bonus it is important to separate the fixed and variable components of compensation.
C. Previous Research
Since the nineties a lot of research has been done on pay-for-performance sensitivity in the corporate sector, presumably because the SEC required enhanced disclosure on executive compensation (Perry and Zenner, 2001). In the last decade executive compensation has been researched even more thorough.
Dutch hospitals are not allowed to distribute a profit; therefore they fall in the category nonprofit hospitals. In general, there are also for-profits hospitals, mainly found in the US. Previous research has studied the difference between CEO compensation in for-profit and nonprofit hospitals. This section first looks at the empirical results of how performance is reflected in pay. Then it provides a comparison between for-profits and nonprofits. A third part looks into Dutch hospitals, while a fourth paragraph presents previous research with respect to past and future performance.
C.1. General
The empirical studies of Barro and Barro (1990), Hall and Liebman (1998), Houston and James (1995), Jensen and Murphy (1990) and Murphy (1985) all found a positive relationship between pay and contemporaneous performance, explaining that CEO compensation is highly responsive to firm performance. These results show that the level of compensation a CEO receives is dependent on how the firm performs, aligning his interests to the interests of the financiers. The empirical studies of Barro and Barro (1990), Hall and Liebman (1998), Houston and James (1995), Jensen and Murphy (1990) and Murphy (1985) were performed on publicly traded companies.
Barro and Barro (1990) researched the pay-performance sensitivity in commercial banks and discovered that lower levels of experience result into higher sensitivity in pay for performance, the
sensitivity declines with tenure. They interpret this effect of a declining sensitivity as a decline in information content of additional observations in performance (Barro and Barro, 1990). Hall and Liebman (1998) found that a pay-performance relationship is generated by stock holdings and options. They researched the period from 1980 until 1994 and state an increase in total pay of 136% at the median, which comes mostly from stock options, since in this period there was a stock option boom (Hall and Liebman, 1998). Additionally they confirm that aligning the incentives of executives with those of owners is the most direct way to mitigate the agency problem and state that if there is no relation between performance and pay it is doubtful that the assets are being managed efficiently (Hall and Liebman, 1998). Jensen and Murphy (1990) also found a positive relationship between pay and performance, although they found it to be much smaller than Hall and Liebman (1998). Jensen and Murphy (1990) had a much longer time period and found a decline in sensitivity over time, which they state is due to political forces. Murphy (1985) wrote that contracts depend on factors such as entrepreneurial ability, managerial responsibility, firm size and past and current performance. Besides, Murphy (1985) also separated fixed and variable compensation to some extent, but reports that at this time firms were not required to report bonus and salary separately. Houston and James (1995) also researched the pay-performance sensitivity in banks and did find a difference in the pay-performance sensitivity in the banking industry and other industries. On the contrary to Jensen and Murphy (1990), who found that in the banking industry less equity-based incentives are used in compensation contracts and cash compensation, on the other hand, is much more sensitive to firm performance than in other industries (Houston and James, 1995).
C.2. The Nonprofit Sector
A nonprofit organization can be described as a firm that is prohibited from distributing of its net earnings to individuals who exercise control over it (Hansmann, 1980). A governmental organization is also subject to additional constraints, such as political and legal (Ballou and Weisbrod, 2003). However, a nonprofit firm is allowed to make a profit, and in fact many firms do have a surplus on their accounting books, the criterion is that the firm is barred from distribution of the profits.
In comparing for-profit hospitals to nonprofit hospitals in the US, total CEO remuneration is about 25% higher in for-profit hospitals than in nonprofit hospitals and bonuses, usually based on different criteria, are much less likely to be rewarded in nonprofit hospitals (Roomkin and Weisbrod, 1999). This could potentially be so because for-profit and nonprofit hospitals differ in their goals, which eventually comes with different types of managers and reward structures (Roomkin and Weisbrod, 1999). Nonprofits may pursue objectives that reflect greater concern about collective goods or outputs that are more difficult to measure and reward (Roomkin and Weisbrod, 1999). Alternatively, due to the nondistribution constraint on the payout of profit, nonprofit hospitals may lack in incentives for efficiency (Roomkin and Weisbrod, 1999). Also government hospitals are much
less likely to have an incentive program in place (Ballou and Weisbrod, 2003); they too have the nondistribution constraint on the payout of profits. On the other hand, both types of hospitals focus on financial performance in designing the CEO remuneration, and CEO turnover and pay are strongly related to firm performance in hospitals according to Brickley and Van Horn, (2002). This is supported by the research of Preyra and Pink (2001), explaining that this is plausible since financial surpluses create new opportunities for growth and development.
Preyra and Pink (2001) made a comparison between CEO compensation in industrial firms and nonprofit hospitals in major cities in Canada. Nonprofit hospital CEOs earn about 50% less than those in the industrial sector, albeit those CEOs have eight times the income variance (Preyra and Pink, 2001). These results are contrasted by the more recent research of Joynt, Le, Orav and Jha (2013). They found that neither financial performance, the process of care, patient outcomes nor community benefit contribute to the remuneration of the CEO in nonprofit hospitals. The size, location, and state of technology and whether or not the hospital is an academic hospital does contribute to the compensation level of the CEO (Joynt et al., 2013).
Most of the research on executive compensation in hospitals has been done in the US. Nonprofit hospitals in the US and in the Netherlands may be in many ways similar to each other. Nevertheless there are some substantial differences. First, in the US not all citizens are obliged to have health insurance, while in the Netherlands they do (Research paper by the NZA, 2007). This makes health care more available in the Netherlands. Second, the larger part of the US hospital market has been without direct price regulation, while in the Dutch hospital market there is large regulation by the government (Research paper by the NZA, 2007).
