Influence of the gender of CEO/CFO on credit rating

(1)

Influence of the gender of

CEO/CFO on credit rating

Amsterdam Business School

Faculty of Economics and Business, University of Amsterdam

Msc Accountancy & Control, Control track

Master thesis Control

Supervisor dr. Bo Qin

12420 words

Tim van Houten

6113958

(2)

Statement of Originality

This document is written by student Tim van Houten who declares to take full responsibility for the contents of this document.

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

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

(3)

Abstract

This study examined if credit rating agencies impound the gender of a CEO/CFO in their risk assessment. The findings provide evidence that there is a significant difference for the credit rating of a company when there is a transition from male to female CEO/CFO. For a transition from female to male CEO/CFO is found that there is almost no difference for the credit rating of a company, but the result is not significant. This study examined also if companies with credit rating

concerns a more likely to have female than male CEO/CFO. However the results for this research were not significant. Thus there is no evidence if companies with credit rating concerns a more likely to have female than male CEO/CFO.

(4)

1. Introduction

In the fall of 2008 the financial crisis began. Major companies were bankrupt. One of the main reasons was that the companies took too much risk (Bullard, Neely and Wheelock, 2009). Also companies went bankrupt with high credit ratings a few months before they went bankrupt (e.g. Lehman Brothers and AIG). While the credit rating of a company is ranked as the second-most important concern in financial decision (Graham and Harvey, 2011). Palvia, Vähamaa and Vähamaa (2014)1_{found evidence that female CEOs (Chief executive officer) and} board Chairs asses risk more conservatively and that smaller banks with female CEOs and board Chairs were less likely to fail during the financial crisis. Thus it seems that the gender of a CEO or CFO (Chief financial officer) have influence on the default risk of a company. The goal of this study is to examine if credit rating agencies take account of the gender of CEO or CFO in their risk assessment. In addition this study examined if companies with credit rating concerns prefer to have a female CEO/CFO instead of a male CEO/CFO after a downgrade of the credit rating.

There is a lot of prior literature about the gender difference in risk taking behaviour. They all have almost the same conclusion, that women take less risk than men. Schubert (1999) and Eckel and Grossman (2008) did an experiment with gambling. In both experiments women were less risk taking than men. However Schubert came also to the conclusion that in taking investment decisions women are not more risk averse than men. But Eckel and Grossman found that in taking investment decisions women are also more risk averse than men. Dwyer (2002) didn’t use experiment, but data of a survey of mutual fund investors. Result was that women are less risk taking than men in their most recent, largest, and riskiest mutual fund investment decisions.

1_{Palvia, A., Vähamaa, E. and Vähamaa, S. (2014) “Are Female CEOs and}

Chairwomen More Conservative and Risk Averse? Evidence from the Banking Industry During the Financial Crisis” is published online on 25 July 2014, Springer

(6)

There is also prior literature about the difference in risk taking between male and female CEOs. Elsaid and Ursel (2011) did research in to the relation between changes in various risk measures for the firm and a change from a male to a female CEO. They found that the change from a male to a female CEO is related with a decrease in several measures of firm risk taking. Faccio et al. (2014) examined the relation between CEO gender and corporate risk-taking with a regression analysis. They found evidence that female CEOs are associated with less risky companies.

There is no prior literature about the influence of the gender of a

CEO/CFO on the credit rating. Ashbaugh-Skaife (2006), Cheng and Subramanyam (2008), Lee (2008), Bradley and Chen (2011) and Kuang and Qin (2013)

research to other variables than gender of CEO/CFO, which may have influence on the credit rating. Ashbaugh-Skaife (2006) and Bradley and Chen (2011) found evidence that firms with strong corporate governance benefit from higher credit ratings relative to firms with weaker corporate governance. Cheng and

Subramanyam (2008) found that financial (equity) analysts following are

negatively related to a firm’s credit rating. Lee (2008) found that employee stock options have effect on credit ratings. Kuang and Qin (2013) found that credit rating agencies take into account managerial risk-taking incentives in their credit risk evaluation.

This thesis is a quantitative research. The research used the two models of Kuang and Qin (2013) to explore the influence of the gender of CEO/CFO on credit rating of a company. For the use of those two models for this research, the variable of managerial risk-incentives was changed in 2 different variables for transition from male to female CEO/CFO (POST and TRANSITION) based on the transition variable of Francis et al. (2014). First is examined if there is a

difference of the credit rating of a company, when there is a male or female CEO/CFO. The first model looks at the question if credit rating takes into account the gender of CEO or CFO in their credit risk evaluation, with as independent variable a dummy what look to the pre- and post-transition period from a male to female CEO/CFO transition (POST). The regression analysis shows that there

(7)

is a significant difference in the credit rating between a female CEO/CFO and male CEO/CFO. To extent this finding the regression model is also done for a sample of transition from female to male CEO/CFO. This regression didn’t extent the prior finding, because there was almost no difference in the credit rating of a company between female and male CEO/CFO.

Next this thesis explores if credit rating have effect on the decision to change from male to a female CEO/CFO, in particular for companies with credit rating concerns. There are two-measures to determine if a company has credit rating concerns: when a company have a downgrade of the credit rating relative to priors year, and when a company have a downgrade of the credit rating to the lower edge of the investment category (i.e. BBB-). This two measure are based on the findings of Kisgen (2006) and Kuang and Qin (2013) who used those 2

variables also. The second model examined if companies with credit rating concerns, when there is a downgrade of the credit rating, prefer to have a female CEO/CFO instead of a male CEO/CFO. This research used as dependent variable a dummy that is one if there was a transition from male to female, zero otherwise (TRANSITION). This measure is used, because when companies have credit rating concerns they then have a transition from male to female CEO/CFO. The regression shows no significant coefficient, thus the findings were not reliable. The sample size of the female to male transition was too small (Flack and Chang, 1987).

There are several contributes for this study. In the first place this is the first study which examined the influence of gender of CEO/CFO on credit rating. In this study a significant difference of the credit rating between the pre- and post-transition period of male to female CEO/CFO was found. Second place this research help explore how credit rating agencies are evaluating the credit rating. There was already prior literature (Ashbaugh-Skaife, 2006; Cheng and

Subramanyam, 2008; Lee, 2008; Bradley and Chen (2011); Kuang and Qin, 2013) which help to explore how credit rating agencies are evaluating the credit rating, but not for the perspective of the gender of CEO/CFO in the rating process. How the credit rating agencies evaluate isn’t exactly know. Understanding of the rating process can help companies to improve their credit rating and help to

(8)

understand for people (e.g. investors) why a credit rating is low or high. In the third place in this research is found that there is a significant difference of the credit rating between the pre- and post-transition period from male to female CEO/CFO. This finding can boost the transition from male to female CEO/CFO. In began 2015 there are 23 female CEO’s out of the S&P 500 companies. This means that 4,6% of the CEO’s are women. This is a very low percentage.

