Master's Thesis Student number: s3488330 Name

(1)

The Effect of General Ability of a CEO on Accruals quality and

Smoothness of Earnings

Master's Thesis

Student number: s3488330

Name: Reinier Meijerink Email: H.R.Meijerink@student.rug.nl

Supervisor: S. Wang Word count: 12,568 University of Groningen Faculty of Economic and Business

(2)

Abstract

This thesis focuses on the influence of Chief Executive Officers (CEOs) with more general skills, collected during a lifetime of work experience on the quality of accruals and the smoothness of earnings. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. The sample included firms from the Standard and Poor's 1500 in the U.S.. First, my results show that CEOs with more general skills generate financial reports with a higher quality of accruals than specialized CEOs. Furthermore, I find a positive relation between the generality of CEOs and the smoothness of the reported earnings.

(3)

1. Introduction

Stock prices fluctuate all the time. These days stock prices fluctuate due the trade war between the United States and China, Chinese firms who are accused of spying, wrong expectations of firms like ASML, etcetera. Because investors base their decisions mainly on the quality of earnings, which is influenced by the volatility in stock prices, the riskiness of investing increases (Francis et al., 2004; Francis et al., 2005). Reporting volatile earnings is of high concern to top management, who therefore will engage in earnings management to decrease the volatility (Graham et al., 2005). Other studies argue that firms mislead investors by engaging in earnings management (Gong et al., 2008).

There is done a lot of research like the influence of CEO turnover and different managerial characteristics on earnings management (El Diri, 2018). Other studies focused for example on equity-based stock option compensation. This will motivate managers to increase firm value by engaging in earnings management, so they can exercise their options (Coles et al. 2006; Ronen and Yaari 2008). Other researchers argue that firm size influences earnings management (Lee and Choi, 2002) others find no influence of firm size (Siregar and Utama, 2008). Most research is done on firm characteristics and earnings management. Less research is done about the relation between CEO characteristics and earnings management. Habib and Hossain (2013) did do some research on the influence of CEO overconfidence, CEO turnover and CEO gender on reporting quality. Other studies on this topic focused on separable characteristics of CEOs, like managerial ability, turnover, gender. However, no research focused on the historical career path of a CEO which also indicates individual characteristics. I want to explore the effect of CEO characteristics further and therefore I focus on the general ability of CEOs.

Custodio et al. (2013) developed an index which measures the general skills of a CEO, called the general ability index (GAI). This index is mainly based on the historical career aspects of these CEOs, which have certain skills attached to it. With their study they found that the importance of general CEO skills has grown over the last decades compared to specific CEO skills. The increase in importance of general skills can be caused by globalization (Cunat and Guadalupe, 2009b), industry deregulation which resulted in product market changes (Hubbard and Palia, 1995; Cunat and Guadalupe 2009b) or the increasing

(5)

effect of CEO talent on firm value due to changes in technology and management practices (Garicano and Rossi-Hansberg, 2006).

The idea that individual characteristics influence decision making derives from the upper echelon theory (Hambrick and Mason, 1984). CEOs have to endure a lot of pressure where individual characteristics are decisive when making decisions (Bonner, 2008). The upper echelon theory states that managers have to make difficult decisions, which concern many details and have high dependency. Those managers have to simplify these situations in order to make decisions, which reflects their individual characteristics (Hambrick and Mason, 1984; Finkelstein et al., 2009).

The quality of accruals is based on the estimations of accruals and the mapping of accruals into cash flows. Dechow and Dichev (2002) explain this in an example:

''Recording a receivable accelerates the recognition of a future cash flow in earnings, and matches the timing of the accounting recognition with the timing of the economic benefits from the sale. However, accruals are frequently based on assumptions and estimates that, if wrong, must be corrected in future accruals and earnings. For example, if the net proceeds from a receivable are less than the original estimate, then the subsequent entry records both the cash collected and the correction of the estimation error. We argue that estimation errors and their subsequent corrections are noise that reduces the beneficial role of accruals. Therefore, the quality of accruals and earnings is decreasing in the magnitude of accrual estimation errors'' (p.36).

As the upper echelon theory explains, the estimations managers make are influenced by their individual characteristics. This has been confirmed by other research. Demerjian et al. (2012) for example find that earnings quality increases when managerial ability is higher, which is in line with Graham et al. (2005) and Demerjian (2013b). When firms use earnings management to mislead investors this normally decreases the quality of earnings (Myers et al., 2007). That is why more able CEOs are better capable of keeping their own personal welfare and the interest of the investors of the firm satisfied (Demerjian, 2012). Graham et al. (2005) state that meeting targets will increase firm value. A CEO who has more general skills has a broader set of skills and more experience in different fields, which makes him more practiced (Custodio et al, 2013) and may increase accruals quality. Hence, my first hypothesis is as follows: 'CEOs with more general skills are associated with higher earnings quality'. CEOs need an increased focus on the investor relation and other external communications. Furthermore, accruals quality has also been included in the investor relation efforts since these external communications are mostly general in nature. General CEOs are more likely to pursue a higher accruals quality (Murphy and Zabojnik, 2007).

(6)

The earnings management levels of firms in the same industry are comparable (Bagnoli and Watts, 2000), because firms pay attention to each other when making decisions (Kallunki and Martikaine, 1999; Popp et al. 2006). According to Mcgahan and Porter (1997) and Popp et al. (2006) firms in different industries take other resources into consideration. Because of the use of other resource considerations, they manage their earnings in different ways. The similarity in making decisions between firms in the same industry and the differences among different industries can be an explanation why generalist CEOs engage more in earnings management than specialized CEOs.

Demerjian et al (2017) described several incentives to engage in earnings smoothing. For example, a generalist CEO with a good reputation has less incentives to engage in earnings smoothing because of the damage this can cause to their reputation (Demerjian et al, 2012; Fee & Hadlock, 2003). However, firms could gain from earnings smoothing due to the improvement of earnings quality by bringing reported earnings nearer to the average earnings (El Diri, 2018). Hence, my second hypothesis is as follows: 'CEOs with more general skills

will engage more in earnings management to smooth earnings'. It may be the case that more

general CEOs have broader knowledge and the experience and therefore may engage earlier in earnings smoothing than the less experienced specialized CEOs who mainly focus on the accepted pursuits in the industry.

