Earnings management : the influence of gender diversity

(1)

Earnings management: The influence of gender

diversity

Bachelor Thesis 07-08-2013

Nina Dréau (6149642)

Specialization Finance & Organisation

Field: Organisational Economics

Supervisor: Eszter Czibor

(2)

Abstract

In this research paper I extend the existing literature on Earnings management by testing for the marginal effect of including more female board members on the level of Discretionary accruals. Using four regression tests, based on the Modified Jones model (1991), no significant change in effect can be proved. Neither is there enough evidence to conclude female participation has an effect on board meeting attendance. Interesting is the significance of the effect of at least three female board members. With a significance level of 5%, the coefficient indicates an gender diversity indeed is related to earnings management when at least three female directors are part of the board. Future research might focus on the participation of at least four or at least five female members, test for a possible declining effect and see what is the perfect gender division on the board.

Key words: Earnings management, Gender diversity, Director attendance, Discretionary accruals and Modified Jones Model

(3)

1. Introduction

In 2003, Norway was the first to set a quota of at least 40% of the board members being of each sex (Doldor et al., 2012). Nowadays more and more governments question whether or not such quotas are a useful way to increase the quality of earnings; based on the assumption more women on the board would have impact on the corporate governance and profitability (Adams & Ferreira, 2009; Campbell & Mínguez-Vera, 2008).

The goal of this paper is to extend the existing evidence on the relation between female board presence and earnings quality by testing on a decreasing marginal effect of gender diversity on earnings quality. Future research might use the results when testing quotas set for the minimal required number of women on the board.

Onwards from the 1990s, much research has been done on earnings management and the question what is the contribution of female presence on the board of directors. Croson and Gneezy (2009) studied the gender differences with respect to preferences and conclude women are more risk averse than men, women’s preferences are adapted more to the situation and men are more competitive than women. The relation of gender diverse boards and firm value, stock returns, or earnings quality can be found in several articles, such as Nguyen, Locke and Reddy (2012). In their article, Nguyen, Locke and Reddy have found that the presence of female directors in the board have significant influence on the company’s performance. These findings raise the question whether a reduction of earnings management is partly the reason for this improvement of performance.

The sample used in this study contains 3946 unique firm years, and while the average number of directors on the board is 9.3 per company, the average female board members is only 1.2 which is a good reference to the scare number of women in top positions. This is in line with earlier studies such as Braakman (2011) and Terjesen, Sealy and Singh (2009). These papers question why women fill only a small percentage of the highest managerial positions. As Adams and Ferreira (2009) state in their article on women in the boardroom and their impact on governance and performance, the lack of women in top positions is going to change because boards around the world are under increased pressure to choose female directors. The study of the European Professional Women's Network (EPWN, 2004) argues firms are more likely to have one female board participant nowadays, but only a small range of firms have more than one.

The Modified Jones model is used to test for the actual existence of a significant relation between the level of earnings management and female participation. The first regression is used to test whether female participation has an effect on earnings management in the first place. The

(5)

second and third regressions are used to test if this effect reduces when replacing female

participation for at least two, or at least three female board members. A final control test is done to check if the board meeting attendance is affected by female presence.

The results indeed show a decreasing level of earnings management with an increase in female presence, but the effect is only considered to be significant with three or more female board members. Since this study focused on the change of the effect of gender diversity on earnings quality for none, at least one, at least two or at least three female board members future research might extend this study for at least 4 and at least 5 female members to test whether there is an actual declining or stagnating effect.

This research paper will continue with a literature review in section 2, the methodology and model in section 3, the data and results in section 4, the discussion of the results in section 5 and finally the discussion and conclusion in section 6.

2. Literature review

In this section the relevant literature will be studied in detail and important terminologies will be explained. Paragraph 2.1 gives a theoretical description of earnings management and earnings quality, followed by some background information on ethical behaviour and gender diversity in paragraph 2.2. Once the theoretical background of these terminologies has been discussed, paragraph 2.3 reviews prior studies and methods comparable to this study on earnings quality and the presence of female board members. Finally the hypotheses are defined in paragraph 2.4.

2.1 Earnings management and earnings quality

Not all components of earnings are reflected in current cash flows and the firms assets not reflected in the current cash flows are called accruals. Partly due to an increase in stock-based CEO

compensation the level of accruals has been increasing significantly since 1995 (Bergstresser and Philippon, 2006). The process of using accruals to increase or reduce the reported income is referred to as earnings management. In their article on accruals quality Francis et al. (2005) mention the difference between accruals quality driven by economic fundamentals versus management choices. Accruals driven by management fundamentals are referred to as discretionary accruals.

Dechow, Sloan and Sweeney (1995) are the first to compare five different models used to calculate discretionary accruals, as did Guay, Kothari and Watts in 1996. The five models evaluated in their research are the models of Healy (1985), DeAngelo (1986), Jones (1991), the Modified Jones model (1995) and the Industry model proposed by Dechow and Sloan (1991). Only the Jones and

(6)

Modified Jones model seems to provide reliable estimates of discretionary accruals (Guay, Kothari and Watts, 1995). The original Jones model attempts to include firm-specific changes due to

economic circumstances in the model to define the non-discretionary accruals. The modified version extends this model by subtracting the change in revenues to control for discretion exercised over the revenue (Dechow, Sloan and Sweeney, 1995).

In 2002 Dechow and Dichev introduced a new measure for the quality of accruals, using the volatility of accruals and the volatility of earnings. They assume accruals to be temporary

adjustments that resolve timing problems in the fundamental cash flows. Accruals quality is based on the component of the total accruals explained by the realized cash flow and this quality is measured using the standard deviation of the error term (Dechow and Dichev, 2002). McNichols suggests extending the Dechow and Dichev model, often referred to as the DD model, by including specifications of the Jones model (1991). The first option is to carefully identify the cash flow generating process, the information available to management, and management’s estimation task including the consequences. The second option suggested by McNichols is to focus on specific accruals rather than aggregate accruals.

Burgstahler, Hail and Leuz (2004) use the proxies for earnings management designed by Leuz et al. (2003), since these proxies take into account the economic situation of the firms. The proxies mentioned are smoothing reported operating earnings, smoothing and the correlation between changes in accounting accruals and operating cash flows, small loss avoidance and the magnitude of accruals. More recent research such as Krishnan and Parsons (2008) used similar attributes to calculate the level of earnings management.

2.2 Ethical behaviour and gender diversity

It is important to define ethical behaviour for which Trevino, Weaver and Reynolds (2006, pp. 953) state: ‘behavioural ethics refers to individual behaviour that is subject to or judged according to generally accepted moral norms of behaviour.’ Studies of Bernardi and Arnold (1997) and Betz et al. (1989) suggest that women are less likely to engage in unethical behaviour to gain financial rewards. In 1994 Ford and Richardson concluded eight out of thirteen studies concerning ethical behaviour and gender suggest there is a relation between these two variables. Furthermore earlier studies, such as Adams and Ferreira (2009), on the link between financial performance of companies and the presence of women on the boards suggest there is a significant effect of gender diversity in the board on the firm performance.

(7)

Corporate governance has become of great importance in the listed companies since stakeholders are better informed and earnings management is more discussed. Ethical behaviour plays a huge role in corporate governance and according to Navran (1997) the challenge for managers and leaders is to conduct ethical behaviour and take responsibility for their decision choices. In 1984 Beltramini et al. suggested that ethical business practices would be significantly affected by the increasing participation of women. The main findings found by Glover et al. (2002) include that women are more likely to make ethical choices than men. Further findings suggest there is a correlation between the years of work experience and a higher level of ethical behaviour, as well as between the high levels of need for achievement and the high level of ethical decision-making.