The structure of the paper is as follow. Section 2 provides prior literature and development of the hypothesis. Section 3 introduces the research method. Section 4 discuss the results of this research. Section 5 presents the results of the robustness check. Section 6 is the conclusion.

2. Literature Review

To examine the research question a literature review was done and a hypothesis was made. The hypothesis is based on the findings in the literature review. This chapter starts with elaborating what risk is, then the agency theory is explored Then the gender difference in risk taking is explored, In the fourth place the information content of credit rating will be studied and at last the hypothesis will be explored.

2.1 What is risk?

What is risk-taking behaviour exactly. When is a decision of CEO’s risky? The paper of March and Shapira (1987) is used to elaborate what risk and risk taking by CEO’s is.

March and Shapira did a research to the relation between decision theoretic conceptions of risk and conceptions of risk held by managers. In

classical decision theory, risk is most commonly defined as reflecting variation in the distribution of possible outcomes, their likelihoods, and their subjective values. Risk is measured by the variance of the probability distribution of

(9)

possible gains and losses. A larger variance means that it is more risky. The choice of taking risk is affected by the expected return of an alternative. Decision makers prefer larger expected return above smaller expected return, if the factors are constant (Lindley, 1971).

The theory assumes that decision makers deal with risk by calculating all the possible outcomes and then choosing one of those outcomes. But human decision makers can define risk significantly different than the theoretical literature. Managers are associating risk with negative outcomes instead of possible outcomes. Risk averse (seeking) managers prefer relative low (high) risks, that result in lower (higher) expected return in order to reduce (increase) the variation in possible outcomes. In the next paragraph the role of the agency theory (Jensen and Meckling, 1976) on the risk taking behaviour of CEO’s will be discussed.

2.2 Agency Theory

Agency theory of Jensen and Meckling (1976) elaborated the relationship between principal and agent (e.g. CEO). Agency theory helps to elaborate the risk-sharing problems as one that arises when cooperating partners have different interest (Eisenhardt, 1989). Jensen and Meckling (1976) define an agency relationship as contract under which one or more persons (principal(s)) engage another person (agent) to perform some service on their behalf which involves delegating some decision making authority to the agent.

The agent based his decisions on what is the best for him, but this decision isn’t always the best for the principal. The agent has different interest (e.g. goals and risk preference) than the principal. A problem for the principal is that the agent has more information (asymmetric information). The more information a principal has, the less the agent has impact on the outcome of the company. The principal used incentives for the agent and is monitoring the agent to diminsh the difference between the interest and information of an agent and a principal.

(10)

Essential part of the agent theory is the decision between measuring the cost of behaviour and the cost of outcomes and transferring risk to the agent. When an agent becomes increasingly less risk averse, it is more attractive to pass the risk to the agent using an outcome-based contract. But if an agent is more risk averse, it is increasingly expensive to pass risk to the agent. It is similar for the principal. When a principal is more risk averse, it is more attractive to pass the risk to the agent (Eigenhardt, 1989). Because the shareholder value of a company can increase harder, when they take more risk. With an outcome-based contract the company trigger the agent to choose the option that increase rather than decrease the shareholder value. An outcome-based contract is an incentive for an agent, because the agent gets higher compensation when the shareholder value increase (Murphy, 1999).

2.3 Relation between gender and risk taking behaviour

There is a lot of prior literature about the relation between gender and risk taking behaviour. The conclusion of the prior literature is almost the same, but there are a few differences. In this paragraph the difference between the results of the prior literature of the risk taking behaviour of women and men will be discussed.

Schubert (1999) did a research to gender-specific risk attitudes. They did an experiment to examine the gender-specific risk propensity in decisions relevant for investors and managers. The experiment was designed in two treatments: context and abstract. In the context treatment the risk behaviour of male and female subjects are directly measured for risky choices in the form of investment and insurance decisions. The abstract treatment is a control

treatment. For the abstract treatment they use the same risky choices, but as an abstract gambling decision.

The results in the context and abstract treatment are different. The context treatment had as result that women didn’t make more risk averse choices than men. But the abstract treatment found that women make more risk

(11)

averse choices than men in the gain-gambling frame. In the loss-gambling frame the result was that women make less risk averse choices than men.

Eckel and Grossman (2008) examined the difference in risk attitudes of female and male university students. They did an experiment with gambling under three frames: an abstract frame where the two highest-payoff gambles carry the possibility of losses, an abstract frame with no losses and an

investment frame. For each frame there are five 50/50 gambles.

The result is that women are significantly more risk averse than men in all the three frames. Because women may feel less comfortable in risk-taking than men and they have less confidence in risk-taking than men. The result of the investment frame is the opposite of Schubert (1999) result for the investment frame, who find that women aren’t more risk averse than men.

Dwyer (2002) did a study into the question if investor gender is related to risk taking in mutual fund investment decision. This research used data from national survey of 2000 randomly selected mutual fund investors. The survey asked question about which kind of mutual funds they have and how they purchased these funds. There are also a series of questions to determine their knowledge of basic financial concepts. The result was that women take less risk than men.

However they made also a regression with other factors (age, education and income) that influence risk taking, because prior literature suggests that those factors also influence risk taking. Dwyer examined this regression model with and without control for investor knowledge. The result is that women take less risk than men, but the gender effect is weaker if the model control for investor knowledge.

Johnson and Powell (1994) did a study in the difference in the nature of decisions taken by males and females. They use literature to examine this. This study used two populations. There is a ‘non-managerial’ population. This is a population of individuals who hadn’t a formal management education. And in contrast there is a ‘managerial’ population. This is a population of potential and

(12)

actual managers who had a formal management education. There were different results for those two populations. For the ‘non-managerial’ population was the result that women are more risk averse then men, but for the ‘managerial’ population was the result that women and men have similar risk propensity and make decisions of equal quality.

There is also recent prior literature, which uses female CEO’s as population. Elsaid and Ursel (2011) did a research to the relation between changes in various risk measures for the company and a change from a male to a female CEO. This is examined with a regression analysis. The dependent variable is the change in the risk profile of the company from pre to post CEO succession. They used the following risk measures: financial leverage, research and

development expenses as a per cent of sales, cash holding as a per cent of sales, operating leverage and the standard deviation of cash flows adjusted for sector average standard deviation. The data is from Standard and Poor’s Execucomp database for CEO successions between 1992 and 2005. They found that the change from a male to a female CEO is related with a decrease in several measures of firm risk taking, because the results of the regression shows that gender related variables (e.g. change in CEO gender) are important key variables for determining the risk of a company and for the decision to take a female as CEO.