In my thesis, I investigated the influence of general ability on the quality of accruals and the smoothness of earnings by manipulating earnings. I used the General Ability Index (GAI) of Custodio et al. (2013) to proxy the general skills of a CEO. The GAI is based on past work experience using multiple aspects of a CEOs career. The GAI consists over 21,000 firm year observations in the period from 1993-2007 including more than 4,500 CEOs and 33,200 former positions. To investigate this, I have formulated two hypotheses. First, I clarify the relation between general ability of CEOs and accruals quality. Secondly, I tested the influence of general CEO skills on the smoothness of earnings by managing earnings. After calculating the different proxies the dataset consist of 4,592 firm-year observations with 1,197 CEOs and 771 firms.

I found the following results: first the differences between generalist CEOs and specialized CEOs. When a more generalist CEO is leading the firm, the firms are mostly larger and also more leveraged. Furthermore, I tested whether the general ability of CEOs has a relation with the quality of accruals. I found that the quality of accruals increases when a

(7)

CEO has more general ability. When including control variables the relation is still holds and the explanatory factor of the model increases to 23,3%.

My subsequent test, examined the influence of the GAI on earnings smoothing where the smoothness is influenced by earnings management. I tested the smoothness based on five measures. The first four proxies include earnings management by the manipulation of real activities and the last proxy of earnings smoothing for the mapping of accruals into cash flows. I cannot conclude that general skills of CEOs influence the smoothness of earnings due to earnings management.

To the best of my knowledge, this study is the first study to look at the influence of the aforementioned CEO characteristics on the quality of accruals and the smoothness of earnings due to earnings management. I add to prior research regarding the effects of CEO skills on accruals quality and earnings smoothing, in respect to their historical career.

The structure of this thesis is as follows. Section 2 develops the hypothesis from existing literature. Section 3 discusses the data collection and methodology. Section 4 provides an overview of the empirical results from the data. Section 5 discusses the results and concludes.

(8)

2. Theoretical framework

This chapter analyzes the existing literature on real earnings management, earnings smoothing, earnings quality and the generality of CEOs.

In this research the focus was on annual data, which for example was done by Roychowdhury’s (2006) about measuring real earnings management. Seasonality leads to the provision of quarterly losses by a number of firms. By looking at annual data these losses are not that important, especially if you look at the zero earnings target at annual level. However, the yearly publication of losses have much more impact on the numerous stakeholders, therefore avoiding the reporting of annual losses are more important than quarterly losses and provides managers with higher incentives (Roychowdhury 2006).

General Ability Index (GAI)

The general ability index (GAI), developed by Custodio et al. (2013), captures the human capital of a CEO on its generalization based on a sample of 4500 CEOs out the S&P 1,500 firms. These 4,500 CEOs have had 32,500 past positions according to their résumés including non-S&P 1500 firms. To determine whether a CEO’s human capital can be defined as generalist or specialist, Custodio et al. (2013) developed an index taking into account five different aspects of their careers:

v Positions v Firms v Industries v CEO experience

v Conglomerate experience

Custodio et al. (2013) argue that these five aspects are positively related to a generalist skill set. First, if a CEO worked in different positions it is likely that this CEO has been in contact with various areas of organizations, which provide the CEO with more general knowledge about the organization. Working in numerous firms induces that a CEO has seen more different situations which had to be handled. This increases the general human capital of a CEO. The third proxy, industries, leads to more experience in different business environments. The CEO experience dummy is based on whether a CEO has worked as a CEO at another firm, because a CEO position requires more general skills. These general skills are necessary to deal with the different departments in an organization. The last proxy is the

(9)

experience in a conglomerate, which is seen as a more complex organization. A conglomerate has several divisions that are not related to each other, but need to be managed together. Working in a more complex organization will therefore increase the development of generic skills, which is why Custodio et al. (2013) added this last proxy to the index.

Earnings Qualiy

Earnings Quality (EQ) can be defined as the provision of information on the performance of a firm, where the provided information is relevant to investors' decisions making (Dechow et al., 2010). According to Dechow and Dichev (2002) the quality of earnings can be derived from the mapping of accruals into cash flows. Accruals are an estimate of future cash flows, better estimates result in a higher quality of accruals. Earnings quality can be classified into two categories, the determinants and the consequences. The researches on consequences aim at the impact earnings quality have on another variables like the cost of capital, where earnings quality is seen as the independent variable. Research on determinants studies how a variable influences earnings quality (Dechow et al., 2010), this paper can be put into this group because I look at the influence of the GAI on earnings quality. Previous research shows that investors use EQ mainly for decision making, thus higher EQ is associated with lower costs of capital (Francis et al., 2004) and the financial information is more relevant for investors (Scott, 2015). This is also in line with Francis et al. (2005) who shows that equity prices are explained by more accounting information. Therefore, it is important to create high quality earnings reports to satisfy current investors and attract new ones.

In this study I followed Dechow and Dichev’s (2002) measurement of earnings quality. They stated that the height of earnings quality depends on right estimations of accruals. For example, when a receivable is recorded there is also recognition of future cash flows involved in the organization, which is in line with the economic benefits that comes along with the sale. When the net gain of the receivable is not estimated correctly, the following entry will record both the incoming cash flow as the error made in the estimation of the original estimation. Following Dechow and Dichev (2002), I will not make a partition in the measurement of estimation errors between intentional and unintentional measurement errors because both errors will decrease the quality of earnings.

CEOs have to endure a lot of pressure in leading a firm and the strategic decision-making that comes with it. I divided this pressure in three groups. First, there is pressure from

(10)

analysts and investors to meet or beat certain targets. This pressure is important because otherwise they would think the stock price is too high and therefore they will insist on selling the stock. Secondly, CEOs experience pressure due the tendency to avoid the violation of debt covenants. According to Francis et al. (2004) analysts and investors base their decisions mainly on earnings quality, therefore it is important to make the correct accruals estimations. This pressure indicates that managers' individual characteristics are important in decision making (Bonner, 2008) and can be explained by the upper echelon theory. The upper echelon theory (Hambrick and Mason, 1984) states that individuals, like CEOs, simplify very complex and difficult situations in order to make decisions. Managers do this because they focus on a certain direction and cannot focus on every aspect. So CEOs limit their focus to that what they selectively perceive as important, instead of focus on every detail because that is not possible. The interpretation of the selected area they pay attention to will further filter the difficult situation to a simplified situation. The decisions made based on these simplifications are a good reflection of the managers’ characteristics and therefore decisions can be predicted by analyzing individual characteristics of managers (Finkelstein et al, 2009). CEOs with more general skills probably have more experience, so they know how to filter this complex information in the right way to make good decisions. CEOs who have more experience understand the economic trends and business of multiple firms and industries better. Hence, they foresee upcoming risks and trends and may anticipate in an appropriate manner, whereby accruals can be better estimated and the quality of accruals will increase. Other studies have confirmed the influence of individual characteristics of managers on decision making and firm performance (Plöckinger et al., 2016). Plöckinger et al. (2016) also indicate that there is inconsistency between different studies how certain characteristics, like experience and knowledge, influence financial reporting decisions and needs to be studied further.