Prior research on gender differences involving ethical behaviour state women have better attendance records and a more gender-diverse board reduces the attendance problems (Adams and Ferreira, 2009). Women are also more likely to join monitoring committees, but even regarding these findings, Adams and Ferreira suggest with their results that introducing gender quotas for directors can reduce firm-value for well-governed companies. Other related literature includes the article on the relation between board gender diversity and the informativeness of stock prices (Gul, Srinidhi and Ng, 2011). Their findings suggest that firms with more women on the boards reflect more firm-specific information and extend earlier literature by suggesting gender-diverse boards could act as a substitute for weak corporate governance.

2.3 Prior research on the relation between board diversity and the earnings quality

In 2008 Krishnan and Parsons were the first to do empirical research on the relation between the quality of reported earnings and the percentage of women in the senior management ranks. The data used came from the companies listed on the Fortune 500, and reflect the period of 1996 to 2000. Besides the higher profitability and higher stock returns related to a more diverse

management board, the earnings quality is also positively correlated to the gender diversity. The independent variables leading to a change in earnings quality according to Krishnan and Persons (2008) are conservatism, which means earnings are more sensitive to bad news than good news, the earnings smoothing, the loss avoidance tendency and the earnings persistence. Earnings smoothing refers to attempts by managers to conceal economic shocks to the firm’s operating cash flows (Krishnan and Parsons, 2008), loss avoidance tendency refers to the ratio of ‘small profits’ to ‘small losses’, and the earnings persistence refers to the sustainability of the earnings.

Srinidhi, Gul and Tsui (2011) have examined whether U.S. corporations with gender-diverse boards exhibit higher quality earnings, using data from 2001 to 2007 available in COMPUSTAT. They detect two ways of measuring the quality of earnings: i) market-based measures such as earnings

(8)

response coefficient that measure the quality of earnings by its association with stock-based measures such as price, return or volume. ii) measure of the ability of current earnings to predict future cash flows and earnings. Srinidhi, Gul and Tsui (2011) suggest including female participation as a control variable for future research on board governance linked to firm performance. They conclude female board participation, in particular on the audit committees, results in better reporting and higher earnings quality.

Terjesen, Sealy and Singh (2009) have also done research on the relation between gender diversity on corporate boards and the firm performance. Despite their conclusion stating women directors contribute to firm performance outcomes, they mention a lack of theoretical development needed to make a strong business case. The literature written by Terjesen, Sealy and Singh (2009) is not of great importance in this study because the paper is written from a theoretical perspective while this paper takes an empirical approach.

2.4 Hypotheses

The main goal of this study is to give an answer to the following research question:

Does the marginal effect of female board members on the level of earnings management maintain constant, or is there a change in effect?

The first hypothesis is bases on the prior studies involving earnings management and state the level of earnings management reduces with female participation.

Bernardi and Arnold (1997) suggest that women are less likely to engage in unethical behaviour to gain financial rewards, and in case male directors do not act differently regardless the number of female board members, a constant marginal effect is plausible. Adams and Ferreira (2009) conclude male directors are less likely to have an attendance record of less than 75% if the board is more diverse. This effect, together with the expected linear effect based on Bernardi and Arnold (1997) indicates an increasing marginal effect of gender diversity on the attendance problem. The second and third hypotheses are based on these prior studies:

The second hypothesis state the level of attendance less than 75% tends to reduce with female participation.

The third and final hypothesis is based on the first and second and extend the expected effect of female participation on earnings management by implicating an increasing marginal effect of gender diversity.

(9)

3. Methodology and model

In this section the methodology and model is discussed using the prior studies and theoretical background. Paragraph 3.1 starts with an in-depth introduction of the Modified Jones model used in this study. The estimation equation and explanation of the variables used in the model are

mentioned in paragraph 3.2 and extended in 3.3, which contains the control variables for the problematic issues. All variables used are explained in Table 1.

3.1 The Modified Jones model

Jones was the first, in 1991, to develop a model estimating the discretionary component of total accruals instead of the discretionary part of just one accrual. She states managers are likely to exert earnings management on more than one component of total accruals in order to reduce the level of earnings (Jones, 1991, pp. 194). To estimate non-discretionary accruals, also referred to as normal accruals which can be fully explained by firm-specific information (Jones, 1991), a regression model is used to estimate the firm-specific parameters. The error-term in this model, shown in Equation 1, can be referred to as the discretionary accruals.

Equation 1

𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 = 𝑏𝑏0+ 𝑏𝑏1 �_𝑇𝑇1

𝑖𝑖𝑖𝑖−1� + 𝑏𝑏2 (𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖) + 𝑏𝑏3 (𝑃𝑃𝑃𝑃𝛥𝛥𝑖𝑖𝑖𝑖) + 𝜀𝜀𝑖𝑖𝑖𝑖

The variable for total accruals (TA) is scaled by total assets and is measured using Equation 2.

Equation 2

𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖

𝑇𝑇𝑖𝑖𝑖𝑖−1=

𝛥𝛥𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖− 𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖

𝑇𝑇𝑖𝑖𝑖𝑖−1

In their article on detecting earnings management (1995), Dechow, Sloan and Sweeney criticize the Jones model by introducing the bias of the earnings management towards zero. They give an example of manipulation of the receivables leading to an increase in non-discretionary accruals, and introduce a modified version of the model distracting the change in receivables from the change in revenue (Dechow, Sloan and Sweeney, 1995, pp. 199).

Dechow, Sloan and Sweeney (1995) mention five models for detecting earnings management and conclude all five models have a relatively low power. In case of extreme

circumstances all models lead to misspecified tests. Nevertheless, the modified Jones model (1991) has the highest power among the five and is therefore suggested as the most reliable one.

(10)

A more extended model is the Dechow and Dichev model (2002), as discussed in appendix A. This model corrects for misspecified tests and assumes the accruals quality, is affected by

measurement errors in accruals. Due to data limitations this study is based on the Modified Jones model (1991) instead of the DD model (2002).

3.2 Estimation equation and explanation of the variables

In the regression model, shown in Equation 3, the female board presence (FP1) and attendance level (AttDir) will function as independent variables and the level of earnings quality as the dependent one. Included control variables are the director independence (DirInd), the directors on the audit committee (DirAud), the size, two dummy variables for year and ten dummy variables for the S&P economic sector. Equation 3 𝛥𝛥𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑚𝑚𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚𝐸𝐸𝑚𝑚 (𝛥𝛥𝐸𝐸) = 𝑏𝑏4+ 𝑏𝑏5 𝐶𝐶𝑃𝑃1 + 𝑏𝑏6𝑇𝑇𝑚𝑚𝑚𝑚𝐴𝐴𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝑏𝑏7 𝐴𝐴𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐷𝐷𝑖𝑖𝑖𝑖+ 𝑏𝑏8𝐴𝐴𝐸𝐸𝐸𝐸𝑇𝑇𝐷𝐷𝐷𝐷𝑖𝑖𝑖𝑖+ 𝑏𝑏9 𝑆𝑆𝐸𝐸𝑆𝑆𝑚𝑚𝑖𝑖𝑖𝑖 + 𝑏𝑏10𝐸𝐸𝐸𝐸𝐷𝐷970 + 𝑏𝑏11𝐸𝐸𝐸𝐸𝐷𝐷925 + 𝑏𝑏12𝐸𝐸𝐸𝐸𝐷𝐷974 + 𝑏𝑏13𝐸𝐸𝐸𝐸𝐷𝐷976 + 𝑏𝑏14𝐸𝐸𝐸𝐸𝐷𝐷978 + 𝑏𝑏15𝐸𝐸𝐸𝐸𝐷𝐷935 + 𝑏𝑏16𝐸𝐸𝐸𝐸𝐷𝐷800 + 𝑏𝑏17𝐸𝐸𝐸𝐸𝐷𝐷905 + 𝑏𝑏18𝐸𝐸𝐸𝐸𝐷𝐷940 + 𝑏𝑏19𝐸𝐸𝐸𝐸𝐷𝐷600 + 𝑏𝑏20𝑌𝑌𝑚𝑚𝐸𝐸𝐸𝐸10 + 𝑏𝑏21𝑌𝑌𝑚𝑚𝐸𝐸𝐸𝐸11 + 𝜀𝜀𝑖𝑖𝑖𝑖

The earnings management refers to the amount of discretionary accrual. The higher the value of earnings management, the lower the level of earnings quality. The nondiscretionary accruals can be measured using equation 4. As explained in section 3.1, the Modified Jones model corrects the original Jones model (1991) by subtracting the change in receivables from the change in revenue. Using the Ordinary Least Squares method (OLS), the estimates of b1, b2 and b3 can be

measured.

