The relation between busy directors and the use of nonfinancial performance measures in annual bonus contracts

1

The relation between busy directors

and the use of nonfinancial

performance measures in annual

bonus contracts

Name: Abdellatif el Bouzidi Student number: 11112506 Thesis supervisor: Mario Schabus Date: 26-06-2017

MSc Accountancy & Control, specialization [Accountancy] Faculty of Economics and Business, University of Amsterdam

2 Statement of Originality

This document is written by student Abdellatif el Bouzidi who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3 Table of Content 1. Introduction ...4 2. Literature review ...7 2.1 Agency Theory ...7 2.2 Incentive plan ...8

2.3 Nonfinancial performance measures ...9

2.4 Busy directors ...12

2.5 Hypothesis development ...14

3. Methodology ...16

3.1 Sample selection ...16

3.2 Nonfinancial performance measures (dependent variable) ...17

3.3 Busy directors (independent variable) ...18

3.4 Control Variables...18 3.5 Empirical model ...19 4. Results ...22 4.1 Descriptive statistics ...22 4.2 Correlation Matrix ...26 4.3 Multivariate Analysis ...29 4.4 Robustness Analysis ...31 Conclusion ...36 References ...39

Appendix 1: Compare means by weighting nonfinancial performance measures ...41

4 1. Introduction

The evaluation of a CEO performance can be based on several performance indicators. These performance measures are mainly categorized into financial – and nonfinancial measures. Financial performance measures such as ROI and stock market returns have been traditionally used as the main measures to evaluate a CEO’s performance (Davila & Venkatachalam, 2004; Dechow & Sloan, 1991; Ittner et al., 1997). Besides, studies have also shown that nonfinancial measures such as customer satisfaction and product quality are increasingly used for evaluating a CEO (Davila & Venkatachalam, 2004; Zuriekat et al., 2011) as they provide information

additional to financial measures (Ittner & Larcker, 1998). Ittner & Larcker (1998) argue that nonfinancial measures focus on long-term performance, deal with potential non-financial objectives and can affect firm’s future financial performance positively. The authors find for example that a higher score on customer satisfaction leads to a customer’s repurchase of the firm’s products or services which results in a higher future revenue and thus financial performance. However, there are also some limitations regarding the use of nonfinancial performance measures. Nonfinancial performance measures are, compared to financial

performance measures, expensive and time consuming to implement (Ittner and Larker, 2002). Moreover, nonfinancial performance measures are rarely audited which makes them very susceptible for manipulation (Ittner et al., 1997) and measured differently across firms and even across departments. This makes them very difficult to compare.

As the board of directors has the responsibility to design the CEO annual bonus contract, monitor and evaluate the CEO, it is important to study which board characteristics affect the design of the annual bonus contracts. Core et al. (1999) conducted a research about the effect of board and ownership structures on the CEO compensation and the subsequent firm performance. They find a positive relationship between the presence of busy directors and the CEO

compensation. This means that the presence of busy directors contributes to higher CEO compensation which implies that busy directors are ineffective monitors and that their presence leads to lower board governance quality. They find also that higher CEO compensation and the presence of busy directors are significantly and negatively associated with the subsequent firm operating and stock return performance. This is in line with the finding of Fich and Shivdasani (2006) that firms with busy directors have a lower market performance than firms without busy directors. Moreover, Fich and Shivdasani (2006) find that directors holding more directorships

5 are less effective monitors and that this ineffectiveness is being strengthened when the majority of the board consist of busy directors. The reason for this ineffectiveness lies in the fact that they have multiple directorships which makes them too busy to observe and monitor their CEO. Given that nonfinancial performance measures are, compared to financial performance measures, expensive and time consuming to implement (Ittner and Larker, 2002), the presence of busy directors who are too busy to observe and monitor their CEO may affect the inclusion of nonfinancial performance measures in annual bonus contracts.

Prior studies have already shown that there is a relationship between certain board

characteristics and the design of the CEO annual bonus contract. Ittner et al. (1997) predicted that CEOs who have power over board are more likely to encourage the use of nonfinancial

performance measures given their susceptibility to manipulation. The authors assumed that CEOs will always look for means to inflate their compensation. In contrast to what they have predicted, they found no evidence for greater weighting on nonfinancial performance measures when the CEO has power over the board. This indicates that CEOs are not likely to influence the design of the annual bonus contract to inflate his compensation.

Overall, existing literature indicates that weak corporate governance, to which the

presence of busy directors contributes (Core et al., 1999), allows CEOs to serve their self-interest and to gain excess benefits. Additionally, busy directors are considered ineffective monitors and are associated with low firm performance. Due to their multiple directorships they are too busy to monitor and evaluate their CEO adequately. Given the time consuming and expensive character of nonfinancial performance measures, the question arises whether firms with busy directors rely on nonfinancial performance measure and to which extent. This study examines the effect of busy directors on the design of the CEO annual bonus contract and in particular the reliance on

nonfinancial performance in these contracts.

This study contributes to the existing literature as the relation between busy directors and the reliance on nonfinancial performance measures in CEO annual bonus contracts has, to my best knowledge, not been studied yet. So, this study extends the existing literature on the influence of corporate governance quality and board characteristics on the design of the CEO annual bonus contract. The rest of this paper is structured as follows. Section two contains the literature review. The sample selection process and the measurements of variables of interest are

6 discussed in section three. Section four discusses the results of the hypothesis testing. The

7 2. Literature review

2.1 Agency Theory

In order to examine the association between the presence of busy directors in the board of directors and the reliance on nonfinancial performance measures in annual bonus contracts, it is useful to point out the agency theory arising from the goal incongruence between the principal and the agent. The agency theory refers to the relationship between the owners of a firm and the CEO. Jensen & Meckling (1976) defines this relationship as a “contract under which one or more

persons (the principals) engage another person (the agent) to perform some service on their behalf which involves delegation some decision making authority to the agent ”. In the

relationship between the principals and the agent the problem of information asymmetry arises. Information asymmetry refers to the phenomenon that some parties have an information

advantage regarding a business transaction. This phenomenon can be divided in two types, namely adverse selection and moral hazard.

Adverse selection can be described as a phenomenon wherein one party has knowledge or information, so called inside information, which the other party does not have. Due to the fact that one party possesses information that the other party does not have, it is possible that

investors take decisions without having relevant information. In the context of the principal and agent relationship, the agent has an information advantage in comparison to the principals. By means of full disclosure this advantage can be lifted as all relevant inside information is disclosed publically. Adverse selection can be therefore controlled by converting inside information to outside information in a reliable manner.