Another paper is the working paper of Faccio et al (2014). They examined the relation between CEO gender and corporate risk-taking with a regression analysis. Corporate risk-taking is measured as financial leverage, the volatility of the firm’s operating return on assets and the likelihood of survival. They use a sample of European companies from Amadeus Top 250000. Result is that female CEOs are associated with less risky firms. They give also a few possible

explanations. First one is self-selection, female (male) CEO’s choose to work for low (high) risk firms and/or firms that have experienced a decline (increase) in risk (Bandiera, 2014). Second, women are less risk tolerant than men in general (Johnson and Powell, 1994). As third, women are less overconfident and less overconfident CEO’s reduce risk. As fourth, compensation and incentives

(13)

structure can also be an explanation. Low risk firms prefer to offer fixed pay contracts and that is maybe more attractive for female CEO’s (Bandiera, 2014). As fifth, if corporate risk-taking is positively correlated with the possibility of losing your job. Because finding a new job is more difficult for women than men, female CEO’s choose to work for low risk firms or to reduce the firms risk. The last one is that high risk firms involves with longer working hours and less flexible schedules, women might choose for low risk firms, because then they have more time for the children and household tasks.

What can be the explanation why there is a gender difference in risk taking? Corson and Gneezy (2009) give three explanations in their research to gender difference in economic experiments. The first one is that women and men have different emotional reactions to risky situations. An emotional experience is more strongly for a woman than men, this can affect the utility of risky choice (Harshman and Piavo, 1987). Emotions can also have an affect on the

perceptions of probability. Women tend to feel fear and men tend to feel anger in identical situations (Grossman and Wood, 1993).

The second one is overconfidence. Women and men are both

overconfident, but men are more overconfident in their success in uncertain situations (Lundeberg, Fox and Punccohar, 1994). Estes and Hosseini (1988) found that for investment decisions women have substantially less confidence than men.

The last one is the different interpretation of risky situations, risk as challenge or threats. Women interpreted risky situations as threats and males are interpreted risky situations as challenge (Arch, 1993).

2.4 Credit rating

A credit rating is a method to evaluate the credit worthiness of a company or a government. The evaluation is done by credit rating agencies. Credit rating agencies use qualitative and quantitative information to evaluate the credit rating. Standard & Poor’s (S&P), Moody’s and Fitch are the major credit rating

(14)

agencies (Lawrence, 2010). The credit rating of credit rating agencies is a reliable measure of risk (Ederington, 1987; Lawrence, 2010).

How the credit rating agencies evaluate isn’t exactly know. Credit rating agencies receive a considerable quantity of strict confidential sensitive

information of a company2_{. The Securities and Exchange Commision in}

December 2008 and in November 2009 decided that rating agencies are required to be more transparent on their methodologies, assumptions, and track records in developing of ratings, because of the Enron, Worldcom and other big scandals (Federal Register, 2009a; Federal Register, 2009b; Lawrence, 2010). Credit rating agencies elaborated that they not only use objective measures of credit risk, but the rating methodologies have also subjective elements (UN General Assembly, 2013).

There isn’t a lot of prior research about which variables have effect on the credit rating. Kuang and Qin (2013) found that credit rating agencies impound managerial risk-taking incentives in their credit risk evaluation. Lee (2008) examined that employee stock options have effect on credit ratings. Ashbaugh-Skaife (2006) and Bradley and Chen (2011) examined that firms with strong corporate governance benefit from higher credit ratings relative to firms with weaker corporate governance. Cheng and Subramanyam (2008) found that financial (equity) analysts following are negatively related to a firm’s credit rating.

Kuang and Qin (2013), Cheng and Subramanyam (2008) and Ashbaugh-Skaife (2006) all made a regression model with credit rating of a firm as

dependent variable. They have only examined for different kind of independent variable. Kuang and Qin (2011) use two different kind of independent variable for managerial risk-taking incentives, the sensitivity of managerial wealth to the volatility of firm performance and the sensitivity of managerial wealth to firm performance. Cheng and Subramanyam (2008) model credit rating as a function of two independent variables, analyst following and institutional holdings.

(15)

Ashbaugh-Skaife (2006) models credit rating as a function of firm characteristics and corporate governance attributes.

Bradley and Chen (2011) made a regression model with cost of debt as dependent variable. They use two proxies for the cost of debt, the credit rating of a firm and the yield spread for it’s publicly traded, senior unsecured bonds.

2.5 Hypothesis

Most of the time the prior literature concludes that women take less risk then men (Schubert (1999); Dwyer (2002); Eckel and Grossman (2008)). Corson and Gneezy (2009) give 3 reasons why women take less risk than men: emotions, less overconfidence and risk as a threat rather than challenge. And there is also recent prior literature that used female CEO’s as research population to examine risk taking by women compared with men (Elsaid and Ursel, 2011; Faccio et al., 2014). Both came to the conclusion that women CEO’s take less risk than men CEO’s. For evaluating the credit risk is important to know which variables have effect on the risk. If prior literature elaborates that women CEO’s take less risk then men CEO’s, you will expect that credit rating agencies take this in to account in their risk assessment.

Hypothesis 1:

Credit rating agencies are impounding the gender of CEO/ in their risk assessment

Female CEO/CFO take less risk then men CEO/CFO. If a company switch from a male CEO/CFO to a female CEO/CFO the expectation is that the company will take less risk after the transition. Taking less risk result in lower credit risk and lower credit risk will lead to a better or more stable credit rating of the company.

Companies which experienced a recently downgrade of their credit rating, will expect to have more credit rating concerns. If a company have credit rating concerns, then they are looking for triggers that can lower their risk. A switch from male to female CEO/CFO can lead to lower firm risk and this is a trigger for

(16)

companies whit credit rating concerns to switch from male to female CEO/CFO. Thus the expectation is that companies with credit ratings concerns are more likely to switch from male to female CEO/CFO, than companies without credit rating concerns.

Hypothesis 2:

Companies that have credit rating concerns, because they recently experienced a downgrade of their credit rating, are more likely to switch from male to female CEO/CFO, than companies without credit rating concerns.

3. Research method

This section explains the research method for this research. The two hypothesis are examined with two different models, both models are based on the model of Kuang and Qin (2013). First the model and the variables of both models are elaborated and as second the sample distribution and the summary statistic are discussed.