Equation 4

𝑁𝑁𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖 = 𝑏𝑏1 �_𝑇𝑇1

𝑖𝑖𝑖𝑖−1� + 𝑏𝑏2 (𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖− 𝛥𝛥𝛥𝛥𝛥𝛥𝐶𝐶𝑖𝑖𝑖𝑖) + 𝑏𝑏3 (𝑃𝑃𝑃𝑃𝛥𝛥𝑖𝑖𝑖𝑖)

To test hypothesis 1, whether female participation on the board affects the level of earnings management a regression is done on Equation 3. The null-hypothesis will be rejected if the

estimated coefficient of b5, measuring the effect of female presence on earnings management, is

(11)

*(in millions of Dollars)

Table 1: All variables used in the given equations

1 _{Where EM = DA = TA – NDA}

Variable Acronym Definition

Dependent variable

- Earnings management1

- Earnings quality

EM EQ

- The level of accruals showing no

firm-specific information, measured as in footnote 1

- The earnings quality measures as natural

logarithm of discretionary accruals divided by natural logarithm of total accruals, see appendix B

Independent variable

- Number of female

directors

- Dummy for presence of

female directors

FP FP1 FP2 FP3

- Total number of female directors

- Dummy variable that takes a value of 1 if there’s at least one (or at least 2, FP2, or at least 3, FP3) female director(s), non-executive director(s) or audit committee member(s), and 0 otherwise.

Control variables - Director independence - Directors on audit committee - Meeting attendance directors

- Dummy for S&P

economic sector

- Dummy for year

DirInd DirAud AttDir Indi

Yearx

- Fraction of independent directors in the board

- Fraction of directors on the audit committee

- Fraction of directors attended < 75% of Meeting

- Dummy variable for ith_{economic sector}

taking the value of 1 or 0. The omitted economic sector is number is 700

- Dummy variable for year x taking the value

of 1 or 0. The omitted year is 2009 Accruals - Size - Discretionary accruals - Total assets - Total accruals - Change in receivables - Change in revenues

- Gross property, plant and

equipment

- Operating cash flow

- Earnings before extraordinary items - Non-discretionary Accruals - Firm-specific parameters Size DA A TA ΔREC ΔREV PPE CFO EXBI NDA b1, b2 and b3

- The size of the firm measured as total assets (in millions of Dollars)

- The level of discretionary accruals* as measured in Equation 7

- The total assets of the firm*

- The total number of accruals* measured in

total assets for a certain period t

- The change in receivables* over a certain period t

- The change in revenues* over a certain

period t

- The gross property, plant and equipment*

at time t

- Operating cash flow at time t

- The earnings before extraordinary items at

time t

- The accruals not to be explained by the firm-specific information, as calculated in equation 1

- Regression coefficient estimates

(12)

3.3 Control variables and additional tests

Prior research, such as EPWN (2004), argue firms are more likely to have one female board participant nowadays, but only a small range of firms have more than one. Therefore two other regressions on earnings quality will be done, Equation 3, one with at least two female board

members, and one with at least three female board members as independent variables. The variable FP1, used in the first regression, will change to FP2 and FP3, and still reflect a dummy variable. A test for non-linearity will be done (Srinidhi, Gul and Tsui, 2011), to test the second hypothesis, whether the marginal effect of female presence declines with more than one female board member.

Earlier studies often face the difficulties of proving the evidence in the relation between gender diversity and reporting quality, due to omitted variables. These variables can cause a reverse causality problem, which indicates the correlation disappears when including corrections for these omitted variables. Earlier studies such Adams and Ferreira (2007) suggest too much board

monitoring can decrease shareholder value, which suggest it is possible gender diversity only increases value when additional board monitoring would enhance firm value.

The omitted variable problem can be split into two components. The first component has to do with endogeneity, which arise when more female board members are participating when the earnings quality is high. Prior research, such as Adams and Fereirra (2009), show that female presence on boards is associated with firm performance. Due to this endogeneity interpreting the results must be done carefully. Findings only suggest an association or correlation, but to speak of a causal link is inappropriate. The second component is board governance. The effect of female participation on earnings quality is only useful when this effect is not already visible in other measures of board governance.

As Klein’s (2002) findings conclude, independent audit committees and boards are better able to monitor the earnings process. This suggests there is a significant relationship between the level of audit committee independence of the board and the level of earnings management. Important is to use a variable in the model to control for this relationship. The proxies used to control for board independence will be DirInd (a fraction of independent directors) and is computed using affiliated directors (COMPUSTAT) and DirAud (a fraction of directors on the audit committee).

(13)

Adams and Ferreira (2009) state in their article, women have better attendance records. To test for the last hypothesis, including more women on the board reduces the low attendance, measured as attendance below 75% of the meetings, is tested using a linear regression on AttDir, as in Equation 5. The null-hypothesis state c1 is equal to zero.

Equation 5

𝑇𝑇𝑚𝑚𝑚𝑚𝐴𝐴𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝑐𝑐0+ 𝑐𝑐1𝐶𝐶𝑃𝑃𝑖𝑖𝑖𝑖+ 𝜀𝜀𝑖𝑖𝑖𝑖

4. Data and results

In this section the sample and the source of the data will be described in the first paragraph. The descriptive statistics discuss the means of the variables in section 4.2, and in the final paragraph contains the results of the tests.

4.1 Descriptive statistics

The data on board characteristics is obtained using the RiskMetrics database, which is the leader in corporate governance data (The Wharton School at the University of Pennsylvania). From 1996 onwards, RiskMetrics contains data from the S&P 1500 companies. Since 2007 RiskMetrics started to collect data following ISS specifications instead of the earlier used IRRC, which is the reason this study will only include data from 2009 till 2011. The available data will be merged with the data from the COMPUSTAT database using the year and CUSIP, which is a 9-charactered number indicating a North-American or Canadian financial security.

After merging the databases, dropping the unmatched values and controlling for all missing values to compute the earnings quality, the final sample exist of 1062 companies and the number of usable company years is 3946. To control for industry, ten dummy variables have been created indicating the S&P economic sector code2_{(COMPUSTAT). The amount of company years for each}

S&P economic sector is listed in table 1 of appendix B.

2_{Where 600 = Transportation} _{935 = Energy}

700 = Utilities 940 = Technology

800 = Financials 970 = Basic Materials

905 = Health Care 974 = Communication Services

925 = Capital Goods 976 = Consumer Cyclical

978 = Consumer Staples

(14)

Other data obtained from COMPUSTAT are variables to compute the accruals; total assets, gross property plant and equipment and total receivables from the balance sheet, total revenue from the income statement and from the cash flow statement income before extraordinary items and net operating activities. A summary of the important data is given in Table 2.

Table 2: Mean estimations of the important variables

*(in millions of Dollars)

To study the impact of female presence on the board three dummy variables are created indicating none, at least one, at least two or at least four female board members. Other variables from RiskMetrics used to control for other board characteristics are attended <75% of meetings, Audit Committee member, and Board independence.

