CFO performance-evaluation and reporting quality, Should we explicitly reward CFOs’ on the basis of their fiduciary duties to strengthen the quality of financial statements?

(1)

CFO performance-evaluation and reporting quality

Should we explicitly reward CFOs’ on the basis of their fiduciary duties to strengthen the

quality of financial statements?

Student name: Dave Simon van Slooten Student number: 10852204

Date of final version: 25 June 2018 Word count: 13.197

Supervisor: dr. P. Kroos

MSc Accountancy & Control, variant Control Amsterdam Business School

(2)

Statement of Originality

This document is written by student Dave Simon van Slooten 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)

Abstract

This study examines the association between explicitly rewarding Chief Financial Officers’ (CFOs) based on their fiduciary responsibilities and accounting manipulations. This executive is provided with the fiduciary responsibility to monitor and report about the firms’ financial performance. Despite their key role in the reporting process, CFOs are primarily rewarded at their “self-reported’ financial performance. This situation is dictated by an asymmetric principal-agent relationship of the firms’ owners and the CFO. This thesis investigates to what degree explicitly incentivizing CFOs at their fiduciary responsibilities could prevent owners from making agency costs caused by accrual-based earnings management. Based on accounting literature, a negative relation is expected. However, regression analysis of fiduciary performance measures on the proxy of earnings management did not show a significant effect on earnings management. The findings in this thesis argue that explicitly incentivizing CFOs at their fiduciary duties has not impacted the degree of accounting manipulation. Therefore, agency costs aren’t alleviated by inclined use of fiduciary performance measures in CFO compensation structures.

(4)

1 Introduction

1.1 Background

The daily management of public firms is often done by a group of specialized managers, also called executives, executive managers or executive directors. Since the ownership of a corporation is dispersed over many shareholders, and mostly don’t engage in managing the firm, control and ownership of the corporation are separated. This separation is beneficial for the owners since it creates the possibility to have an equity portfolio that is more diversified, such is explained by the Modern Portfolio Theory (Markowitz, 1952). Secondly, the owners profit from delegating the daily management to people who are highly qualified for the job.

On the contrary, the shareholders do face a negative side effect of the separation of ownership and control, which is referred to the agency problem. Information asymmetry between managers and shareholders is widely recognized as the main cause of an agency problem. Information asymmetry arises because managers possess private information of the firm (e.g. actual financial performance, market conditions and existing organizational barriers) which is essential to potential and current shareholders for making well informed decisions of whether they should buy, sell or hold their stock. However, they do not always possess this information. This problem is referred as the adverse selection problem. Another existing asymmetric information problem regards moral hazard. Here, managers aren’t always behaving in the shareholders’ best interest but act in a way that is self-interested. For example, the consumption of perquisites at the expense of the firm, like buying luxurious airplanes that are used rarely. This behavior harms value maximization of the shareholders’ investment and is

therefore the managerial behavior that shareholders want to avoid.

These problems are mitigated by the periodical disclosure of information to investors. The mandatory annual reports in combination with supplemental information assist investors about the current and future conditions of firms. Supplemental information is provided by the firms through disclosure of, for example quarterly reports, management forecasts and conference calls. The quality of the

information is of major importance to investors because they base their decisions on this information. Therefore, a higher financial reporting quality is deemed to alleviate adverse selection and moral hazard concerns. Breaches in the financial reporting quality are perceived to be harmful. For example, an important stream of accounting research looked at the effect of earnings restatements. Earnings restatements typically lead to a drop in firms’ stock price (Palmrose, Richardson, & Scholz, 2004) and are also associated to an increased turnover (dismissal) of executive managers that are responsible for those disclosures (Hennes, Leone, & Miller, 2008, pp. 1515–1516).

(6)

Prior research points out that the CFO carries main responsibility towards activities of controlling and recording the financial performance of the firm (Geiger & North, 2006; Mian, 2001, p. 144). Geiger and North (2006) demonstrate this by finding changes in the degree of the firms’ reported discretionary accruals during the appointment and dismissal of CFOs. This finding emphasizes the amount of

influence a CFO can exercise on reporting practices and the way this elaborates on the perceived financial reporting quality. In this vein, Indjejikian and Matĕjka (Indjejikian & Matĕjka, 2009) make notice of the possible threat of using bonus incentives for CFOs when these are tied to financial performance measures. That is, rewarding CFOs on what is “self-reported” performance induces the likelihood to misreporting. Their research data showed that bonuses in CFO compensation plans are also the most used incentives, which on average are based for 50% on accounting-based financial performance measures (Indjejikian & Matĕjka, 2009). The primary financial responsibilities of the CFO seems to be the explanation that incentivizing CFOs have a stronger effect on earnings management than CEOs’ incentives have (Jiang, Petroni, & Wang, 2010).

Evaluating and incentivizing CFOs’ performance using accounting measures induces managers to an increased likelihood of earnings management. However, little is known about the effect of evaluating and incentivizing CFOs on their fiduciary responsibilities and the likelihood of earnings management. For example, the adoption and reinforcement of internal control should decrease the likelihood of earnings management (Friedman, 2014). On the account of internal control structures, the quality of information is improved and better controlled. The investigation of Hoitash, Hoitash and Johnstone (2012) showed that CFOs who were responsible for internal control material weakness disclosures of previous year were restrained in their compensation this year. This is suggesting that the supervisory board attaches significant value towards performance criteria indicative of their fiduciary

responsibilities. It remains unclear whether measures, indicative of financial reporting quality, were explicitly incorporated in the compensation plan or that firms exploited them ex-post subjectively and reduced compensation subsequent to internal control breaches. The objective of this thesis is to

determine if a relationship can be found between explicitly rewarding CFO on their fiduciary duties and the degree of accounting manipulations.

1.2 Research question

In order to address the topic which is posed in the introduction, the following research question is formulated: what is the relation between explicitly rewarding CFOs based on their fiduciary responsibilities and accounting manipulations?

It should be noted that there is tension in this research question. When relying on the premise, “what you measure and reward is what you get”, one would expect that when firms explicitly reward CFOs on

(7)

the basis of their fiduciary responsibilities, this should be associated with less accounting manipulations. However, note that, this question is examined in a cross-section of firms where only the firms that face a higher likelihood of accounting manipulation may self-select into explicitly incorporated fiduciary measures in CFO bonus plans. Furthermore, I’ve to allow for the possibility that the implicit incentives (dismissal, ex-post subjective cuts in compensation) are so powerful that explicit incentives have no supplemental disciplinary effect.

1.3 Relevance

Most of the literature on incentive compensation design looked at CEOs (e.g. Bergstresser &

Philippon, 2006; Bertrand & Mullainathan, 2001; Ertimur, Ferri, & Muslu, 2010). This thesis add value to the accounting literature in multiple ways. At first, this research shifts the scope from CEOs’

incentive compensation design towards CFOs’ incentive compensation design regarding the prevention of accounting manipulation. Most of the prior research has focused on CEO compensation.

Secondly, most of the CFO incentive compensation literature examined agency problems in relation to CFO incentives. These studies were often pointed towards the financial performance measures that were used in the CFO incentive plans (Chava & Purnanandam, 2010; Mergenthaler, Rajgopal, & Srinivan, 2009). This thesis will fulfill an empirical investigation to what extend CFO incentives are tied to a different set of performance measures. In specific, the non-financial measure regarding the

fiduciary responsibilities of the CFOs.