3.1 Empirical models

This research uses the empirical models of Kuang and Qin (2013) to examine the hypothesis. They use regression analysis for their model. Regression analysis is a quantitative research method. The variable of managerial risk-incentives in the model of Kuang and Qin will change in a variable of the pre- and post-transition period for transition from male to female CEO/CFO. We get the following

regression model for hypothesis 1:

𝑅𝐴𝑇𝐼𝑁𝐺𝑖,𝑡 = 𝛽0+ 𝛽1∗ 𝑃𝑂𝑆𝑇𝑖,𝑡+ 𝛾 ∗ 𝐶𝑂𝑁𝑇𝑅𝑂𝐿1𝑖,𝑡+ 𝜀𝑖,𝑡 (1)

RATING proxies for the credit risk assessment of company i in year t, POST

(17)

CEO/CFO of company i in year t, CONTROL1 is a vector of control variables and ε is an error terms.

This research uses a sample of CEO and CFO from companies of the United States. The data is from Compustat, ExecuComp Executives and Director Ownership and CRSP. The research period is from 1992 till 2014. For this period is data available.

The regression model has independent, dependent and control variable. First the dependent variable will be discussed, then the independent variable and as last the control variables.

The dependent variable is the credit rating of companies (RATING), because credit rating is endogenous for the hypothesis. This research follows Kuang and Qin (2013) and focus on Standard & Poor’s Long Term Domestic Issuer Credit Rating. Following Kuang and Qin (2013) and Cheng and

Subramanyam (2008) and the Standard & Poor’s ratings get a value from 1 to 20. A higher value corresponds to higher credit risk.

The independent variable for model (1) is pre- and post transition period for transition from male to female CEO/CFO (POST). POST is a dummy variable that equals zero if a year is before the CEO/CFO transition and is one if a year is after the CEO/CFO transition year. This variable is used, because the hypothesis expects that the gender of CEO’s is impounded in the risk assessment. The gender of the CEO/CFO has to have influence on the credit rating. The POST variable is exogenous. The effect of POST has to be significant in the regression model. The data for the gender of the CEO and CFO is from ISS database.

Gender of CEO/CFO isn’t the only variable that can have an effect on the credit rating. If the regression model only use POST as independent variable, then the result can be biased, because other variables have also effect. Other variables that have effect on risk measures are included in regression model as control variable. There are a number of control variables. The control variables are selected from prior literature (Kuang and Qin (2013); Ashbaugh-Skaife

(18)

(2006); Cheng and Subramanyam (2008)). This is a set of variables that measure the companies financial risk and information risk.

Financial risk variables are return on asset (ROA), leverage (LEV), interest coverage (COVER), loss of a company during the current year (LOSS), investment in intangible assets (INTAN) and new equity capital raised during the year (EQ).

Higher return on asset, higher interest coverage, or a lower leverage ratio often results in lower credit risk. Expectation is then that return on asset and interest coverage has a negative relationship with credit rating and that leverage ratio has a positive relationship. The loss of a company during the current year is included as dummy variable. This has a positive relationship with the credit rating, because companies with a loss have greater likelihood of default. Other control variable is investment in intangible assets, because creditors evaluate intangible assets differently then tangible assets. Investment in intangible assets has no expectation relate to the credit rating. Higher investment in intangible assets can have a positive effect on the future profit, but increases the likelihood of default. New equity capital raised during the year is included as dummy variable. This variable has also no expectation regarding the credit rating. Because when it is positive, it mitigates agency problem between shareholders and debtholders, and when it is negative, it is more difficult to raise debt capital.

Market-based measures of financial risk variables are stock return (RET), volatility of net income (SDNI), volatility of monthly return (SDRET), book-to-market ratio (BTM), firm size (MV) and financial industry (FIN). SDNI and SDRET expect to have a positive relationship on the credit rating. There is no

expectation of the relationship between stock return and credit rating and between book-to-market ratio and credit rating. Because Bhojraj and Sengupta (2003) show arguments for a positive and negative relationship between stock return and credit rating. Bohjarj and Sengupta also hypothesized a negative relation between default risk and market ratio, but lower book-to-market ratio can also indicate increased growth opportunities, which indicate decreased default risk. Larger firms and firms in the financial industry expect to have a negative relationship with the credit rating, because this variables are less riskier.

(19)

Information risk variables are financial transparency (TRANSP) proposed by Gu (2007) and abnormal accruals (ABSACC). TRANSP expect to have a

negative relation with credit rating and ABSACC expect to have a positive relation with credit rating.

The second model of Kuang and Qin (2013) examined the effect of a downgrade of the credit rating on the managerial risk-taking incentives. This equation can be used to examine the second hypothesis. We get the following regression model for hypothesis 2:

𝑇𝑅𝐴𝑁𝑆𝐼𝑇𝐼𝑂𝑁𝑖,𝑡 = 𝛽0+ 𝛽1∗ 𝐼𝑁𝐷𝑖,𝑡+ 𝛾 ∗ 𝛥𝐶𝑂𝑁𝑇𝑅𝑂𝐿1𝑖,𝑡+ 𝜀𝑖,𝑡 (2)

TRANSITION is a proxy for the year that a company change from male to female

CEO/CFO. TRANSITION is a dummy variable. TRANSITION is one if in the year t-1 there was a male CEO/CFO and in year t a female CEO/CFO, and zero if there is no transition from male to female CEO/CFO. TRANSITION is the dependent variable. OLS regression with a dummy variable as dependent variable will lead to biased and inconsistent estimates. Logit model are used for models in which the dependent variable is a dummy variable. Thus for model (2) a logit model is used.

IND proxies for an downgrade of the credit rating (DOWN) or for an downgrade of the credit rating to the lower edge of the investment category (DOWNBBB-), for instead BBB- . If the credit rating of a company in year t is worse than in year t-1, then DOWN is one and zero if the credit rating is better or stay constant. The expectation is that companies have credit rating concerns, if there was recently a downgrade of the credit rating. Such concerns will trigger companies to change their CEO/CFO from male to female to get a better or stable credit rating, because female CEO/CFO take less risk than male CEO/CFO (e.g. Elsaid and Ursul, 2011).

The second independent variable is DOWNBBB-. If the credit rating of a company downgrade to the lower edge of the investment category from year t-1 to year t, then is DOWNBBB- one and otherwise zero. We measure also for this variable, because companies with a downgrade of their credit rating to the lower

(20)

edge of the investment category will be most concerned about their credit rating (Kisgen, 2006). When companies are most concerned about their credit rating is the expectation that they switch from male to female CEO/CFO, because female CEO/CFO take less risk than male CEO/CFO (e.g. Elsaid and Ursul, 2011).

