The effect of CEO Masculinity on Earnings Management
Name: Almir Nikocevic Student number: 11143428 Date: 28th June 2017
Word count: 8348
Supervisor: Mario Schabus MSc Accountancy & Control
Statement of Originality
This document is written by student Almir Nikocevic who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
The facial width to height ratio (fWHR) is an indicator for the testosterone levels of individuals. Masculine behavior is linked with testosterone and results in aggression, risk seeking behavior, ego centric behavior and increased risk tolerance in males. This study is the first study that investigates CEO masculinity with earnings management of companies. Using manual collected data from 153 companies from Euronext Paris 100 and FTSE 100 England, I find some evidence for a positive relationship between CEO masculinity and earnings management measured through fWHR and discretionary accruals. This study found some support for a positive association between CEO masculinity and corporate financial decision making. A possible explanation is that more masculine males are induced by their biological marker to pursue more risky decision making.
Key words: CEO, accrual-based earnings management, masculinity, fWHR, testosterone, behavior
Data availability: Data used in this study are obtainable from sources that are mentioned in this research.
C
ontents
Abstract ... 2
1 Introduction ... 4
2 Literature review ... 6
2.1 CEO characteristics and Earnings Management ... 6
2.2 Facial structure, testosterone, masculinity and behaviour ... 7
2.3 Masculine behaviour and earnings management ... 9
3 Methodology ... 10
3.1 Sample ... 10
3.2 Dependent variable ... 11
3.2.1 Discretionary accruals ... 11
3.3 Independent variable ... 12
3.3.1 Facial metrics measurement theory ... 12
3.3.2 Collection of photographs ... 14
3.4 Empirical model ... 16
3.5 Control variables ... 16
4 Descriptive statistics and Empirical Results ... 18
4.1 Descriptive Statistics ... 18
4.2 Multivariate Analysis ... 20
4.3 Robustness Analysis ... 22
5 Conclusion ... 23
5.1 Limitations and further research ... 24
References ... 25
1 Introduction
There is a large body of research available on the influence of Chief Executive Officers (CEOs) characteristics on companies. CEOs have in various ways effect on e.g. firm performance, company profits, tax evading decisions and earnings management. Earnings management is reasonable, legal management decision making and reporting, intended to achieve and disclose stable and predictable results (Healy and Wahlen, 1999, p. 368). It has substantial effect on firms since it influences the company’s financial statements and therefore their appearance and worth. It’s a risky activity since earnings management could be misleading, harmful, lead to misreporting penalties and fraud allegations if done incorrectly (Healy and Wahlen, 1999). Accruals
management and real earnings management are legal activities and are not heavily penalized. The penalties are usually restatements of the annual report or a warning of the auditor of the
company. Accounting fraud however, is illegal and can lead to fines or even a prison sentence. Stakeholders as auditors, shareholders, society, government, NGO’s etc. have a deep interest in representative results, while companies have interest in increasing the profits and growth in order to increase company value through increased stock prices. CEOs have a lot of decision making power within a company and different managerial styles of CEOs, have also different effects on corporate decision making (Borghans et al., 2008; Davidson, Malmendier and Tate, 2005). Therefore CEOs could, based on their experience and background, differ in their preferences regarding the level of earnings management within a company.
Bertrand and Schoar (2003) found that biological factors such as age and background have influence on the managerial style of managers. They also found that managers differ on the grade of risk-aversion, motivation and personal preference based on biological factors. One of the biological factors is masculinity. Masculine behaviour in males is associated with: egocentrism, risk seeking, aggression and maintenance of social status. A proxy for masculine behaviour is the hormone testosterone. Testosterone levels differ between individuals; therefore there is a
difference in risk-seeking, masculine behaviour to be found between managerial styles of CEOs. As mentioned earlier, earnings management can be seen as a risky activity and individuals that have more masculine traits, show more risk-seeking behaviour. This study investigates whether there is a positive relation between CEOs masculinity and the level of earning
management in a company. The results would indicate that biological factors have influence on the managerial style and predict more risk seeking behaviour in more masculine CEOs. In turn these differences in managerial style influence the performance of a company.
Internationalization of the globe has led to increased attention to biological and cultural differences between people and societies. Multicultural environments have become cornerstones
of organisational policies and personal development has become a priority for employees. Even though equal chances and mutual treatment of people exist in western countries, there remains the fact that individual behaviour and performance differ between individuals. In this respect the results of this study will give more insight in the effect of biological factors that predict
individual behaviour. The aforementioned results could be of value for board members,
shareholders, authorities and other stakeholders when appointing a new CEO. Furthermore, the results will contribute to the growing amount of scientific material on the influences of earnings management.
This study is mainly motivated by the studies of Jia, van Lent, and Zeng (2014) and Wong, Ormiston, and Haselhuhn (2011).These papers showed a new variable for CEO characteristic. This study extends to the literature as it is the first study that relates facial masculinity with earnings management in Europe. Lastly, whereas previous studies have investigated masculine behaviour and direct financial misreporting or financial performance in the US, this study will look at company’s earnings management.
2 Literature review
2.1
CEO characteristics and Earnings Management
There is vast amount of research available on the influence of CEOs attributes and characteristics as e.g. CEO gender, nationality, experience, age, skill level and background. Every individual can deliver something a bit differently than another. This is caused by different characteristics that are related to a person. Therefore research also finds that CEO characteristics have in various ways effect on e.g. firm performance, company profits, tax evading decisions and earnings management (Bertrand and Schoar, 2003; Fee, Hadlock, and Pierce 2013; Dyreng, Hanlon and Maydew, 2010; Rule and Ambady, 2008).
CEOs have substantial influence in their companies, especially in small and informal companies. Therefore the characteristics, experience and background of CEOs have an influence on various aspects of companies. An example of CEO characteristics is that newly appointed CEOs appearances can predict companies’ profits (Rule and Ambady, 2008). Furthermore, Bertrand and Schoar (2003) found a relationship between CEO characteristics and financial reporting and decision making. These findings are similar to other research that found a positive relation between CEO characteristic and earnings management (Bamber, Jiang, and Wang, 2010; Brochet, Faurel, and McVay 2011). Malmandier and Tate (2009) found that some CEOs tend to go far to maintain their social status, especially after achieving “Superstar” status. Jia, van Lent, and Zeng (2014) found a positive relationship between masculine behaviour in CEOs and CFOs on financial misreporting. Lastly, Wong, Ormiston, and Haselhuhn (2011) found a positive relationship between masculine behaviour and financial performance.