Despite the earlier described premise - what you measure and reward is what you get - regarding the tension that exists in the research question, this thesis contributes the accounting literature by examining the relation between the degree in which CFOs are rewarded based upon their fiduciary responsibilities and the degree of accounting manipulation. As described earlier, it is not a priori clear that the documented relation will hold as the implicit incentives that originate from dismissal may in many cases be sufficient to deter executives from manipulation. The findings of this research can provide new insights in finding a better optimal equilibrium in the benefits and costs regarding the contractual arrangements that are described by Jensen and Meckling (1976).

1.4 Structure

The structure of this thesis is organized in the following way. The second chapter contains the literature review and hypotheses. The research method is explained in chapter three and section four will

describe the results from the analyzed data. The fifth section will cover the conclusion drawn from the analyzed data.

(8)

2 Literature Review

2.1 Separation of ownership and control

Large public firms often have a capital structure that has separated the owners from the controlling activities of the firm. The owners transferred the decision rights of day-to-day management to a group of professionals (managers) that have the right set of qualities. The problems that arise through the separation of ownership and control are known as agency problems and are explained by the agency theory. Jensen and Meckling (1976) appoint the relationship between the owners and managers as an agency relationship. The authors define agency relationships as follows:

“An agency relationship as a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent.” (Jensen & Meckling, 1976, p. 308).

2.1.1 Agency theory

In this thesis, the focus is on the decision-making authority delegated to the Chief Financial Officer (CFO). This provides the CFO to private firm-specific information unlike to the owners/shareholders1

and forms an agency problem. The cause of agency problems, originates in differing perceptions of “value maximization” by both actors. That is, the CFO (agent) wants to be rewarded (maximize value) for the effort and related risk. The shareholder (principal) wants an accounting performance (maximize value) at least equal to the earnings forecast. Hereby, assuming both the agent and principal are utility maximizers. The assumption that all parties are utility maximizers but both have different interests causes conflicts about what actions create most value. To mitigate these conflicts, the principal could induce (motivate) the agent to take actions in their advantage by offering them proper incentives. Besides different interests, both parties may also have access to different information.

The decision-making authority of the CFOs gives them access to private information of the firm relative to shareholders. Shareholders are therefore dependent of information that is offered by, amongst others, the CFOs. This potentially influences the shareholders’ ability to make well informed investment decisions. These problems are called adverse selection problems.

1_{There are other stakeholders who are influenced and interested in private firm-specific information such as}

debtholders. Despite that, this thesis is primarily interested in the agency relationship between executives and shareholders (owners). That’s why only owners/shareholders are mentioned to avoid an inflation of information

(9)

Another type of an asymmetric information problem, is called a moral hazard problem. Such problems are characterized through self-interested managerial actions/decisions that decrease shareholders’ investment returns.

It is the contractual arrangement that captures what managerial behavior is needed to contribute the owners’ interests against predefined considerations towards the CFOs for mitigating agency problems. The expenditures composed of developing agency contracts and the costs - resulting from inefficiencies in the organizational and contractual design - cause a negative deviation of optimal firm value. These expenditures are defined as agency costs. Jensen and Meckling (1976) differentiate three categories of agency costs: monitoring expenditures, bonding expenditures and the residual loss. Their aim is finding the optimal contract design that has an equilibrium between the three categories and maximum firm value. This approach forms the foundation for answering the research question of this thesis. That is, applying their concept of agency costs as firms try to find a compensation structure for CFOs that minimizes agency costs by using performance measurement criteria.

2.2 Financial reporting

The monitoring expenditures are among others incurred for making it possible to produce financial statements. This specifically mitigates the agency problem for shareholders because it provides them with information about the business. In general, it is argued that an increased financial reporting quality should improve the information asymmetry between corporate insiders (executive managers) and outsiders (shareholders).

The accuracy of input data and the design of the transformation/storage process determines quality of the output. The output contains information that forms the foundation of a firms’ annual report. The CFO carries the responsibility to control these procedures and produce financial reports. It’s assumed that the CFO will act in the best interest of the firm and (among others) produce financial reports that are reliable and relevant to their users. The duties concerning the design, implementation and

monitoring of the qualitative aspects of firm performance are captured by the term fiduciary duties/responsibilities.

Since shareholders depend on the disclosed information, the quality of this information is deemed important. If disclosed information holds irregularities, shareholders make costs due to suboptimal investment decisions. Likewise, owners use financial reports to infer which actions have been taken by executive managers, and to exert judgment about the appropriateness of those actions. Because of the relevance of these disclosures, GAAP prescribes a certain level of disclosure in terms of which

information should be disclosed and the level of detail wanted to present the information. Despite the regulation, discretion remains. That is, executive managers in general, and CFOs specifically, exercise

(10)

discretion while producing financial reports. This discretion refers to estimates that have to be made and judgments that has to be exercised to produce accounting numbers. For example, the publication of the line item of accounts receivables includes the estimation of the expected costs associated with the uncollectable accounts. Likewise, valuation of different classes of assets also imply the estimation of the future cash flows to be generated by those assets as impairment of assets may be necessary. Prior research shows the impact that CFOs have on accrual policies. For example, Geiger and North (2006) showed how discretionary accruals change with the replacement of old CFOs by new ones. Likewise, Ge, Matsumoto, and Zhang (2011) report significant CFO effects for, amongst others, the choice of assumptions underlying the pension expense. Overall, the accruals that are an inevitable part of financial reporting may increase the informativeness of financial reports but may also be exploited by CFOs to serve their own interests.

2.2.1 Financial Reporting Quality

As is explained in the paragraphs’ introduction: the quality of financial reports is relevant to shareholders. Some studies provide evidence of the consequences of poor financial reporting. Palmrose, Richardson and Scholz (2004) showed a drop in stock price (average of -9 percent over a two day period) when firms had to do a restatement. This finding coincides with the association of disclosures quality and cost of capital (Botosan, 1997; Easley & O’hara, 2004; Francis, Nanda, & Olsson, 2008). High-quality disclosures, on average, lower the cost of capital (Francis et al., 2008). The questions that remain are: what defines quality, how is it accomplished and what tradeoff has to be made to realize high quality. The following section addresses these questions.

The term quality is a comprehensive concept and is represented by different notations (e.g. earnings quality) with different meanings. Dechow et al., (2010, p. 344) attributes this to the fact that quality is dependent on the decision-context. With respect to the research question in this paper, quality is limited to the qualification of an entities’ reporting practices. DeHaan, Hodge & Shevlin (2013, p. 1031) clearly defined financial reporting quality as: “the extent to which financial statements represent a diligent and unbiased application of financial reporting standards”. While American listed firms are compiled to US GAAP, high quality financial reporting is accomplished through good application of these standards. The Financial Accounting Standards Board (FASB) described qualitative characteristics for providing guidance towards the interpretation of the standards. The characteristics include relevance, faithful representation (same as reliability) and understandability.

(11)

Economic principles are the foundation of accounting standards. The qualitative characteristics are used to translate the economic principles into accounting standards of sufficient quality. Accrual accounting is one of those principles that’s interwoven in accounting standards. Although, the FASB kept the characteristics relating quality in mind at the time of developing the standards, good

application of the standards remains contingent upon how well producers of financial statements apply them. Accrual accounting is therefore, a principle that needs/asks for professional judgement of managers. The implementation of the manager regarding the standards is therefor a factor that determines the reporting quality.