Over Mean Std. Dev Min Max

Total number of directors

2009 2010 2011 9.26 9.33 9.34 2.31 2.28 2.26 4 4 4 34 34 34 Number of female directors

2009 2010 2011 1.19 1.21 1.27 1.04 1.03 1.06 0 0 0 6 6 7 Fraction of directors on the Audit Committee

2009 2010 2011 0.43 0.43 0.43 0.11 0.11 0.11 0.18 0.18 0.15 0.89 0.89 0.88 Fraction of independent directors

2009 2010 2011 0.78 0.79 0.80 0.12 0.11 0.11 0.14 0.33 0.43 1.00 1.00 1.00 Fraction of directors attending <75%

2009 2010 2011 0.0006 0.0007 0.0005 0.008 0.009 0.008 0.00 0.00 0.00 0.13 0.17 0.13 Size * 2009 2010 2011 7.87 7.94 8.06 1.58 1.61 1.61 4.40 4.05 4.33 13.65 13.72 13.74 Total accruals* (scaled by total assets)

2009 2010 2011 -0.081 -0.049 -0.045 0.068 0.057 0.059 -0.46 -0.39 -0.39 0.32 0.24 0.21 Discretionary accruals*(scaled by total assets)

2009 2010 2011 -0.030 -0.005 -0.003 0.068 0.056 0.059 -0.39 -0.37 -0.37 0.40 0.27 0.35

(15)

Figure 1.1 shows the average discretionary accruals per year for each number of female board members. In line with the first hypotheses, an increase in female board members seems to be followed by a decrease in discretionary accruals indicating a positive relation between the number of female board members and the earnings quality. In appendix B the average total accruals,

discretionary accruals and average earnings quality are shown per S&P economic sector.

Figure 1.2 indicates no clear relation between the attendance <75% and the number of female directors. The average fractions of directors attending less than 75% of the board meetings is in all cases less than 0.005, which indicated the sample used is not so appropriate to test the

possible relational effect.

Figure 1.1: Per year the average discretionary accruals

Figure 1.2: Per year the average fraction of directors’ attendance < 75% -. 4 -. 2 0 .2 .4 -. 4 -. 2 0 .2 .4 0 2 4 6 8 0 2 4 6 8 2009 2010 2011 D iscre ti o n a ry Accru a ls

Number of female board members

Graphs by Data Year

The average Discretionary Accruals

0 .00 05 .00 1 .00 15 0 .00 05 .00 1 .00 15 2009 2010 2011 2009 2010 2011 2009 2010 2011 2009 2010 2011

At least 1 female board member At least 2 female board members

At least 3 female board members None female board members

F ra c ti on o f t o ta l di rec to rs w it h an a tt en da nc e < 7 5% Graphs by FP0, FP1, FP2, FP3

(16)

To check for correlations between the variables used to measure the total accruals a Spearman correlation matrix is used, see appendix C.

4.2 Using the Modified Jones Model

As mentioned in section 3, this study will use the modified Jones model (1991) to calculate the non-discretionary accruals, which was first introduced by Dechov, Sloan and Sweeney (1995). In

COMPUSTAT the operating cash flow, total assets and the earnings before extra-ordinary Items are given and used to calculate total accrual (TA) as in Equation 2. Using OLS the estimators of the regression used to calculate non-discretionary accruals are measured as is shown in Table 3. The model has an adjusted R-squared of 45.49%, which refers to the percentage of total accruals explained by the non-discretionary accruals.

This regression test has a confidence interval of 95%, and looking at the p-values in table 4 all variables are significant. Now the estimates of the coefficients are known, the non-discretionary accruals can be measured using Equation 6.

Equation 6 𝑁𝑁𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖 𝑇𝑇𝑖𝑖𝑖𝑖−1 = −8.261917 � 1 𝑇𝑇𝑖𝑖𝑖𝑖−1� + 0.0426629 � 𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖 − 𝛥𝛥𝛥𝛥𝛥𝛥𝐶𝐶𝑖𝑖𝑖𝑖 𝑇𝑇𝑖𝑖𝑖𝑖−1 � − 0.0697716 � 𝑃𝑃𝑃𝑃𝛥𝛥𝑖𝑖𝑖𝑖 𝑇𝑇𝑖𝑖𝑖𝑖−1�

The discretionary accruals can be calculated subtracting the non-discretionary accruals from the total accruals, Equation 7.

Equation 7

𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖 = 𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖− 𝑁𝑁𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖

These discretionary accruals are a proxy for the level of earnings management and to estimate the earnings quality the natural logarithm of non-discretionary accruals is divided by the total accruals,

Table 3: The regression on total accruals, scaled by total assets

ppegat -.0697716 .0021527 -32.41 0.000 -.0739954 -.0655478 RevRecat .0426629 .0126215 3.38 0.001 .0178977 .0674282 scaledat -8.261917 1.542311 -5.36 0.000 -11.28816 -5.235676 TA Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust

(Std. Err. adjusted for 1089 clusters in cusip) Root MSE = .06288 R-squared = 0.4555 Prob > F = 0.0000 F( 3, 1088) = 616.99 Linear reg ression Number of obs = 2767

(17)

as shown in table 2 of appendix B. The averages of the level of discretionary accruals for each number of female board members can be found in Figure 1.1.

4.3 Results

The regression model of Equation 3 has an R-squared of 0.1139, which means 11.39% of the discretionary accruals can be explained by this model.

The first hypothesis to be tested is the effect of female presence on the level of earnings management. The proxy for earnings management is discretionary accruals and therefore a

regression is done to test on the null-hypothesis of a5 being zero. This coefficient indicates the effect

of FP1, the dummy equalling 1 if there is at least one female board member and 0 otherwise. To focus on the relation of female presence on earnings management, the control variables have to be included. Figure 3 shows a p-value for the FP1 of 0.502, which means there is not enough evidence to reject the null-hypotheses. These regression results indicate there is no significant effect of female board members on the level of discretionary accruals.

The regression in Table 4 is repeated twice, for at least two female board members, as shown in appendix D, and at least three female participants, as shown in table 5.

Table 4: The linear regression on discretionary accruals including all control variables _cons .0553221 .0145786 3.79 0.000 .0267168 .0839274 Year10 -.0024197 .0021236 -1.14 0.255 -.0065865 .0017471 Year09 -.028771 .0029525 -9.74 0.000 -.0345643 -.0229777 Ind600 -.0241506 .0064746 -3.73 0.000 -.0368548 -.0114464 Ind940 -.0550833 .0053175 -10.36 0.000 -.065517 -.0446496 Ind905 -.0441789 .0058832 -7.51 0.000 -.0557227 -.0326351 Ind800 -.0462447 .0072128 -6.41 0.000 -.0603972 -.0320922 Ind935 -.0278768 .0067204 -4.15 0.000 -.0410631 -.0146904 Ind978 -.0309859 .0054484 -5.69 0.000 -.0416764 -.0202953 Ind976 -.0403887 .0051876 -7.79 0.000 -.0505676 -.0302099 Ind974 -.034758 .0091016 -3.82 0.000 -.0526166 -.0168994 Ind925 -.0264717 .0049278 -5.37 0.000 -.0361408 -.0168027 Ind970 .0042433 .0052498 0.81 0.419 -.0060576 .0145441 Size -.0018267 .0009879 -1.85 0.065 -.0037651 .0001118 DirAud .0081562 .0130009 0.63 0.531 -.0173535 .0336658 DirInd -.0139898 .0132748 -1.05 0.292 -.0400369 .0120573 AttDir -.0794977 .0878078 -0.91 0.365 -.2517895 .0927941 FP1 -.0024145 .0035975 -0.67 0.502 -.0094733 .0046442 DA Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust

(Std. Err. adjusted for 1089 clusters in cusip) Root MSE = .05839 R-squared = 0.1139 Prob > F = 0.0000 F( 17, 1088) = 21.50 Linear regression Number of obs = 2767

(18)

No significant effect is found for FP1 and FP2, but for FP3, at least three female board members, the coefficient is significant at 5%. The discretionary accruals are scaled by the total assets, which means the coefficient for at least three female board members should be multiplied by the average of total assets to imply what is the actual effect. On average, the presence of at least three female board members result in a decrease of discretionary accruals of 119.61 million dollar3_.