Moral hazard reflects the phenomenon that some parties cannot observe the action of others, while these actions affect the interest of all parties. This phenomenon refers to the reliability of information. Because of the before mentioned separation of ownership and control in firms, it is for the owners (principals) difficult to verify the information presented by the agent as the agent can have manipulated this information and consciously created noise. The

information provided in these cases can lose its reliability. So, moral hazard arises because the agent possesses information about how a certain result or performance is achieved and whether it is a high or a low quality result, while the principals do not have this information. As in general the principals and agent are utility maximizers, there is a huge possibility that the agent will not always act in the best interests of the principals (Jensen & Meckling, 1976). This divergence of

8 interest can be limited by creating incentives for the agent and monitor his actions (Jensen & Meckling, 1976).

2.2 Incentive plan

A tool that firms use to incentivize agents is the incentive plan (Banker et al., 2000). According to Murphy (1999) these incentive plans typically consist of five components. The first component is the base salary which is a fixed compensation. The second component is the annual bonus plan. These bonus plans are based on the short-term performance and are paid when the CEO achieve a certain threshold performance (Murphy, 1999). The third component is stock options. Stock options give CEOs the right to purchase a share or option when a pre-determined performance is attained (Murphy, 1999). The fourth component is restricted stock. Restricted stock is granted to a CEO as compensation but cannot be fully transferred until a certain

condition is met (Murphy, 1999). The last component is the long-term incentive plan. Long-term incentive plans have approximately the same structure as the annual bonus plan. The difference between them is that the long-term incentive plans are based on a cumulative long-term

performance threshold (Murphy, 1999) while annual bonus plans are based on annual performance.

Compensation plans are constructed in order to incentivize the CEO to act in the best interest of the firm. An example is the incentive role of the restricted stock. When CEOs are granted restricted stock they are incentivized to increase the future stock price as they will benefit from this increase at the moment the restriction term ends. However, CEOs can also be punished for not achieving a certain threshold by cutting their compensation. Because all individuals in a firm are self-interested (Jensen & Meckling, 1992), the CEOs would always take actions to avoid the punishment even though the actions would negatively affect the principals. Jensen &

Meckling (1992) state that a control system is needed to align the actions and interest of a CEO with that of the organization and the principals. They describe the control system as a system that specifies the performance measurement and evaluation system for the agent (CEO) and the reward and punishment system that relates to the agent’s individual performance.

9 2.3 Nonfinancial performance measures

In order to measure the performance and to observe the actions of the CEO, performance measures are used. According to Barkema & Gomez-Mejia (1998) there are three principles that need to be considered when choosing performance measures for performance evaluation. One of these principles is informativeness. Informativeness means that the performance measure has to reflect the actions of the CEO (Hölmstrom, 1979; Ittner & Larcker, 2002). Besides, a

performance measure has to be accurate, timely and consider the effect of the CEO’s current actions on the firm future profitability (Gibbs et al., 2004; Hölmstrom, 1979).

Firms have traditionally used financial performance measures such as ROI and stock market returns to evaluate and reward the CEO’s performance (Davila & Venkatachalam, 2004; Dechow & Sloan, 1991; Ittner et al., 1997). However, the financial measures do not always meet the requirements for an adequate performance measurement. Financial performance measures provide for example information about the performance in the past (Ittner & Larcker, 2003). This means that financial performance measures only reflects the CEO’s actions that took place in the past and doesn’t reflect the current actions and their effect on the firm’s future profitability (i.e., they are backward looking performance measures. Kihn (2007) shows that the use of financial performance measures incentivizes managers to improve the short-term profitability.

However, this feature has a downside as managers who are incentivized by financial performance measures are inclined to take decisions that will lead to short-term profitability but that are not aligned with the shareholder’s interest (Bushman et al., 1996). Examples are R&D investments and long-term investments where expenses related to such investments are

immediately recognized and expected future benefits are recognized when realized. The effect of the latter is that CEOs and managers that are evaluated on the basis of financial performance measure may postpone these investments in order to meet the financial performance objectives, even though the investments have positive NPVs (Dechow & Sloan, 1991). Besides CEOs may manage earnings to report better earnings and to meet the financial threshold when their

evaluation is based on short-term financial performance measures (Burgstahler & Dichev, 1997; Healy, 1985; Watts & Zimmerman, 2003). Although the actions of the CEOs in these particular situations are not aligned with the shareholder’s interest, the CEOs are still rewarded when the financial threshold is met. The reason is that the actions of the CEO are not fully observable for the board and the shareholders as the financial performance measures used do not capture all the

10 CEO’s actions. Therefore using financial performance measures alone in annual bonus contracts may not provide efficient incentives for CEOs to act in the best interest of the shareholders (Ittner et al., 1997).

According to Ittner et al. (1997) a non-zero weight should be place on performance measures that gives incremental insight on the dimensions of managerial actions and the relative weighting of the performance measures should be in accordance with the informativenes

regarding the actions of the CEO. As the use of nonfinancial performance measures provides additional information about the CEO’s actions (Ittner et al., 1997), nonfinancial performance measures should be included in annual bonus contracts. Studies have shown that nonfinancial performance measures are increasingly used as a supplement to the financial performance measures in annual bonus contracts (Davila & Venkatachalam, 2004; Zuriekat et al., 2011) as they can reinforce the weak and noisy financial performance measures (Feltham and Xie, 1994). A second advantage of nonfinancial performance measures is that they reduce information asymmetry and particularly the moral hazard problem (Hölmstrom, 1979; Davila &

Venkatachalam, 2004) which enhances the monitoring and evaluation quality. This is consistent with the finding of Ittner et al. (1997) that nonfinancial performance measure provide additional information about the CEO’s actions and especially about actions that affects aspects other than financial aspects, such as environment or employee satisfaction. The additional information about the CEO’s actions that is obtained from nonfinancial performance measures subsequently reduces the information asymmetry as CEO’s actions are more observable. The board’s ability to make right decision regarding the CEO compensation increases.

The third advantage is that nonfinancial performance measures focus, in contrast to financial performance measures, on long-term performance (Ittner & Larcker, 1998). Using metrics such as customer satisfaction, employee satisfaction and product quality can provide information about future financial performance. Ittner & Larcker (1998) investigated the

association between customer satisfaction and firm performance on customer, business unit and firm level. They find that an improvement of customer satisfaction leads to a customer’s

repurchase of the firm’s products or services which results in a higher future revenue/financia l performance. Besides, they find that a higher customer satisfaction results also in an increase in the number of customers which subsequently leads to higher revenues. Furthermore, the authors argue that customer satisfaction is also useful for investors as investors predict a favourable

11 future financial performance for firms with a high customer satisfaction. Thus, the use of

nonfinancial performance measures in annual bonus contracts positively affects the future performance of firms (Banker et al., 2000; Said et al., 2003). Said et al. (2003) claims that firms using nonfinancial performance measures in annual bonus contracts have a better future

accounting and stock market performance. According to Banker et al. (2000) are managers in hospitality incentivized to make long-term decisions when nonfinancial measures such as customer satisfaction are used. Based on the latter it can be inferred that nonfinancial

performance leads to the alignment of the CEO’s actions to the firm strategy and subsequently result in better future financial performance and thus the firm value.