Financial reports pose to inform shareholders about the actual firm performance. Performance of organizations is reflected by their reported earnings. Earnings are more accurate in measuring

performance as a result of accrual accounting. Dechow (1994) labelled earnings as a summary measure of firm performance. That is, earnings are formed by adding up the companies recorded accruals and cash flows. Several prior studies (Dechow, 1994; Dechow & Dichev, 2002) argue earnings to be a good performance measure because accruals mitigate timing and matching problems of cash flows at revenue recognition. Dechow (1994) explained accruals to be more capable in estimating future cash flows which is useful to shareholders while it has predictive value of the firms’ future performance. The predictive value is derived from the principle that cash is withdrawn from an entity (e.g. capital expenditures) when it serves the goal to maintain or increase the firm value. That is, investments are assumed to have a net present value that is equal or higher than zero. At time of making a decision to invest capital, managers assume the investment to be profitable. Though, these assumptions are based at managerial expectations and estimations. Managers are required to use discretion in these cases. This discretion is expounded by accruals and drive future performance. Cash flows on the other hand, are not susceptible to managerial judgement because they are characterized as real transactions. Firms that aren’t compelled to the codification of US GAAP (or other accounting regulation) could only report firm performance by using the recorded cash flows. This accounting treatment is called cash

accounting. Dechow (1994) performed his investigation by comparing the association of both accounting treatments to a benchmark (stock returns) of firm performance. Barth, Cram and Nelson (2001) complements the study of Dechow (1994). They proved that accrual components to be mutually different in predictive ability. Changes in the accrual components accounts receivable, inventory,

accounts payable, depreciation, amortization are appointed as accruals that significantly enhances the predictive value of earnings. This paragraph induces the following. Earnings are

better in measuring firm performance while earnings are based on accrual accounting. Organizations are required to transfer decision-rights to executive managers by allowing them to have discretion. In

(12)

response to this, managerial discretion signals private information to shareholders. Financial reports are now a better representation of actual performance. It’s hereby assumed that managers act in the best interest of the firm.

If managers are acting in the best interest of the firm than the discretion is used from an informational perspective. Though, managers could be driven by behavior and provide users of financial reports with biased information of company performance. Such activities are labeled under earnings management. Healy and Wahlen (1999) clearly describe this:

“Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” (Healy & Wahlen, 1999, p. 368) The effect of earnings management to firm value strongly depends on the motives for manipulation of accounting numbers. The purpose of allowing executives to have discretion is to improve information that is signaled to shareholders. These motives improve company value. Though, if executives behave from an opportunistic point of view, it’s destructive to firm value (Badertscher, Collins, & Lys, 2012; Healy, 1996).

The communication of shareholders and the organization strongly depends on reporting practices. That is, financial reporting serves multiple goals to shareholders (Dechow, 1994, p. 7). At first, the disclosure of financial information is used to monitor a companies’ performance. In second, the same disclosed information is used to evaluate the involvement that management held in achieving the reported

performance. This data is often closely connected the managers’ payout rate. Earnings management can therefore be redirected to the, in section 2.1 described, agency problem.

Specifically, the CFO is incentivized to more opportunistic behavior while he or she is evaluated on self-reported performance. Jiang et al., (2010) reports CFOs using a higher amounts of accruals while beating earnings forecasts. Mergenthaler, Rajgopal and Srinivasan (2011) have discovered that CFO’s were cut in bonus payments or were asked to resign when missing an analyst forecasts. Likewise, a period with extensive losses or that is effected by a decline in operating return on assets, is associated to increased turnovers of CFOs as well (Mian, 2001).

Exclusion of opportunistic earnings management enhances the predictive ability of future cash flows. The majority of studies use accrual-based measures for detection of earnings management (e.g.

(13)

The general approach within all these models measuring accrual quality is finding the part that lacks any association to economic reasoning representing shareholders best interest. Exercised discretion can be derived into accrual components (e.g. receivables) that each substantiate and verify a part of the

discretion (Barth et al., 2001). The difficulty arises in verifying and qualifying the appropriateness of the exercised managerial discretion while not all components are well traceable to a benchmark.

The accounting literature begins by separating accruals in two basic components: ‘non-discretionary accruals’ and ‘discretionary accruals’. The first one represents the amount that’s clearly traceable to industry benchmarks and this verifies the economic reasoning for exercising that part of the accrual. Discretionary accruals are the partition which are more difficult in assigning their relevancy to the firm. What drove the manager to that choice of accrual? The prior literature in this topic widely agreed that the majority of accrual choices by managers are backed up by the informational perspective, contractual perspective and opportunistic meat-or-beat perspective (Badertscher et al., 2012). The first two

perspectives are motivated to increase firm value. Badertscher et al., (2012) endorses the opportunistic perspective to executives attempting to meat-or-beat earnings benchmarks. Their findings showed that approximately two third of the restatements are opportunistically motivated which leaves the remaining restatements to be based on the informational perspective. The blurred line between the perspectives cause measurement error or ‘noise’ and harms the reliability of accruals quality, earnings management and earnings as performance measure (Francis, LaFond, Olsson, & Schipper, 2005; Guay, Kothari, & Watts, 1996; Healy, 1996; Kothari, Leone, & Wasley, 2005).

(14)

Earnings management compromises the reporting quality but the underlying discretion that can be exploited is also a necessary evil for accomplishing financial reports that are relevant to users. Organizations are able to mitigate the costs of earnings management through the implementation of solid management control systems. Systems containing well designed structures of internal controls is fundamental for producing trustworthy information and constrains attempts of earnings management (Friedman, 2014). Organizational procedures and processes need to be controllable and consist documented audit trails. The organizational structure therefore consists segregation of duties. Doyle, Ge and Mcvay (2007) examined the relation between the reporting quality and use of internal controls. They found a relation between having a weak control environment and a greater likelihood of having material weaknesses. The internal control system could prevent earnings management by mitigating the possibilities for intentional errors (misreporting) and unintentional errors (mistakes). A strong control environment influences the reporting quality in a positive way.

The definition of reporting quality is of importance for answering the research question. The essence of reporting quality is captured by qualitative characteristics: reliability and relevancy. Standard setters are forced into making a trade-off between both elements when designing accounting regulation that enhances optimal reporting quality. Although regulation prescribes accrual accounting, regulation imposes restrictions to secure the element of reliability.