Besides that the characteristics of managers have an influence on the height of EQ, higher earnings quality may also be lucrative for CEOs due the incentives they may receive when they meet certain targets. These incentives are secured in contracts to limit the agency problem. An agency problem exists when top management does not act in the best interest of the shareholders (Berle & Means, 1932; Jensen & Meckling, 1976; Zajac & Westphal, 1994). Compensation plans are invented to reduce the information asymmetry between top management and the shareholders (Davidson et al, 2004). Shareholders base their decisions mainly on the quality of the earnings (Francis et al., 2004). Thus, compensation plans incentivize CEOs to create high earnings quality to fulfil the targets set and satisfy the shareholders. Since CEOs with more general skills have broader knowledge their estimation

(11)

of accruals and the corresponding cash flows may be better, which automatically increases the earnings quality of those firms.

H1: CEOs with more general skills are associated with higher earnings quality.

2.2 Earnings smoothing

According to Gao and Zang (2015) the definition of earnings smoothing is as follow:

'' Smoothing is the reduction of variability in reported earnings that would otherwise exist in the absence of some action. Direct actions that smooth earnings commonly take the form of real strategic business decisions and cost management. The type of intentional income smoothing, depends on managerial intent. Managers application of available accounting discretion through the use of estimates, assumptions, and alternative choices, is an effective way to smooth earnings."

CEOs engage in earnings smoothing to influence investor's risk perception by decreasing the fluctuations in earnings over several years (Walker, 2013). Graham et al. (2005) shockingly discovered that managers would manipulate activities to achieve earnings targets. In their chase to achieve these targets and earn the associated compensation, it is interesting for managers to engage in earnings management (Myers et al., 2007). Ronen and Yaari (2008) stated that CEOs engage in earnings smoothing due the manipulation of real activities as well as accruals. There are three categories of earnings smoothing according to Ronen and Yaari (2008) which are: (i) beneficial, (ii) neutral and (iii) pernicious smoothing. Earnings smoothing is beneficial if the earnings are high in one period or year and the CEO expects a decrease in the future. The CEO will decrease the earnings in this period and increase it when earnings are expected to be lower, and contrariwise. This will increase the usefulness of earnings to investors because they can predict future earnings better.

Smoothing of earnings can also occur when firm performance is not as good as it should be according to the target set. Resulting in increased earnings to hide the negative news from shareholders. Necessary for this kind of earnings smoothing is the improvement in performance in the following years; otherwise earnings will be worsening (Marciukaityte and Park, 2009). This kind of earnings smoothing is called pernicious smoothing and decreases of firm value can only be averted if it is in combination with other good news (Yaari, 2005).

(12)

The third category of earnings smoothing occurs when the market is well informed and rational and the smoothing does not influence cash flows, it is then seen as the neutral form of earnings smoothing (Goel and Thakor, 2003).

In the first two categories managers make use of the information asymmetry between them and the investors. In earnings management literature it is stated that earnings management, which is not telling the truth, can only take place when the revelation principle (where no agency problem exist) is violated (El Diri, 2018). Due to the information asymmetry between the manager (agent) and the shareholders (principals) there is a demand for contracts to control the managers' decision-making behaviour. The demand for contracts are high when there is information asymmetry between the agent and the principals. The shareholders do not know if the information they perceive is correct and the managers do not know if the false information they release, if it is indeed not the whole truth, will be detected (El Diri, 2018). Especially the manipulation of real activities to smooth earnings is complex and therefore difficult to detect this kind of smoothing (Ewert and Wagenhofer, 2005). Beneficial earnings smoothing will increase stock prices and because the earnings will be smooth in the next couple of years the misrepresentations do not have to be negative to shareholders either (Sankar and Subramanyam, 2001). CEOs with more general skills have seen more diversity, like different industries, firms, operating cycles, economic fluctuation etcetera. Because they foresee operating cycles and trends in the economy they can adjust the strategy of the company in a way that positively influences the smoothness of earnings. They can create slack by, for example, reserve earnings when they foresee a decrease in earnings in the following years and a high earnings figure in the current year. The slack they have created will be released in the next years where earnings are low, so earnings are smooth. Due to their general skills they may have complicated strategies they can use to accomplish this. Therefore, to my perspective general CEOs, who have more experience and bring more knowledge into the company, will engage more in earnings smoothing. They may have more ways and ideas which are not known in that industry to smooth earnings.

H2: CEOs with more general skills will engage more in earnings management to smooth earnings.

As stated in the previous paragraph, CEOs endure a lot of pressure. There is pressure from analysts, investors and the pressure to adhere debt covenants. To avoid violations of debt covenants CEOs are likely to engage in earnings smoothing (Demerjian et al., 2017). Lastly

(13)

the investors also demand firm growth in a continually and preferably smooth way. Smooth earnings increase the predictability of the earnings which satisfies investors and analysts and volatile earnings are associated with risk because future earnings cannot be predicted (Chaney and Lewis, 1995).

In measuring earnings smoothing, I followed the method of Jung et al. (2013), however this research used other measures for earnings management. Where Jung et al. (2013) used discretionary accruals as earnings management measure; I followed Roychowdhury (2006) who made a model that measures real earnings management. Besides real earnings management as a proxy for earnings smoothing I additionally used the model developed by Dechow and Dichev (2002), who look at the mapping of accruals into cash flows. The mapping of accruals into cash flows is explained in the previous paragraph.

Real earnings management is the manipulation of real activities, like operational activities, by management to achieve certain earnings targets. In accrual based earnings management it is assumed that accruals do not affect the free cash flows. In real earnings management the assumption is that free cash flows are affected because of the changes in real activities to achieve targets (Walker, 2013).

Roychowdhury (2006) developed a measure which includes the abnormalities in three measures:

v Productions costs

v Cash flow from operations (CFO) v Discretionary expenses

To include the non-manufacturing companies in the calculations, the production costs consist of the cost of goods sold (COGS) and the change in inventory. This production cost calculation is a good reflection of the real activities because it is prevented from the manipulation of the inventory to lower the COGS. The Cash Flows from Operations are stated in the cash flow report and can be taken over. The discretionary costs exist out of three elements: (i) advertising expenses, (ii) R&D expenses and (iii) selling, general and administrative (SG&A) expenses. Which can be found in the annual report.