C.3. Hospitals in the Netherlands
Most hospitals in the Netherlands are private nonprofit organizations providing in public services, only academic hospitals are public organizations (Kocsis, Koning, and Mot, 2011; Canton et al., 2010). Usually in the case of the agent-principal problem in corporate governance there is a conflict between the manager and the shareholders. In case of hospitals in the Netherlands the principal is not represented by a group of shareholders, creating a lack of ownership incentives (Brickley and Van Horn, 2002). On the other hand, there are the stakeholders, all who can affect of be affected by the actions of the business as a whole. In Dutch hospitals stakeholders are the executive and supervisory boards, the doctors, the nurses, other employees, insurance companies, ministries and the patients (Adamini, Nelissen and Oortwijn, 2010). Hospitals are not allowed to distribute their profits to any shareholders. Hospitals in the Netherlands have been in the spotlight quite some time, because of bankruptcies, mergers and medical failures. The level of remuneration of employees funded with public funds has also caused commotion.
Ikkersheim and Koolman (2012) studied the performance of hospitals in the Netherlands, their findings point out that according the consumer the performance of the hospital increased over the
past few years. When hospitals are forced to publicly announce information regarding their performance (WOPT), they will try to improve the experience of patients (Ikkersheim and Koolman, 2012).
Van Ees, Postma and Sterken (2003) performed a research in the Netherlands about the relation between CEO compensation and firm performance and did not find evidence for any relation. On the other hand, they found negative effects for various settings of supervisory boards, which are commonly used in the Netherlands.
The most recent report of the government about executive compensation in the semi-public sector shows that only 8% of the executives received a bonus for the year 2011, with an average of €15.200 on top of their average salary of €186.000 (WOPT report 2012). The average salary is below the new standard of 130% of a minister’s salary (WOPT report of 2011), however 55% of all the executives of the 110 hospitals at that time were compensated above the norm. The latest regulation, which started at January 2013, prohibits any executives, working in either the public or semi-public sector, to earn above the norm of 130% of the salary of a minister, which is €228,599 in 2013. The current cabinet, in 2013, is planning to decrease the norm even further to 100% of the salary or a minister.
C.4. CEO compensation and performance in time
Most research in the field of corporate governance, more specifically in incentive contracts, has been performed in trying to explain CEO compensation in terms of firm performance. This research additionally tries to explain CEO compensation in term of past firm performance, since the level of the agent’s ability can be explained by previous performance (Murphy, (1986), Banker et al., (2012)). Moreover, this research focuses on the effectiveness of CEO compensation regarding future firm performance.
The research continues on the model of Banker, Darrough, Huang and Plehn-Dujowich (2012). Their findings indicate that current CEO salary is positively associated to future performance, while bonuses are not. They specify that salary is adjusted to meet the reservation utility and information rent, and is positively correlated over time to reflect ability. Bonuses, on the other hand, serve to address moral hazard and adverse selection by separating high ability agents in to riskier contracts (Banker et al. (2012)). Additionally, they find that salary is positively, while bonus is negatively, associated with past performance. Banker et al. (2012) also stress the importance of separating the fixed and variable components of compensation. This research continues to build on their model, as explained before.
Murphy (1986) studied performance sensitivity contracts with respect to past performance. He starts by saying that compensation depends on past performance and that anticipation of higher future compensation provides incentives in earlier years (Murphy, 1986). The research by Murphy (1986) confirms that past performance is used to estimate ability levels and levels of contemporaneous
performance. Boards can create more optimal contracts with the use of the information on the CEO gathered over the years. The empirical evidence of the paper by Murphy (1986) shows that there is a positive pay for performance, with a stronger relation in the earlier years of being at a firm.
Mishra, McConaughy and Gobeli (2000) studied the effectiveness of CEO performance. They found that the future performance of firms is positively related with CEO pay-for-performance sensitivity at low levels of sensitivity. Their results show this weakens as sensitivity rises and eventually at high levels of sensitivity, future performance is negatively associated with CEO pay-for-performance (Mishra et al., 2000). This is consistent with the managerial power view of Bertrand and Mullainathan (2001).
Abowd (1990) found that rewarding a CEO with an incremental 10% bonus after a year of good firm performance is associated with a 30 to 90 basis point increase in the expected after-tax gross economic return in the following year. Additionally, Abowd (1990) found that rewarding a CEO with an incremental raise of 10% after a year of good stock market performance is associated with a 400 to 1200 basis point increase in expected total shareholder return.
Hayes and Schaefer (2000) found that if firms use observable and unobservable executive performance measures, and observable performance measures are correlated with future firm performance, then the unexplained variation in executive compensation should predict future variation in firm performance. Additionally, Hayes and Schaefer (2000) state that if the firm shifts more weight onto unobservable performance measures, compensation becomes more positively related to future earnings.
The next section outlays the hypotheses and the coherent methodology. The sections after that continue with the data, the results and the conclusion.
II. Hypotheses and Methodology
The research question of this paper is: what is the relation between CEO compensation and the past, contemporaneous and future performance of Dutch hospitals? The purpose of this research is to find a causal relation between the level of compensation of the CEO and the performance of the hospitals, controlling for firm and CEO characteristics. First, the hypotheses are formed based on the model by Banker, Darrough, Huang and Plehn-Dujowich (2012). In the second part of this section the variables and the regression model are defined.