2.3 The role of CFOs in financial reporting

The responsibilities of CFO’s include managerial duties and fiduciary duties. The managerial duties cover the decision-making authority regarding operational performance (Mergenthaler et al., 2011). The fiduciary duties are clearly described by Hoitash et al., (2012, p. 772) as: “Fiduciary duties involve

maintaining high quality internal controls (COSO 1992), which help ensure that a firm’s financial statements accurately represent the firm’s financial condition”. They refer to COSO 1992, but still apply for the current Integrated Framework (draft version) of COSO (2011). A primary fiduciary responsibility is therefore the

production of financial statements that are a fair representation of the firms’ actual financial condition. The study of Geiger and North (2006) showed that CFOs not only have the fiduciary responsibility regarding reporting practices but have the power to make reporting choices that have a strong influence on the financial reporting quality. The findings showed that the amount of discretionary accruals

decreased after the appointment of a new CFO. This makes the involvement of the CFO very likely, because the size of accruals changed at the time there was a change of guard. Dichev et al., (2013, p. 2) supports this finding by adding the argument that the CFOs have the knowledge to make these key

(15)

reporting decisions. Mian (2001, p. 174) pointed out the effect that CFOs have on shareholder profitability. That is, downward stock-prices preceded the two-year period of CFO dismissal. So, it’s clear that the CFO has severe influence on financial reporting quality. However, the other executive managers can also directly influence the quality, but they are less influential than the CFO. Jiang, Petroni, and Wang (2010) show how especially CFO incentives, and not so much CEO incentives, are associated with accrual-based earnings management. Kim, Li, and Zhang (2011) show how primarily CFO incentives, instead of CEO incentives, are positively associated with firms’ future stock price crash risk. The study of Friedman (2014) examines if and how the CEO can (indirectly) influence reporting quality by pressuring the CFO to manipulate accounting performance. The author cites the following about the key tasks of the CFO: “His primary action is “reporting” effort to set up and maintain a system with effective controls that produces a performance measure useful for contractually motivating the officers.” (Friedman, 2014, p. 118). The fact that the CFO is held accountable by the CEO, gives the CEO power over the CFO. This power can be used to pressure the CFO in making loopholes that allow manipulations (Friedman, 2014). Feng, Ge,

Lou and Shevlin (2011) show evidence that suggests that powerful CEOs pressure CFOs into

accounting manipulation. Firms that want to emphasize financial reporting quality can provide stronger incentives to CFOs tied to their fiduciary duties. In this way, the misreporting costs of CFOs to engage in accounting manipulation increases.

2.4 CFO incentive compensation design and financial reporting quality

Previous paragraphs explained that financial reporting quality positively influences shareholder value. Higher reporting quality decreases the agency costs with regard to the firms’ owners. The CFO is the executive manager whose primary responsibility is controlling the firms’ financial system. Therefor he/she has been awarded to special responsibilities regarding financial reporting, also referred to as the fiduciary duties. Prior research has shown how increased financial reporting quality can translate into a greater predictive usefulness and representational faithfulness of reported accounting numbers. This can ultimately translate into a lower cost of capital (Francis et al., 2008). Other studies (e.g. Geiger & North, 2006; Mian, 2001), made it plausible to assume that CFOs are most influential in making key reporting decisions/choices that affect the financial reporting quality. Firms that encounter breaches in the financial reporting quality try to restore the confidence of investors by, for example, replacing the involved CFOs by new CFOs (Hennes et al., 2008; Mian, 2001).

Firms can support the financial reporting quality through the incentive compensation design of their CFOs. The CFOs’ bonuses are often associated with CFOs’ managerial duties (Gore, Matsunaga, & Yeung, 2011). The performance criteria tied to these incentives are on average for 50% based on

(16)

financial accounting numbers (Indjejikian & Matĕjka, 2009). It is argued (Indjejikian & Matĕjka, 2009) that stronger bonus incentives lead to inclined engagement of earnings management. According the survey of Dichev et al., (2013) - handed out to CFOs to quantify the degree of earnings management - 20% of firms engage in earnings management and includes a size of approximately 10% of the reported EPS. However, a sole focus on the degree to which CFO bonus incentives are tied to accounting performance measures is incomplete. Firms can rebalance an otherwise strong inclination to manage what is effectively perceived to be self-reported performance by also providing incentives contingent on their fiduciary duties. That is, when incentives are explicitly tied to fiduciary duties, firms can

support and strengthen their fiduciary duties. For example, firms can include a qualitative assessment of the quality of internal controls in the CFO incentive compensation design. The findings of Indjejikian and Matĕjka (2009) described a decline in incentive plans tied to financial performance after the arrival of SOX. The new regulation increased the costs of misreporting and caused firms to reconsider their priorities towards value creation and optimal compensation structures. The costs of misreporting and the inclined availability of information about internal control system is a logical explanation for the declined weight of financial evaluation criteria within bonus structure. The study of Hoitash et al., (2012) found a decline in CFO compensation the year after there was a disclosure of an internal control material weakness (ICMW). This implies that there are implicit incentives for CFOs to reinforce the financial reporting quality. Specifically, firms may use ex-post discretion to decrease CFO

compensation when financial reporting quality is impaired through a breach in the quality of the internal controls.

However, firms can also decide to provide incentives for CFOs beforehand to safeguard reporting quality by explicitly rewarding their fiduciary duties. That is, ex-ante incentives provided to CFOs’ fiduciary duties will redirect CFOs effort towards those activities that are expected to reinforce financial reporting quality. Given that investments in the internal control quality decreases the susceptibility of the firm’s accounting system to manipulation, I expect a negative association between the extent that CFOs’ bonus compensation is tied to fiduciary duties and accounting-based earnings manipulation. So, my hypothesis is formulated as follows:

Hypothesis 1: CFO bonus compensation tied to fiduciary duties is negatively associated with accrual-based earnings

(17)

3 Method

3.1 Sample

To perform the research, data about reporting quality and CFO compensation structures is required. I’ve chose for firms listed at the S&P 500 stock exchange given that disclosure quality what is imperative for the hand-collection of data is better for larger firms. The implementation of SOX in 2002 led to more sophisticated reporting practices regarding internal control structures and executive compensation disclosures (Hoitash et al., 2012). Information about CFO incentive compensation is obtained from DEF 14A filings, stored in the Edgar database. Financial statement information from the Compustat North American database is used to estimate entities’ reporting quality.

At first, the annual fundamentals are extracted for calendar years 2014-2016. The firms that weren’t listed in the period from January 2015 till December 2015, had been removed. Duplicates or missing values are deleted. This led into an initial sample size of 469 firms. Subsequently, this sample must be divided into firms (i.e. a treatment group) which evaluated CFOs on fiduciary measures and firms (i.e. a control group) who didn’t do this. Given the time-consuming nature of hand-collecting data, I

randomly selected firms (in an alphabetical order) and checked their proxy statements for fiduciary measures in CFO bonus plans until the point that I retrieved 74 treatment firms. The framework of Indjejikian and Matĕjka (Indjejikian & Matĕjka, 2009) is used as guidance to identify CFO fiduciary metrics. See appendix 1 for detailed information. I checked 428 firms to retrieve the treatment group of 73 firms where the remaining 354 firms that did not explicitly evaluate CFOs on fiduciary duties

formed the control group.

The research model requires that firms in the treatment group are matched to a comparable firm from the control group. Matching is based on Standard Industry Classification (SIC) and total asset value. With respect to industry classification, a match is only accepted when the match is based on a minimum of 2-digit SIC. The second criterion requires that the control firms’ total assets value lies within 70-130% of the treatment firms’ total asset value. Some treatment firms could not be matched with a control firm and were therefor dropped from the sample. At total of 21 firms of the treatment group were dropped because of finding no match. The final sample contained 53 matched pairs (106 firms in total) of treatment firms and control firms. Extra control variables are added to the regression model to mitigate any remaining bias caused through imperfect matching. I will elaborate on this in section 3.3.

(18)

Sample Overview N Initial Sample 469 Not Examined 41 Final Sample 428 Matching N

Total No Match Matched Pairs

Total Treatment Firms 74 21 53

Total Control Firms 354 301 53

Final Sample 428

3.2 Regression model

The investigation is separated into a main model and an auxiliary model. Both models use an ordinary least squares regression. The auxiliary model produces a proxy of accrual-based earnings management which forms the dependent variable in the main regression model. The following subsections contain descriptions of the models.