On the other hand, it could be argued that managing style and decision making is not influenced by personal characteristics. Some research suggests that companies policies and incentives determine decision making instead of the CEO (Chava and Purnanandam, 2010). Other studies find that integrity, companies’ overconfidence, morale and environment influence corporate decision making and therefore earnings management (Dikolli, Mayew, and Steffen, 2012).
Schipper (1989, p. 82) defines earnings management as: “A purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain.”. Another definition is: “Reasonable and legal management decision making and reporting, intended to achieve and disclose stable and predictable results” (Healy and Wahlen (1999, p. 368). These actions are not to be confused with fraud, “cooking the books” or misreporting since earning management is based on legal and acceptable decisions. Earnings management is thought of as
important because it influences the financial statements and in turn the financial statements which have an impact on company value.
According to Gunny (2010), earnings management has two classifications, accrual-based earnings management and real activities management. Accrual-based earnings management is based on accounting policy choices and Judgements (Scott, 2014, p. 445). It translates into companies choice on e.g. amortization, revenue recognition, credit losses, provisions, write offs and inventory. Real earnings management is based on real variables, as R&D, advertising expenses, production and disposal of capital assets (Scott, 2014, p. 446). By making operating decisions the actual cash flow changes instead of that is being reported. This study will look into accrual-based models because it is better measurable, accessible and suitable for this study.
There are two forms of accruals. Firstly, discretionary accruals, that indicate accrual-based earnings management. And secondly nondiscretionary accruals, that indicate real earnings management. Non-discretionary accruals are based on the operating cycle of a business. An example is the growth of the firm. This component creates ‘naturally’ accruals. Discretionary accruals are created by management choices. This means that there is no actual business operating reason to have an accrual other than management involvement. Therefore discretionary accruals are a better proxy for accrual-based earnings management.
In summary, prior literature shows that CEO characteristics and behaviour influence through different managerial styles, corporate actions. Earnings management, influenced by corporate and financial decision making, is also subjected to CEOs characteristics. CEO characteristics are always measured in an indirect way as it is not simple as a financial measure.
2.2
Facial structure, testosterone, masculinity and behaviour
The characteristics of a person are defined by their background, education, experience, age, culture and much more (Huang and Kisgen, 2013; Yim, 2012; Malmanier and Tate 2009; Graham and Narasimhan, 2004). The richness of the literature on this subject is immense. One factor that determines individual’s behaviour and characteristics is the biological background of a person. All individuals differ in appearance and biological composition from each other. There is a vast amount of research available on the relationship between the hormone testosterone and masculine behaviour. Penton-Voak and Chen (2004) found that male individuals were perceived as more masculine when they measured high testosterone through saliva samples as opposed to low level testosterone males.
Testosterone is an anabolic steroid hormone that plays a key role in the development and maintenance of male characteristics as behavior, wellbeing, health, muscle, bones, body hair and
body structure. Research on this subject varies from psychological to molecular level of influence. This includes the explanation that testosterone predicts aggression, risk-seeking behaviour, enhanced competitiveness, desire for social status, deception, dominance and lack of fear (Eisenegger et al., 2010; Pound, Penton-Voak, and Surridge, 2009). Carré and McCormick (2008) found a positive relation by examining the amount of penalty minutes hockey players obtain and the fWHR ratio of the players.
Aggression in the corporate world can be seen as more financial risk taking behaviour (Apicella, Dreber, and Mollerstrom 2014). This also translates into various studies like Apicella et al. (2008) who found a direct positive relationship between testosterone, measured through both the fWHR ratio and saliva and the level of risk preference in an investment set up with true monetary rewards. An example is that more overconfident CEOs tend to be more often involved in value decreasing mergers and acquisitions (Malmendier and Tate, 2008). Other behaviours have influence on empire building and perceived egocentrism of managers (Huang and Kisgen, 2013).
Coates and Herbert (2008) measured systematically each morning at 11:00 in a period of 8 days the testosterone levels of male traders through their saliva. Next to that they measured the financial performance of the traders. They found that traders earned higher profits on days where their morning testosterone levels were high. They concluded that the short-term financial success was related to the testosterone levels of the individuals. They also noted that it’s not necessary related to long term success.
Egocentrism, also associated with testosterone, can lead to principal – agent problems. Becker (1968) argues that individuals base their decision making on a costs and benefits analysis of the situation. A more masculine CEO, that wants to secure his social position could therefore act differently based on his cost and benefits analysis. The weight on social status for example would be more than the weight a less masculine CEO places on his reputation. Masculine individuals also have lower emotional cost from misreporting or other punishments. This causes that the costs in the analysis don’t outweigh the benefits in their opinion, which makes the risky decision more appealing to them. It should be noted that on the other hand, aggressiveness and being able to cope with risks, are found attractive traits in CEOs in tough times of economic disturbance and bad firm performance. A strong CEO might push the company into better market positions and increased performance.
The craniofacial growth of individuals is influenced by the testosterone hormone in conjunction with other growth hormones during the growth. Carré, McCormick, and Mondloch, (2009) used the facial bone sizes to successfully predict behaviour. In this study photographs of
males with neutral expressions were shown to subjects. The subjects were asked to estimate the aggression of the person based on the photograph shown. After that, the males of the
photographs were tested separately for their actual level of aggression. Both the estimation of the subjects and the actual level of aggression were positively associated with the fWHR ratio of the males.
Carré, Murphy and Hariri (2013) found that there is a direct relationship between facial structure and masculine behaviour though the use of MRI brain scans. Prior studies in biology have shown that the facial structure can be used as a biological marker and predictor of testosterone levels, and therefore masculine behaviour (Stirrat and Perrett, 2012). As nature is very complex, this subject is not fully understood. Despite that fact it remains a predictable measure for masculine behaviour.