Besides the advantages, nonfinancial performance measures may also have some negative implications that have to be addressed. According to Ittner and Larker (2002) the use of

nonfinancial performance measures can be costly as it is expensive and time consuming to implement and measure the nonfinancial aspects. Financial performance measures on the other hand are less expensive as information on them is widely available within the organization and can be implemented easily. They also state that nonfinancial performance measures are only useful if the measures are relatively more informative when there is a moral hazard problem. Additionally, Ittner et al. (1997) state that nonfinancial enables CEOs to inflate their

compensation, given that nonfinancial performance measures are, relatively to financial performance measures, susceptible to manipulation and rarely audited by external auditors. Consistent with the latter, the susceptibility of the nonfinancial performance measures to manipulation causes measurement problems, because financial performance measures are less difficult to measure and quantify than nonfinancial performance measures. Nonfinancial performance measures can be measured by managers differently across time and entities or business units which reduces the comparability. Furthermore, including different nonfinancial performance measures that are measured on different levels in annual bonus contracts

complicates the evaluation of the CEO’s performance and the analysis of how this performance can be translated into the realization of the nonfinancial targets. Moreover, some nonfinancial performance measures have more than one dimension which makes it difficult to use these particular measures effectively. Customer satisfaction for example contains elements such as quality, efficiency and response time. Actions undertaken to improve product quality will

12 enhance the overall customer satisfaction, but can affect the other elements negatively (Lillis, 2002).

After having discussed the benefits and implications of nonfinancial performance, it is useful to point out the determinants of the use of nonfinancial performance measures in annual bonus contracts. Ittner et al. (1997) examined this by taking organizational strategy, the adoption of strategic quality initiative, regulatory environment, exogenous noise in short-term financial performance measures and the relative power of the CEO over the board of directors into account. They find that firms with a prospector strategy1 are more likely to use nonfinancial performance measures than firms with a defender strategy2. Additionally, they find that firms adopting a strategic quality initiative put a greater weight on nonfinancial performance measures which is in line with the strategic alignment discussed before. Moreover, the authors find a positive relation between the exogenous noise in short-term financial performance measures and the weight put on nonfinancial performance measures. This means that when the noise in short-term financial performance measures increase, firms will tend to put more weight on nonfinancial performance measures. Besides, they find that firms that face external regulation put relatively more weight on nonfinancial measures. Furthermore, they find no evidence for greater weighting on nonfinancial performance measures in annual bonus contracts (because of the manipulation susceptibility of these measures) when the CEO has power over the board of directors by means of appointing internal and external board members. Instead, the authors show that there is a negative relation between these two variables. Based on the latter it can be inferred that board characteristics can affect the financial and nonfinancial weighting in annual bonus contracts. 2.4 Busy directors

One of the board of directors’ responsibilities is to delegate the decision making process to the CEO. A second responsibility is to monitor and evaluate the decisions made and the execution of the firm strategy by the CEO. Besides, the board determine the performance measures to be used for the monitoring and evaluation of the CEO’s performance. Prior studies have already shown that board characteristics such as board composition are determinants for the design of the CEO’s annual bonus contract (Core et al., 1999; Ryan & Wiggins, 2001). For

1

Prospect strategy: identifies new products and services and adapt quickly to market environment (Ittner et al., 1997)

2

Defender strategy: provides a stable set of products and services and attempts to improve the current products and services (Ittner et al., 1997)

13 instance, Ittner et al. (1997) show that the presence of board members appointed by the CEO put significantly less weight on nonfinancial performance in annual bonus contracts.

Another board mechanism that may influence the design of an annual bonus contract and thus the weight put on financial and nonfinancial performance measures is the presence of busy directors. Busy directors are directors who have multiple directorships. According to Ferris et al. (2003) directors holding three or more directorship are considered to be busy. Besides, they find that these directors are more likely to be appointed to the board if the CEO can influence the selection process. The authors argue that busy directors become inefficient to observe and

monitor the CEO’s actions as they have multiple directorships and thus busy to monitor the CEO. Core et al. (1999) find for example that the CEO compensation is an increasing function of directors with three or more directorships. This means that the monitoring of the CEO and the governance structures in controlling agency problems are less effective when there is a high percentage of busy directors and will subsequently lead to poorer firm performance (Core et al., 1999). The latter is in line with the finding of Fich and Shivdasani (2006) that firms with busy directors have a lower market performance than firms without busy directors. Moreover, Fich and Shivdasani (2006) find that directors holding more directorships are less effective monitors and that this ineffectiveness is being strengthened when the majority of the board consist of busy directors.

However, note that there are some studies that suggest that director busyness captures merely director’s ability or expertise. In contrast to the before outlined negative effects of busy directors on the monitoring quality and firm performance, Ferris et al. (2003) find no evidence that busy directors are ineffective monitors and the likelihood that they will be charged with fraud. Additionally, they find no evidence for their predicted negative effect of busy directors on firm performance as market participants do not react negatively to the presentation of a multiple director. The authors believe that directors sitting on the board of reputable corporates and on relatively larger boards are more likely to be appointed in multiple directorships as they have expertise, experience and considered to be effective monitors. Moreover, Masulis and Mobbs (2011) find that market participants react positively to the first appointment of an additional directorship which is in line with the finding of Ferris et al. (2003). However, Masulis and Mobbs (2011) show that the market participants react negatively to the appointment of subsequent

14 2.5 Hypothesis development

As argued before, the board is responsible for the delegation of the decision making process to the CEO. The delegation of the decision making process leads to a control problem. In order to solve this control problem, boards monitor and evaluate the decisions made and the execution of the firm strategy by the CEO. Subsequently, performance measures are defined and used in the monitoring and evaluation process. Prior studies have shown that board characteristics such as board composition are determinants for the design of the CEO annual bonus contract (Core et al., 1999; Ryan & Wiggins, 2001). Another board characteristic that may affect the design of the annual bonus contract is the presence of busy directors. Busy directors are directors who hold multiple directorships. This type of directors is considered inefficient and ineffective in observing and monitoring CEO behaviour (Ferris et al, 2003). Because of their multiple

directorships, they are compared to directors with one directorship too busy to observe and monitor their CEO. Besides, busy directors are positively associated with excessive CEO compensation and negative firm performance which indicate ineffective monitoring (Core et al., 1999; Ferris et al., 2003; Fich & Shivdasani, 2006). Since nonfinancial performance measures (compared to financial measures) are expensive and time consuming to implement and to measure (Ittner & Larker, 2002), it is expected that the implementation of these measures is too costly for busy directors and hence they are less likely to recommend the use of nonfinancial performance measures. Moreover, incorporating nonfinancial performance in annual bonus contracts is a sign of an effective monitoring and evaluation process. Since busy directors are considered inefficient and ineffective in observing and monitoring CEO behaviour (Ferris et al, 2003), it is predicted that they are less likely to recommend the use of nonfinancial performance measures.