These proxies are positively related with the manipulations of real activities. The production costs can be manipulated to avoid decreases in earnings by sale price reductions in the fourth quarter (Jackson and Wilcox, 2000; Roychowdhury, 2005; Cohen et al, 2008) or increase production to decrease COGS (Roychowdhury, 2005; Cohen et al, 2008). The

(14)

actions taken to increase earnings in one period influences the cash flows in later periods in a negative way (Gunny, 2005), which is also in line with beneficial earnings smoothing where the one period earnings need to be increased and in another period a decrease is needed (El Diri, 2018). Bens et al. (2002) find that discretionary is a good proxy because CEOs are cutting into the R&D expenses to achieve earnings targets.

(15)

3. Data and Methodology

3.1 Sample selection

Most data used in this study is retrieved from the Compustat annual database from 1993 until 2007. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). The GAI is publicly available1. The purpose of this index is to show the generality of managerial skills of a CEO, obtained by their professional career in the period before their current CEO position. As said, the GAI index contains information about 4,500 CEOs between 1993-2007 and includes more than 21,000 firm year observations from the S&P 1,500. All together, these CEOs have more than 33,200 former positions according to their résumés (Custodio et al., 2013).

For measuring the variables earnings smoothing, earning quality and control variables I used balance sheet and income statement items from Compustat. I also retrieved the SIC (Standard Industry Classification) codes from Compustat to define the different industries. By following Roychowdhury (2006) I eliminated the firms in regulated industries and banks and institutions with SIC codes respectively 4400-5000 and 6000-6500. Furthermore Roychowdhury (2006) also requires ''at least 15 observations for each industry-year grouping'' which I followed.

After creating the different variables mentioned above, I constructed the data sample by merging all these different datasets into one to create a workable data sample. The complete dataset after merging and eliminating consist of 4,592 firm-year observations with 1,197 CEOs and 771 firms.

3.2 Methodology

In this research there are two measures I focused on. The first measure is, the quality of the earnings statements and the second measure is the manipulation of earnings, whereby smoothing is the goal. This section provides information about these dependent variables.

1

The General Ability Index is downloaded as .dta file from the following site: https://sites.google.com/site/claudiapcustodio/research

(16)

3.2.1 Accruals quality

Accruals quality is the first dependent variable that needs to be measured. I measure accruals quality based on the model developed by Dechow and Dichev (2002). This model looks at the estimation error in the expected cash flows:

(1) Where the total current accruals are represented by TCA, CFOt-1 are the lagged operating cash flows, CFOt are the operating cash flows of current year and CFOt+1 are the operating cash flows of the following year. The measurement of the total current accruals is extended by McNichols (2002) which also includes REVt and PPEt. Where REVt is the change in sales in the current year and PPEt are the values of Plant, Property and Equipment in the current year. To be able to measure accruals quality I took the standard deviation over a period of four years (from t-4 to t) of the residuals from equation (2). The higher the standard deviation, the lower the earnings quality, therefore I multiplied the standard deviation with minus one to create a situation where a higher standard deviation is indicated with higher earnings quality.

I would like to show that CEOs with more general skills truly have higher earnings quality. To test my first hypothesis I use the following regression equation:

(2)

Where AQ is the accruals quality of firm i in year t; GAI is the general ability index of firm i in year t; and

are the control variables of firm i in year t. Using this model I controlled for firm size, market to book ratio, leverage, return on assets, board independence, big four auditor, industry and year.

3.2.2 Earnings smoothing

Earnings smoothing is the manipulation of earnings to avoid volatile earnings statements over the years. If the earnings become smoother over the years, stakeholders know what to expect and this will decrease the riskiness (Chaney and Lewis, 1995; Walker, 2013). I measure earnings smoothing in five ways. In four measurements I include the proxies of real earnings management and in the last one I use the proxy of accruals quality. The accruals quality measure is explained in the previous paragraph, equation (2). To measure earnings smoothing first the proxies of real earnings management had to be constructed.

(17)

Real earnings management

First, I examined the three measurements of real activity manipulations by following Roychowdhury (2006). Roychowdhury (2006) uses three real activities and looks at the abnormalities of these activities in comparison with the normal levels. The activities included in this study are (i) production costs, (ii) discretionary expenses and (iii) cash flow from operations. The production costs are included because the costs of goods sold can be reduced due to overproduction (ABN_PROD). Earnings can be increased (decreased) by cutting in (making extra) expenses (ABN_DISEXP) and increasing (decreasing) sales by extreme discounts (no discounts) can influence cash flow from operations (ABN_CFO).

The first measurement of real activities manipulation is the abnormal level of production costs. By following Roychowdhury (2006), I define production costs as: where the normal level of cost of goods sold (COGS) and normal growth in inventory (INVT) are respectively estimated as:

(3) (4)

Where COGS are the cost of goods sold in year t, At-1 is the total assets in year t-1, S is the sales in year t, S is the change in sales between year t-1 and t and lastly St-1 is the change in sales from year t-2 to year t-1. For measuring the normal level of production costs I use the following regression:

(5)

To measure the abnormal level of production cost (ABN_PROD) I used the estimated residuals from the regression. Overproduction is more present when residuals are higher, and the decrease in COGS will increase the reported earnings.

The second measurement of real activities manipulation is the abnormal level of discretionary expenses. Following Roychowdhury (2006), discretionary expenses are the sum of R&D expenses, advertising and selling, general and administrative expenses (SG&A). Measuring the normal level of discretionary expenses (DISEXP) is done by the following regression:

(18)

In this regression I used sales (S) in year t-1 because the residuals of firms that manage sales (not even the discretionary expenses) to increase earnings are extremely low if the sales in year t are used. The abnormal level of discretionary expenses (ABN_DISEXP) are the estimated residuals from the regression. To create a situation where higher abnormal discretionary expenses are indicated with higher reported earnings I multiplied the residuals with minus one.

The third measure of real earnings management is the cash flow method (CFO). To measure the normal level of CFO I used the sales in year t and the change in sale between year t and year t-1, where regression (7) is made of. The abnormal level of CFO (ABN_CFO) is calculated by estimating the residuals of the regression.

(7)

The last measurement of real earnings management is calculated by the sum of the abnormal production costs and the abnormal discretionary expenses.

(8)

Overall, I used four different kind of measurements of real earnings management: abnormal production costs, abnormal discretionary expenses, abnormal cash flow out of operations and the abnormal productions costs plus the abnormal discretionary expenses.

Earnings smoothing

To measure the smoothing of earnings I followed the method of Jung (2013) who measured smoothing of earnings due to the manipulation of earnings (SMOOTHACT) by subtracting the smoothness based on earnings (SMOOTHXEM) from smoothness based on reported earnings (SMOOTH). Smoothness based on earnings is adjusted for performance matched discretionary accruals. SMOOTH is measured by taking the standard deviation of earnings (ibc) scaled by the standard deviation of the cash flow out of operations (CFO).