2.3
Masculine behaviour and earnings management
Biological factors determine partly the characteristics and behaviour of a person. Masculine behaviour is associated with risk-seeking, less emotional cost, lack of fear and much more attributes as mentioned in the previous paragraph. One hormone especially emphasizes masculine behaviour; the hormone testosterone. Therefore it could be hypnotised that masculine CEOs, measured through craniofacial metrics, tend toward more risky behaviour. This translates into the tasks CEO’s have to complete regarding corporate decision making, tax (evasion) and (mis) reporting.
One of the main tasks of CEOs is to generate company value. This can be done through various ways as increasing cash flows, acquisitions, mergers or other strategic decision. Company value is based on the financial statements of a company. These statements are subjected to corporate decision made internally in conjunction with CEO. Society and stakeholders are interested in correct financial statements, therefore society has set in place controls as rules, regulations, penalties and auditors to ensure companies report correct information (Xie et al., 2003; Rahman & Ali, 2006). CEOs have interest in reporting figures that match pre-determined targets or shareholders expectations to not lose company value or their own social status. An example is that reported losses influence shareholders immensely negatively compared to reporting profit (Burgstahler & Dichev, 1997; Leuz, Nanda & Wysocki, 2003). There can be also an egocentric reason for earnings management by the incentive structure of the company. CEOs can choose to use discretionary accruals to change the perception of financial statements.
Previous research found a link between masculine behaviour and facial measures of individuals. The facial width-to-height ratio is positively associated with masculine behaviour.
This study keeps in mind that risk seeking, masculine behaviour can also be caused by several other factors due to the broadness and richness of the field of biology. It could also well be that there are differences between CEOs preferences for more risky earnings management, but that rules and regulations withhold materialization of actions.
Discretionary accruals are created by management choices. This means that there is no actual business operating reason to have an accrual other than management involvement (unlike real earnings management). Therefore, in this study accruals management is used as a proxy for earnings management. With that in mind, the following hypothesis is formulated:
Hypothesis 1: CEO masculinity is positively related to earnings management.
3 Methodology
3.1
Sample
Bureau van Dijk Financials, Bureau van Dijk Auditors, Compustat Fundamentals Annual database and Datastream are used for the necessary variables to calculate earnings management and the control variables. Bureau van Dijk AMADEUS Managers is used for determining the independent variable because it contains CEO names. The data from the sources are matched with the ISIN number (International Security Identification Number), to ensure that CEO names align and financial data of companies are present. The facial width-to-height measures of CEOs are gathered manually. This will be discussed in Chapter 3.3.
The initial sample was created by retrieving the Euronext Paris (NXT) 100 and London FTSE 100 companies of 2014. These indexes were chosen because data on these companies is readily available and represents the countries companies well. Furthermore, it benefits the data gathering process because of the popularity and readily available pictures of CEOs of listed companies. This decision is also based considering the size of this study. The choice for the year 2014 is based on the preference of a recent year, and that the data of this year is sufficiently available (unlike 2016 data). It is common that for international studies the literature uses original U.S. measures for earnings management (e.g. Burgstahler, Hail & Leuz, 2006; Maijoor & Vanstraelen, 2006; Van Tendeloo & Vanstraelen, 2008).
After this, the CEO names were retrieved and linked to the companies. After that, earnings management data was retrieved from Datastream. Datastream didn’t have the financial data for the complete sample. This resulted in a sample of 168, consisting of 85 companies from
England and 83 companies from France. England and France are used because there is much data available on listed companies in these countries, and as discussed in Chapter 3.5, I can control for country heterogeneity. The last reason for choosing these countries is because CEO masculinity and misreporting has been researched already in the U.S. As this study is focused on European counties, it will test difference in geography and add to the empirical knowledge on earnings management and CEO masculinity (Burgstahler, Hail & Leuz, 2006; Maijoor & Vanstraelen, 2006).
From this sample of 168 companies the process of data collection of the independent variable is started; the facial metrics are collected and documented. Of the total sample of 168, 10 good quality pictures of CEOs could not be found or the companies had female CEO’s. Females are not included in this study since the independent variable measurement is only valid for measuring masculine behaviour in males. This resulted in a remaining sample of 158. Lastly, after retrieving control variable data 4 companies were deleted from the sample due to incomplete data. The end sample is 153, 76 companies from England and 77 from France. Table 1 shows the sample.
3.2
Dependent variable
3.2.1 Discretionary accruals
There are different well-known models as DeAngelo Model (DeAngelo, 1986), Healy Model (Healy, 1985) and the Indistry Model (Dechow & Sloan, 1991). This study uses the Modified Jones Model developed by Dechow et al. (1995) because in empirical research it is seen as one of the most accurate models for determining the level of earnings management in a company (Chen, 2010). This model operates with the formula below.
Variable N Total N England N France Total FTSE 100 and NXT 100 200 100 100 Less
Missing EM data 32 15 17
Excluded CEO's (Picture Quality)
7 3 4
Excluded CEO's (Female) 4 2 1 Missing control variable data 4 4 1 Subtotal excluded pictures -47 -24 -23
End total pictures 153 76 77
Table 1
NOA / ATA = β0 + β1(1/LTA) + β2(ΔSales – ΔRec / LTA) + β3(GPPE / LTA ) + |ε| Where:
ATA: Average total accruals NOA: Net operating accruals LTA: Lagged total assets ΔSales: Change in sales
ΔRec: Change in accounts receivable GPPE: Gross PP&E
The data is retrieved from Compustat Fundamentals Annual database and Datastream. One of the most reliable measures for determining earnings management is through discretionary accruals. Since discretionary accruals are not based on real asset generated cash flow movements, it is linked with earnings management. It is calculated through the difference of actual discretionary and non-discretionary accruals. The amount of non-discretionary accruals is estimated with the revenue, receivables and PPE of the company. These indicate the operations of a company and are scaled by total assets. Earnings management is then determined by the difference of total accruals and non-discretionary accruals.