Furthermore, Ittner et al. (1997) examined the relationship between the power of CEOs over the board of directors by means of appointing internal and external board members and the use of nonfinancial performance measures. The authors predicted that due to the susceptibility of nonfinancial performance measures to manipulations, CEOs would use their power over the board of directors to encourage the use of nonfinancial performance measures. Nonfinancial performance measures can be manipulated easily and they are rarely audited. According to the authors, using nonfinancial performance measures in bonus contracts would enable the CEO to inflate his compensation. However, their results depict the opposite of what they have predicted.

15 The authors found a negative relationship between the power of a CEO and the use of

nonfinancial performance measures. This suggest that CEOs are less likely to encourage the use of nonfinancial performance measures and that they may use other means to inflate his

compensation. Additionally, Davila & Penalva (2006) studied the relationship between the governance structure and the design of CEO contracts and found that firms where the CEO has a strong influence on governance decisions place a higher weighting on accounting-based

performance measures in CEO compensation contracts instead of stock-based performance measures. The authors argue that CEOs with power over board influence the design of

compensation contracts in order to place more weight on performance measures which they can easily control/manipulate and with lower variance. Additionally, the authors state that placing more weight on performance measures that can be easily controlled increases the likelihood of achieving the performance targets, reduce the variability in the actual compensation and hence increase their utility. In this particular case a higher weight is placed on accounting-based

performance measures instead of stock-based performances measures, because accounting-based performance measures are less noisy than stock-based performance measures. Less noisy

performance measures can be controlled and does not lead to variability in actual compensation. Inconsistent with the finding of Ittner et al. (1997), this finding suggest that CEO s may encourage the use of nonfinancial performance in annual bonus contracts given that nonfinancial

performance measures are less noisy than accounting-based performance measures (Ittner et al., 1997).

Based on the above, the following research hypothesis is formed.

Hypothesis: There is a negative association between the presence of busy directors and the use of nonfinancial performance measures.

16 3. Methodology

3.1 Sample selection

To test the hypothesis mentioned before, a database research will be conducted. The firms that will be included in my starting sample are part of the Standard & Poor 1500 index. In order to do so the directors covered by ISS over the period 2012-2015 are examined first. The data covers 6016 firms across four years and has an average of approximately 1504 firms a year. Table 1 presents the numbers of the firms, board position and directors included in the first selection. Remarkable is that the average of directors per firm increased from 9.38 to 9.41 between 2012 and 2015. This increase in board size may indicate a less effective monitoring as a larger board size is associated with weak performance (e.g. Guest, 2009). However, it is not clear whether this increase is significant or not.

Table 1: Numbers of firms, directors and board positions.

Besides, the average directorships held per director increased also over time. In 2012 the average directorships held by a director was 1.21 increasing in 2015 to 1.22. This increase is presented in more detail in table 2. As shown in this table, the majority of the directors are holding only one directorship. This fact remains over time. Moreover, an increase in the number of directors holding two to four directorships can be noticed. The number of directors holding three directorships increased from 348 in 2012 to 378 in 2014. Consistently, the number of directors holding four directorships increased from 45 in 2012 to 62 in 2015. However, there is also a decrease in the number of directors holding five directorships. Taking table 1 and 2 into consideration, the figures suggest that directors became busier which may indicate that they have become less effective monitors. This may suggest that directors became busier.

Year # Firms

# Director

Positions # Directors Mean

Standard deviation Mean Standard deviation 2012 1497 13993 11590 9,35 2,32 1,21 0,50 2013 1515 14204 11787 9,38 2,35 1,21 0,50 2014 1499 14065 11527 9,38 2,29 1,22 0,52 2015 1505 14162 11652 9,41 2,27 1,22 0,51 Total 6016 56424 46556 9,38 2,31 1,21 0,51

# Board positions per director # Directors per firm

17

Table 2: Numbers of multiple directorships over time

As a next step, I hand-collected data on performance measures using the proxy statements available on the EDGAR database. This is done for 50 companies and for the period 2012-2015, and hence reduces my sample to 200 observations. However, due to missing values I was obliged to reduce even more to 195 observations. Moreover, the hand collected performance measures data consist of weightings of both financial and nonfinancial performance measures.

3.2 Nonfinancial performance measures (dependent variable)

The main dependent variable that is used to test the research hypothesis, is NFPMUSE. This is an indicator variable coding firms using nonfinancial performance measures in annual bonus contract as one and zero if this is not the case. The alternative dependent variable that is use in this research is NFPMWEIGHT and contains the weight, relative to financial performance measures, placed on nonfinancial performance measures in annual bonus contracts. In this

variable a distinction is made between placing a high weight on nonfinancial measures and a low weight. The cut-off that is used to distinguish high from low weighting is 10%. Firms placing a weight on nonfinancial measures below the 10% are considered low weighting firms, while firms placing a weight beyond the 10% are considered high weighting firms. Low weighting firms are coded 0 and high weighting firms are coded 1.

The nonfinancial performance information is hand collected by reviewing and reading the proxy statements, especially the compensation discussion and analysis section, provided in the EDGAR database. In case the weights placed on nonfinancial performance measures are not provided, a dummy variable is used where firms using both financial and nonfinancial

performance measures are coded as one. Firms using only financial performance measures are coded as zero (Ittner et al., 1997). To identify whether firms use nonfinancial performance measures in annual bonus contracts, keywords such as “nonfinancial measures”, “non-financial measures”, “customer satisfaction”, “customer service”, “employee satisfaction”, “leadership”, “innovation”, “safety”, “product/service quality”, “employee engagement”, “customer

complaints”, “market share”, “returned products”, “customer response time”, “capacity

Year Total 2012 964383,20% 154813,36% 3483,00% 450,39% 60,05% 11590 2013 983583,44% 155113,16% 3452,93% 480,41% 80,07% 11787 2014 949982,41% 158613,76% 3783,28% 600,52% 40,03% 11527 2015 963482,68% 159213,66% 3623,11% 620,53% 20,02% 11652 Total 3861182,93% 627713,48% 14333,08% 2150,46% 200,04% 46556 Number of Directorships

18 utilization”, “passenger load factor”, “defect-free rate” and “R&D productivity”. This approach is derived from the study of Ittner et al. (1997). The keywords used were dependent on the firm type and the industry in which the firm is operating. For example “passenger load factor” is used as a search term for firms in the airline industry, because this nonfinancial performance measure is more applicable on firms in the airline industry than firms operating in the manufacturing industry.