𝛥𝐶𝑂𝑁𝑇𝑅𝑂𝐿1 is a vector of the first difference of CONTROL1 and ε is an error terms. The first difference of control variable is taking, because for

example the difference of return on ROA of year t with year t-1 have effect on the decision to switch from CEO/CFO for the company. A CEO/CFO is responsible for the ROA. A decrease of ROA relative to last year, have a positive influence on the decision to switch from CEO/CFO, because the performance of the CEO/CFO is worse (Denis and Denis, 1995).

3.2 Sample distribution and summary statistics

The sample consists of the firm-year observations from 1992 to 2014 of the years that there was a male CEO and the years after transition to a female CEO. The data are received from three different databases. The accounting data is from COMPUSTAT Industrial Annual. Data of the gender of the CEO is from Execucomp, when there information about who was the CEO for a year was missing, then the company’s websites and other business websites (e.g. google.com and bloomberg.com) were used for further searching. Data of the market return is from CRSP.

To examine the gender effect on credit rating, the sample has to consist of a pre- and post-transition periods for male to female CEO. The research of

Francis et al. (2014) is the base for filtering our sample. In our sample the following filters were constructed: Male and female CEO must be in office

consecutively for at least one year excluding the transition year, if a firm changes its CEO from male to female more than once, then only count the most recent change for each firm. A filter which Francis et al. (2014) used is that male and female CFO must be in office consecutively for at least four years including the transition year, but if this filter is used for this research the sample will be very small (30 transition cases). To get a bigger sample the years of being in office consecutively is lowered from four to two years including transition year, this

(21)

change result in a sample with 47 transition case. This is still a small sample (Flack and Chang, 1987).

Not only data of CEO gender are used, but also data of CFO gender are used to increase the sample. For instead Francis et al. (2014) are also examined the impact of CFO gender on financial reporting decision making. Data of the gender of the CFO is from Execucomp from 2006 to 2014, if there is missing information about who was the CFO for a year, then the company’s websites and other business websites (e.g. google.com and bloomberg.com) were used for further searching. The following filters were used to add CFO in to the sample: Male and female CFO must be in office consecutively for at least one year

excluding the transition year, if a firm changes its CFO from male to female more than once, then only count the most recent change for each firm, if a firm changes its CEO and CFO from male to female, then only count the most recent change for each firm. The sample of CEO and CFO consist of 84 cases of male to female transitions. This sample is then merged with the COMPUSTAT and CRSP sample. To mitigate the influence of extreme values on the analyses, all continuous variables are winsorized at the 1 percent and 99 percent levels. The final sample consists of 1050 firm-year observations with 84 cases of male to female

transitions.

For the robustness check the sample female to male transitions are needed. The sample of female to male transition consists of 345 firm-year observations with 50 cases of female to male transitions.

The final sample shows that from 1992 until 1996 there are no cases of male to female transition. From 1997 until 2006 there were most of the time 1 case of male to female transitions for each year, with 2000 (4 cases) and 2006 (5 cases) as outliers. In 2007 (10 cases) there is an increase of male to female transitions, this stay around the 9 to 10 cases till 2011. 2012 is the year with the highest amount of cases (16) of male to female transitions. One reason of the higher amount of cases in 2007 and after 2007 is because the CFO data is from 2006 till 2014. Before 2007 there are only CEO transitions from male to female and in 2007 and after 2007 there are CEO and CFO transitions from male to female. Between 1997-2006 there is a total of 16 cases of CEO male to female transitions. There were in total 47 cases of CEO male to female transitions, this

(22)

means that from 2007until 2014 the cases of CEO male to female transitions doubled relative to the period from 1997 until 2006.

Table 1

Coding credit rating, sample frequency and percentage S&P debt rating Credit rating score (RATING) #S&P debt rating #Male (Pre-transition period) #Female (Post-transition period) Total Male sample percentage Female sample percentage Full sample percentage AAA 1 1 1 0 1 0.1% 0.0% 0.1% AA+ 2 0 0 0 0 0.0% 0.0% 0.0% AA 3 27 22 4 26 3.1% 1.5% 2.6% AA- 4 53 49 2 51 7.0% 0.8% 5.0% A+ 5 77 59 15 74 8.4% 5.7% 7.3% A 6 161 114 36 150 16.2% 13.7% 15.3% A- 7 122 82 32 114 11.6% 12.2% 11.6% BBB+ 8 141 92 37 129 13.1% 14.1% 13.4% BBB 9 149 101 35 136 14.3% 13.3% 14.2% BBB- 10 106 69 32 101 9.8% 12.2% 10.1% BB+ 11 46 24 16 40 3.4% 6.1% 4.4% BB 12 51 41 6 47 5.8% 2.3% 4.9% BB- 13 37 13 18 31 1.8% 6.8% 3.5% B+ 14 29 13 11 24 1.8% 4.2% 2.8% B 15 29 14 13 27 2.0% 4.9% 2.8% B- 16 14 7 4 11 1.0% 1.5% 1.3% CCC+ 17 0 0 0 0 0.0% 0.0% 0.0% CCC or CC 18 5 3 1 4 0.4% 0.4% 0.5% C 19 0 0 0 0 0.0% 0.0% 0.0% D or SD 20 2 0 1 1 0.0% 0.4% 0.2% Total 1050 704 263 967 100% 100% 100%

Table 1 shows the recoded ratings, including the number of observations and percentage of the total observation for total, male (pre-transition period) and female (post-transition period) observations. The distribution of the full sample is similar with Kuang and Bo (2013) and Cheng and Subramayan (2008). For all three is A (34.3 percent in our research) and BBB (37.7 percent for this research) category the biggest. Also Credit rating A (15.3percent) and BBB (14.2 percent) are the largest weight in the full sample for all three.

Table 1 shows also that for female the total percentage of credit rating below the investment rating (38.8 percent), BBB- and lower, is much bigger than

(23)

for male (26.1 percent). The credit rating with the largest weight in the sample is for female BBB+ (14.1 percent) and for male A (16.2) percent. This means that male has more observations of lower credit rating score (lower credit risk).

Table 2 presents the frequency and percentage of the credit rating for male (pre-transition period), female (post-transition period) and full sample for each fiscal year. Table 2 shows also, as table 1 show before, that more female CEO/CFO results in higher mean of credit ratings (higher credit risk). Credit rating was first around the A- (credit rating score of 7) for the period with no female CEO/CFO, after that period the percentage of female increase and the credit rating slowly increase to BBB (credit rating score of 9). But for the period with almost or no male CEO/CFO (2012-2014) the credit rating decrease to BBB+ (credit rating score of 8).