The dependent variable is the discretionary accruals, calculated through the Modified Jones Model based. In literature the model is usually based on fiscal year and SIC codes whereas at least e.g. 10 companies per SIC group are needed. Now it’s only based only on one year, year 2014. The name of the dependent variable is DA. It’s calculated by the formula above. The total accruals are calculated through the net income before extraordinary items – cash flow from operations. Lagged total assets are based on previous year Total assets. Inverse total assets are calculated through 1/lagged total assets. After that, the ΔSales is subtracted by ΔRecievables and then divided by lagged total assets. For the calculation of delta ΔSales and ΔRecievables the data of 2013 was obtained and subtracted from 2014 Sales and Recievables. Lastly the Gross Property, Plant and Equipment (GPPE) is divided by lagged total assets. This study considers earnings management in both ways since the goal can be both income increasing and income decreasing and therefore, both ways of earnings management are considered.
3.3
Independent variable
3.3.1 Facial metrics measurement theory
Masculinity can be measured through brain scans, blood, saliva, questionnaires etc. As mentioned earlier in this study, measurement of masculinity in empirical research is mostly performed through the use of upper face lengths of individuals. This is because craniofacial
growth is determined significantly by testosterone in various growth stages in humans (Lindberg et al., 2005; Vanderschueren and Bouillon, 1995).
The independent variable is called the fWHR (face width height ratio) and is calculated by dividing the width by the height of the face. The width is measured by the distance between the two zygions (cheekbones) of the face (bizygomatic diameter/width). The length is measured from the upper lip to the highest point of the eyelid. With the individual ratio’s it is possible to distinguish high and low masculine CEOs and compare this with earnings management data of the firms. Figure 1 shows an example of the fWHR ratio.
Figure 1: Upper face: Facial width and height ratio (fWHR)
Lefevre et al. (2013) use a model to objectively measure the level of masculine behaviour by measuring the face width and height of the upper face of a person. After measurement the width is divided by the height by which a ratio results. Even though some studies doubt the accuracy of this measure in some circumstances, studies are mainly in favour of this measure. Jia, van Lent, and Zeng (2014) found, using the same method, a positive relationship between the face width-to-height ratio of CEOs and CFOs and financial misreporting. Wong, Ormiston, and Haselhuhn (2011) also used this method successfully to find a positive relationship between and facial masculinity of managers and financial performance.
3.3.2 Collection of photographs
Based on the sample the CEO names are retrieved from the only database containing personal data of CEOs and that is Bureau van Dijk Amadeus – Managers. The dataset is selected on function of “CEO”. Similar functions but named differently (French) are also selected. Results with no CEO names are filtered out and females are not included in this study since the facial width-to-height ratio (fWHR) measurement is only valid for measuring masculine behaviour in males. After that, the right CEO of the firm is carefully extracted from the company’s website and annual report.
The second step is to construct a pictorial database. This is done by the measurement of the facial width-to-height ratio of the CEOs. Therefore, high quality pictures are needed of the individuals. Quality pictures are considered by:
1. Face forward looking: The cheekbones are essential for measuring the correct width. If the face is looking to the side a few degrees, then the width of the face can be incorrectly measured because of the depth difference of one cheekbone to the other. 2. Neutral expression: Lips can go up when smiling, decreasing face height and
increasing the ratio.
3. Resolution: The measurement is done in Adobe Photoshop CS6 (full explanation is found below). Because of the use of a feature “Rulers” with measurement based on pixels, at least 300 x 400 dpi resolution was necessary for accurate measurement. Pictures are obtained from the company website, company’s annual report, website or YouTube videos. 7 CEOs didn’t have an adequate picture or video and were excluded from this research. Pictures were saved and named consistent with their names and company ID. To be sure that the best quality picture is used for measurement, the CEO name is entered in Google and verified if it’s the best one available. Based on be distorted by depth differences. In total for 7 CEOs correct pictures for measurement could the search results the best quality picture is determined and used for this study. A neutral expression is required since the upper face can differ with different emotions. The picture must give a clear angle of the face for measurement. The face needs to be forward looking or else the measurement will not be found.
After storage the pictures are imported in Adobe Phtoshop CS6 through the “Browse in Bridge” function which automatically opens the JPEG file format pictures in the right size. Pictures that were downloaded in PNG format had to be gathered again. After the import and setting the “Ruler” function on, which enables measurement, the canvas (the platform on which the picture is placed) is adjusted by arbitrary image rotation to the right degree of the face (neutral). This is a trial and error process. After positioning the image to the right degree, the
function of “Guides” is enabled. This is a function that locks an area for measurement. By creating the straight lines for the fWHR ratio over the face, the canvas is set up for marking and measurement. The canvas looks similar to Figure 1 now. The next step is the marking by using the locks. A dark blue line is painted across the measurement area. The purpose of this is to be able to retrieve and recollect measured pictures if need be. An example is given below in Figure 2. After the marking the measurement takes place. The function “Line Tool” is used. It’s placed over the markers at the intersection area of the width and height of the face. It shows the width and height of the area highlighted. This is a semi-manual process that is precise on the level of pixels. The width and height is noted and the picture is stored in JPEG format again under a new name based on a document number system related with the ISIN of the company for increased traceability. Of each picture a psd (Photoshop) file format has been stored for when re-measurement was needed. After the gathering the width was divided by the heights of all individuals and the fWHR was obtained.
Figure 2: Example Measurement
To be sure that the measurements are accurate, the second best qualified picture available is measured when the face is slightly tilted or if not looking forward. If the difference in the ratio differs less than 5%, the best quality picture is used as the fWHR value. If there is more than 5% difference, another picture is measured and the fWHR result is compared with the first two. If it is close to one of the first two measured (<5% difference), then the average of these pictures will be used. If all three measures differed, the measurements will were judged and the reason investigated. The course of action was decided per case.