3.3 Busy directors (independent variable)

To test the research hypothesis, two independent variables of interest are used. Following Ferris et al. (2003) I use the number of busy directors (NBUSYDIR) as the main variable and the proportion of busy directors in a board (PBUSYDIR) as the alternative independent variable of interest. The number of busy directors is measured at firm level. As mentioned before in paragraph 2.4, directors are considered to be busy when they hold two or more directorships (Ferris et al., 2003; Fich & Shivdasani, 2006). So, to get the total busy directors sitting on a firm’s board, I have summed up all the directors who have two or more directorship at firm level.

The alternative independent variable that is used, is the proportion of busy directors in a board (PBUSYDIR). This variable is measured by dividing the number of busy directors (this is the main independent variable) by the board size. According to Fich and Shivdasani (2006), the PBUSYDIR identifies the busyness of a certain board. This makes the PBUSYDIR variable interesting to use when analysing the extent to which busy board contributes to the use of nonfinancial performance measures. The PBUSYDIR is used in the robustness analysis. 3.4 Control Variables

In this paragraph the control variables that will be included in the empirical model are outlined. As discussed in the literature review Ittner et al. (1997) find a positive relationship between the exogenous noise in short-term financial performance measures and the weight put on nonfinancial performance measures. This means that when the noise in short-term financial performance measures increase, firms will tend to put more weight on nonfinancial performance measures. Therefore the variable for the exogenous noise in short-term financial performance measures, MNOISE, is included in the model. Following Ittner et al (1997) this variable is measured at firm level by using the Fisher z-scores for the correlation between return on assets and stock market returns.

19 The second control variable that is included in the model is financial performance.

According to Ittner et al. (1997) is the use of nonfinancial performance measures less likely in firms with a poor financial performance. They state that firms with poor financial performance tend to evaluate CEO’s on the basis of short-term financial performance measures in order to incentivize the CEO to improve the poor financial performance. This variable, FDIST, is measured by computing the probability of bankruptcy using the Altman’s (1968) model. Subsequently, firms exceeding the cut-off point of 2.675 are not likely to go bankrupt. These firms are coded 1 and firms having a score below the 2.675 are coded zero.

The third variable that is included in the model is the strategy of a firm, STRATEGY. As discussed in the literature review, firms with a prospector strategy are more likely to use

nonfinancial performance measures than firms with a defender strategy (Ittner et al., 1997). Following Ittner et al. (1997), this variable is measured by computing the ratio to employees to sales, market-to-book-ratio and ratio of R&D to sales. Firms with higher scores on these indicators are considered firms with a prospector strategy. Furthermore, firm size and leverage are also included in the model following Said et al. (2003).

3.5 Empirical model

The main objective of this study is to examine the relationship between the presence of busy directors and the reliance on nonfinancial performance measures in annual bonus contracts. Based on the literature review the following hypothesis is developed: “There is a negative

association between the presence of busy directors and the use of nonfinancial performance measures”. To test this hypothesis, the following model is estimated:

Multivariate analysis (main analysis)

- NFPMUSE = β0 +β1 NBUSYDIR +β2 MNOISE + β3 FDIST + β4 FSIZE + β5

EMP_SALES + β6 MTB + β7 R&D_SALES + β8 LEVERAGE +

ε

Since the outcome variable of this model is an indicator variable, this test is carried out by using binary outcome models, namely the logit and probit model. The binary outcome models limit the predicted probabilities between 0 and 1, while the regression model does not limit the probabilities. Therefore binary outcome models are more suitable for this empirical model. However, the regression model is also used for comparison purposes.

20 In order to support the research hypothesis, the independent variable of interest

(NBUSYDIR) has to be significant and negative. A significant and negative NBUSYDIR would mean that the presence of a number of busy directors in the board would affect the use of nonfinancial performance measures negatively and hence reduce the likelihood that a firm will use these types of measures.

Robustness analysis (additional analysis)

For the robustness analysis, the following empirical models are used. Given that the main (NFPMUSE) and the alternative (NFPMWEIGHT) dependent variables are both indicator

variables, the tests are carried out by using binary outcome models.

- NFPMUSE = β0 +β1 NBUSYDIR +β2 BOARDSIZE + β3 NBUSYDIR*BOARDSIZE +

β4 FSIZE +β5 MNOISE + β6 FDIST + β7 FSIZE + β8 EMP_SALES + β9 MTB + β10

R&D_SALES + β11 LEVERAGE +

ε

-

EMP_SALES + β6 MTB + β7 R&D_SALES + β8 LEVERAGE +

ε

-

EMP_SALES + β6 MTB + β7 R&D_SALES + β8 LEVERAGE +

ε

The variables included in the empirical models above can be defined as follows.

- NFPMUSE: main dependent variable and a dummy variable coding firms that use

nonfinancial performance measures in annual bonus contract as one and zero if this is not the case.

- NFPMWEIGHT: alternative dependent variable and contains the weight, relative to financial performance measures, placed on nonfinancial performance measures in annual bonus contracts. Firms that place a weighting on nonfinancial measures below the 10% are considered low weighting firms and coded 0. Firms that place a weighting beyond the 10% are considered high weighting firms and coded 1.

- NBUSYDIR: the number of busy directors (directors holding two or more directorship) in a board. This is the main independent variable

- PBUSDIR: the proportion of busy directors in a board (number of busy directors/board size). This is the alternative independent variable.

21 - BOARDSIZE: total board members.

- MNOISE: exogenous noise in short-term financial performance measures. Measured at firm level by using the Fisher z-scores for the correlation between return on assets and stock market returns.

- FDIST: stands for financial distress and measured by computing Altman’s (1986) probability of bankruptcy. Firms are coded as 1 when they exceed the cut-off point of 2.675. Firms that have a score below the 2.675 are coded zero.

- FSIZE (firm size): the natural logarithm of total assets.

- EMP_SALES: ratio to employees to sales, computed as follows; number of employees/sales.

- MTB: market-to-book ratio, computed as follows; market value/equity book value. - R&D_SALES: ratio of R&D to sales, computed as follows; R&D expense/sales. - LEVERAGE: (total long term debt + debt in current liability)/stockholders equity.