A. Hypotheses
According to the previous literature on CEO compensation, there is a positive pay-for-performance sensitivity regarding contemporaneous performance (Barro and Barro, 1990, Hall and Liebman, 1998, Houston and James, 1995, Murphy, 1985). When a CEO is rewarded with an increase in compensation, such as a bonus or a substantial raise, but not severance payments, there should be a positive balance in firm performance in the related period (Mishra et al., 2000). Additionally, previous research has shown that there is a positive relation between past performance and CEO compensation (Banker et al. (2012), Murphy (1986)) and that there is a positive relation between CEO compensation and future performance (Mishra et al. (2000), Abowd (1990), Hayes and Schaefer (2000))
The models used in this research are based on the model of Banker, Darrough, Huang and Plehn-Dujowich (2012). They have developed a two-period principal-agent model with moral hazard and adverse selection. Moral hazard defines the unobservable effort of the agent and adverse selection defines the unobservable ability of the agent, as described in the previous section. The model contains the purpose of using bonus to induce effort and rewarding better ability agent types. An optimum compensation structure trades off productive efficiency with information rent extraction (Banker et al., 2012).
The two-period model predicts that salary increases with past performance, since past performance can give an indication of the ability level of the agent. The principal is cautious in the first period on setting the salary, but in the second period it is easier to set salary according to ability. Since bonus is designed to extract more effort, bonus is expected to be positively associated with contemporaneous performance and to be negatively associated with past performance. For example, an agent with a superior past performance, is already known for his high ability, the principal does not need to offer a high-powered incentive contract to solve the agency problem (Banker et al., 2012).
Hypothesis 1a: Fixed compensation is positively associated with contemporaneous performance. Hypothesis 1b: Fixed compensation is positively associated with past performance.
Hypothesis 2b: Variable compensation is negatively associated with past performance.
On the same line of reasoning, higher current salary predicts higher future performance (Banker et al., 2012). Since in the subsequent period the principal has attained more information about the agent’s ability, the principal does not need to use as much bonus as before to induce more effort, but can substitute this for a higher salary. Bonus, as said before, is expected to be only positively associated with contemporaneous performance, and negatively associated with future performance.
Hypothesis 3: Future performance is positively associated with current fixed compensation, after controlling for current and past performance.
Hypothesis 4: Future performance is negatively associated with current variable compensation, after controlling for current and past performance.
Because the predictions differ between salary and bonus it is important to separate the fixed and variable components of compensation. Fixed compensation consists only of salary and variable compensation consists of bonus, severance and car payment.
These hypotheses are empirically tested by examining the extent to which past performance measures influence current fixed and variable components of compensation and the extent to which current compensation predicts future performance (Banker et al., 2012).
B. Methodology
The type of data in this research is parted into hospital performance and CEO compensation. Both hospital performance and CEO compensation consist of various contributors, which are explained below.
B.1 Variables for Hospital Performance
Performance of a hospital must reflect the values of various stakeholders, since it is a non-profit firm (Roomkin and Weisbrod, 1999). For most of the previous research, the principal is defined as the shareholders. In this case it is appropriate to use shareholders returns as performance measure (Murphy, 1985). For the non-profit hospitals in the Netherlands, there are no shareholders, just stakeholders. Therefore, the performance variable is measured in financial and qualitative performance, through a number of components. These components consist of financial ratios such as ROA, for financial performance and rankings for qualitative performance.
Financial performance of a hospital is measured in net income. Even though Dutch hospitals do not have any stock returns, earnings have demonstrated to be more sensitive to firm-specific changes than to market-wide changes (Sloan, 1993), thus making accounting measures sufficient.
Table 1. An explanation of the set of variables
This table describes each variable used for this research. Its name is stated along with its abbreviation in the regression. Then it shows a description of the variable and the source.
Variable Description Source
ROA ROB PM SB RANK Hospital performance ROA
Return per Beds Profit Margin Staff per Bed Ranking
Performance based on ROA
Performance based on return on beds, Performance based on profit margin Performance based on staff per bed Performance based on ranking
Annual report of hospital Annual report of hospital Annual report of hospital Annual report of hospital National newspaper results
Y F B S V CEO Total Salary Fixed Salary Bonus Severance Vehicle
Fixed salary plus bonuses, severance payments, other expenses.
Salary on a yearly basis Bonus payment Severance payment
Vehicle value (hospital contribution)
Annual report of hospital Annual report of hospital Annual report of hospital Annual report of hospital Annual report of hospital
G T Br A Characteristics Gender Tenure 1P Board Academic
Dummy (1=male, 0=female) Number of years as a boardmember Dummy (1= one-man board, 0= more board members)
Dummy (1=academic hospital, 0=non)
Annual report of hospital Annual report of hospital Annual report of hospital Annual report of hospital
Previous research has also shown that tying nonprofit hospital CEO compensation to financial performance can potentially improve organizational performance (Reiter, Sandoval, Brown and Pink, 2009). This research takes net income over assets (ROA). ROA is a financial performance measure, where the performance is controlled for the size of the hospital (Brickley and Van Horn, 2002). Another measurement is net income over number of beds, which is also used in order to assess the profitability of the hospital. A third financial measurement is the profit margin, with net income over revenues. This research uses profit margin to compare the different hospitals.
Qualitative performance is measured through ranking and capacity measurements. Berg et al. (2005) and Goodall (2011) find that using rankings is convenient to measure qualitative performance. The yearly status of a Dutch hospital is collected for this research. The quality of a hospital is also measured through the staff-bed ratio. The staff-bed ratio represents the number of employees per bed. Table 1 gives an overview of the variables and indicates their sources.