The second hypothesis states the effect of female participants declines with the number of female members. When comparing the coefficients, it shows the coefficients for at least 1, at least 2 and at least 3 female board members are respectively -0.0024145 (0.502)_{, -0.0027006}(0.363)_and

-0.0086312 (0.032)_{. Hypothesized was an absolute decrease in coefficient, but the tests result in an}

absolute increase in coefficient. This indicates the marginal effect of women on the board of directors increases.

For testing the third hypothesis the regression in Table 6 is done. The total number of female board members acts as the independent variable, while the fraction of directors with an attendance < 75% acts as the dependent variable. The same control variables are added to the model.

3_{-0.0086312 * 13857.95 = 119.61 million (see appendix D for average total assets)}

Table5: The linear regression on discretionary accruals including all control variables for at least three female board members

_cons .0570438 .014599 3.91 0.000 .0283984 .0856892 Year10 -.0024464 .002124 -1.15 0.250 -.0066139 .0017211 Year09 -.0288601 .0029516 -9.78 0.000 -.0346516 -.0230687 Ind600 -.0229693 .0064984 -3.53 0.000 -.03572 -.0102186 Ind940 -.0537727 .0052214 -10.30 0.000 -.0640178 -.0435275 Ind905 -.0435387 .0059232 -7.35 0.000 -.0551608 -.0319166 Ind800 -.0446925 .0072452 -6.17 0.000 -.0589086 -.0304764 Ind935 -.0254342 .0065261 -3.90 0.000 -.0382394 -.012629 Ind978 -.0313903 .0053939 -5.82 0.000 -.0419739 -.0208068 Ind976 -.0403421 .0051722 -7.80 0.000 -.0504908 -.0301934 Ind974 -.033635 .0090462 -3.72 0.000 -.0513849 -.0158851 Ind925 -.0249398 .0049041 -5.09 0.000 -.0345624 -.0153171 Ind970 .0048406 .0052568 0.92 0.357 -.0054741 .0151553 Size -.0024507 .0010028 -2.44 0.015 -.0044183 -.000483 DirAud .0115061 .0130948 0.88 0.380 -.0141878 .0372001 DirInd -.0160134 .0133431 -1.20 0.230 -.0421945 .0101676 AttDir -.0894922 .0867642 -1.03 0.303 -.2597363 .0807519 FP3 .0086312 .004014 2.15 0.032 .0007551 .0165073 DA Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust

(19)

With a significance level of 5% the null-hypothesis stating a parameter c1 of zero cannot be rejected.

This means there is not enough evidence to indicate the existence of a relation between the female presence on the board of directors and the fraction of directors attending less than 75% of the meeting.

5. Discussion and conclusion

This final section gives a recapitulation of the paper and a brief summary of the results. The primary goal was to extend the existing literature with evidence on a possible reduction of the effect of gender diversity on earnings management with each extra female board member.

The sample contains 3946 unique firm year of 1062 S&P 1500 companies with on average 9.3 directors on the board of which 1.2 are women. To test for a significant effect of female

participation on the board of directors on earnings management, hypotheses 1, the Modified Jones model (1991) is used. The null-hypotheses, stating there is no significant effect, cannot be rejected. To test the main hypotheses the regression is done two more times, replacing at least one female board member by at least two female members and at least three female members. There is not

Table 6: Regression on the director attendance <75% including all control variables

_cons .0012641 .0016981 0.74 0.457 -.0020675 .0045957 Year10 -.0000207 .0003432 -0.06 0.952 -.0006941 .0006527 Year09 -.0001535 .0002995 -0.51 0.608 -.0007412 .0004342 Ind600 .0036658 .0021099 1.74 0.083 -.0004738 .0078054 Ind940 .0006141 .0003367 1.82 0.068 -.0000465 .0012746 Ind905 .0005362 .0003725 1.44 0.150 -.0001947 .001267 Ind800 .0005271 .0005316 0.99 0.322 -.0005158 .00157 Ind935 .0006656 .0005353 1.24 0.214 -.0003846 .0017158 Ind978 .0005492 .0003623 1.52 0.130 -.0001618 .0012601 Ind976 .0010833 .0004192 2.58 0.010 .0002609 .0019058 Ind974 -.000034 .0001024 -0.33 0.740 -.0002349 .000167 Ind925 .0006539 .0003033 2.16 0.031 .0000589 .001249 Ind970 .0008519 .000526 1.62 0.106 -.0001801 .0018839 Size .0000176 .0000944 0.19 0.853 -.0001677 .0002028 DirAud -.0020007 .0012879 -1.55 0.121 -.0045275 .000526 DirInd -.000855 .0016035 -0.53 0.594 -.0040011 .002291 FP .0000999 .0001918 0.52 0.602 -.0002764 .0004763 AttDir Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust

(Std. Err. adjusted for 1175 clusters in cusip) Root MSE = .00855 R-squared = 0.0056 Prob > F = 0.0569 F( 16, 1174) = 1.62 Linear regression Number of obs = 3946

(20)

enough evidence to reject the null-hypotheses, the marginal effect of gender diversity increases with each extra woman on the board, using at least two and at least three female board members.

Interesting are the results of the regression on at least three female board members. The coefficients of FP1 and FP2 are far from significant, but the coefficient for at least three female board members is significant at 5%. This result indicates that gender diversity only effects earnings management when at least three female directors are part of the board. For future research it might be useful to study the change in effect from at least three female members onwards. As Adams and Ferreira (2009) conclude, men are more likely to have attendance problems than women and a greater fraction of female directors reduces the attendance problems of male. This could probably be an explanation why the gender diversity only effect earnings management starting from three female directors.

The third question concerns the relation between the attendance level and female

participation. Based on Adams and Ferreira (2009) hypothesized was a reduction in attendance less than 75% with an increase in female participation. The test does not give enough evidence to conclude this to be true, which is possibly due to the fact the regression model has an R-squared of 0.0056. The model does not explain the director attendance, which means the attendance is not a necessary control variable needed to answer the main question.

In section 3.3, the second component for omitted variable bias is mentioned being board governance. The effect of female participation on earnings quality is only useful when this effect is not already visible in other measures of board governance. In future research the board governance variable, CGboard, and an index of CEO power, CEOPower (Larcker et al. 2007) together with the inverse Mills ratio (IMR) can be used to control for this endogeneity. The inverse Mills ratio is referred to as the Heckman correction and can be computed from a probit model (Srinidhi, Gul and Tsui, 2011). The model can be found in appendix A.