22 4. Results

4.1 Descriptive statistics

Table 3 shows the descriptive statistics of the sample that is used to test the research hypothesis. The table presents the mean, standard deviation and the ranges of the dependent and independent variables for observations between 181 and 195 observations. The sample was intended to cover 200 observations. However, due to missing data the sample is reduced to the aforementioned number of observations.

As shown in table 3 panel A, the mean of the dependent variable NFPMUSE is 0.415. This means that 41.5% of the sample consists of firms that use both financial and nonfinancial performance measures in annual bonus contracts. The remaining 58.5% of the firms rely solely on financial measures. Besides the table shows that the mean of the relative weight placed on nonfinancial performance measures in annual bonus contracts is approximately 12%. Firms that use nonfinancial performance measure in annual bonus contracts place, relatively to financial measures, 12% weight on nonfinancial performance measures when evaluating their CEO. The mean of NBUSYDIR indicates that an average board consist of three directors who hold two or more directorships and thus consist of three busy directors. Variable PBUSYDIR is used in the robustness analysis and expresses this number as a proportion of the board size. The mean of the variable PBUSYDIR shows that approximately 33% of an average board is considered busy as 33% of the board members hold two or more directorships.

Another variable that is shown in table 3 panel A is FDIST, which defines firms with poor and strong financial performance. The mean of this variable is 0.595 and can be interpreted as the proportion of the firms in the sample that have a strong financial performance. In this case 59.5% of the firms included in the sample have a strong financial performance. As stated in the previous section, it can be expected that the remaining 40.5% of the firms are less likely to rely on

nonfinancial performance measures when evaluating their CEO. The ratios of employees to sales, market-to-book and R&D to sales are measures to determine whether a firm has a prospector strategy or a defender strategy. Firms with high scores on these variables are considered

prospector strategy firms and tend to put more weight on nonfinancial measures. By considering the means and the medians of these variables it can be stated that the strategy of the firms in this sample varies substantially. The mean and median of these variables are respectively 0.003 /

23 0.002, 2.488 / 2.286 and 0.034 / 0.006. Moreover, the average leverage of the firms is 0.675. This means that the debt of an average firm in this sample is equal to 67.5% of the total equity.

Table 3 Panel A: Descriptive statistics

Variables N Mean SD P25 Median P75

NFPMUSE 195 0,415 0,494 0,000 0,000 1,000 NFPMWEIGHT 181 0,116 0,182 0,000 0,000 0,200 NBUSYDIR 195 3,369 2,457 1,000 3,000 5,000 PBUSYDIR 195 0,334 0,227 0,125 0,333 0,500 FDIST 195 0,595 0,492 0,000 1,000 1,000 EMP_SALES 195 0,003 0,002 0,001 0,002 0,004 MTB1 195 2,488 1,534 1,162 2,286 4,489 RD_SALES 195 0,034 0,051 0,000 0,006 0,054 F_SIZE 195 3,678 0,694 3,072 3,702 4,244 LEV 195 0,675 0,499 0,162 0,722 1,284 M_NOISE 195 0,141 0,625 -0,370 0,273 0,767

Descriptive statistics for the dependent variables nonfinancial measures (use and weight), the variables of interest busy directors (the number of busy directors in a firm’s board and the busyness of a board) a nd the control variables.

In order to analyse the descriptive statistics even more, the means of the independent variables and the control variables are compared for the use of nonfinancial performance measures. Table 3 panel B reports the results of this comparison by showing the means of the variables for firms using nonfinancial measures and firms that do not use nonfinancial measures. Besides panel B shows whether the difference in the means between these groups is significant under a significance rate of 0.05. As mentioned before 41.5% of the firms included in the sample use both financial and nonfinancial performance measures, which equals 81 firms. Panel B indicates that there is a significant difference in the means of almost all variables. Remarkable is that the mean of the number of busy directors is higher for firms that use nonfinancial measure. The latter may indicate that firms with more busy directors are more likely to use nonfinancial measures when evaluating their CEO. This is the opposite of what is hypothesized in section two. Additionally, the mean of the variable PBUSYDIR is also higher for firms using nonfinancial measures which may indicate that boards, where a higher proportion of the board members are busy, are more likely to include nonfinancial measures in the annual bonus contract. This is also the opposite of what is predicted.

Moreover, the table shows that the mean of financial distress is significantly higher for firms using nonfinancial measures. This in line with Ittner et al. (1997) as they state that firms

24 with poor financial performance are more likely to use financial performance measures instead of nonfinancial performance measures. The difference in the means of the STRATEGY ratios does not totally reflect the prediction of Ittner et al. (1997) and the prediction made in section 3. It was expected that firms with higher scores on these ratios are firms with a prospector strategy and hence more likely to use nonfinancial performance measures. Table 3 shows that the R&D to sales ratio does not have a significant effect on the use of nonfinancial measures. Besides the mean of the employee to sales ratio for firms using nonfinancial performance measures is

significantly lower than for firms that do not use nonfinancial measures. This indicates an adverse effect which is the opposite of what is predicted. However, the mean of the market-to-book ratio is significantly higher for firms using nonfinancial performance which indicates a significant effect on the use of nonfinancial measures as predicted. Furthermore, a higher mean of the exogenous noise in financial performance can be observed for firms using nonfinancial measures. This indicates that when the noise in financial performance measures increase, firms will be more likely to use nonfinancial measures. This is in line with the prediction of Ittner et al. (1997).

Finally, the same comparison as in table 3 panel B is done for the weight placed on nonfinancial performance measures (see Appendix 1). Specifically, a comparison of the variable means is made between firms placing high weight on nonfinancial performance measures and firms placing low or no weight on nonfinancial performance measures. The cut-off that is used to distinguish high from low weighting is 10%. Firms that place a weighting on nonfinancial

measures below the 10% are considered low weighting firms, while firms that place a weighting beyond the 10% are considered high weighting firms. Similarly to the results shown in table 3 panel B, the mean of the number of busy directors is higher for firms that place a high weight on nonfinancial measure (so > 10%). The latter may indicate that firms with more busy directors are more likely to place a high weight on nonfinancial measures when evaluating their CEO. This is the opposite of what is hypothesized in section two. Additionally, the mean of the variable PBUSYDIR is also higher in comparison with firms that place a low weight on nonfinancial measures. This may indicate that boards, where a higher proportion of the board members are busy, are more likely to place a high weight on nonfinancial measures. This is also the opposite of what is predicted. The results of the remaining variables are quite similar as the results shown in table 3 panel B, except for the variables market-to-book ratio, firm size and leverage. For these variables there’s no significant difference found in the means for firms using nonfinancial

25 performance measures. This may indicate that these variables do not have a significant effect on the weight placed on the nonfinancial performance measures. In order to determine the specific relationship between the variables a correlation test is conducted in the next paragraph.