B.2 Variables for CEO compensation
CEO compensation consists of the following factors: gross compensation, severance payments, bonuses and vehicle value. Gross compensation is the base salary the CEO receives, also referred to as the fixed component of compensation. Severance payments are optional one-time only payments, and are only applicable when the CEO is leaving the company. Bonuses are optional incentive payments, where the CEO is rewarded for his good work. Vehicle value represents the value of the CEO’s car paid by the hospital, not all CEOs have this, while others have even put in additional money themselves for the car. The additional cash put in by the CEO himself is not taken into account, for this is not compensation from the hospital. Since a CEO of a smaller hospital will be likely to earn less than a CEO of a larger hospital the change in compensation is being used.
The first model (1) examines the pay-for-performance sensitivity to past and contemporaneous performance measures for each component of CEO compensation. The various performance measures are return on assets (ROA), return on beds (ROB), profit margin, number of staff per bed and a national ranking. Financial performance measures are separated into three regressions, since they are based on net income and therefore correlated. Components of compensation are fixed salary, bonus payment, severance payment and vehicle value. In addition, the regression is performed on total compensation. Also, packages of the variable pay were made. First, bonus and severance were combined and later, bonus, severance and vehicle value were put together. This was done to create a clear distinction between fixed and variable pay.
(1a) ln (Compensationi,t) = β0 + β1ROAi,t + β2ROAi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7Gi,t + β8Ti,t + β9Bri,t + β10Ai,t + ei,t
(1b) ln (Compensationi,t) = β0 + β1ROBi,t + β2ROBi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7Gi,t + β8Ti,t + β9Bri,t + β10Ai,t + ei,t
(1c) ln (Compensationi,t) = β0 + β1PMi,t + β2PMi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7Gi,t + β8Ti,t + β9Bri,t + β10Ai,t + ei,t
*eit = the statistical error *βk = parameters of the future performance relation
For these three regressions, hypothesis 1 (A en B) predicts that βK for k =1- 6, is positive for fixed compensation. Hypothesis 2a predicts that βk, for k = 1,3,5, is positive for bonus compensation, while Hypothesis 2b predicts that βk, for k = 2,4,6, is negative for bonus compensation.
The second model (2) examines whether current compensation is associated with future performance, after controlling for current and past performance. Per variable of future firm performance, three regressions take place. First, bonus, severance and vehicle value are separated in
the regression. Second, bonus and severance are combined and third, bonus, severance and vehicle value are combined. The second model also performs another regression solely based on total compensation.
(2a) Pi,t+1 = β0 + β1ln(Fi,t) + β2ln(Bi,t) + β3ln(Si,t) + β4ln(Vi,t) + β5ROAi,t + β6ROAi,t-1 + β7ROBi,t + β8ROBi,t-1 + β9PMi,t + β10PMi,t-1 + β11SBi,t + β12SBi,t-1 + β13RANKi,t + β14RANKi,t-1 + β15Gi,t + β16Ti,t + β17Bri,t + β18Ai,t + ei,t
(2b) Pi,t+1 = β0 + β1ln(Yi,t) + β2ROAi,t + β3ROAi,t-1 + β4ROBi,t + β5ROBi,t-1 + β6PMi,t + β7PMi,t-1 + β8SBi,t + β9SBi,t-1 + β10RANKi,t + β11RANKi,t-1 + β12Gi,t + β13Ti,t + β14Bri,t + β15Ai,t + ei,t
*eit = the statistical error *βk = parameters of the future performance relation
For regression (2a), hypothesis 3 predicts that β1 > 0, and hypothesis 4 predicts that β2, β3, β4 ≤ 0. To address the correlated omitted variable problem this research uses feasible generalized least squares to make estimations of the unknown parameters (Banker et al., 2012).
III. Data and Descriptive Statistics
This section contains a description of the data collection and an overview of the summary statistics. The research uses only publicly available data.
A. Data Collection
According the Health Care Inspectorate (2014), the IGZ (Inspectie voor de Gezondheidszorg), there are 84 hospitals and 8 teaching hospitals in the Netherlands. Most of the hospitals are standalone entities and some are part of a group. The sample consists of only 82 hospitals including academic hospitals. Ten hospitals are removed from the sample, a couple of them merged during the sample period, one is counted as a military hospital and others had not delivered a complete data set. Table 1 in the appendix provides a complete list of the hospitals in this sample. For every hospital the CEO’s total compensation is retrieved from either the annual report or the DigiMV document. DigiMV documents are, next to the annual reports, mandatory quality reports delivered by the hospitals, verified by Mediquest (CIBG, 2014). This report contains total salary, gross salary, bonus payment, severance payment, vehicle value, gender and tenure. In the annual report of a hospital, the number of beds, number of staff and the number of board members are given. On the balance sheet in the annual report the total value of assets is given and on the income statement EBIT and net income are published. Whether or not a CEO has a medical background can be found either in the annual report or in the BIG register. All measures of compensation are inflation adjusted to 2012 based on the CPI, the CPI for the Netherlands can be attained through the central bureau of statistics (CBS).
The sample period is from 2007-2012 to have a sufficient number of observations. Another reason for this sample period is that from 2007 all health care facilities were required to publish an annual report, which complies with the rules of the Ministry of Health, Welfare and Sport. These annual reports are publicly available. The sample period ends in 2012, since from January 1st 2013 there has been a cap on remuneration in the semi-public sector.