3.4
Empirical model
I run the following model to examine the effect of masculinity of CEOs on earnings management:
DA = b0 + b1*fWHR + b2*BIG4 + b3*LEV_FIRMS + b4*lnTA + b5*FIN_DIS + b6*COUNTRY
Initial checks on the data will be conducted before the estimation of the model. There were no abnormalities found and the initial sample was used. Since the sample is small (N = 154), there is no winsorizing performed. I expect a positive relation between earnings management and CEO masculinity which should be reflected in a significant positive beta 1. Table 2 shows an overview.
3.5
Control variables
In order to control for other influences that could potentially affect the results control variables are chosen. Badertsche et al. (2010) found that companies have a higher reporting quality when audited by one of the Big 4 accounting firms. The companies are also supposed to engage less in earnings management. On the other side, Cohen et al. (2008) found that earnings management is not necessary influenced by having a big 4 auditor, because it’s possible that self-selection of auditors has influence. The first control variable is BIG4 auditor or not. The proxy is called BIG4 and is defined by the auditor in the year of 2014. The data is extracted from Bureau van Dijk Auditors and an dummy variable is created. If the firm has a big 4 auditor it’s coded with 1, otherwise it’s coded as 0. As this research is based on European firms, there was not
Variable Definition Measurement
Dependent variable
DA Discretionary accruals Discretionary accruals calculated with the Modified Jones Model
Independent variable
fWHR Face Width to Height Ratio The width between the cheekbones divided the height of the upper face of a CEO
Control variables
BIG4 Auditor reputation Coded as 1 if auditor is Big4 firm and 0 otherwise LEV_FIRMS Leverage ratio Total liabilities divided by total assets
lnTA Firm size ln of total assets
FIN_DIS Financial distress Dummy variable; if a company reported a consolidated loss, then coded as 1, otherwise a 0
COUNTRY Country name Dummy variable; encoded as 1 if England and 0 if France
Table 2
always full availability of this data. Therefore the auditor of the firm was manually gathered by retrieving the annual reports from the website and investigating who their auditor was in 2014.
Second, we use the leverage of firms as a control variable because higher assets may be an incentive to manage earnings. Firms also want to prevent situations of not meeting debt
covenants (Becker, et al., 1998; Press & Weintrop, 1990; DeFond & Jiambalvo, 1994) because debt covenants can be very costly. The proxy is called LEV_FIRMS and is calculated by dividing the total liabilities by total assets.
The third control variable will be the size of the firm measured by the ln of the total assets. It’s named lnTA. More assets give more opportunity for earnings management. On the other side, larger firms tend to be more scrutinized by rules, regulations and auditors. This makes earnings management harder (Xie et al., 2003; Rahman & Ali, 2006).
The fourth control variable is whether the firm is in financial distress. The proxy for this control variable is FIN_DIS and is encoded as a dummy variable. If the company reports a financial loss then it gets a 1, otherwise a 0. When a firm has a loss it impacts the company value greatly. Therefore, firms with negative net income have greater incentives to use earnings
management (Burgstahler & Dichev, 1997; Leuz et al., 2003).
The last control variable is the country. Lefevre et al. (2012) and Gomez-Valdes et al. (2013) warn for social-cultural controls that would influence this study because masculinity levels can differ between social environments. By taking two different big European countries it can be controlled for social cultural controls and also for the difference in regulations on earnings management between the two countries. The proxy for this control variable is COUNTRY and is coded as 1 for English firms and 0 for firms from France.
4 Descriptive statistics and Empirical Results
4.1
Descriptive Statistics
Table 3 gives an overview of earnings management and the control variables. This includes the whole sample.
The mean for discretionary accruals is less than 0.01%, this indicates that the companies in the sample manage on average less than 0.01% earnings by use of discretionary accruals. This value is in line with prior research. The median is not negative and has a value, common in other literature (Badertscher, 2011; Cohen et al., 2008,). Furthermore, the standard deviation is below average. This is likely due to the small sample size in comparison with usual bigger samples. In addition outliers play a role in the sample as is noticeable with the 10% and 90%.
The mean of the fWHR ratio is in line with prior literature (Jia, van Lent, and Zeng, 2014; Wong, Ormiston, and Haselhuhn, 2011) and deviates approximately 0.05% less than from their mean fWHR. The standard deviation, Q1 and Q3 is a little lower but still approximately in line with prior literature. The reason for the small difference can be due to the difference in sample size. However, this has no significant effect on the research.
The BIG4 control variable shows, as expected, that almost all firms have a big 4 auditor. 97% of the firms in the sample are audited by a Big 4 auditor. The results for the leverage of the firm (LEV_FIRMS) are in line with prior literature. The mean leverage of 61% shows that the firms consist of more debt than assets. A small percentage of the sample reported a loss (9%). The table below shows the Pearson Correlation Matrix. The bold coefficients are significant at 10%.
Table 3
Descriptive statistics overall sample
Variable N Mean Median Std. Dev Q1 Q3 10% 90%
DA 153 0.00000289 0.004 0.017 0.001 0.007 -0.013 0.012 fWHR 153 1.996 1.982 0.155 1.894 2.086 1.822 2.16 BIG4 153 0.973 1 0.160 1 1 1 1 LEV_FIRMS 153 0.614 .603 0.199 0.485 0.728 0.373 0.840 lnTA 153 9,150 9,069 1,346 8,222 10,062 7,662 10,858 FIN_DIS 153 0.098 0 0.298 0 0 0 0 COUNTRY 153 0.496 0 0.501 0 1 0 1
Cohen (1988) and Rosenthal (1996) stated benchmarks for identification of the effect size of correlations and biases: small, medium, large and very large for respectively the values 0.1, 0.3, 0.5 and 0.7. There are no very large correlations, therefore the matrix shows no multicollinearity issues.
The biggest correlation in the tables is as expected, a significant negative correlation between DA and lnTA. Because bigger companies have more scrutiny from the capital markets and have better corporate governance, which makes it harder for CEOs to manipulate accruals (Francis et al. 2009).
For the correlation between discretionary accruals and LEV_FIRMS there is significant negative correlation. This means that there is less earnings management when the leverage is bigger. This makes sense since more debt can be related to more involvement and scrutiny of debtholders, resulting in less possibilities for earnings management.