Multiplying this result by minus one, creates a situation where earnings are smoother if the value become higher. Earnings before extraordinary items is used for measuring the standard deviation of earnings and both the earnings as well as the CFO are scaled by total assets.

(19)

(10)

SMOOTHXEM of firm i in year t is measured by the standard deviation of earnings

management (EM) of firm i in year t divided by the standard deviation of CFO (CFO) of firm i in year t.

(11)

The standard deviation of earnings management (EM) of firm i in year t in equation

(11) is made in five different ways. The first one is the standard deviation of

ABN_PROD_DISEXP from equation 8. The second standard deviation derives from the

abnormal productions costs (ABN_PROD), which is calculated as the residuals out of equation (5). The third standard deviation derives from the abnormal discretionary expenses (ABN_DISEXP), which is calculated as the residuals out of equation (6). The fourth standard deviation derives from the abnormal cash flow out of operations (ABN_CFO), which is calculated as the residuals out of equation (7). The last standard deviation derives from mapping of accruals into cash flows (AQ), which is calculated from the result of equation (2). The standard deviations are calculated from the years t-4 until t. Overall I used five different measurements of SMOOTHACT, which are labelled as follows:

SMOOTH_ACT1 where the earnings management proxy is production costs plus discretionary

expenses, SMOOTH_ACT2 where the earnings management proxy is production costs,

SMOOTH_ACT3 where the earnings management proxy is discretionary expenses, SMOOTH_ACT4 where the earnings management proxy is CFO, and SMOOTH_ACT5 where

the earnings management proxy is the mapping of accruals into cash flows.

My intent was to show that CEOs with more general skills have smoother earnings due to the manipulation of earnings. To test whether the second hypothesis is true I made use of the following regression equation:

(12)

Where SMOOTHACT is the earnings smoothing of firm i in year t. As stated above I used four different kind of calculations to measure real earnings smoothing; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t.

(20)

4. Empirical results

All tables I discuss in this chapter can be found in Appendix I.

4.1 Descriptive statistics and comparison of means

I started from the Compustat Merged Database 1993-2007 and excluded regulated industries (SIC 400-4400) and the financial institutions (6000-6999) just like Roychowdhury (2006). First, I touched upon the descriptive statistics in table 2, the definitions of the variables can be found in table 1. To begin with the generality of the CEOs in this sample, the average firm has a CEO with a generality value of -.063. Besides, the median of the whole general ability index is -0.1823 and in this sample the median is -0.266, which shows that more than half of the CEOs in this sample has low general managerial skills. As it should be the mean and median of the accruals quality variable is close to zero, because this variable is the residual from equation (2).

(21)

Table 2 - Summary Statistics

Mean Median Standard

Deviation

Minimum Maximum 10th percentile 90th percentile Number of Observations General Ability Index -0.063 -0.266 0.958 -1.504 5.782 -1.168 1.235 4592 AQ -0.060 -0.046 0.061 -1.316 -0.000 -0.117 -0.017 4428 SMOOTHACT1 14.2 1.7 54.5 -58.3 1225.4 -0.259 30.625 3913 SMOOTHACT2 1.7 0.5 29.3 -1709.1 206.3 -0.824 4.029 3913 SMOOTHACT3 14.0 1.2 57.0 -97.7 1572.9 -0.596 30.955 3913 SMOOTHACT4 3.7 0.8 18.6 -300.5 811.7 -0.663 9.269 3913 SMOOTHACT5 0.1 0.1 4.0 -76.2 93.6 -1.143 1.501 3795 size 7.2 7.0 1.6 -5.0 12.8 5.357 9.202 4589 opercycle 81.0 65.7 76.5 -59.9 2032.1 14.685 158.7 3210 leverage 0.199 0.180 0.199 0.0 4.9 0.000 0.426 4577 MTB -2.057 0.378 158.6 -10737.7 12.1 0.125 0.853 4589 ROA 0.053 0.066 0.212 -7.8 1.2 -0.058 0.177 4592 ned_n 1.356 1.000 0.479 1.0 2.0 1.0 2.0 4592 bigN 0.699 1.000 0.459 0.0 1.0 0.0 1.0 4592

Table 2 - This table reports the summary statistics of the GAI, accruals quality, earnings smoothing measures and controlling variables. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I.

(22)

In table 3 I show the differences between generalist CEOs and specialist CEOs. Firms are significantly bigger in size when a generalist CEOs is appointed in comparison to firms with specialized CEOs (7.6 and 6.8). There is also a significant difference in leverage, which is calculated by the amount of total interest-bearing debt scaled by total assets (0.210 for generalist CEOs and 0.189).

The Market-To-Book value (MTB) of generalist CEOs is 0.332, versus -4.157 for specialized CEOs. The MTB value of generalist CEOs is positive which means that the average firm where a CEO has more general skills have a higher market value than the book value represents. There is also a difference in the average operating cycle, which indicates that on average a generalist CEO needs 3.2 days less to get cash, put into its operations, returned to the firm's cash account. Though, these last two differences are not significant.

In table 1 I give a definition of all the variables used. I made a separation between the accruals quality variables, real earnings management variables, smoothing variables and control variables.

(23)

Table 3 - Accruals quality, Earnings Smoothing and earnings Management: comparison of Means Generalist CEOs Specialist CEOs

D1ifference t-statistic p-value

General Ability Index 0.744 -0.772 1.516 87.192 0.000 AQ -0.056 -0.064 0.008 4.080 0.000 SMOOTH_ACT1 14.5 13.9 0.570 0.3 0.7 SMOOTH_ACT2 1.1 2.2 -1.071 -1.1 0.3 SMOOTH_ACT3 14.4 13.7 0.749 0.4 0.7 SMOOTH_ACT4 3.8 3.7 0.029 0.0 1.0 SMOOTH_ACT5 0.1 0.2 -0.109 -0.8 0.402 size 7.6 6.8 0.769 17.1 0.000 opercycle 79.309 82.524 -3.215 -1.190 0.234 leverage 0.210 0.189 0.021 3.546 0.000 MTB 0.332 -4.157 4.5 0.957 0.339 ROA 0.047 0.058 -0.011 -1.700 0.089 ned_n 1.347 1.365 -0.018 -1.292 0.197 bigN 0.702 0.696 0.007 0.491 0.623

Table 3 - This table compares the means of accruals quality, earnings smoothing measures and controlling variables between generalist and specialized CEOs. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I.