B. Descriptive Statistics
In this section descriptive statistics are displayed in order to present a first view upon the data. Table 2 presents the descriptive statistics. The median values for total CEO compensation, change in compensation and percent change in compensation are €246,205, €5,956 and 2.41% respectively. Fixed salary averages at €207,955. The median CEO drives a car of € 44,610 or more (it is optional for the CEO to make a contribution to the car), but does not receive a bonus and receives no severance payment. In fact, in the entire sample it was only 65 (out of 492) times a bonus was given and 19 times a severance was rewarded. The median CEO is a male with tenure of 7.69 years on the board and is not the only member on the board. Being a member of a board with multiple board members can be considered positive, since it is less considered less of a risk that the CEO can have influenced
Table 2. Descriptive Statistics
This table displays summary statistics for every variable given in table 1.
N Mean Standard Deviation 25th percentile Median 75th percentile
CEO compensation total 492 € 257,046 € 68,983 € 216,482 € 246,256 € 289,458 change 486 € 7,088 € 74,164 -€ 2,090 € 5,963 € 15,099 change in percentage 486 8.61% 39.97% -0.94% 2.42% 6.59% salary 492 € 207,955 € 51,365 € 181,759 € 207,084 € 231,126 bonus 65 € 2,982 € 8,900 0 0 0 severance 19 € 7,716 € 43,551 0 0 0 car value 370 € 39,883 € 28,413 € 22,863 € 44,730 € 54,398 CEO characteristics 492 tenure 5.42 4.75 2.22 3.86 7.11 gender 0.912 0.285 1 1 1 1-person board 0.193 0.397 0 0 0 Hospital characteristics 492 Total assets € 245,978,204 € 214,649,344 € 106,506,281 € 186,745,514 € 297,901,272 Number of beds 529 277 315 455 689 Return on Assets 1.82% 4.31% 0.63% 1.36% 2.43% Profit Margin 2% 4.93% 0.71% 1.44% 2.50%
Return per bed € 7,845 € 18,280 € 2,583 € 5,308 € 10,102
Staff per bed 3.67 1.27 2.88 3.46 4.07
Ranking 67.46 7.09 62.81 67.30 72.18
his own compensation. Only 9% of the CEOs are female. The median hospital has €186 million in assets and earns a 1.36% return on assets. In comparison, in the same period medium and small companies in the Netherlands averaged on a return on assets of 4% - 5% (CPB, 2014). The median hospital has 456 number of beds and earns a €5,312 per bed. Its median profit margin is 1.45%. The median number of personnel per bed is 3.46. Academic hospitals are expected to have more assets and a higher number of personnel per beds. Students are not counted as personnel, nevertheless more personnel is required for teaching and administration in academic hospitals.
Figure 1 shows that over the years the total compensation has gone up to rise from just below €250.000 to €275.000, while gross compensation had stayed around €200.000. This could be explained by larger increases in the variable payments, such as bonuses or severance payments, though there is also a slight increase in vehicle value, especially in 2012. Since bonus is not a common tool in the hospital industry, it is averaged very low and does not show an increase in figure 1. The new regulation, starting from January 1st 2013, poses a limit on gross compensation at €228.599, prohibits bonuses and limits severance payments at €75.000 (Rijksoverheid, 2014). To these means, gross compensation could continue to grow, but variable payments are restricted. The restriction of variable payments limits the use incentive contracts.
Figure 1. Mean CEO compensation
This figure displays the mean CEO compensation in term of gross salary, total salary and vehicle value.
In table 3 results of the correlation matrix are shown. Table 3 shows that some of the variables are strongly correlated. The variables for which the correlation is high have bold printed correlation values. First, the financial performance measures are strongly correlated, which makes sense since they are all based on net income. It is however important, to exclude the other two financial performance measures when one of the financial performance measurements is used. Furthermore, there is a high correlation between assets and staff per bed, and academic hospitals and staff per bed. Additionally, there is a high correlation between assets and the number of beds, and assets and academic hospitals. Because of this, the variable Assets is excluded from the regression.
Table 3
Cross Correlation Matrix
ROAt ROBt PMt SBt Rankt Gender Tenure 1p
Board
Assets Beds Academic
ROAt 1.000 ROBt 0.941 1.000 PMt 0.952 0.936 1.000 SBt 0.013 0.110 0.004 1.000 Rankt 0.098 0.111 0.085 0.122 1.000 Gender 0.022 0.028 0.019 -0.068 0.011 1.000 Tenure -0.032 -0.007 -0.005 0.125 0.091 0.013 1.000 1p Board 0.033 -0.025 0.024 -0.171 0.001 0.064 -0.010 1.000 Assets -0.016 0.066 -0.031 0.658 0.019 -0.013 0.102 -0.311 1.000 Beds 0.001 0.022 -0.022 0.417 0.080 -0.016 0.072 -0.382 0.764 1.000 Academic -0.018 0.066 -0.031 0.674 -0.080 -0.016 0.097 -0.162 0.816 0.535 1.000
Bold = strongly correlated
€ - € 50,000 € 100,000 € 150,000 € 200,000 € 250,000 € 300,000 2012 2011 2010 2009 2008 2007 Fixed Salary Total Compensation Vehicle Value Bonus Severance
IV. Results
In this section the results are shown and discussed. The section is divided into two parts, starting with the first model about current and past performance and continues with the second model about future performance. A brief summary concludes the section.
A. Model 1
The first model (1) examines the pay-for-performance sensitivity to past and contemporaneous performance measures for each component of CEO compensation. For both the first and the second model the estimates are derived from a Generalized Least Squares (GLS) regression. GLS is used to control for relations between the right-hand side variables and their previous values inside a group.