Table 3 Panel B Means comparison

Variables Use of NFPM N Mean T-statistic P-value

NBUSYDIR No 114 2,825 -3,797 0,000 Yes 81 4,136 PBUSYDIR No 114 0,279 -4,193 0,000 Yes 81 0,411 FDIST No 114 0,509 -2,955 0,004 Yes 81 0,716 EMP_SALES No 114 0,003 4,016 0,000 Yes 81 0,002 MTB1 No 114 2,288 -2,174 0,031 Yes 81 2,768 RD_SALES No 114 0,034 -0,098 0,922 Yes 81 0,034 F_SIZE No 114 3,587 -2,213 0,028 Yes 81 3,808 LEV No 114 0,607 -2,258 0,025 Yes 81 0,769 M_NOISE No 114 0,036 -2,828 0,005 Yes 81 0,289

Comparison of the variable means between firms using nonfinancial performance measures and firms not using nonfinancial performance measures.

26 4.2 Correlation Matrix

In this paragraph the correlation between dependent variable, the independent variables of interest and the control variables. Consistent with the results of the mean comparison in table 3 panel B, the spearman correlation matrix of table 4 shows that the use and the weight placed on nonfinancial performance measures correlates positively and significantly (at a significance level of 0.01) with the number of busy directors. This means that the use and the weight placed on nonfinancial performance measures increases when the number of busy directors increases. The same relationship is also found for the proportion of busy directors. Table 4 shows that there is a significant positive relationship between the proportion of busy directors and the use and weight placed on nonfinancial measures. It was initially predicted that due to the busyness of directors and costly character of nonfinancial measures, the use of nonfinancial will decrease when a board of directors contains busy directors. However, the correlation matrix depicts the opposite of the hypothesized negative relationship between busy directors and the use and weighting of

nonfinancial measures. Additionally, table 4 shows a significant (significant level of 0.05) positive relationship between the use of nonfinancial measures and financial distress. This is consistent with the prediction of Ittner et al. (1997) that the use of nonfinancial performance measures is less likely in firms with a poor financial performance. However, this relationship is not significant for the weight placed on nonfinancial performance measure.

As outlined in section three, the ratios employees to sales, market-to-book and R&D to sales are used as a measure to determine whether a firm has a prospector or a defender strategy. Firms with a higher score on these ratios are considered firms with a prospector strategy and firm with a low score as firms with defender strategy. Ittner et al. (1997) state that firms with a prospector strategy are more likely to use nonfinancial performance measures. Inconsistent with this prediction, table 4 depicts a negative relationship between the ratios R&D to sales and employees to sales and the use/weighting of nonfinancial performance measures. Observable is that the R&D to sales ratio does not significantly correlate with the use of nonfinancial

performance measures. However, the market-to-book ratio correlates positively and significantly with the use and weight placed on nonfinancial measures which is consistent with prior

expectation. Taking these results together, it can be concluded that firms having a lower score on the ratios employees to sales and R&D to sales are more likely to use nonfinancial performance measures and firms with a higher score on market-to-book ratio are more likely to use

27 nonfinancial performance measures.

Moreover, the correlation matrix shows, consistent with prior literature and the expectations outlined in section three, a positive relationship between firm size, leverage and exogenous noise in financial measures and the use of nonfinancial performance measures. This association is only significant for leverage and exogenous noise in financial measures. Similarly, the weight placed on nonfinancial performance is also positively associated with firm size, leverage and exogenous noise in financial measures. However, this association is only significant for exogenous noise in financial measures. Concerning the independent variable of interest, namely the number of busy directors, there is a significant and positive association between this variable, firm size and the proportion of busy directors in a firm’s board.

28 Table 4 Spearman Correlation Matrix

NFPMUSE NFPMWEIGHT NBUSYDIR PBUSYDIR FDIST EMP_SALES MTB1 RD_SALES F_SIZE LEV M_NOISE

NFPMUSE 1 NFPMWEIGHT 0,9660*** 1 NBUSYDIR 0,2971*** 0,3156*** 1 PBUSYDIR 0,2823*** 0,3214*** 0,9629*** 1 FDIST 0,1588** 0,122 0,1244 0,1675** 1 EMP_SALES -0,2286*** -0,2064*** 0,0068 -0,0178 -0,2909*** 1 MTB1 0,2138*** 0,1887** 0,0853 0,0485 -0,0164 0,0573 1 RD_SALES -0,0248 0,0011 0,0299 0,0029 -0,1822** 0,3582*** 0,2745*** 1 F_SIZE 0,1051 0,064 0,1702** 0,1352 0,5301*** -0,4373*** 0,2023*** -0,1780** 1 LEV 0,1544** 0,0876 0,0783 0,054 0,3380*** -0,3526*** 0,2115*** -0,2888*** 0,4904*** 1 M_NOISE 0,1961*** 0,1675** -0,0116 -0,0553 -0,0964 0,0727 -0,0506 -0,0179 -0,3560*** -0,0147 1 ***P<0.01, **P<0.05, *P<0.1

29 4.3 Multivariate Analysis

As indicated in section three, NFPMUSE is used as a dependent variable and the NBUSYDIR as the independent variable of interest to conduct a regression analysis. Since the dependent variable NFPMUSE is a dummy variable (where 1 stands for the use of nonfinancial performance measures and 0 for not using nonfinancial measures), binary outcome models such as logit and probit are used to estimate the probability that firms use nonfinancial performance as a function of the number of busy directors. The regression model is less suitable as the predicted probabilities by the regression model are not limited between 0 and 1. However, the regression test is also carried out in order to compare the outcome. The results are shown in table 5 Panel A. Table 5 Panel A shows that R-squared under the regression, logit and probit model are

respectively 0.221, 0.188 and 0.185. This means that, for the regression model, 22% of the variance in the use of nonfinancial performance measures is explained by the model. For the logit and probit this is approximately 19%.

Concerning the hypothesis that firms with busy directors are less likely to use

nonfinancial performance measure, table 5 Panel A shows that the opposite of what predicted. The three models show that for each additional busy director on the board of directors, firms are more likely to use nonfinancial performance measures. This effect is significant (at 0.01) under all the models. Besides, the results of the models show a significant effect of the employee to sales ratio, market-to-book ratio and the noise in financial performance measures on the use of nonfinancial performance measures. The effect of the employee to sales ratio on the use of nonfinancial measures is negative which means that each percentage increase in the ratio, firms will be less likely to use nonfinancial performance measures. Given that firms with a defender strategy have a tendency to maximize efficiency which results in lower employees to sales ratio (Ittner et al., 1997), it can be concluded that firms with defender strategy are more likely to use nonfinancial performance measures. This is contrary to what was predicted in section. It was expected that prospector strategy firms are, relatively to defender strategy firms, more likely to use nonfinancial performance measures. Furthermore, table 5 Panel A shows that for each percentage point increase in the market to book ratio, firms are more likely to use nonfinancial performance measure. This indicates that firms that have an increasing growth and investment opportunities are more likely to use nonfinancial measures. Consistent to the expectation indicated in section 3, table 5 Panel A confirms that an increase in exogenous noise in financial

30 measures increases the likelihood that firms use nonfinancial performance measures. The latter is in line with prior literature showing that nonfinancial measures are increasingly used as a

supplement to the financial performance measures (Davila & Venkatachalam, 2004; Zuriekat et al., 2011) as they can reinforce the weak and noisy financial performance measures (Ittner et al., 1997; Feltham and Xie, 1994).