Table 2

Credit rating frequency and percentage for each fiscal year Fiscal

year Frequency Mean Median Male Female Total

Percentage female (Pre-transition period) (Post-transition period) 1992 18 5.555556 4.5 18 0 18 0.0% 1993 33 6.848485 7 33 0 33 0.0% 1994 32 6.71875 7 32 0 32 0.0% 1995 34 6.852941 7 34 0 34 0.0% 1996 34 7.029412 7 34 0 34 0.0% 1997 35 7.228571 7 33 0 33 0.0% 1998 37 7.432432 7 34 2 36 5.6% 1999 41 7.829268 8 38 3 41 7.3% 2000 40 8.075 8 35 1 36 2.8% 2001 40 8.3 8 35 4 39 10.3% 2002 40 8.4 8 35 4 39 10.3% 2003 39 8.410256 8 33 5 38 13.2% 2004 38 8.342105 8 33 5 38 13.2% 2005 40 8.425 8 34 5 39 12.8% 2006 65 9.015385 9 55 5 60 8.3% 2007 73 9.068493 9 54 10 64 15.6% 2008 73 9.438356 9 46 17 63 37.0% 2009 69 9.449275 9 35 25 60 41.7% 2010 69 9.391304 9 29 31 60 51.7% 2011 63 9.095238 9 20 33 53 62.3% 2012 58 8.827586 8 4 38 42 90.5%

(24)

2013 55 8.527273 8 0 51 51 100.0%

2014 24 8.166667 8 0 24 24 100.0%

Total 1050 8.401905 8 704 263 967 27.2%

The mean of the credit rating score of the 704 male CEO/CFO observation is 8.00, which corresponds to a credit rating of BBB+. The mean of the credit rating score of the 263 female CEO/CFO observations is 9.12, which corresponds to a credit rating of circa BBB. Mean difference is then 1.12 with a t-statistic of -5.24. The mean difference is significant for a significant level of 5 percent. These univariate comparison support our first hypothesis that credit rating agencies impound the gender of CEO/CFO in their risk assessment, because the mean of the credit rating score is significant higher for female CEO/CFO than for male CEO/CFO.

Table 3 presents summary statistics, panel A is the distributional

characteristics of variables and panel B is the correlation matrix. Panel A of table 3 shows that the mean (median) of the credit rating score of the full sample is 8.40 (8), which corresponds to a credit rating of BBB or BBB+. The mean (median) of the full sample is comparable with the mean (median) of prior research (Kuang and Qin, 2013; Cheng and Subramanyam 2008; Ashbaugh-Skaife, 2006). The most control variables are comparable with prior

research(Kuang and Qin, 2013; Cheng and Subramanyam 2008; Ashbaugh-Skaife, 2006), only not the control variable TRANSP and ABSACC. The standard deviations of TRANSP (5.57) is very big compare with prior research and the number of observations of TRANSP (274) and ABSACC (311) are small compare to the number of observation of the full sample (1050).

Panel B of table 3 reports the Pearson and Spearman correlation matrix between variables. RATING and POST are significantly positively correlated (Pearson is 0.17 and Spearman is 0.27). This means that female have a positive effect on the credit rating score and higher credit risk. A correlation between RATING and POST show us that there is a difference between the credit rating score of companies with male CEO/CFO and with female CEO/CFO and that is in line with the prediction of hypothesis one. Correlations between the control variables and RATING are in the predicted direction. For instead the variable ROA is expected to have a negative correlation with RATING. Panel B shows that

(25)

RATING and ROA are significantly negatively correlated. Only the variable FIN raise collinearity concerns in our regression model, because Pearson correlation test give no result between FIN and ABSACC and Spearman correlation test give no result between FIN and all other variables. Because of collinearity FIN is dropped out in the regression model.

Table 3

Descriptive statistics

Panel A: Distributional Characteristics of Variables Variable N Mean Std dev

25th percentile 50th percentile 75th percentile RATING 1050 8.40 3.05 6.00 8.00 10.00 POST 967 0.27 0.45 0.00 0.00 1.00 DOWN 1050 0.13 0.34 0.00 0.00 0.00 DOWNBBB 1050 0.03 0.16 0.00 0.00 0.00 ROA 1050 0.05 0.06 0.02 0.04 0.08 LEV 1050 0.28 0.14 0.18 0.28 0.37 COVER 998 13.96 22.45 4.15 6.94 13.70 LOSS 1050 0.12 0.32 0.00 0.00 0.00 INTAN 1050 0.03 0.04 0.00 0.00 0.04 EQ 1050 0.79 0.41 1.00 1.00 1.00 RET 1009 0.07 0.41 -0.14 0.05 0.25 BTM 1021 0.47 0.45 0.23 0.45 0.67 MV 1021 8.57 1.49 7.53 8.47 9.57 FIN 1050 0.13 0.33 0.00 0.00 0.00 SDRET 873 0.58 2.09 0.20 0.30 0.44 SDNI 967 0.20 0.73 0.05 0.10 0.17 TRANSP 274 0.00 5.57 -1.90 0.00 2.47 ABSACC 311 0.00 0.06 -0.03 0.00 0.03

Panel B: Correlation Matrix

RATING POST ROA LEV COVER LOSS INTAN EQ

RATING 1.00 0.27 -0.68 0.19 -0.70 0.42 -0.28 -0.26 POST 0.17 1.00 -0.22 0.07 -0.13 0.09 -0.05 -0.07 ROA -0.43 -0.05 1.00 -0.13 0.82 -0.56 0.42 0.20 LEV 0.29 0.05 -0.27 1.00 -0.36 0.09 0.04 0.09 COVER -0.23 0.09 0.42 -0.45 1.00 -0.40 0.38 0.15 LOSS 0.41 0.08 -0.64 0.22 -0.16 1.00 0.01 -0.22 INTAN 0.05 0.00 0.17 0.01 0.04 0.04 1.00 0.22 EQ -0.02 0.04 0.22 0.01 0.04 -0.14 0.14 1.00 RET 0.10 0.09 0.16 -0.07 0.04 -0.14 -0.02 0.01 BTM -0.01 -0.01 -0.28 -0.06 -0.11 0.05 -0.24 -0.16 MV -0.60 0.06 0.45 -0.22 0.34 -0.31 0.11 -0.04

(26)

FIN -0.12 0.04 -0.18 -0.15 -0.06 0.03 -0.23 -0.12 SDRET 0.27 0.04 -0.10 0.01 -0.06 0.15 -0.06 -0.10 SDNI 0.26 0.05 -0.10 0.01 -0.06 0.16 -0.06 -0.08 ABSACC 0.07 0.08 -0.17 0.17 -0.17 0.10 -0.01 0.00 TRANSP -0.19 -0.02 0.07 0.02 0.08 0.00 0.08 -0.01