(21)

6. References

Adams, R.B., Ferreira, D (2009). ‘Women in the boardroom and their impact on governance and performance’. Journal of Financial Economics, V. 94, Issue 2, pp. 291–309

Bergstresser, D. and Philippon, T. (2006). ‘CEO incentives and earnings management’. Journal of Financial Economics, V. 80, Issue 3, pp. 511-529

Bernardi, R. A. and Arnold, D. F. (1997). ‘An Examination of Moral Development within Public Accounting by Gender, Staff Level, and Firm’. Contemporary Accounting Research, V. 14, pp. 653-668

Burgstahlera, D. C., Hailb, L. and Leuz, C. (2006). ‘The Importance of Reporting Incentives: Earnings Management in European Private and Public Firms’. The accounting Review, V. 81, No. 5 pp. 983–1016

Croson, R. and Gneezy, U. (2009). ‘Gender Differences in Preferences’. Journal of Economic Literature, V. 47, No. 2, pp. 448-474

Dechow, P. M., Sloan, R.G. and Sweeney, A. P. (1995). ‘Detecting Earnings Management’. The Accounting Review, V. 70, No. 2, pp. 193-225

Dechow, P. M., Dichev, I. D. (2002). ‘The Quality of Accruals and Earnings: The Role of Accrual Estimation Errors’. The Accounting Review, Vol. 77, pp. 35-59

Doldor, E., Vinnicombe, S., Gaughan, M. and Sealy, R. (2012). ‘Gender Diversity on Boards: The Appointment Process and the Role of Executive Search Firms’. Equality and Human Rights Commission, Research report 85

EPWN (European Professional Women's Network), 2004. ‘The European PWN board women monitor 2004’. www.europeanpwn.net/index.php?article_id=8

Feltham G. A. and Xie, J. (1994). ‘Performance Measure, Congruity and Diversity in Multi-Task Principal/Agent Relations’. The Accounting Review, V.69, No. 3, pp.429-453

Ford, R.C. and Richardson W.D. (1994). ‘Ethical decision making: A review of the empirical literature’. Journal of Business Ethics, V.31, pp. 205-221

Francis, J., LaFond, R., Olsson, P. And Schipper K. (2005). ‘The market pricing of accruals quality’. Journal of Accounting and Economics, V.39, pp. 295-327

Glover, S.H., Bumpus, M.A., Sharp, G.F., Munchus, G.A. (2002) ‘Gender differences in ethical decision making’. Women In Management Review, V. 17, Issue5, pp. 217-227

Guay, W.R., Kothari, S.P. and Watts, R.L. (1996). ‘A market-based evaluation of Discretionary Accruals Models’. Journal of accounting research, V. 34

(22)

Gul, F.A., Srinidhi, B. and Ng, A.C. (2011). ‘Does board gender diversity improve the informativeness of stock price?’ Journal of Accounting and Economics, V. 51, Issue3, pp. 314-338

Klein, A. (2002). ‘Audit committee, board of director characteristics, and earnings management’. Journal of Accounting and Economics, V.33, Issue 3, pp. 375-400

Krishnan, G. V. and Parsons, L.M.(2008). ‘Getting to the Bottom Line: An Exploration of Gender and Earnings Quality’. Journal of Business Ethics, V. 78, pp. 65-76

McNichols, M.F. (2002). ‘Discussion of the Quality of Accruals and Earnings: The Role of Accrual Estimation Errors.’ The Accounting Review, V. 77, pp. 61-69

Nguyen, T., Locke, S. and Reddy, K. (2012). ‘Do Female Directors Add Value? Evidence from an Emerging Market’. Emerging markets review

Rose, C. (2007). ‘Does female board representation influence firm performance? The Danisch evidence. Corporate Governance, V. 15, pp. 404-413

Srinidhi, B., Gul, F.A. and Tsui, J. (2011). ‘Female Directors and Earnings Quality’. Contemporary Accounting Research, V. 28, Issue 5, pp. 1610-1644

Terjesen, S., Sealy, R. and Singh, V.(2009). ‘Women Directors on Corporate Boards: A Review and Research Agenda’. Corporate Governance: An International Review, V. 17, Issue 3, pp. 320-337

Trevino, L.K., Weaver, G.R. and Reynolds, S.J. (2006). ‘Behavioral Ethics in Organizations: A review’. Journal of management, V. 32, No. 6, pp. 951-990

(23)

7. Appendixes

Appendix A.

In the perfect situation the model used to measure earnings quality would be based on McNichols’ combination of the Jones model (1991) and the DD model (Dechow and Dichev, 2002). A correction for reflections of prior changes in diversity the absolute value of the residual would be made. The five discussed models all lead to misspecified tests (Dechow, Sloan and Sweeney, 1995) and the abnormal accruals, as explained by the modified Jones model, are the accruals not explained by the fundamentals such as changes in revenues (Dechow and Dichev, 2002). Following the assumption of Francis et al. (2005) such abnormal accruals contain a certain amount of uncertainty and therefore reduce the link to information risk. Francis et al. (2005) state the Dechow and Dichev model is based on the idea that accruals quality is affected by measurement errors in accruals, regardless the management intentions. To study the actual effect of female board members on the level of earnings management the best model to be used is an extended version of the Dechow and Dichev model (DD model). The regression model should be comparable to the one shown in equation 1.

𝑨𝑨𝑨𝑨𝑨𝑨 = 𝒃𝒃𝟎𝟎 + 𝒃𝒃𝟏𝟏 𝑭𝑭𝑭𝑭 + 𝒃𝒃𝟐𝟐 𝑫𝑫𝑫𝑫𝑫𝑫𝑫𝑫𝑫𝑫𝑫𝑫 + 𝒃𝒃𝟑𝟑𝑪𝑪𝑨𝑨𝑪𝑪𝑭𝑭𝑷𝑷𝑷𝑷𝑫𝑫𝑫𝑫 + 𝒃𝒃𝟒𝟒𝑫𝑫𝑫𝑫𝑷𝑷𝑫𝑫𝑫𝑫 + 𝒃𝒃𝟓𝟓𝑳𝑳𝑫𝑫𝑳𝑳𝑳𝑳𝑨𝑨

+ 𝒃𝒃𝟔𝟔 𝑨𝑨𝑨𝑨𝑨𝑨│𝜟𝜟𝑾𝑾𝑪𝑪│ + 𝒃𝒃𝟕𝟕 𝑰𝑰𝑫𝑫𝑰𝑰𝑫𝑫𝑫𝑫𝑫𝑫 + 𝒃𝒃𝟖𝟖𝑫𝑫𝑫𝑫𝑫𝑫𝑨𝑨𝑫𝑫𝑰𝑰 + 𝒃𝒃𝟗𝟗𝑨𝑨𝑨𝑨𝑨𝑨𝑫𝑫𝑫𝑫𝑫𝑫 + 𝒃𝒃𝟏𝟏𝟎𝟎𝝀𝝀

+ ∑ 𝒉𝒉𝒌𝒌𝒀𝒀𝑫𝑫𝒀𝒀𝑫𝑫𝑫𝑫𝒌𝒌 + 𝜺𝜺4

4_{In equation 1 the AEE represents the absolute value of the accruals estimation error, FP represents a dummy}

variable for the presence of at least 1 female board member, IndDir indicates the percentage of independent directors in the board, DirAud the number of directors on the audit committee, AffDir indicates the affiliated directors which means the nonexecutive outside related, CEOpower reverse to the CEO’s power over the decisions of the board and impairs oversight, DirTen is the average number of years the director has been on the board and λ is referred to as the inverse Mills ratio for predicting the presence of female directors on the board. The lnMEV indicated the natural logarithm of the market value of equity, Dloss a dummy variable indicating a net loss for the firm, STDsales is the standard deviation over the firm sales, AvgOC the average operating cycle and Avg│ΔWC│the average change in annual working capital.

(24)

Appendix B.

Table 1 shows the number of unique company years per S&P Economic sector5_{. In the tables 2.1, 2.2}

and 2.3 the average earnings quality per sector is shown for each year.6

5_{Where 600 = Transportation} _{935 = Energy}

700 = Utilities 940 = Technology

800 = Financials 970 = Basic Materials

905 = Health Care 974 = Communication Services

925 = Capital Goods 976 = Consumer Cyclicals

978 = Consumer Staples

6_{EQ = lnDA / lnTA, the higher the level of earnings quality, the lower the actual quality. Since the natural}

logarithm is only available for positive numbers, the sample exists of only 337 unique company years.