A medium positive significant relationship is found between COUNTRY and earnings management. This means that there is an influence of a country and their laws on discretionary accruals. Since England and France have different financial regulations and laws this is possible.
Furthermore, a significant positive relationship is found between lnTA and Big 4 auditor. This is logical because a big company is obliged to have a Big 4 auditor for audits. Smaller companies are not obliged to have a big 4 auditor. Regulations are in place for each country to ensure clear guidelines for audits and auditors e.g. when to have a big 4 auditor.
Lastly, there is a significant positive relationship between lnTA and the LEV_FIRMS. This is also logical since the bigger a firm, the more likely it will make use of debt as leverage to increase their return on assets (ROA) since it has more creditability for obtaining liabilities.
Table 4
Pearson R Correlation Matrix
DA IDV_fWHR BIG4 LEV_FIRMS lnTA FIN_DIS COUNTRY
DA 1 fWHR -0,061 1 BIG4 -0,045 0,03 1 LEV_FIRMS -0.136* 0,022 0.132 1 lnTA -0.457* 0,086 0.147* 0.225* 1 FIN_DIS -0,039 0.103 0,054 0.117 0.116 1 COUNTRY 0.322* 0,05 -0,001 -0,017 0,013 0,019 1 * The coefficient is significant at the 10% level.
4.2
Multivariate Analysis
The results of the multivariate analysis will be explained. Table 5 reports two regressions: the results of the OLS regression (I) and Robust regression (II). I chose to run additionally the robust regression as it is considered in literature to generally outperform winsorization and focusses more on overall model fit to deal with influential observations (outliers) (Leone, Minutti-meza, and Wasley, 2015).
The R-squared for the OLS (I) is 0.321 and for the robust regression (II) 0.221. This means that respectively 31% and 22% of the variation in the dependent variable is explained by the
independent variables.
With the OLS regression (I) there is no significant relationship (p value of 0.550) found between the fWHR and earnings management, for which hypothesis 1 with this regression model is not supported. The robust regression however shows a significant positive relationship (p = 0.067) between fWHR and the level of earnings management. This is likely due to the fact that the robust regression is not highly sensitive to outliers (contrary to the OLS regression) (Leone, Minutti-meza, and Wasley, 2015). The relationship is positive, as expected with the
Table 5
Regression results
VARIABLES OLS (I) Robust (II)
fWHR -4.632 6.250* t statistic -0.60 1.85 BIG4 2.859 1.067 t statistic 0.38 0.32 LEV_FIRMS -2.921 -5.834** t statistic -.47 -2.15 lnTA -5.923*** -1.406*** t statistic -6.45 -3.49 FIN_DIS 1.596 -102 t statistic 0.39 -0.06 COUNTRY 11.472*** 4.289*** t statistic 4.84 4.13 Constant 56.602*** 4.212 t statistic 3.15 0.54 Observations 153 153 R-squared 0.321 0.221 *** p<0.01, ** p<0.05, * p<0.1
hypothesis. Based on the robust regression I find support for hypothesis 1; accepting that masculinity is positively related with earnings management.
To check for the effect of outliers I winsorized the data at 1% and 99% and at 5% and 95%. Then I run both type of regressions again. The OLS regression remains insignificant at both levels of winsorizing. For the robust regressions; there is no more significant support for hypothesis 1 when the data is winsorized at 5% and 99%. The robust regression still shows a significant positive relationship when the data is winsorized at 1% and 99%. There is a small improvement of the p value (0.061).
Regarding the control variables there is a significant negative relationship at significance level of <0.05 with the robust regression between the leverage of a firm and discretionary accruals. This suggests that higher leverage of a firm suggests that there is less earnings management performed. The OLS regression doesn’t show a significant relation (0.637). The control variable lnTA, controlling for firm size, shows in both regressions a significant negative relationship with discretionary accruals at p < 0.01. This suggests that a bigger firm is associated with less earnings management. Lastly, a significant positive relation at p < 0.01 is found for the control variable country in both regressions. This suggests that the country of the firm influences the level of earnings management. This is possible as the countries have different regulations regarding the financial statements and accounting rules.
Concluding, there is some support found for a positive relationship between masculinity and earnings management. It’s depending on which regression is used. Mostly, performing a robust regression results in a significant result that suggests support for hypothesis 1, confirming the positive relation of facial masculinity and earnings management. The OLS regression shows no significant p value, not supporting the hypothesis. Therefore, I conclude that there is some support, but the results are sensitive to the way that the data is treated. The results are in line with prior literature that confirms a positive relationship with masculinity, measured through the fWHR, and risk- seeking behavior of CEOs (Wong, Ormiston, and Haselhuhn, 2011; Jia, van Lent, and Zeng, 2014).
4.3
Robustness Analysis
I run additional tests in order to obtain a better understanding of the sample. As mentioned in the previous paragraph, the data was winsorized at 1% and 99%, and then at 5% and 95% to check for the effect of outliers. I found that the robust regressions keep showing significant results with a significance level of 10%, while the OLS regressions remain insignificant. I expected that winsorizing the data would lead to significant results with the OLS regressions. This was not the case. As my data was not normally distributed, I prefer to base my conclusions on the robust regression.
I performed a check for multicollinearity. This happens when at least two or more variables strongly correlate with each other. The Pearson correlation already showed no very large correlations, so that’s the first indication of no multicollinearity. Second, I performed the Variance Inflation Factor (VIF) for all the variables in my model. To reject multicollinearity a value less than 5 is needed. But according to Doane and Seward (2013), a value up until 10 is still acceptable. Above 10 needs further investigation. Table 6 shows the VIF values of the sample. No VIF value of the tested variables shows a greater number than 10. All values are below 5, therefore there is no sign of multi collinearity.