(24)

4.2 Correlation Matrix

Table 4 shows the correlations between all variables used in the regressions. Some of the dependent variables are correlated with each other, but since every variable is separately used in the regressions this causes no problems.

The variable accruals quality (AQ) is significantly positively correlated with the GAI on the 1% level with  = 0.060. This means that an increase of one standard deviation of GAI results in an increase of 6% of accruals quality.

Not all variables that aim at earnings smoothing through the manipulations of real activities are significant, but positive related to general ability, with  = 0.300,  = 0.912,  = 0.387 and  = 0.367. I hypothesize that there would be a significant relationship between real earnings smoothing and GAI, these results are not in line with my expectations.

In the correlation matrix you can see that SMOOTHACT5 have a significant relationship with the GAI and is negatively correlated with  = -0.035. This means that the smoothness of earnings in this case will decrease when a CEO has more general skills.

When looking at the control variables, I found that firm size has a significant correlation with the general ability index, AQ and earnings smoothing due to the manipulation of productions costs. The positive correlation of firm size with the GAI means that CEOs with more general skills are more often active at larger organisations. The positive correlation of firm size with the AQ indicates that bigger firms have a higher quality of accruals. The negative correlation with earnings smoothing indicates that larger firms engage less into earnings smoothing due to the manipulation of production costs and the positive correlation with SMOOTHACT5 indicates that larger firms engage more in earnings smoothing by mapping of accruals into cash flows.

There are also other control variables that have a significant correlation with the dependent variables earnings smoothing. One example is the operating cycle. The operating cycle has a negative correlation with SMOOTHACT1 (3), SMOOTHACT2 (4) and

SMOOTHACT3 (5). Variables (3), (4) and (5) are earnings smoothing variables and the longer

it takes to return cash, put into operations, back into the cash accounts, the firms will engage less in earnings smoothing due to the manipulation of production costs plus discretionary expenses.

Lastly, there is a positive and significant correlation between the return on assets (ROA) and the quality of accruals with  = 0.034. When the return on assets is higher the quality of accruals also increases.

(25)

Table 4 - Correlation Matrix Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (1) gai 1.000 (2 AQ 0.060* 1.000 (3) SMOOTH_ACT1 0.017 -0.005 1.000 (4) SMOOTH_ACT2 0.002 0.010 -0.259* 1.000 (5) SMOOTH_ACT3 0.014 -0.000 0.992* -0.337* 1.000 (6) SMOOTH_ACT4 0.014 0.055* 0.287* 0.333* 0.240* 1.000 (7) SMOOTH_ACT5 -0.035* 0.045* 0.210* -0.300* 0.253* 0.064* 1.000 (8) size 0.266* 0.116* -0.000 -0.029* 0.003 0.013 0.055* 1.000 (9) opercycle -0.021 -0.009 -0.052* -0.102* -0.050* -0.009 -0.021 -0.032* 1.000 (10) leverage 0.074* 0.100* -0.041* -0.017 -0.043* -0.013 0.001 -0.049* -0.073* 1.000 (11) MTB 0.007 0.022 0.006 0.006 0.005 0.008 0.021 0.116* 0.026 -0.032* 1.000 (12) ROA -0.034* 0.403* 0.002 0.016 0.007 0.027* 0.144* 0.155* -0.158* -0.167* 0.027* 1.000 (13) ned_n -0.003 -0.009 0.018 -0.015 0.021 0.005 0.018 -0.059* -0.009 0.015 -0.020 -0.035* 1.000 (14) bigN 0.023 0.043* 0.017 -0.005 0.017 0.004 0.033* 0.003 -0.034* 0.033* -0.010 0.000 -0.001 1.000

Table 4 - In this table the correlation coefficients are displayed between all variables. The symbol * shows significance at the .10 level. Further details about the variables can be found in table 1 in Appendix I. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI).

(26)

4.3 Accruals Quality and General Managerial Ability

Table 5 shows the results of the regression model of general ability on Accruals Quality. Reported in different columns are the regressions without and with control variables in three steps. The first column (1) shows the simple regression results of general ability on Accruals Quality. In columns (2) and (3) I included control variables. The difference between column (2) and (3) is that in column (3) I also included control variables for industry and fiscal year. As expected, all coefficients are positive and statistically significant. Column (1): =0.004 and p = 0.000, Column (2): =0.002 and p = 0.026, Column (3) =0.002 and p = 0.004. The coefficients confirm the positive influence of the general ability of CEOs on the

quality of accruals.

When adding controlling factors that influence the quality of accruals, the coefficient becomes smaller and the effect of general ability on accruals quality slightly decreases. The explanatory power of this model increases, despite of the declining effect of the control variables. The R-squared is 0.34%, 8.29% and 23.26% respectively for columns 1 to 3. When looking at the control variables, firm size influences the quality of accruals positively and significantly, just as the operating cycle and auditor controlling variables. In table 3 there is a lower value of return on assets (ROA) displayed for generalist CEOs. Despite of that given in table 3, the accruals quality is positively and significantly related with the ROA, which I had expected. To conclude, I can accept hypothesis 1, CEOs with more general skills are associated with higher accruals quality.

(27)

Table 5 - General Managerial Ability and Accruals Quality

(1) (2) (3)

AQ AQ AQ

b/p b/p b/p

General Ability Index 0.004*** 0.002** 0.002***

(0.000) (0.026) (0.004) size 0.003*** 0.003*** (0.000) (0.000) opercycle 0.000** 0.000*** (0.025) (0.000) leverage 0.049*** 0.031*** (0.000) (0.000) MTB -0.000 -0.000 (0.306) (0.500) ROA 0.058*** 0.048*** (0.000) (0.000) ned_n -0.001 0.000 (0.525) (0.993) bigN 0.005*** 0.005*** (0.002) (0.001) Industry No No Yes Year No No Yes Observations 4428 3207 3207 R2 0.004 0.083 0.233

Table 5 - This table reports the estimates of regression:

Where AQ is the earnings quality of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(28)

4.4 General Managerial Ability and Earnings Smoothing

I used five different measures of earnings smoothing, the results of the regression models of GAI on earnings smoothing are visible in tables 6-10. In tables 6-9 earnings smoothing is measured by making use of real earnings management proxies. In table 10 earnings smoothing is measured by making use of the mapping of accruals into the cash flow.