The performance measures are return on assets (ROA), return on beds (ROB), profit margin, number of staff per bed and a national ranking. Components of compensation are fixed salary, bonus payment, severance payment and vehicle value. In addition, the regression is performed on total compensation. Also, packages of the variable pay were made. First, bonus and severance are combined into bonus2 and second, bonus, severance and vehicle value are put together into bonus3. The first table, table 4, presents the results on the following regression:
ln (Compensationi,t) = β0 + β1ROAi,t + β2ROAi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7G + β8T + β9Br + β10A + ei,t
Table 4 presents the empirical result on H1 and H2, and uses return on assets (ROAt) as a firm performance measure. Column 1 shows that the coefficients on both contemporaneous and past performance measures are all positive, but not all are significant. Only contemporaneous ranking shows a significant positive coefficient of 0.003 (t-stat = 1.73). This indicates that a higher current position at the ranking by the national newspaper AD, more specifically an increase of the rating by one, increases the CEO’s salary by 0.3%. This supports H1a, in that fixed salary is positively correlated with contemporaneous performance. Banker et al. (2012) show similar results, while they only use return on equity (ROE) and stock return (RET). Their research shows an increase in salary of 1% and 0.6% for ROE and RET, respectively.
The second column starts with variable pay, which ends at the sixth column. In the first column bonus, on its own, is tested. There are no significant coefficients found in this column. In the second column the results on severance are shown, and the results on contemporaneous and past return on assets (ROA) seem to be significant. First, contemporaneous ROA has a coefficient of -5.922 (t-stat = -2.79), indicating that an increase in contemporaneous performance leads to a decrease in severance of close to 600%, which opposes the literature and the hypothesis. On the other hand, there were so few cases of severance that the sample is too small to be indicative in this case. Second,
Table 4
Regression of Compensation Components on Past and Contemporaneous Performance Independent
Variables
ln(sal) ln(bonus1) ln(severance) ln(vehicle) ln(bonus2) ln(bonus3) ln(total) Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat ROAt 0.148 1.323 -5.922*** 0.162 -4.549 -4.146 -0.143 (0.64) (0.55) (-2.79) (0.06) (-1.42) (-0.93) (-0.64) ROAt-1 0.156 3.330 -4.714* 0.050 -1.794 -1.010 0.075 (0.52) (1.07) (-1.72) (0.01) (-0.43) (-0.17) (0.26) SBt 0.000 0.186 0.160 0.183 0.374* 0.635** -0.005 (0.01) (1.22) (1.18) (1.00) (1.84) (2.22) (-0.32) SBt-1 0.002 0.010 -0.147 0.282 -0.071 0.250 0.014 (0.14) (0.06) (-1.12) (1.59) (-0.36) (0.90) (0.99) RANKt 0.003* -0.025 -0.022 0.038* -0.044 -0.013 0.002 (1.73) (-1.52) (-1.42) (1.94) (-1.94) (-0.40) (1.30) RANKt-1 0.000 -0.025 0.007 -0.027 -0.022 -0.058* 0.002 (0.09) (-1.46) (0.46) (-1.33) (-0.95) (-1.80) (1.38) Gender 0.136** 0.289 -0.127 0.564 0.666 1.510 0.176*** (2.40) (0.41) (-0.30) (0.57) (0.77) (1.09) (3.00) Tenure 0.053* 0.420 -0.494** 1.045*** 0.275 1.257** 0.050* (1.86) (1.37) (-2.01) (2.79) (0.68) (2.17) (1.81) Pboard -0.065 -0.856 0.319 1.083 -0.545 0.676 -0.050 (-1.53) (-1.57) (1.03) (1.35) (-0.83) (0.63) (-1.12) Academic 0.193** -1.561 -0.194 -4.304*** -2.111* -6.769*** 0.171** (2.40) (-1.53) (-0.32) (-2.67) (-1.73) (-3.35) (2.06) R2 0.021 0.029 0.022 0.055 0.040 0.047 0.054 n 405
*, **, *** Denote significance, based on two-tailed tests, at or below 10 percent, 5 percent and 1 percent, respectively. The estimation is based on a generalized least squares estimation.
the coefficient on past ROA is also significant at -4.714 (t-stat = -1.72), which would indicate that an increase in past performance decreases severance by 471%. Banker et al. (2012) report on a decrease of 44.8%. The result is in line with the hypothesis, stating that the slope of bonus decreases over time, however there could be a bias because of the small amount of observations in severance payments.