RET BTM MV FIN SDRET SDNI ABSACC TRANSP

RATING 0.07 0.57 -0.64 . 0.40 0.24 0.05 -0.12 GENDER 0.14 0.17 -0.11 . -0.00 0.00 0.15 -0.00 ROA -0.00 -0.70 0.59 . -0.18 -0.01 -0.14 0.09 LEV -0.10 -0.10 -0.06 . 0.05 0.14 0.17 -0.02 COVER 0.02 -0.64 0.61 . -0.25 -0.06 -0.09 0.07 LOSS -0.22 0.26 -0.38 . 0.18 0.25 0.11 0.00 INTAN 0.01 -0.62 0.44 . 0.00 0.13 -0.05 0.12 EQ 0.03 -0.33 0.11 . -0.22 -0.09 -0.03 -0.03 RET 1.00 -0.05 0.08 . 0.08 -0.04 -0.04 -0.06 BTM -0.21 1.00 -0.54 . 0.13 -0.08 -0.04 -0.03 MV 0.07 -0.18 1.00 . -0.19 -0.13 -0.09 0.06 FIN -0.03 0.14 -0.09 1.00 . . . . SDRET 0.15 0.01 -0.06 -0.03 1.00 0.36 0.04 -0.06 SDNI 0.16 -0.01 -0.06 -0.03 0.99 1.00 0.02 -0.03 ABSACC -0.07 -0.10 -0.11 . 0.04 0.05 1.00 0.00 TRANSP -0.11 -0.06 0.10 0.00 -0.01 -0.01 0.07 1.00

*The upper triangle presents the Spearman correlation, and the lower triangle the Pearson correlation. Boldface text indicates significance at p-value<0.05.

4. Results

This research examined if credit rating agencies impounding the gender of CEO’s in their risk assessment. This chapter presents the results of the research. First the results of hypothesis one will be discussed. Then the results of the second hypothesis will be discussed.

4.1 Results of hypothesis one

The first hypothesis focuses on impounding the gender of a CEO/CFO in the risk assessment of credit rating agencies. Regression model (1) examined this hypothesis. Table 4 presents the regression results for (1) without control

variable FIN. The dependent variable for this model is RATING. RATING increases if the credit risk increases. The F-statistics of this model is 26.29, which is

(27)

significant for significant level of 5 percent. The regression model does have some fitting in the data. The independent variable for this model is POST, which measure the effect of the gender of a CEO/CFO to the credit rating score.

The regression results suggest that POST has a positive (0.86)

relationship with RATING and is significant (3.51 for p-value <0.05). This mean that if a company has a female CEO/CFO the credit rating score of the company is 0.86 higher, compared with companies with a male CEO/CFO. For instead, if a company has a credit rating of BBB+ and change from male CEO/CFO to female CEO/CFO, then the credit rating will downgrade to BBB, the credit risk is then higher for a company. Thus there is a significant difference between the credit rating of companies with male CEO/CFO and with female CEO/CFO, which suggest that hypothesis one is true.

The result of the analysis of most of the control variables is what was expected and also consistent with the prior literature (Kuang and Qin, 2013; Cheng and Subramanyam 2008; Ashbaugh-Skaife, 2006), but they are not all significant (e.g. ROA and LEV). The control variable of the standard deviation of stock return over the prior 60 months for the firm has a not significant

negatively sign (-0.50), instead of the expected positive sign. However it is not significant, this means that the result isn’t reliable. Also the control variable of the level of financial transparency has a negatively sign (-0.06) and is significant, instead of the expected positive sign. But panel A of table 3 shows already that this variable has a different distribution compare with the prior literature (Kuang and Qin, 2013; Cheng and Subramanyam 2008; Ashbaugh-Skaife, 2006).

The number of observation for this regression is 227, which is circa one fifth of the full sample (1050 observation). A lot of observations are not taking into account, which can lead to spurious result (Frank and Chang, 1987). Panel A of table 3 reports the number of observation for each variable. The number of observation of ABSACC (311)and TRANSP (274) are very low compare with the other variables. When these two variables are omitted out of the model, the model will regress with more observations and this give a more reliable result.

Table 4 present the result of the regression model (1) without the control variables FIN, ABSACC and TRANSP. The F-statistics of this model is significant (80.26) for significant level of 5 percent. The number of observation is 760,

(28)

which is more than 3 times higher as the number of observation of the first regression model (222).

The effect of POST on RATING is still positive (0.65) and significant , but the effect is a little bit lower than in regression model (1) without FIN (0.86). A Female CEO/CFO will lead to a downgrade of circa a half of the credit rating, compared with male CEO/CFO. The results of both regression models indicate that credit rating agencies impound the gender of CEO/CFO in their risk assessment. This finding supports the first hypothesis, however it was not expected that male to female CEO/CFO transition has a positive effect on the credit rating score, because prior literature shows that women take less risk than men (e.g. Elsaid and Ursul, 2011).

For the control variables COVER, BTM and SDRET is sign different

compared with regression model (1) without FIN. COVER change from -0.025 to 0.005, and the expectations was a negative sign. SDRET change from -0.49 to 0.17, and the expectations was a positive sign. But COVER an SDRET are not significant in regression model (1) without FIN and model (1) without FIN, ABSACC and TRANSP. This means that the result of the coefficient isn’t reliable. BTM changed from 1.72 to -0.35 and is significant for both. There was no expectation for BTM, because Bohjarj and Sengupta (2003) found a negative relation between default risk and market ratio, but lower

book-to-market ratio can also increase the opportunities for a company, which indicates a decrease in default risk.

Table 4

OLS estimation of regression model (1) without FIN and regression model (1) without FIN, ABSACC and TRANSP

(1) without FIN

(1) without FIN, ABSACC and TRANSP RATING Predicted sign Coefficient t-statistics Coefficient t-statistics POST 0.861 3.51 0.651 4.24 ROA - -3.581 -0.99 -10.170 -5.49 LEV + 2.000 1.59 0.649 1.15 COVER - -0.025 -3.29 0.005 1.28 LOSS + 1.273 2.26 1.018 3.16 INTAN ? 5.117 1.69 6.722 3.66

(29)

EQ ? -1.336 -3.58 0.108 0.56 RET ? 1.264 3.7 0.962 5.47 BTM ? 1.719 3.46 -0.350 -1.97 MV - -0.637 -5.49 -0.992 -18.59 SDRET + -0.496 -1.11 0.170 0.86 SDNI + 1.320 1.08 0.237 0.44 ABSACC - -2.012 -0.97 TRANSP + -0.060 -2.95 Intercept 13.016 10.25 16.779 30.19 Observation 227 760 F-statistic 28.01 80.26

* Boldface text indicates significance at p-value<0.05.