The coefficients of the simple regressions in tables 6 until 9 are all positive but not significant. As mentioned in the second hypothesis I expected that the GAI would influence the smoothness of earnings, which seems not to be the case. When adding control variables, year and industry controls included, the coefficient becomes negative for SMOOTHACT1, SMOOTHACT3 and SMOOTHACT4, displayed in table 6, 8 and 9 respectively and is still not significant. It could be that the GAI has an indirect effect on earnings smoothing and a direct effect on one or more control variables, which may explain the change from a positive to a negative coefficient. Looking at the R-squared, the explanatory factor of the correlation, of the simple regression which is 0.00% for tables 6, 7, 8 and 9. When adding control variables this increases to 17.7%, 18.3% 10.8% respectively. To conclude, the GAI has direct relation with the first four measures of earnings smoothing due to the manipulation of real activities.

When looking at the influence of the GAI on earnings smoothing due to the mapping of accruals into cash flow, showed in table 10, I see a negative and significant correlation. The coefficient is -0.144 with p = 0.032 and an explanatory factor of 0.1%. When adding control variables the coefficient becomes positive and without the control variables industry and year the correlation becomes insignificant. Including industry and year as control variable ensures that the relation is significant again and positive. The coefficient after controlling for other variables becomes 0.120 with p = 0.065. In view of the results, I cannot conclude that general ability of CEOs have an influence on earnings smoothing by the use of real earnings management or by mapping accruals into cash flows.

(29)

Table 6 - General Managerial Ability and Earnings Smoothing by production costs and discretionary expenses

(1) (2) (3)

SMOOTHACT1 SMOOTHACT1 SMOOTHACT1

b/p b/p b/p

General Ability Index 0.933 -0.009 -1.172

(0.300) (0.993) (0.198) size 0.749 0.317 (0.189) (0.569) opercycle -0.032** -0.027** (0.011) (0.041) leverage 1.823 12.998*** (0.700) (0.006) MTB -0.032 -0.012 (0.847) (0.938) ROA 4.934 12.712* (0.508) (0.074) ned_n 0.612 0.893 (0.727) (0.581) bigN 0.478 0.884 (0.790) (0.605) Industry No No Yes Year No No Yes Observations 3913 2954 2954 R2 0.000 0.004 0.177

Where SMOOTHACT1 presents the earnings smoothing due to the manipulation of production costs plus the manipulation of discretionary expenses of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(30)

Table 7 - General Managerial Ability and Earnings Smoothing by Production Costs

(1) (2) (3)

b/p b/p b/p

General Ability Index 0.054 0.141 0.171

(0.912) (0.360) (0.271) size 0.045 -0.016 (0.628) (0.866) opercycle -0.010*** -0.003 (0.000) (0.235) leverage -0.578 -0.051 (0.451) (0.949) MTB 0.021 0.016 (0.446) (0.529) ROA 2.892** 3.154*** (0.017) (0.009) ned_n 0.029 -0.033 (0.920) (0.905) bigN 0.382 0.216 (0.189) (0.459) Industry No No Yes Year No No Yes Observations 3913 2954 2954 R2 0.000 0.015 0.099

Where SMOOTHACT2 presents the earnings smoothing due to the manipulation of production costs of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(31)

Table 8 - General Managerial Ability and Earnings Smoothing by Discretionary Expenses

(1) (2) (3)

b/p b/p b/p

General Ability Index 0.816 0.078 -1.135

(0.387) (0.935) (0.213) size 0.788 0.353 (0.169) (0.526) opercycle -0.031** -0.027** (0.015) (0.044) leverage 0.714 12.558*** (0.881) (0.007) MTB -0.041 -0.020 (0.805) (0.897) ROA 5.619 13.438* (0.453) (0.059) ned_n 0.499 0.721 (0.777) (0.656) bigN 0.585 1.058 (0.746) (0.536) Industry No No Yes Year No No Yes Observations 3913 2954 2954 R2 0.000 0.004 0.183

Where SMOOTHACT3 presents the earnings smoothing due to the manipulation of discretionary expenses of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(32)

Table 9 - General Managerial Ability and Earnings Smoothing by CFO

(1) (2) (3)

b/p b/p b/p

General Ability Index 0.277 0.135 -0.059

(0.367) (0.731) (0.880) size 0.313 -0.141 (0.184) (0.556) opercycle -0.001 -0.003 (0.775) (0.621) leverage 0.009 1.457 (0.997) (0.470) MTB 0.001 0.011 (0.985) (0.866) ROA 1.952 3.424 (0.527) (0.263) ned_n 0.389 0.469 (0.592) (0.501) bigN 0.402 0.506 (0.589) (0.492) Industry No No Yes Year No No Yes Observations 3913 2954 2954 R2 0.000 0.001 0.108

Where SMOOTHACT4 presents the earnings smoothing due to the manipulation of cash flows out of operations of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(33)

Table 10 - General Managerial Ability and Earnings Smoothing by mapping accruals into cash flow

(1) (2) (3)

b/p b/p b/p

General Ability Index -0.144** 0.064 0.120*

(0.032) (0.316) (0.065) size 0.024 0.006 (0.538) (0.873) opercycle 0.000 -0.001 (0.826) (0.291) leverage 0.032 -0.536 (0.921) (0.109) MTB 0.007 0.008 (0.555) (0.491) ROA 3.570*** 3.172*** (0.000) (0.000) ned_n -0.085 -0.098 (0.475) (0.398) bigN 0.331*** 0.363*** (0.007) (0.003) Industry -0.899*** (0.001) Year -0.519* (0.067) Observations 3795 2954 2954 R2 0.001 0.024 0.102

Where SMOOTHACT5 presents the earnings smoothing due to the mapping of accruals into cash flows of firm i in year t; GAI is the general ability index of firm i in year t; and X are the control variables of firm i in year t. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). Further details about the variables can be found in table 1 in Appendix I. p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(34)

4.5 Robustness check

I performed an additional test to determine the robustness of the results. The results can be found in table 11 in Appendix I.

I tested whether the results differ when I look at the differences between year t+4 and

t-1 when a CEO turnover takes place. I first made a dummy variable for CEO turnover equal

to one when a CEO is replaced in a firm. Thereafter I made difference variables by subtracting the value in year t-1 from the value in year t+4. I took a standard deviation of year

t-4 till t so I should take t+4 to measure all the years of the new CEO and prevent the measure

from overlapping. I did this for the main dependent variable as well as for the control variables.

As the table shows, the results for earnings quality are not the same as the regression model shows in table 5. Table 11 shows that when a CEO is replaced by another CEO with more general ability the quality of accruals decreases, with a coefficient of -0.001 when all control variables but year and industry variables are included. The coefficient is -0.001 when all control variables are included. The robustness test on the variables of earnings smoothing are mostly positive and some are significant. The positive correlation means that earnings smoothing may increase when a CEO with more general skills is appointed. When a new CEO is appointed with more general skills, the earnings will be smoother due to the manipulation of real activities.