3.2.1 Main regression model

The main model examines the presence of the expected negative association of the dependent variable, accrual-based earnings management (|AB_EMi,t|) by CFO’s and the use of fiduciary measures (FidMt,i)

in CFO incentive plans. Several control variables (CVi,t) are added which are discussed in section 3.3.

The scope of the research question is narrowed down to the year 2015. Performance criteria are determined at the start of years and are evaluated at the end of a year. The expected association of the independent variable on the dependent variable lies within a one-year period.

The size of earnings management is estimated by the model Kothari et al. (2005). The authors improved the Modified Jones model (Dechow et al., 1995) by matching firms on the basis of their current year ROA performance. The authors describe this as ‘performance matched discretionary accruals’. Their model starts by measuring the size of discretionary accruals for firms without matching them at performance. Modified Jones (Dechow et al., 1995) is used to determine the size of each firms’ discretionary accruals (|DA_MDJi,t|). Subsequently, firms are matched at similar ROA performance.

Both values of (|DA_MDJi,t|) are subtracted from each other which results in the size of earnings

(19)

earnings management2_(EM

i,t) is irrelevant to this investigation. That is, both directions are the result of

manipulating activities (Bergstresser & Philippon, 2006, p. 521).

The main regression model holds control variables that are transformed for improved fit.

!𝐴𝐵_𝐸𝑀_',)_{!= 𝛼 +} (−) 𝛽₁𝐷𝑢𝑚𝑚𝑦_',)₊ (−) 𝛽₆𝐿𝑛 𝐵𝑀_',)₊ (−) 𝛽_:𝐿𝑛 𝐸𝑃_',)₊ (+) 𝛽_<𝐿𝑛 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_',)_+𝜀_',)

So, on the basis of my hypothesis, I expect a negative and significant coefficient on β1. With respect to

the controls, also expectations are formulated and will be further discussed in section 3.3. With respect to the measurement of the independent variable of interest (the reliance on fiduciary measures in CFO bonus plans), two alternative measures are used.

As the main model, I use a dichotomous specification where the value is 1 if the firm uses fiduciary measures in CFO incentive plans and 0 if the firm does not. So, here a dummy variable is used to separate treatment firms (1) from control firms (0). The dummy variable only appoints firms with FidMi,t by 1 without taking the size of FidMi,t in consideration. For example, firms that have 10 percent

of their incentive plan tied to fiduciary duties or firms that only have 0,5 percent FidMi,t in their

incentive plans, are valued both equally (i.e., 1). The detection of FidMi,t in DEF-14A filings is

performed by doing accurate estimations/interpretations about occurrence. The dummy model probably less subjective as no size estimations are needed.

As a robustness analysis, I also make use of an alternative specification that besides the use of fiduciary measures in CFO incentive plans, also incorporates the magnitude of the incentives that are involved. Here, executive incentive plans of CFO’s in DEF-14A filings were inspected at presence of fiduciary metrics. The degree of performance measures that specifically evaluates a CFOs’ fiduciary duties, is recorded as percentage of the incentive plan it belongs to3_{. The definition of FidM}

i,t and the DEF-14A

structure made it impossible to capture absolute values of fiduciary measures which led into making interpretations/estimations about occurrence and size. Managerial compensation that is tied

performance is called performance-based-pay (PBP). Remuneration of PBP plans contains both

2_{For the sake of readability, the abbreviation ‘EM’ is used to indicate earnings measurement in general and}

prevent excessive use of each individual measure (AB_EMi,t, DA_MDJt). Subsequently, the vertical lines that

appoint variables/integers as absolute values are sometimes left away. Nevertheless, all earnings management metrics should be interpreted as absolute values.

3_{Executive compensation structures are often divided into a fixed (e.g. base salary) and variable component which is}

(20)

term performance (STI) and long-term performance (LTI). Both have different structures. The existence of FidMi,t in PBP plans are computed as weighted average of STI and LTI.

𝑊𝑒𝑖𝑔ℎ𝑡𝑒𝑑 𝑎𝑣𝑒𝑟𝑎𝑔𝑒_',) = 𝐹𝑖𝑑𝑀_𝑃𝐵𝑃',) 𝑃𝐵𝑃_',) = 𝐹𝑖𝑑𝑀_𝑆𝑇𝐼',) ∗ 𝑆𝑇𝐼_',) 𝑃𝐵𝑃_',)+ 𝐹𝑖𝑑𝑀_𝐿𝑇𝐼',) ∗ 𝐿𝑇𝐼_',) 𝑃𝐵𝑃_',) FidM_STI/LTI/PBP: Degree of fiduciary measures used in STI/LTI/PBP

STI: Short-term incentive plan

LTI: Long-term incentive plan

PBP: Performance-based-pay

The relation of FidMi,t to AB_EMi,t is examined for each compensation term (i.e. STI, LTI & overall PBP). I expect to find the strongest negative association within short-term incentive plans and a less pronounced effect for long-term incentive plans. Few fiduciary measures were detected in LTI plans of DEF-14A filings.

3.2.2 Auxiliary model

The auxiliary model provides a proxy of EMi,t through measurement of discretionary accruals. This

enables me to compare reporting quality of treatment and control firms. The model developed by Kothari, Leone and Wasley (2005) is used to estimate accrual-based earnings management.

The following part amplifies basic considerations towards choosing an accrual model. Followed by the argumentation for choosing the model of Kothari et al., (2005). The final part explains how the model is applied.

Managerial discretion is translated through accrual accounting into accruals. The degree in which firms employ accruals is captured as the difference between the reported earnings minus cash earnings. Only a partition of these total accrual (TA) is motivated by opportunistic meat-or-beat perspectives and are harmful to shareholder value (Badertscher et al., 2012). Mcnichols and Wilson (1988, p. 5) partition accounting accruals into discretionary accruals (DA) and non-discretionary accruals (NA):

(21)

Different approaches can be obtained by estimating DA and NA. The partition of DA assigns earnings

management. An accurate estimation/measure of DA improves the power of detecting EM4_{. The}

explanatory power of accrual models relies on multiple aspects (e.g. research setting and/or purpose for which the proxy is used).

I’ve chosen to use the performance matched discretionary accrual measure of Kothari, Leone and Wasley (2005). In contrast to the original model, I compute TA by using line items of the cash flow statement5_{. It’s argued that the calculating accruals from statement of cash flow mitigates noise. That is,}

the cash flow statement is less effected by articulation problems than the balance statement (Hribar & Collins, 2002).

Kothari et al., (2005) adjusts the Modified Jones model (Dechow et al., 1995) for improved accuracy. Firms of the treatment and control groups are matched by two-digit SIC codes and corresponding values of return on assets (ROA). The Modified Jones provides the first estimates of discretionary accruals for firms in both groups. The DA of a control firm is subtracted from the treatment firms’ DA. The part that remains, is associated to opportunistic earnings management (Kothari et al., 2005). Abbreviated as ‘AB_EMi,t’. Performance matched discretionary accrual measurement (Kothari et al.,

2005) controls for measurement imperfections in earnings management detection caused by problems of extreme financial growth.