Table 5 Panel A Regression, logit and probit analysis NFPMUSE, NBUSYDIR and control var.

Regression model Logit model Probit model

NFPMUSE Coef. t Coef. z Coef. z

NBUSYDIR 0,045*** 3,31 0,221*** 3,20 0,131*** 3,23 FDIST 0,131 1,64 0,732 1,79 0,437 1,79 EMP_SALES -68,104*** -3,19 -412,110*** -3,21 -241,266*** -3,22 MTB1 0,047** 2,02 0,281** 2,28 0,164** 2,22 RD_SALES 0,172 0,25 0,973 0,27 0,424 0,19 F_SIZE -0,002 -0,03 0,019 0,05 -0,010 -0,05 LEV 0,005 0,07 -0,003 -0,01 -0,020 -0,08 M_NOISE 0,178*** 3,15 0,926*** 2,93 0,528*** 2,89 _cons 0,236 0,93 -1,448 -1,09 -0,750 -0,94 R-squared/Pseudo R2 0,221 0,188 0,185 Adj R-squared 0,187 N 195 195 195 ***P<0.01, **P<0.05, *P<0.1

In order to estimate the size of the effect the independent variables have on the use of nonfinancial performance measures the marginal effects are included in table 5 Panel B

(Appendix 2). Observable is that the marginal effects under all the models are approximately the same, except for the marginal effects of the employee to sales ratio. Table 5 Panel B shows that for each additional busy director in the board of directors firms are 4% more likely to use

nonfinancial performance measures. Furthermore, Panel B indicates that firms that increase their market-to-book ratio with one percentage point are 5% more likely to use nonfinancial measures. For an increase in exogenous noise in financial measures the likelihood of using nonfinancial measures increase with 17%.

Based on the above, it can be concluded that the research hypothesis is not supported as the correlation, regression, probit and logit test have shown that there is a significant positive relationship between the number of busy directors and the use of nonfinancial performance measures. The research hypothesis predicted an opposite relationship to what is found. An

31 explanation for this is that busy directors possess, due to their multiple directorships, expertise and experience regarding the CEO’s annual bonus contract design. Besides a director holding multiple directorships may implement the design of an annual bonus contracts based on the annual bonus contract design of another firm where the particular director holds a directorship. Hence, nonfinancial performance measures leading to the desired outcomes for first firm can be a reason for the director to implement the same nonfinancial performance measures in the second firm where he holds also directorship in order to achieve the same outcome. This is merely an intuitive argumentation. To examine whether this is correct, further research on this aspect has to be conducted. However, this is out of my research scope. Furthermore, the result that the

presence of busy directors positively affects the use of nonfinancial performance measures

suggests that busy directors indirectly enhance the monitoring effectiveness of the board. The use of nonfinancial performance measures, which is driven by the number of busy directors, provides incremental information about CEO’s actions (Ittner et al., 1997). This information makes the CEO’s action more observable for the directors and thus strengthens the monitoring effectiveness of the board.

4.4 Robustness Analysis

In this section the conducted sensitivity test is presented. In the multivariate analysis it is shown that firms with busy directors in their board are more likely to use nonfinancial

performance measures in the CEO’s annual bonus contract. However, it may be that this

likelihood is moderated by another board characteristic. To examine the latter, the variable board size is included in the models and tested for interaction with the number of busy directors

(NFPM). Core et al., (1999) state that board size is a determinant for the design of the CEO’s contract. Beside they provide evidence that the CEO compensation pay is a decreasing function of board size. Therefore it is interesting to examine whether this specific board characteristic has a moderating effect on the positive relationship between busy directors and the use of

nonfinancial performance measures. Table 6 reports the result on this test and shows that there is an interaction between the number of busy directors and the board size. This interaction is statistically significant with a p-value below the 0.05. As the coefficient of the interaction is negative, it can be stated that the interaction is negative. This means that the larger the board size the less the effect of the busy directors on the use of nonfinancial performance will be. This is illustrated in graph 1. The dependent variable (NFMPUSE) is an indicator variable where 1

32 0 .2 .4 .6 .8 1 Predic ted us e N F PM 5 10 15 20 25 30 35 BOARDSIZE NBUSYDIR=2 NBUSYDIR=4 NBUSYDIR=6 NBUSYDIR=8 NBUSYDIR=10 Predictive Margins

stands for the use of nonfinancial performance measure and 0 for not using nonfinancial

performance measures. The graph shows that firms with a small board size are more likely to use nonfinancial performance measures when they have a bigger number of busy directors in their board and vice versa. Besides the graph shows that firms with a board size of 15 are less likely to use nonfinancial measures no matter what the number of the busy sitting on their board is.

Table 6 Regression, logit and probit analysis NFPMUSE, NBUSYDIR, interaction and control var.

Regression model Logit model Probit model

NFPMUSE Coef. t Coef. z Coef. z

NBUSYDIR 0,177*** 3,48 0,2730*** 3,00 0,5395*** 3,00 BOARDSIZE 0,001 0,10 0,014 0,17 0,006 0,13 NBUSYDIR#BOARDSIZE -0,011** -1,55 -0,063** -2,21 -0,035** -2,18 FDIST 0,017 0,21 0,130 0,28 0,102 0,38 EMP_SALES -66,648*** -3,18 -417,875*** -3,19 -236,581*** -3,14 MTB1 0,059*** 2,60 0,358*** 2,76 0,197*** 2,60 RD_SALES -0,029 -0,04 -0,099 -0,03 -0,128 -0,06 F_SIZE 0,060 0,89 0,365 0,97 0,188 0,84 LEV 0,027 0,36 0,152 0,37 0,052 0,21 M_NOISE 0,220*** 3,87 1,229*** 3,53 0,681*** 3,50 _cons -0,024 -0,08 -3,003 -1,74 -1,599 -1,57 R-squared/Pseudo R2 0,273 0,235 0,228 Adj R-squared 0,233 N 195 195 195 ***P<0.01, **P<0.05, *P<0.1