The fourth column reports on vehicle value and shows only a significant value on contemporaneous ranking with a coefficient of 0.038 (t-stat = 1.94). This indicates that a higher position in the ranking increases the amount paid for the car by the hospital by 3.8%. Column five and six both report on the combinations of variable pay. The first, bonus2, is a combination of bonus and severance and has a significant coefficient of 0.374 (t-stat = 1.84) on contemporaneous staff per bed. This indicates that an increase in the number of staff per bed increases the variable pay combination by 37.4%. This is in line with H2a, variable compensation is positively associated with contemporaneous performance. The result is also comparable to the result of Banker et al. (2012) reporting that an increase on stock return (RET), shows an increase of 33.8% in variable pay. Additionally, bonus3 seems to confirm this result. The column of bonus3 shows a significant coefficient of 0.635 (t-stat = 2.22) on staff per bed and thus also supports H2a. Column six of bonus3 shows another significant coefficient namely on past ranking. The coefficient of -0.058 (t-stat = -1.80)
Table 5
Regression of Compensation Components on Past and Contemporaneous Performance Independent
Variables
ln(sal) ln(bonus1) ln(severance) ln(vehicle) ln(bonus2) ln(bonus3) ln(total) Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat ROBt 0.000 0.000 0.000*** 0.000 0.000* 0.000 0.000 (-0.22) (0.23) (-2.97) (0.15) (-1.87) (-1.14) (-1.25) ROBt-1 0.000 0.000 0.000* 0.000 0.000 0.000 0.000 (0.33) (1.28) (-1.65) (0.11) (-0.36) (0.01) (0.56) SBt 0.000 0.183 0.169 0.183 0.383* 0.646** -0.004 (0.00) (1.20) (1.28) (1.00) (1.89) (2.25) (-0.30) SBt-1 0.003 0.002 -0.127 0.279 -0.055 0.260 0.014 (0.19) (0.01) (-0.97) (1.56) (-0.28) (0.93) (1.01) RANKt 0.003* -0.025 -0.022 0.038* -0.043 -0.012 0.002 (1.80) (-1.49) (-1.43) (1.93) (-1.89) (-0.39) (1.33) RANKt-1 0.000 -0.025 0.007 -0.027 -0.020 -0.057* 0.002 (0.15) (-1.46) (0.50) (-1.34) (-0.88) (-1.78) (1.41) Gender 0.138** 0.291 -0.100 0.566 0.687 1.529 0.177*** (2.43) (0.41) (-0.24) (0.57) (0.81) (1.11) (3.03) Tenure 0.051* 0.416 -0.489** 1.046*** 0.257 1.239** 0.050* (1.80) (1.35) (-2.00) (2.79) (0.64) (2.14) (1.79) Pboard -0.063 -0.841 0.267 1.088 -0.594 0.633 -0.050 (-1.50) (-1.55) (0.86) (1.35) (-0.92) (0.59) (-1.13) Academic 0.190** -1.586 -0.154 -4.304*** -2.089* -6.767*** 0.170** (2.37) (-1.57) (-0.25) (-2.68) (-1.73) (-3.37) (2.06) R2 0.020 0.032 0.024 0.055 0.039 0.046 0.057 n 405
*, **, *** Denote significance, based on two-tailed tests, at or below 10 percent, 5 percent and 1 percent, respectively. The estimation is based on a generalized least squares estimation.
indicates that a higher position on the ranking of last year decreases this year’s bonus by 5.8%. This result is in line with H2b, past performance decreases this year’s variable pay.
The last column of table 4 is of total pay. Total pay shows no significant results, which was expected. It confirms the expectation that it would be best to separate the fixed and variable components of pay.
Table 5 presents the empirical result on H1 and H2, and uses return on beds (ROBt) as a firm performance measure. It is based on the following regression:
ln (Compensationi,t) = β0 + β1ROBi,t + β2ROBi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7G + β8T + β9Br + β10A + ei,t
Likewise as in table 4 and supporting H1a, contemporaneous ranking shows to have a significant coefficient of 0.003 (t-stat = 1.80) on salary. Additionally, current return on beds seems to have a negative effect on severance, the coefficient is 0.000 (t-stat = -2.97) and past return on beds has a coefficient of 0.000 (t-stat = -1.65). Contemporaneous ranking has a positive performance on vehicle value, with a coefficient of 0.038 (t-stat = 1.93). This indicates a 3.8% increase in vehicle value when the ranking grade has rising by 1, which is equal to the result in table 4. In column 5, the current staff-
Table 6
Regression of Compensation Components on Past and Contemporaneous Performance Independent
Variables
ln(sal) ln(bonus1) ln(severance) ln(vehicle) ln(bonus2) ln(bonus3) ln(total) Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat PMt 0.050 0.451 -4.702*** 0.335 -4.395 -3.943 -0.158 (0.25) (0.22) (-2.55) (0.14) (-1.59) (-1.05) (-0.83) PMt-1 0.054 2.226 -3.602* -0.134 -1.704 -1.370 -0.012 (0.22) (0.89) (-1.63) (-0.05) (-0.51) (-0.30) (-0.05) SBt 0.000 0.185 0.157 0.184 0.375* 0.637** -0.005 (0.00) (1.21) (1.17) (1.00) (1.85) (2.22) (-0.32) SBt-1 0.003 0.012 -0.154 0.282 -0.044 0.251 0.014 (0.18) (0.08) (-1.16) (1.59) (-0.36) (0.90) (1.00) RANKt 0.003* -0.025 -0.023 0.039* -0.044 -0.013 0.002 (1.78) (-1.47) (-1.51) (1.95) (-1.95) (-0.40) (1.32) RANKt-1 0.000 -0.024 0.007 -0.027 -0.021 -0.057* 0.002 (0.13) (-1.41) (0.45) (-1.34) (-0.92) (-1.78) (1.42) Gender 0.137** 0.291 -0.132 0.565 0.669 1.512 0.176*** (2.42) (0.41) (-0.31) (0.57) (0.78) (1.10) (3.01) Tenure 0.052* 0.411 -0.461** 1.044*** 0.290 1.264** 0.050* (1.81) (1.34) (-1.88) (2.79) (0.72) (2.19) (1.82) Pboard -0.064 -0.854 0.304 1.082 -0.548 0.672 -0.050 (-1.51) (-1.57) (0.98) (1.35) (-0.84) (0.63) (-1.12) Academic 0.191* -1.559 -0.206 -4.305*** -2.131* -6.793*** 0.170** (2.39) (-1.53) (-0.34) (-2.67) (-1.75) (-3.37) (2.05) R2 0.020 0.028 0.019 0.055 0.041 0.047 0.054 n 405
*, **, *** Denote significance, based on two-tailed tests, at or below 10 percent, 5 percent and 1 percent, respectively. The estimation is based on a generalized least squares estimation.
bed ratio has a coefficient of 0.383 (t-stat = 1.89) on bonus2. The sixth column shows a coefficient of 0.646 (t-stat = 2.25) for the current staff per bed ratio and a coefficient of -0.057 (t-stat = -1.78) for the past ranking. These results are similar to the results in table 4. Again, none of the coefficients on total compensation is significant.