Table 6
Variance inflation factor (VIF) Variable VIF 1/VIF fWHR 1.02 0.980 BIG4 1.03 0.967 LEV_FIRMS 1.07 0.931 lnTA 1.08 0.923 FIN_DIS 1.03 0.967 COUNTRY 1 0.996 Mean VIF 1.04
5 Conclusion
This study investigates the relation between CEO masculinity and earnings management. This is measured through the width and height ratio (fWHR) of the face of the face and the level of discretionary accruals of a firm. It is hypothesized that more masculine behavior is associated with risk-seeking and ego centric behavior (Eisenegger et al., 2010; Pound, Penton-Voak, and Surridge, 2009). This behavior influences corporate decision making of the CEO and can be translated to the level of earnings management a firm has. Other studies found that CEO characteristics influence in various ways e.g. firm performance, company profits, tax evading decisions and earnings management (Bertrand and Schoar, 2003; Fee, Hadlock, and Pierce 2013; Dyreng, Hanlon and Maydew, 2010; Rule and Ambady, 2008). This study investigates if
masculinity of CEOs affects earnings management.
Prior literature relates masculine behavior to testosterone and uses a measurement method to determine the approximate level of testosterone in a person (Carré, McCormick, and Mondloch, 2009; Stirrat and Perrett, 2012). It implies that the level of testosterone determines the craniofacial growth of the skull and therefore, the skull could be used as a measure. This study also uses this method by measuring and dividing the facial width and height of the face in order to obtain a ratio that says something about the level of testosterone a person has.
In order to reject or accept my hypothesis I created a sample of companies from France and England. Companies from Euronext Paris (NXT) 100 and London FTSE 100 were used for the sample. I gathered the earnings management data from Datastream and Compustat. Van Dijk Bureau – Managers was used for collection of CEO names. The fWHR’s was obtained by measuring high quality pictures of CEO’s through a raster graphics editor. The program was precise on pixel level.
The results show some evidence for a significant positive relationship between
masculinity and the level of earnings management. The data is sensitive to the way it is treated as robust regressions show significant results, whereas OLS and winsorizing fail to find a significant relationship. The findings are in line with other research that found a positive relationship
between facial masculinity and misreporting (Jia, van Lent, and Zeng, 2014) and facial
masculinity and financial performance (Wong, Ormiston, and Haselhuhn, 2011). This study adds to the literature in a new way by relating facial masculinity with earnings management. It is an addition to the growing literature that is confirming the relation between CEO characteristics and earnings management (Bamber, Jiang, and Wang, 2010; Brochet, Faurel, and McVay 2011). In addition, the samples of prior literature are based on US companies. This study is confirming
that identical results can be found in European countries as French and English companies ware used for this study.
5.1
Limitations and further research
A limitation of this study is that I did not control for CEO tenure, board size or weak corporate governance. The reason for this is the unavailability of European firm data. These variables are directly linked with CEO’s and therefore would have been useful to add.
Furthermore, in the process of gathering the fWHR’s I noted that obesity plays a role in the measurement. Prior literature found the obesity and testosterone are related to facial shape (Mayew, 2013; Osuna et al., 2006). I expected by running a robust regression that these observations were taken care for. This study didn’t control more than this for this phenomenon, but a further improvement would be to increase to overall size of the sample as winsorizing didn’t show a significant improvement to the OLS regression. Lastly, I note that the intelligence of CEO’s is an important factor that influences his corporate decision making (Azurmendi et al., 2005). This is unmeasurable through the metric used in this study.
As facial masculinity is a new subject in literature, there is much to relate to and investigate. Further research can e.g. investigate the relation between facial masculinity and insider trading or stock option back trading in Europe. As data availability and picture quality is more available in western countries, I advise to select those firms. E.g. pictures of CEOs in foreign countries or names in different alphabets (languages) are harder to retrieve.
References
Apicella, C. L., Dreber, A. , Campbell, B., Gray, P. B., Hoffman, M. & Little, A. C. (2008). "Testosterone and Financial Risk Preferences.” Evolution and Human Behavior 29, 384–90. Apicella, C. L., Dreber, A., & Mollerstrom, J. (2014). “Salivary Testosterone Change Following Monetary Wins and Losses Predicts Future Financial Risk-Taking.” Psychoneuroendocrinology 39, 58–64.
Azurmendi, A., F. Braza , A. Sorozabal, A. Garcia, P. Braza, M. R. Carreras, J. M. Munoz, J. Cardas, and J. R. Sanchez-martin. “Cognitive Abilities, Androgen Levels, and Body Mass Index in 5-Year-Old Children.” Hormones and Behavior 48 (2005): 187–95.
Bamber, L. S., Jiang J., & Wang, I. Y. (2010). “What’s My Style? The Influence of Top Managers on Voluntary Corporate Financial Disclosure.” The Accounting Review 85, 1131–62.
Becker, G. S. (1968). “Crime and Punishment: An Economic Approach.” Journal of Political Economy 76, 169–217.
Bertrand, M., & Schoar, A. (2003). “Managing with Style: The Effect of Managers on Firm Policies.” Quarterly Journal of Economics 118, 1169–208.
Borghans, L., Duckworth, A., Heckman, J. J., & Ter Weel, B. (2008). “The Economics and Psychology of Personality Traits.” The Journal of Human Resources 43, 972–1059.
Brochet, F., Faurel, L. & Mcvay, S. (2011). “Manager-Specific Effects on Earnings Guidance: An Analysis of Top Executive Turnovers.” Journal of Accounting Research 49, 1123–62.
Carré, J. M., & McCormick, C. M. (2008). In your face: facial metrics predict aggressive behaviour in the laboratory and in varsity and professional hockey players. Proceedings of the Royal Society B: Biological Sciences, 275(1651), 2651-2656.
Carre, J. M., McCormick, C. M., & Mondloch, C. J. (2009a). “Facial Structure Is a Reliable Cue of Aggressive Behavior.” Psychological Science 20, 1194–98.
Carre, J. M., Murphy, K. R., & Hariri, A. R. (2013). “What Lies Beneath the Face of Aggression?” Social Cognitive and Affective Neuroscience 8, 224–29.
Chava, S., & Purnanandam, A. (2010). “CEOs Versus CFOs: Incentives and Corporate Policies.” Journal of Financial Economics 97, 263–78.
Chen, T. (2010). Analysis on accrual-based models in detecting earnings management. Lingnan Journal of Banking, Finance and Economics. 2.