The model is implemented in following way. At first, the total accruals are set as the difference between earnings before extraordinary items and discontinued operations (IBCi,t) and operating cash flows

(OANCFi,t). Line items are retrieved from cashflow statements (Hribar & Collins, 2002; Jones,

Krishnan, & Melendrez, 2008). TAi,t = IBCi,t – OANCFi,t

Next, a first estimate of DA is produced by using Modified Jones (Dechow et al., 1995). Accrual components are scaled by lagged total assets. Data analysis is performed cross-sectional. The sub samples for each two-digit SIC code holds a minimum sample-size of 10 observation of the same year. The final step contains matching firms by current years’ ROA. Return on assets are computed by

4_{Inaccurate estimations of DA could be caused by imprecise NA estimations. Mcnichols and Wilson (1988, p. 5)}

appoints this measurement bias as “white noise”. Their finding ascribes the importance of accurate measurement of DA and NA.

(22)

dividing income before extraordinary items (IBC) by total assets (AT) (Jones et al., 2008; Kothari et al., 2005).

Modified Jones (Dechow et al., 1995) 𝑇𝐴_',) 𝐴𝑇_',()Y1) = 𝛼1+ 𝛽1 1 𝐴𝑇_',()Y1)+ 𝛽6 𝛥𝑆𝐴𝐿𝐸_',)− 𝛥𝑅𝐸𝐶𝑇_',) 𝐴𝑇_',()Y1) + 𝛽: 𝑃𝑃𝐸𝐺𝑇_',) 𝐴𝑇_',()Y1) + 𝐷𝐴_𝑀𝐷𝐽',)

TAi,t Total accruals in current year

ATi,(t-1) Prior year total assets

∆SALEi,t Change sales in current year

∆RECTi,t Change receivables in current year

PPEGTi,t Property, plant & equipment in current year

DA_MDJi,t Modified Jones residual

Performance matched discretionary accrual measurement (Kothari et al., 2005)

ROATREATMENT » ROACONTROL

AB_EMTREATMENT = DA_MDJTREATMENT – DA_MDJCONTROL

3.3 Control variables

The aim of the study is exploring a potential association of AB_EMi,t and FidMi,t. Though, the

explained variance in the regression model could be caused by other factors as well. Additional

variables are therefore added to the main analysis to improve explanatory power of the research model and to reduce bias caused by omitted variables.

Imperfect earnings management detection could among others occur because of undetected growth effects, leading into TYPE I errors (e.g. Dechow et al., 1995). Besides undetected effects, adoption of extra variables could provide an additional explanation to the interpretation of the results. That is, market expectations, abnormal employee reductions or growth expectations could assign possible misstatements (Dechow, Ge, Larson, & Sloan, 2011, pp. 20–21).

This model controls for undetected growth effects by adding the book-to-market (BMi,t) ratio and of

earnings-price ratio (EPi,t). BMi,t and EPi,t are indicators of expected growth (Dechow et al., 2011, p. 21;

Larcker & Richardson, 2004, p. 634).

BMi,t is implemented as book value to common equity (CEQ) to market value of common equity

(CSHO * PRCC_F). The regression model uses a natural logarithm transformation (Ln) of BMi,t. A

(23)

EPi,t is implemented as the annual fiscal closing price of stocks (PRCC_F) to earnings per share (basic)

including extraordinary items (EPSPI). EPi,t is also transformed by a natural logarithm. A negative

effect of LnEPi,t to AB_EMi,t is expected (b4 < 0).

The last control variable is leverage (Leveragei,t). Leverage controls for market debt expectations.

Leverage is implemented as long-term debt plus debt in current liabilities (DLTT + DLC) to long-term debt plus debt in current liabilities plus ordinary equity (DLTT + DLC + CEQ). Increased market expectations could motivate manipulative behavior though, such is argued by Badertscher et al., (2012), it isn’t driven by opportunism. Contractual perspectives are more likely to increase shareholder value. So, a positive effect of LnLeveragei,t to AB_EMi,t is expected (b5 > 0). The natural logarithm

transformation is also implemented for Leveragei,t. Summary regression models

Main analysis (1) !𝐴𝐵_𝐸𝑀_',)_{!= 𝛼 +}𝛽₁𝐷𝑢𝑚𝑚𝑦(−) _1,)₊ (−) 𝛽₆𝐿𝑛 𝐵𝑀_',)₊ (−) 𝛽_:𝐿𝑛 𝐸𝑃_',)₊ (+) 𝛽_<𝐿𝑛 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_',)_+𝜀_',) Additional robust analysis

(2) !𝐴𝐵_𝐸𝑀_',)_{!= 𝛼 +}𝛽₁𝐹𝑖𝑑𝑀(−) _',)₊ (−) 𝛽₆𝐿𝑛 𝐵𝑀_',)₊ (−) 𝛽_:𝐿𝑛 𝐸𝑃_',)₊ (+) 𝛽_<𝐿𝑛 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_',)_+𝜀_',)

(24)

4 Results

4.1 Descriptive statistics

This chapter starts by illuminating the descriptive statistics (Table 1, panel A, B, C & D), group descriptive statistics (Table 2) and Pearson correlation matrix (Table 3).

An overview of all variables is presented by four panels (i.e. panels A, B, C & D) in Table 1. That is, panel A describes raw data; panel B is about earnings management; panel C contains the degree of fiduciary performance measures detected in DEF-14A filings (independent variable #1); and panel D describes the statistics of control variables (independent variables #2 #3 #4).

Panel A shows that the average value of (prior year) total assets of firms within the sample is ($74.830.000.000) $76.831.000.000. The return on assets is approximately 4,4%. The cash flow of operating activities has a mean value of $3.552.000.000 and result in an average income before

extraordinary items of $2.069.000.000. Managerial judgement is represented by total accruals that have a value of circa -$1.483.000.000. Panel B reports that 8,4% of these accruals are driven from an

opportunistic perspective. When using Modified Jones, I found a larger number of discretionary accruals (DA_MDJi,t=12,8%). The size of DA_MDJi,t and AB_EMi,t are in line to the numbers

documented in Kothari et al., (2005, p. 176). Hereby emphasizing, that the differences6_{of the results}

between this thesis and Kothari et al., (2005) could be explained by empirical evidence of Cohen, Dey and Lys (2008). They detected a significant decrease in discretionary accruals after Sarbanes-Oxley Act (2002) was implemented. Kothari et al., (2005) examined pre-SOX data. Post-SOX earnings

management studies had a much closer match to results in panel B (Doyle et al., 2007; Jiang et al., 2010).

There are 53 firms that use fiduciary performance criteria in CFO incentive plans (panel C). Fiduciary measures are best represented in short-term incentive plans. That is, all 53 treatment firms use fiduciary measures in STI plans while only 6 treatment firms adopted the criteria in LTI plans. The weighted average of FidM_STIi,t and FidM_LTIi,t is documented by FidM_PBPi,t. Where FidM_STIi,t weights

most heavy in the degree of FidM_PBPi,t. The mean values of FidMi,t in STI, LTI and PBP plans are

15%, 0% and 5%. The size of FidM_LTIi,t approaches zero because of winsorizing observations

smaller than the 5th_{percentile and larger than the 95}th_{percentile. Therefore, I argue that it’s plausible to}

6_{Kothari et al., (2005) found a proxy of DA_MDJ (½DA_MDJ½ ) = -0,29 (½0,29½) and a proxy of AB_EM}

(25)

assume that there is no association between rewarding CFOs at their fiduciary duties in long-term incentive plans.