Table 1: Frequency of firms per S&P Economic sector Tabel 2.3: 2011 Tabel 2.2: 2010 Tabel 2.1: 2009 . Total 3,946 100.00 978 394 9.98 100.00 976 770 19.51 90.02 974 30 0.76 70.50 970 275 6.97 69.74 940 721 18.27 62.77 935 217 5.50 44.50 925 587 14.88 39.00 905 388 9.83 24.13 800 196 4.97 14.29 700 255 6.46 9.33 600 113 2.86 2.86 Sector Code Freq. Percent Cum. Economic S&P 978 .7471143 .0555037 .6374617 .8567669 976 .7819483 .0216701 .7391372 .8247595 974 .6285168 . . . 970 .6519699 .0403631 .5722288 .7317109 940 .7886261 .0389934 .7115911 .8656611 935 .7773152 .0459202 .6865957 .8680348 925 .7494038 .0231712 .703627 .7951806 905 .7111867 .0458643 .6205777 .8017958 800 .9041084 .0197942 .8650031 .9432136 Earningsquality Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 154 978 .8377871 . . . 976 .690934 .0682875 .5532192 .8286488 970 .6006195 .0441884 .5115051 .6897339 940 .7859818 .0443062 .6966298 .8753337 935 .5824201 . . . 925 .6324642 .0549155 .5217165 .7432119 905 .7463285 .0717854 .6015595 .8910974 800 .9060319 .0337141 .8380409 .974023 700 .483954 .081265 .3200674 .6478405 600 .554307 . . . Earningsquality Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 44

978 .778497 .0374657 .7044158 .8525781 976 .8182544 .0297856 .7593593 .8771495 970 .6326275 .0460039 .5416639 .7235911 940 .8155273 .0406234 .7352025 .8958521 935 .7500904 .0886817 .5747398 .925441 925 .7367276 .0318221 .6738056 .7996496 905 .736945 .0547997 .6285894 .8453006 800 .8985238 .0298661 .8394695 .957578 700 .6252161 .0379071 .5502623 .7001699 600 1.461246 .5907573 .2931399 2.629353 Earningsquality Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 139

(25)

The tables 2.1, 2.2 and 2.3 give a summary of the total accruals and the discretionary accruals, both scaled by t-1, are defined for 2009, 2010 and 2011.

Tabel 3.1: 2009 Tabel 3.2: 2010 Tabel 3.3: 2011 978 -.0001748 .0051797 -.0103393 .0099898 976 -.0001807 .0046384 -.009283 .0089216 974 -.0162815 .0170711 -.0497817 .0172188 970 .0340584 .0055979 .0230731 .0450437 940 -.0210619 .0049298 -.0307362 -.0113877 935 .0044724 .0056738 -.0066618 .0156067 925 .007242 .0037962 -.0002076 .0146916 905 -.0214982 .0081165 -.037426 -.0055704 800 -.0307637 .0070542 -.0446068 -.0169207 700 .0211778 .0033385 .0146264 .0277292 600 -.0064026 .0062174 -.0186037 .0057984 DA 978 -.043732 .0053583 -.0542472 -.0332168 976 -.0432478 .0048793 -.0528229 -.0336727 974 -.0953922 .0174421 -.1296205 -.0611639 970 -.027582 .0057356 -.0388376 -.0163265 940 -.0485976 .0046589 -.0577401 -.039455 935 -.084415 .0087472 -.1015805 -.0672494 925 -.0248162 .0037553 -.0321855 -.0174469 905 -.0519081 .0071102 -.065861 -.0379551 800 -.0388859 .0072779 -.0531679 -.0246039 700 -.0518013 .0032014 -.0580837 -.0455189 600 -.0721981 .0079856 -.0878689 -.0565272 TA Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 979 978 -.0278781 .0067164 -.0410624 -.0146937 976 -.0475233 .0056017 -.0585195 -.0365271 974 .0048158 .0176612 -.0298534 .0394851 970 .0074476 .0070949 -.0064798 .021375 940 -.055207 .0057062 -.0664084 -.0440056 935 -.0298093 .0118201 -.0530123 -.0066062 925 -.0312446 .0059796 -.0429826 -.0195066 905 -.0200669 .0091903 -.0381076 -.0020262 800 -.0318148 .0092845 -.0500405 -.0135891 700 .0104989 .0059034 -.0010895 .0220873 600 .0064082 .0071138 -.0075564 .0203728 DA 978 -.0843845 .0077818 -.0996603 -.0691088 976 -.0983594 .0056143 -.1093803 -.0873385 974 -.1029348 .0168017 -.1359168 -.0699528 970 -.0652246 .0068542 -.0786795 -.0517696 940 -.0939352 .0057526 -.1052276 -.0826427 935 -.1234083 .014052 -.1509926 -.0958241 925 -.078615 .0058239 -.0900474 -.0671826 905 -.0574371 .0083325 -.0737939 -.0410803 800 -.0395261 .0096801 -.0585283 -.0205239 700 -.0624784 .0062775 -.0748011 -.0501556 600 -.076855 .0086509 -.093837 -.0598731 TA Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 779

978 -.0033301 .0044516 -.0120655 .0054053 976 -.0094346 .004652 -.0185633 -.0003059 974 -.0168879 .0065921 -.0298238 -.0039521 970 .031314 .0059593 .0196199 .043008 940 -.0226426 .0046779 -.0318222 -.013463 935 .0005666 .0076816 -.0145072 .0156404 925 .0071764 .0040868 -.0008432 .015196 905 -.0232311 .0051627 -.033362 -.0131002 800 -.0175947 .0061114 -.0295873 -.0056022 700 .0230151 .0040778 .0150132 .0310171 600 -.0012161 .005969 -.0129292 .0104971 DA 978 -.0537368 .0050244 -.0635962 -.0438773 976 -.0536255 .0047029 -.0628541 -.044397 974 -.1110619 .0171863 -.144787 -.0773368 970 -.0299336 .0061676 -.0420363 -.0178309 940 -.0518693 .0045899 -.0608762 -.0428624 935 -.091401 .0088887 -.1088434 -.0739585 925 -.0273503 .0039872 -.0351745 -.0195261 905 -.0561876 .0052178 -.0664266 -.0459486 800 -.0252061 .0068796 -.0387061 -.0117062 700 -.0500861 .0039947 -.0579249 -.0422474 600 -.0648209 .0082025 -.0809168 -.0487251 TA Over Mean Std. Err. [95% Conf. Interval] Mean estimation Number of obs = 1009

(26)

Appendix C.

The correlation matrix for all included variables in the regression on DA.