Coates, J. M., & Herbert, J. (2008). Endogenous steroids and financial risk taking on a London trading floor. Proceedings of the National Academy of Sciences, 105(16), 6167-6172.
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences, 2nd Edition. Hillsdale: Lawrence Erlbaum
Davidson, R., Dey, A., & Smith, A. (2013). “Executives’ ‘Off-the-Job’ Behavior, Corporate Culture, and Financial Reporting Risk.” Journal of Financial Economics.
Dechow, M., Sloan, G. & Sweeney, P. (1995). Detecting Earnings Management. The Accounting Review, 193-225.
Dikolli, S. S., Mayew, W. J., & Steffen, T. D. (2012). “Honoring One’s Word: CEO Integrity and Accruals Quality,” Working paper, Duke University.
Doane, D. and Seward L. (2013). Applied statistics in business and economics. New York: McGraw-Hill Irwin
Dyreng, S. D., Hanlon, M., & Maydew, E. L. (2010). “The Effects of Executives on Corporate Tax Avoidance.” The Accounting Review 85, 1163–89.
Eisenegger, C., Naef, M., Snozzi, R., Heinrichs, M., & Fehr, E. (2010). “Prejudice and Truth About the Effect of Testosterone on Human Bargaining Behaviour.” Nature 463, 356–59.
Fee, C. E., Hadlock, C. J., & Pierce, J. R. (2013). “Managers with and Without Style: Evidence Using Exogenous Variation.” Review of Financial Studies 26, 567–601.
Francis, J., Richard, C., & Vanstraelen, A. (2009). Assessing France’s Joint Audit Requirement: Are Two Heads Better than One? Auditing: A Journal Of Practice & Theory, 28(2), 35-63
Gomez-Valdes, J., Hunemeier, T., Quinto-Sanchez, M., Paschetta, C., De Azevedo, S., Gonzalez, M. F., Martinez-Abadias, N., Esparza, M., Pucciarelli, H. M., Salzano, F. M., Bau, C. H. D., Bortolini, M. C., & Gonzalez-Jose, R. (2013). “Lack of Support for the Association Between Facial Shape and Aggression: A Reappraisal Based on a Worldwide Population Genetics Perspective.” PLoS One 8.
Graham, J. R., & Narasimhan, K. (2004). Corporate survival and managerial experiences during the Great Depression. manuscript, Duke University.
Gunny, K., (2010). The Relation Between Earnings Management Using Real Activities Manipulation and Future Performance: Evidence from Meeting Earnings Benchmarks. Contemporary Accounting Research, Vol. 27 No. 3
Healy, P.M., & Wahlen, J.M. (1999). "A review of the earnings management literature and its implications for standard setting." Accounting Horizons.
Huang, J., & Kisgen, D. J. (2013). Gender and corporate finance: Are male executives overconfident relative to female executives?. Journal of Financial Economics, 108(3), 822-839. Jia, Y., van Lent, L., & Zeng, Y. (2014). Masculinity, Testosterone, and Financial Misreporting. Journal of Accounting Research, Vol. 52 No. 5
Lefevre, C. E., Lewis, G. J., Bates, T. C., Dzhelyova, M., Coetzee, V., Deary, I. J., & Perrett, D. I. (2012). “No Evidence for Sexual Dimorphism of Facial Width-to-Height Ratio in Four Large Adult Samples.” Evolution and Human Behavior 33, 623–27.
Lefevre, C. E., Lewis, G. J., Perrett, D. I., & Penkle L. (2013). “Telling Facial Metrics: Facial Width Is Associated with Testosterone Levels in Men.” Evolution and Human Behavior 34, 273–79.
Leone, A., M. Minutti-Meza, and C. Wasley. 2015. Influential observations and inference in accounting research. Working paper.
Lindberg, M. K., Vandenput, L., Moverare Skrtic, S., Vanderscheuren, D., Boonen, S., Bouillon, R., & Ohlsson, C. (2005). “Androgens and the Skeleton.” Minerva Endocrinologica 30, 15–25.
Malmendier, U., & Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market's reaction. Journal of Financial Economics,89(1), 20-43.
Malmendier, U., & Tate, G. A. (2009). “Superstar CEOs.” Quarterly Journal of Economics 124, 1593–638.
Mayew, W. J. “Reassessing the Association Between Facial Structure and Baseball Performance: A Comment on Tsujimura & Banissy (2013).” Biology Letters 9 (2013): 1–3.
Osuna C., J.; R. Gomez-Perez; G. Arata-Bellabarba; and V. Villaroel. “Relationship Between BMI, Total Testosterone, Sex Hormone-Binding-Globulin, Leptin, Insulin and Insulin Resistance in Obese Men.” Systems Biology in Reproductive Medicine 52 (2006): 355–61.
Penton-Voak, I. S., & Chen, J. Y. (2004). High salivary testosterone is linked to masculine male facial appearance in humans. Evolution and Human Behavior,25(4), 229-241.
Pound, N., Penton-Voak, I. S., & Surridge, A. K. (2009). “Testosterone Responses to Competition in Men Are Related to Facial Masculinity.” Proceedings of the Royal Society B: Biological Sciences 276, 153–59.
Rosenthal, J.A. (1996). Qualitative descriptors of strength of association and effect size, Journal of Social Service Research, 21(4), pp.37-59.
Rule, N. O., & Ambady, N. (2008). “The Face of Success.” Psychological Science 19, 109–11. Schipper, K. (1989). Commentary on Earnings Management. Accounting Horizons, 91-102. Stirrat, M., & Perrett, D. I. (2010). “Valid Facial Cues to Cooperation and Trust.” Psychological Science 21, 349–54.
Vanderscheuren, D., & Bouillon, R. (1995). “Androgens and Bone.” Calcified Tissue International 56, 341–46.
Wong, E. M., Orminston M. E., & Haselhuhn, M. P. (2011). “A Face Only an Investor Could Love.” Psychological Science 22, 1472–83.
Yim, S. (2013). The acquisitiveness of youth: CEO age and acquisition behavior. Journal of Financial Economics, 108(1), 250-273.