Table 1: Descriptive variables

Panel A: Raw data

N MEAN SD 25% MEDIAN 75% IBCi,t $mln 106 2069 3388 413 866,5 2408 OANCFi,t $mln 106 3552 5063 796 1576 3843 ATi,t $mln 106 76831 130761 8967 27964 94408 ATi,t-1 $mln 106 74830 132590 8569 27724 86814 SALEi,t $mln 106 21943 36057 3838 7746 20563 RECTi,t $mln 106 15621 34277 594 1788 9835 PPEGTi,t $mln 106 17249 56175 760,7 2132 12383

Panel B: Measuring earnings management

N MEAN SD 25% MEDIAN 75%

Total Accruals $mln 106 -1483 2726 -1660 -618 -255 TAi,ta Lag 106 -0,04109 0,06766 -0,06574 -0,02854 -0,00783

Inverse_ATi,t-1 Lag 106 0,00008 0,00010 0,00001 0,00004 0,00012

∆SALEi,t $mln 106 -2054 14352 -488 142 446

∆RECTt $mln 106 321 3563 -129 21 270

(∆SALEi,t-∆RECTi,t) $mln 106 -2375 13772 -1044 -7 319

(∆SALEi,t-∆RECTi,t) a Lag 106 -0,00086 0,17570 -0,03237 -0,00039 0,02501

PPEGi,t a Lag 106 0,32810 0,38390 0,03537 0,20140 0,43740

ROAi,t 106 0,04350 0,08471 0,01156 0,04209 0,06771

DA_MDJi,ta Lag 106 0,12770 0,13510 0,01451 0,08732 0,19050

AB_EMi,ta Lag 106 0,08399 0,10990 0,01589 0,04146 0,10370 Panel C: Degree fiduciary metrics in incentives plan

N MEAN SD 25% MEDIAN 75%

Dummyi,t 53 1,00 0,00 1,00 1,00 1,00

FidM_STIi,t 53 0,15 0,07 0,10 0,15 0,20

FidM_LTIi,t 53 0,00 0,00 0,00 0,00 0,00

FidM_PBPi,t 53 0,05 0,04 0,02 0,04 0,06

Panel D: Control variables

N MEAN SD 25% MEDIAN 75% BMi,t 106 0,47 0,34 0,18 0,37 0,71 EPi,t 106 0,05 0,04 0,03 0,05 0,07 Leveragei,t 106 0,47 0,20 0,33 0,47 0,56 Ln BMi,t (+1) 106 0,36 0,22 0,17 0,32 0,54 Ln EPi,t(+2) 106 0,72 0,02 0,71 0,72 0,73 Ln Leveragei,t (+1) 106 0,37 0,13 0,29 0,38 0,44 a _{All values are absolute numbers.}

(26)

Several variables appear to have a somewhat skewed distributions because there is a difference between the mean and median values. Despite that, the central limit theorem describes sample data to be normally distributed if samples hold at least 30 observations (N>30) (Boston

University School of Public Health, 2016)7_{. This study has 53 observations in each group which}

allows me to assume that the sample data has no severe violations regarding normality.

Table 2: Descriptive variables for individual groups

Group descriptives Treatment Firms Control Firms Mean Difference

N MEAN SD Q2 N MEAN SD Q2 b𝜎2₌𝜎2 _T-stat

DA_MDJi,ta Lag 53 0,12590 0,13240 0,08929 53 0,12960 0,13900 0,06487 YES 0,14

AB_EMi,ta Lag 53 0,09360 0,11850 0,04649 53 0,07439 0,10080 0,03332 YES -0,90

Dummyi,t 53 1,00 0,00 1,00 53 0,00 0,00 0,00 - - FidM_STIi,t 53 0,15 0,07 0,15 53 0,00 0,00 0,00 NO -15,32*** FidM_LTIi,t 53 0,00 0,00 0,00 53 0,00 0,00 0,00 - - FidM_PBPi,t 53 0,05 0,04 0,04 53 0,00 0,00 0,00 NO -9,76*** BMi,t 53 0,47 0,33 0,38 53 0,46 0,35 0,34 YES -0,15 EPi,t 53 0,05 0,04 0,05 53 0,06 0,04 0,05 YES 1,09

Leveragei,t 53 0,43 0,18 0,39 53 0,50 0,21 0,49 YES 1,90**

Ln BMi,t (+1) 53 0,36 0,21 0,32 53 0,35 0,23 0,29 YES -0,24

Ln EPi,t(+2) 53 0,72 0,02 0,72 53 0,72 0,02 0,72 YES 1,08

Ln Leveragei,t (+1) 53 0,35 0,12 0,33 53 0,40 0,14 0,40 YES 1,86**

* Significant at 0,10, ** Significant at 0,05; *** Significant at 0,01

a _{All values are absolute numbers;}b_{Pretest equal variances at α = 0,10}

The descriptive statistics of all variables in the regression model are also documented by group (i.e. treatment group & control group) in Table 3. This provides a first impression about the differences between both groups. Logically, the fiduciary performance measures in CFO

incentive plans differ significantly (i.e. FidMi,t is set as indicator variable). However, the earnings

management proxies are not deviating between groups. Which suggests that rewarding CFOs at their fiduciary duties isn’t affecting the average reporting quality in both groups. Regarding the influence of the additional variables (BMi,t, EPi,t and Leveragei,t), only a companies’ leverage

differs significantly between groups. That is, 43% of treatment firms are financed with debt in contrast to 50% of debt in control firms.

(27)

Table 3: Control variables and their behavior to AB_EMi,t

Control variables AB_EMi,t Mean Difference

N Low N High a𝜎2₌𝜎2 T-stat

Book-to-market (BMi,t) 53 0,11759 53 0,05040 NO -14,83***

Earnings-to-price (EPi,t) 53 0,11213 53 0,05586 YES -11,44***

Leverage (Leveragei,t) 53 0,07739 53 0,09060 NO -13,24***

Besides comparing variables between groups, more insight in the behavior of the control

variables and AB_EMi,t is provided in Table 3. The data within the table is stratified8 in ‘Low’ and

‘High’ scores of BMi,t, EPi,t and Leveragei,t. As can be seen in the table, the degree of earnings

management deviates within each control variable when these variables are stratified in “Low” and “High” groups. The findings describe that low BMi,t ratios and low EPi,t ratios have higher

values of AB_EMi,t then high ratios. The behavior of Leveragei,t is as expected. High leveraged

organizations are more susceptible to accrual-based earnings management. These findings

correspond with findings of other earnings management studies (Badertscher et al., 2012; Larcker & Richardson, 2004) arguing that discretionary accrual choices are affected by a firms’ debt structure and growth effects.

Another aspect is visualized by Table 3 as well. That is, bias is caused by measurement error in

AB_EMi,t. When taking the group means of AB_EMi,t of both the LOW and HIGH the group,

together, you see that the mean value of AB_EMi,t deviates from zero and thus violates the

condition for unbiased regression output (i.e. non zero mean error). Values are biased upward which is referred as type I errors. This means that a partition of discretionary accruals is unfairly allocated as earnings management (wrongly reject the hypothesis of ‘no earnings management’). The findings correspond with the paper of Kothari et al., (2005).

The output in Table 3 substantiates the different behavior of AB_EMi,t along effects of growth

(debt structures) and pinpoints the limitation of the used proxy. The readers of this paper should take notice of this when interpreting the answer to the hypothesis.