Y e a r 1 1 -0 . 0 0 7 2 -0 . 0 1 2 3 0 . 0 0 4 4 -0 . 4 6 3 2 -0 . 5 6 0 6 1 . 0 0 0 0 Y e a r 1 0 0 . 0 0 5 8 -0 . 0 0 3 3 0 . 0 0 0 4 -0 . 4 7 4 2 1 . 0 0 0 0 Y e a r 0 9 0 . 0 0 1 5 0 . 0 1 6 6 -0 . 0 0 5 1 1 . 0 0 0 0 I n d 6 0 0 -0 . 0 5 7 2 -0 . 0 8 2 2 1 . 0 0 0 0 I n d 9 4 0 -0 . 1 5 4 0 1 . 0 0 0 0 I n d 9 0 5 1 . 0 0 0 0 I n d 9 0 5 I n d 9 4 0 I n d 6 0 0 Y e a r 0 9 Y e a r 1 0 Y e a r 1 1 Y e a r 1 1 0 . 1 0 3 8 0 . 0 2 9 9 -0 . 0 0 2 2 0 . 0 4 1 6 -0 . 0 1 6 2 0 . 0 5 3 7 -0 . 0 0 5 9 0 . 0 0 9 4 0 . 0 2 0 8 0 . 0 0 5 1 0 . 0 0 0 9 0 . 0 0 8 6 0 . 0 0 0 6 Y e a r 1 0 0 . 0 8 0 2 0 . 0 0 0 8 0 . 0 0 5 2 0 . 0 1 5 6 -0 . 0 0 3 1 0 . 0 0 3 5 -0 . 0 0 0 3 -0 . 0 1 0 6 0 . 0 0 0 7 0 . 0 0 1 5 0 . 0 0 3 4 0 . 0 0 6 4 -0 . 0 0 4 6 Y e a r 0 9 -0 . 1 9 6 2 -0 . 0 3 2 7 -0 . 0 0 3 1 -0 . 0 6 0 9 0 . 0 2 0 4 -0 . 0 6 0 8 0 . 0 0 6 6 0 . 0 0 1 3 -0 . 0 2 2 9 -0 . 0 0 7 0 -0 . 0 0 4 6 -0 . 0 1 6 0 0 . 0 0 4 2 I n d 6 0 0 0 . 0 2 8 8 -0 . 0 4 3 2 0 . 0 6 9 2 -0 . 0 5 3 1 0 . 0 0 5 6 0 . 0 0 6 0 -0 . 0 4 8 1 -0 . 0 7 3 6 -0 . 0 1 4 5 -0 . 0 8 5 9 -0 . 0 5 8 2 -0 . 0 4 2 0 -0 . 0 4 0 0 I n d 9 4 0 -0 . 1 5 6 1 -0 . 1 8 7 2 -0 . 0 1 3 1 0 . 0 0 1 0 -0 . 0 0 7 9 -0 . 1 5 9 4 -0 . 1 2 9 5 -0 . 1 9 8 1 -0 . 0 3 9 1 -0 . 2 3 1 0 -0 . 1 5 6 5 -0 . 1 1 3 0 -0 . 1 0 7 7 I n d 9 0 5 -0 . 0 5 5 5 0 . 0 0 9 0 -0 . 0 2 3 6 0 . 0 1 5 6 -0 . 0 3 0 7 -0 . 0 3 9 4 -0 . 0 9 0 2 -0 . 1 3 7 9 -0 . 0 2 7 2 -0 . 1 6 0 9 -0 . 1 0 9 0 -0 . 0 7 8 7 -0 . 0 7 5 0 I n d 8 0 0 -0 . 0 5 6 1 0 . 0 3 8 1 0 . 0 0 4 2 -0 . 0 4 9 4 0 . 0 2 3 5 0 . 2 0 4 7 -0 . 0 6 3 1 -0 . 0 9 6 5 -0 . 0 1 9 1 -0 . 1 1 2 6 -0 . 0 7 6 3 -0 . 0 5 5 0 1 . 0 0 0 0 I n d 9 3 5 0 . 0 2 1 9 -0 . 1 2 5 6 0 . 0 0 4 8 0 . 0 0 6 2 0 . 0 4 1 9 0 . 1 3 3 2 -0 . 0 6 6 2 -0 . 1 0 1 2 -0 . 0 2 0 0 -0 . 1 1 8 0 -0 . 0 8 0 0 1 . 0 0 0 0 I n d 9 7 8 0 . 0 1 2 3 0 . 1 8 7 5 -0 . 0 1 2 4 -0 . 0 0 8 4 -0 . 0 5 4 7 0 . 0 1 7 9 -0 . 0 9 1 7 -0 . 1 4 0 2 -0 . 0 2 7 7 -0 . 1 6 3 5 1 . 0 0 0 0 I n d 9 7 6 -0 . 0 4 2 4 0 . 0 6 0 4 0 . 0 1 2 2 -0 . 1 5 6 8 -0 . 0 4 2 7 -0 . 1 5 4 2 -0 . 1 3 5 3 -0 . 2 0 7 0 -0 . 0 4 0 9 1 . 0 0 0 0 I n d 9 7 4 -0 . 0 0 2 6 0 . 0 4 4 5 -0 . 0 0 6 0 0 . 0 6 4 5 -0 . 0 2 3 5 0 . 0 8 4 9 -0 . 0 2 2 9 -0 . 0 3 5 0 1 . 0 0 0 0 I n d 9 2 5 0 . 0 5 1 7 -0 . 1 2 6 3 0 . 0 0 8 4 0 . 0 4 1 9 0 . 0 5 5 5 -0 . 0 3 7 1 -0 . 1 1 6 0 1 . 0 0 0 0 I n d 9 7 0 0 . 1 6 2 7 0 . 0 5 2 4 -0 . 0 0 2 2 0 . 1 0 6 8 0 . 0 6 3 6 0 . 0 3 7 6 1 . 0 0 0 0 S i z e 0 . 0 0 5 7 0 . 4 3 5 2 0 . 0 0 7 4 0 . 2 5 1 4 -0 . 1 8 7 7 1 . 0 0 0 0 D i r A u d 0 . 0 3 7 1 -0 . 2 5 4 9 -0 . 0 1 8 3 0 . 0 0 9 3 1 . 0 0 0 0 D i r I n d 0 . 0 1 4 1 0 . 2 1 5 1 -0 . 0 2 4 6 1 . 0 0 0 0 A t t D i r -0 . 0 0 8 3 0 . 0 0 5 3 1 . 0 0 0 0 F P 0 . 0 3 8 5 1 . 0 0 0 0 D A 1 . 0 0 0 0 D A F P A t t D i r D i r I n d D i r A u d S i z e I n d 9 7 0 I n d 9 2 5 I n d 9 7 4 I n d 9 7 6 I n d 9 7 8 I n d 9 3 5 I n d 8 0 0 ( o b s = 2 7 6 7 ) . c o r r e l a t e D A F P A t t D i r D i r I n d D i r A u d S i z e I n d 9 7 0 I n d 9 2 5 I n d 9 7 4 I n d 9 7 6 I n d 9 7 8 I n d 9 3 5 I n d 8 0 0 I n d 9 0 5 I n d 9 4 0 I n d 6 0 0 Y e a r 0 9 Y e a r 1 0 Y e a r 1 1

(27)

Appendix D.

Table 1: The linear regression on discretionary accruals including all control variables

Table 2: The mean estimation of total assets

_cons .0566439 .0145719 3.89 0.000 .0280517 .0852362 Year10 -.0024021 .002123 -1.13 0.258 -.0065677 .0017635 Year09 -.0287623 .0029557 -9.73 0.000 -.0345617 -.0229629 Ind600 -.0229604 .0065197 -3.52 0.000 -.0357531 -.0101677 Ind940 -.0537267 .0053481 -10.05 0.000 -.0642203 -.043233 Ind905 -.0435921 .0059484 -7.33 0.000 -.0552637 -.0319204 Ind800 -.04545 .0071715 -6.34 0.000 -.0595215 -.0313784 Ind935 -.0256643 .0067654 -3.79 0.000 -.038939 -.0123896 Ind978 -.0310386 .0054055 -5.74 0.000 -.041645 -.0204323 Ind976 -.0401903 .0051793 -7.76 0.000 -.050353 -.0300277 Ind974 -.033788 .0091058 -3.71 0.000 -.0516549 -.015921 Ind925 -.0252758 .0050035 -5.05 0.000 -.0350934 -.0154582 Ind970 .0046094 .0052893 0.87 0.384 -.0057689 .0149878 Size -.0022964 .0010477 -2.19 0.029 -.0043522 -.0002407 DirAud .0111082 .0131162 0.85 0.397 -.0146277 .036844 DirInd -.0169577 .0133609 -1.27 0.205 -.0431737 .0092583 AttDir -.0789165 .0887381 -0.89 0.374 -.2530336 .0952007 FP2 .0027006 .0029689 0.91 0.363 -.0031249 .0085261 DA Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust

at 13857.95 1731.969 10459.78 17256.12 Mean Std. Err. [95% Conf. Interval] Robust

(Std. Err. adjusted for 1152 clusters in cusip) Mean estimation Number of obs = 3140

Earnings management : the influence of gender diversity