(35)

Table 11 - Robustness check Accruals Quality (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) b/p b/p b/p b/p b/p b/p b/p b/p b/p b/p b/p b/p dif_gai -0.001 -0.001 9.448 15.113* -0.205 -0.693 9.712 15.604* 1.616 2.710** 0.109 0.096 (0.748) (0.801) (0.156) (0.089) (0.839) (0.565) (0.155) (0.087) (0.147) (0.045) (0.787) (0.850) dif_size -0.008* -0.011* 6.944 0.446 -4.769** -2.043 7.155 -0.173 -1.802 -1.496 0.401 0.611 (0.085) (0.058) (0.575) (0.978) (0.013) (0.358) (0.573) (0.992) (0.384) (0.544) (0.568) (0.516) dif_MTB 0.003* 0.004** 13.062 20.805 -4.526* -0.993 14.012 19.533 -3.276 1.556 0.834 1.001 (0.079) (0.030) (0.449) (0.417) (0.088) (0.777) (0.429) (0.458) (0.256) (0.689) (0.394) (0.500) dif_leverage 0.036 -0.028 19.151 -60.109 -43.569** -18.143 17.728 -64.624 -44.339** -48.078** -4.823 -5.084 (0.532) (0.667) (0.873) (0.704) (0.019) (0.403) (0.886) (0.691) (0.029) (0.048) (0.477) (0.579) dif_ROA 0.061*** 0.041 -49.330 -11.602 13.888 1.024 -47.211 -8.280 1.598 8.760 0.323 -1.187 (0.003) (0.216) (0.432) (0.878) (0.150) (0.921) (0.464) (0.915) (0.879) (0.445) (0.927) (0.786) dif_ned_n 0.001 0.004 14.261 22.330 -1.640 -0.850 14.254 22.932 1.876 4.401* -0.992 -1.109 (0.929) (0.535) (0.266) (0.178) (0.402) (0.706) (0.279) (0.178) (0.380) (0.082) (0.172) (0.248) dif_bigN -0.025 -0.024 13.980 21.343 0.832 3.132 11.716 19.368 0.212 4.717 -1.856 -2.188 (0.107) (0.133) (0.704) (0.616) (0.882) (0.591) (0.757) (0.658) (0.972) (0.467) (0.373) (0.377)

Industry No Yes No Yes No Yes No Yes No Yes No Yes

year No Yes No Yes No Yes No Yes No Yes No Yes

Observations 141 141 108 108 108 108 108 108 108 108 107 107 R2 0.114 0.377 0.046 0.171 0.108 0.378 0.045 0.169 0.075 0.334 0.037 0.105 (1) AQ (2) AQ (3) SMOOTHACT1 (4) SMOOTHACT1 (5) SMOOTHACT2 (6) SMOOTHACT2 (7) SMOOTHACT3 (8) SMOOTHACT3 (9) SMOOTHACT4 (10) SMOOTHACT4 (11) SMOOTHACT5 (12) SMOOTHACT5

(36)

Table 11 - This table presents the estimates of regressions:

Where AQ is the accruals quality of firm i in year t+1 minus year t-1; GAI of firm i in year t+1 minus year t-1; and

are the control variables of firm i in year t+1 minus t-1; SMOOTHACT1 is the earnings smoothing of firm i in year t+1 minus t-1; SMOOTHACT2 is the earnings smoothing of firm i in year t+1 minus t-1; SMOOTHACT3 is the earnings smoothing of firm i in year t+1 minus t-1; SMOOTHACT4 is the earnings smoothing of firm i in year t+1 minus t-1; and SMOOTHACT5 is the earnings smoothing of firm i in year t+1 minus t-1. Further details about the variables can be found in table 1 in Appendix I. The sample consists of data retrieved from the Compustat annual database from 1993 until 2007 and board independence data from Boardex. This dataset contains the same years as used by Custodio et al. (2013) to develop the General Ability Index (GAI). p- values are reported in parentheses. The significance levels are designated by stars (* is the 10% level, ** is the 5% level and *** is the 1% level).

(37)

5. Discussion

This thesis aimed to give a better understanding of managerial ability on firm performance. In more detail this thesis explored the relationship between the general ability of CEOs and accruals quality and the smoothness of earnings.

Previous research, which looked at the influence of managerial ability, already found that individual characteristics are important to decision making and influence firm performance (Bonner, 2008; Plöckinger et al., 2016). According to Plöckinger et al. (2016) there is inconsistency between the results of the influence of individual characteristics, like experience and knowledge, on firm performance. In this research I investigated these characteristics further. I first examined the influence of general ability on accruals quality. General ability is represented in the General Ability Index (GAI) developed by Custodio et al. (2013). I found evidence that a CEO with more general skills is more able to generate accruals with a higher quality. An explanation for this could be that generalist CEOs have broader knowledge which helps them to better estimate the mapping of accruals into cash flows. Hence, CEOs with more general skills are more able to see trends in the economy due to their broader vision, which helps to better estimate accruals.

Furthermore, I investigated if earnings are smoother when a firm has a CEO with more general skills. I hypothesized that CEOs with more general skills can provide smoother earnings due to the manipulation of earnings because CEOs with more experience have seen more different situation where they have learned from. They can put these lessons into practice in other industries, companies etc. Also smoother earnings due to earnings management does not have to be negative for shareholders, when using the beneficial way of earnings smoothing stock price will even increase (Sankar and Subramanyam, 2001). My findings indicate no relation between the general ability of CEOs and the smoothness of earnings due to real earnings management.

Comparing tables 6 till 9 with table 10 I found that generalist CEOs are not associated with earnings smoothing due to the manipulation of real activities but there is somehow a relation with earnings smoothing due to the mapping of accruals into cash flows. An explanation can be that generalist CEOs may have more advanced strategies to smooth earnings because they have seen more diverse environments. Most of the earnings smoothing due to the manipulation of real activities are negatively correlated with general ability which points out that more generalist CEOs are not manipulating real activities or have unexplored options which cannot be measured with the Dechow and Dichev (2002) method.

Master's Thesis Student number: s3488330 Name

The Effect of General Ability of a CEO on Accruals quality and

Smoothness of Earnings

Master's Thesis

Abstract

Table of contents

1. Introduction

2. Theoretical framework

2.2 Earnings smoothing

3. Data and Methodology

3.1 Sample selection

3.2 Methodology

(2)

4. Empirical results

4.1 Descriptive statistics and comparison of means

4.2 Correlation Matrix

4.3 Accruals Quality and General Managerial Ability

4.4 General Managerial Ability and Earnings Smoothing

4.5 Robustness check

5. Discussion