(28)

Table 4: Pearsons' Correlation Matrix Variables AB_EMt i, DA_MDJi, t Dummyi,t FidM_STIi, t FidM_LTIi, t FidM_PBPi,

t LnBMi,t LnEPi,t

LnLeveragei, t AB_EMi,t 1,00 DA_MDJi,t 0,38*** 1,00 Dummyi,t 0,09 -0,01 1,00 FidM_STIi,t 0,00 -0,08 0,83*** 1,00 FidM_LTIi,t - - - FidM_PBPi,t -0,10 -0,19* 0,69*** 0,89*** - 1,00 LnBMi,t -0,32*** -0,36*** 0,02 0,11 - 0,21** 1,00 LnEPi,t -0,37*** -0,17* -0,11 -0,04 - 0,10 0,34*** 1,00 LnLeveragei,t 0,16 0,10 -0,18* -0,24** - -0,17* -0,22** 0,19* 1,00

The Pearson correlation matrix is described in Table 4. The dependent variable has a weak correlation with the main independent variables: r(Dummyi,t)=0,09 and r(FidM_PBPi,t)=-0,10.

The direction of the association with the dummy variable differs from earlier made predictions. The reported positive correlation in Table 4 was expected to be negative. The sign of

FidM_PBPi,t is in line to what was expected. No correlation with FidM_STIi,t is detected et al.

Not one of the main independent variables is tested significant.

A moderate and significant association between AB_EMi,t and the control variables is measured.

That is, r(LnBMi,t)=-0,32*** and r(LnEPi,t)=-0,37***. The direction of the observed correlation is

in line with earlier made predictions. No significant correlation is reported for LnLeveragei,t

(r=0,16).

Visual inspection of the matrix signals possible threats of multicollinearity. Collinearity between LnBMi,t and Ln EPi,t is r=0,34***. Other explanatory variables have coefficients of approximately

r=+/- 0,20. Subsequent tests such as the variance inflation factor (VIF) and tolerance (1/VIF) are computed to rule out remaining treats for multicollinearity. The average VIF won’t transcend values of 1,19. Tolerance is approximately 0,85. These test results show no severe problems of multicollinearity (Bedard, Hoitash, Hoitash, & Westermann, 2012, p. 40).

4.2 Outcomes hypothesis testing

The hypothesis tested in this thesis is formulated as: CFO bonus compensation tied to fiduciary duties is negatively associated with accrual-based earnings management. Which means, it’s expected that earnings management will reduce when firms increase evaluation of their CFO’s at fiduciary

(29)

regression results are described in Table 5. Additional information about the regression diagnostics is included in appendix 2.

4.2.1 Results from main regression model

The dummy model was predicted to have a negative effect to accrual-based earnings

management (AB_EMi,t) when Dummyi,t=1. Thus, b1Dummy1,t<0. Table 5 shows that the model

is able to explain approximately 22% (R2_{»22%; F-stat=7,13***). The model is tested significant at}

a confidence level of 99%. The slope coefficient of b1Dummy1,t=0,02 (t=1,02) which is a

surprising outcome. The direction of the association is opposed to the earlier made prediction. In spite of that, the effect of b1 isn’t significant. Meaning that, when Dummyi,t=1, no changes are

explained in AB_EMi,t because b1=0. The degree of misreporting is equal for both the treatment

group and the control group. The output of the independent t-test in Table 2 regarding AB_EMi,t

corresponds with the findings in Table 5.

The expected side effects from the control variables are supported by the regression output. Each control variable in the main analysis reports a significant – though at different confidence levels – association with AB_EMi,t. The signs are in line with expectations as well. The

documented coefficients are: b2LnBMi,t=-0,08, b3LnEPi,t=-2,03 and b4LnLeveragei,t=0,17. These

results correspond with accounting literature. Higher market expectations (i.e. low book-to-market ratios and low earnings-price ratios) increases the likelihood of CFOs to inclined

accounting manipulations (Badertscher et al., 2012; Larcker & Richardson, 2004). Badertscher et al., (2012) refers this by ‘meet-or-beat behavior’. This paper also argues that managerial discretion which is contractually motivated is performed in favor of shareholder value (e.g. achieve optimal debt covenants). The leverage ratio in Table 5 shows an incline (decline) in earnings management (reporting quality) however, this discretion is contractually motivated and will therefor benefit shareholders.

When relying on the output, the adoption of FidMi,t in CFO incentive plans is not causing a

change in AB_EMi,t. All control variables are causing a change in AB_EMi,t. So, based on this

dummy model, the hypothesis is rejected. Though, there are some aspects constrain the validity of the answer.

At first, it isn’t clear if these side-effects (i.e. control variables) interact with the main independent variable. For example, firms with low BM ratios have more earnings management then high BM firms have. That is, low BM firms must meet higher market expectations. This makes it plausible to suggest that these low BM firms have self-selected fiduciary evaluation criteria because they are

(30)

were argued to have stronger governance (Dey, 2008). The sample of firms in my investigation could therefore be limited by reverse causality. A similar argumentation is applicable to the EP and Leverage variable. The power to measure the actual/real relation between FidMi,t and

AB_EMi,t could therefore be restricted (type II error).

At second, the earnings management proxy contain some measurement error (i.e. noise) that can lead into type I errors by allocating discretionary accruals towards opportunistic earnings

management (Kothari et al., 2005). Noise within AB_EMi,t limits the quality for the examined

association between FidMi,t and AB_EMi,t.

Table 5: Results main regression model AB_EMi,t

F-statistic 7,13***

R2 _22,02%

adj-R2 _18,93% _{AB_EM}_i,t

Variables Sign N df 𝛽 St. error T-stat 95% C.I. St.beta

Constant 106 4 1,5027 0,4018 3,74*** 0,7056 2,2998 - Dummyi,t - 106 4 0,0200 0,0196 1,02 -0,0189 0,0588 0,0913

Ln BMi,t - 106 4 -0,0818 0,0488 -1,68* -0,1786 0,0149 -0,1644

Ln EPi,t - 106 4 -2,0335 0,5760 -3,53*** -3,1760 -0,8910 -0,3445 Ln Leveragei,t + 106 4 0,1666 0,0791 2,11** 0,0097 0,3235 0,2002

4.3 Robustness analysis

The dummy in the main regression model is able to provide a rough answer about the hypothesis. That is, a dummy can only bear values of 0/1 and aren’t applicable to explain

different gradations of the underlying variable (i.e. FidMi,t). While the variable Dummyi,t isn’t able

to mimic the position of fiduciary responsibilities within compensation structures of CFO’s, I perform an additional, robust analysis. Hereby, the estimated weightings of FidMi,t within

short-term incentive plans (STI), long-short-term incentive plans (LTI) and performance-based-pay plans (PBP) are added to the regression. The predicted effects of FidM_STIi,t, FidM_LTIi,t and

FidM_PBPi,t in AB_EMi,t are similar to that of dummy variable in the main regression model.

Thus, the more CFO’s are rewarded for their fiduciary responsibility, the less practices of

earnings management are conducted. The following three sub-sections describe the outcomes of AB_EMi,t and FidMi,t in STI and PBP plans. Regression results of these robust analysis are

CFO performance-evaluation and reporting quality, Should we explicitly reward CFOs’ on the basis of their fiduciary duties to strengthen the quality of financial statements?