Table 6 presents the empirical result on H1 and H2, and uses profit margin (PMt) as a firm performance measure. The table is based on the following regression:
ln (Compensationi,t) = β0 + β1PMi,t + β2PMi,t-1 + β3SBi,t + β4SBi,t-1 + β5RANKi,t + β6RANKi,t-1 + β7G + β8T + β9Br + β10A + ei,t
The results in table 6 are again fairly similar to the results in table 4 and 5. The first column shows a positive coefficient of 0.003 (t-stat = 1.78) for current ranking on salary. Contemporaneous and past profit margin seem to have negative coefficients of -4.702 (t-stat = -2.55) and -3.602 (t-stat = -1.63) on severance. Current ranking shows a coefficient of 0.039 (t-stat = 1.95). Current staff per bed ratio has positive coefficients of 0.375 (t-stat = 1.85) and 0.637 (t-stat = 2.22) for both bonus2 and bonus 3, respectively. Also, as in table 4 and 5, past ranking has a negative coefficient of 0.057 (tstat = -1.78).
B. Model 2
Table 7 presents empirical results from testing on H3 and H4. The second model (2) examines whether current compensation is associated with future performance, after controlling for current and past performance. Per variable of future firm performance, three regressions take place. First, bonus, severance and vehicle value are separated in the regression. Second, bonus and severance are combined and third, bonus, severance and vehicle value are combined. The second model also performs another regression solely based on total compensation. Table 7 shows the results on the following regression:
Pi,t+1 = β0 + β1ln(Fi,t) + β2ln(Bi,t) + β3ln(Si,t) + β4ln(Vi,t) + β5ROAi,t + β6ROAi,t-1 + β7ROBi,t + β8ROBi,t-1 + β9PMi,t + β10PMi,t-1 + β11SBi,t + β12SBi,t-1 + β13RANKi,t + β14RANKi,t-1 + β15G + β16T + β17Br + β18A + ei,t
The first three columns report the result for current compensation in relation to future return on assets (ROA). The first column takes the variable components of compensation all separately, while the second column combines bonus and severance compensation into Bonus2 and the third column combines bonus, severance and vehicle value in Bonus3. For every performance measure three regressions are done in the same order as just described.
In the first column, none of the current compensation components is associated with future ROA, since none of them is significant. The signs of every coefficient for the different components of compensation are all negative. This would indicate that an increase in salary, bonus, severance or vehicle value would lead to decrease in future performance, but because of their insignificance it is assumed that there is no relation at all between compensation and future performance. The second column does not report significant coefficient estimates either. Though the overall results are similar to column 1. Fixed salary again has, though not significantly different form zero, a negative coefficient, which would oppose H3. Variable pay also has a negative coefficient, which would support H4. The third column of future ROA shows the same results as column 1 and 2. Though these results are insignificant they would support H4, since the variable components of compensation have a negative coefficient, but they would not support H3.
Column 4, 5 and 6 show the results for future return on beds. None of the three columns show any results significantly different from zero, though all three columns show similar results. Salary has a negative coefficient in all three columns, which would again oppose H3. The variable components and combinations of compensation show negative results, which would support H4. Column 7,8 and 9 show the results for future profit margin. However, none of these results are significantly different from zero either. Once more, salary shows negative coefficients, which would oppose H3. While the variable components and combinations of compensations show negative results that support H4. The
results from the financial performance measures indicate that the components of compensation do not give any prediction about future performance.
Until column 10 there are no significant coefficients that could indicate an association between current compensation and future performance. In column 10, the results indicate that current salary is positively correlated with the staff-bed ratio in the future with a coefficient of 0.311 (t-stat = 1.68). It seems that when the CEO receives a higher salary in one period, the CEO will hire more staff to attain to the patients. This is clearly the case in all three tests on the staff-bed ratio, with the coefficients being 0.330 (t-stat = 1.80) and 0.314 (t-stat = 1.71) in the second two columns. These results provide evidence to support H3. On variable pay no significant coefficients are found in the staff-bed ratio columns. In these columns, significant coefficients on the academic variable are found. This is fairly intuitive since academic hospitals need a high number of staff to help and teach the student doctors.
In the columns on ranking no significant coefficients are found either, not on salary nor on variable pay.
Table 8 shows the results on H3 and H4 with only total compensation as the independent variable. The second model (2) in this case examines whether current total compensation is associated with future performance, after controlling for current and past performance. Table 8 shows the results on the following regression:
Pi,t+1 = β0 + β1ln(Yi,t) + β2ROAi,t + β3ROAi,t-1 + β4ROBi,t + β5ROBi,t-1 + β6PMi,t + β7PMi,t-1 + β8SBi,t + β9SBi,t-1 + β10RANKi,t + β11RANKi,t-1 + β12G + β13T + β14Br + β15A + ei,t
None of the variables of total compensation show significant coefficients. This confirms the idea that total compensation would not be a great indicator for a pay-for-performance sensitivity. The hypotheses made in this paper predict that variable and fixed pay can have opposite, positive and negative, coefficients. When you put the variable and fixed components of compensation together this would thus lead to an insignificant coefficient for total pay.