4.2 Results of hypothesis two

The second hypothesis focuses on the effect of a downgrade of the credit rating on the transition of a male to female CEO/CFO. This is examined with a logit regression with DOWN as independent variable and examined with DOWNBBB- as independent variable. Table 5 presents the results of the logit regression model (2). DOWN is the independent variable in columns 1 and 2 while DOWNBBB- is the independent variable in columns 3 and 4.

For the logit regression (2) with DOWN is the log likelihood of the logit regression -229.6, this is not significant for a significant level of 5 percent. This means that the model can give spurious result. Table 5 shows that DOWN has a coefficient of 0.14. This mean that companies with a downgrade of their credit rating are more likely to switch from male to female CEO/CFO. The expectation was that a transition from male to female CEO/CFO is a trigger for companies with credit rating concerns to get better or more stable credit rating. This is the same as what the result suggest. But the coefficient is not significant (0.38) for significant level of 5 percent. And also almost all the control variables are not significant for significant level of 5 percent. If the model, the independent variable and almost all the control variables are not significant, then the results of the regression can be spurious. Thus the results of this regression are not reliable.

Table 5 presents also the result of the logit regression model (2) with DOWNBBB-. This regression model has a log likelihood of -228.83, which is not

(30)

significant for a significant level of 5 percent. This model can also give spurious result, the same as logit regression model (2) with DOWN did. Table 5 reports that the coefficient of DOWNBBB- is -1.17. This suggest that companies with a downgrade to the lower edge of the investment category are less likely to switch from male to female CEO/CFO. This is the opposite of what was expected.

Because companies with a downgrade of their credit rating to the lower edge of the investment category will be most concerned about their credit rating

(Kisgen, 2006), and this should be a bigger trigger for them to switch from male to female CEO/CFO. But the coefficient of DOWNBBB- is not significant (-1.11), and also almost all the control variables are also not significant for significant level of 5 percent. The result can be spurious. This means that the result of logit regression model (2) with DOWNBBB- is not reliable. Thus there is no answer of hypothesis two is valid.

Table 5

Logit estimation of model (2) with DOWN and model (2) with DOWNBBB- (2) with DOWN (2) with DOWNBBB-

TRANSITION Coefficient z-statistic Coefficient z-statistic DOWN 0.145 0.38 DOWNBBB- -1.17496 -1.11 ROA -0.374 -0.1 -0.66571 -0.17 LEV -0.957 -0.41 -0.88789 -0.37 COVER 0.003 0.32 0.003071 0.33 LOSS 0.168 0.36 0.18233 0.39  0.121 0.05 -0.14407 -0.06 EQ 0.652 1.66 0.652764 1.65 RET 0.187 0.66 0.229332 0.82  0.035 0.1 0.047361 0.14 MV -0.427 -1.12 -0.54808 -1.45 SDRET 1.009 1.14 1.126836 1.31 SDNI -2.513 -2.23 -2.52642 -2.27 Constant -2.325 -14.69 -2.27553 -15.84 Observations 750 750 Log Likelihood -229.60 -228.83

(31)

5. Robustness check

The regression presented in table 4 examined for the possibility that credit rating agencies impound the gender of CEO/CFO in their risk assessment. This is a regression for a sample with transition from male to female CEO/CFO.

However, there are also unobservable time series of transitions for female to male CEO/CFO, which could also affect the credit rating of a company. This makes it possible that the results of table 4 are spurious. Francis et al. (2014) did a regression with a sample for transitions from male to female CFO and did the robustness check with a sample for transitions from female to male CFO. In this study the same is done. The regression model (1) without FIN, ABSACC and TRANSP, regression model (2) with DOWN and with DOWNBB- are regress for a sample of transitions from female to male CEO/CFO. The construction of the sample of transitions from female to male CEO/CFO use the same criteria as for the sample of transitions from male to female CEO/CFO. Final sample consist of 345 firm-year observations and 50 cases of female to male CEO/CFO transitions.

The expectation of the regression with a sample of transition from female to male CEO/CFO is still the same as for a sample of transition from male to female CEO/CFP . Credit rating agencies will impound the gender of a CEO/CFO in their risk assessment, and a transition of female to male CEO/CFO are less likely for companies with credit rating concerns. The expectations of the control variables are also still the same.

Table 6 presents the result of the regression of model (1) without FIN, ABSACC and TRANSP with a sample of transitions from female to male CEO/CFO. The F-statistic of the regression model is significant (46.98). The number of observations is only circa three times as low (280 against 760) compared with regression of male to female CEO/CFO transition sample. In the regression is POST a dummy variable that equals zero if a year is before the CEO/CFO

transition and is one if a year is after the CEO/CFO transition. The coefficient of POST is 0.06 and is significant. The credit rating score increase with 0.06 in the post-transition period when there is a male CEO/CFO. An increase of 0.06 of the credit rating score is very small, this will most of the time not directly result in a

(32)

downgrade of the credit rating. Thus the influence on the credit rating if a CEO/CFO is a female or male CEO/CFO is almost nothing. This result is different than the result of the sample with a male to female transition sample and also different with was expected. But the sample of female to male CEO/CFO transition has high amount of cases of transition relative to firm-year

observations (1/7 against 2/25) and the sample is three times smaller, compare with the male to female transition sample. This can result in spurious result for the regression with female to male transition sample (Franck and Chang, 1987).

Table 6

Robustness check of regression model (1) without FIN, ABSACC and TRANSP

Female to male transition sample RATING Predicted sign Coefficient t-statistic POST 0.064 0.27 ROA - -5.033 -2.96 LEV + 2.106 2.72 COVER - 0.004 0.92 LOSS + 1.122 2.96 INTAN ? 2.295 0.58 EQ ? 0.416 1.37 RET ? 0.526 3 BTM ? 0.061 0.71 MV - -1.215 -15.11 SDRET + 0.887 2.1 SDNI + -1.299 -1.57 Constant 18.821 22.83 Observations 280 F-statistic 46.98

* Boldface text indicates significance at p-value<0.05.

Table 7 presents the result of the logit regression of model (2) with DOWN and with DOWNBBB. For both models is the loglikelihood (113.09 and -111.85) not significant for significant level of 5 percent. Thus the result can be spurious. Table 7 shows that the coefficient of DOWN is 0.37. This means that companies with credit rating concerns are more likely to do a transition from female to male CEO/CFO. This is the opposite as what was found for logit model

(33)

(2) with DOWN with sample of male to female CEO/CFO transition. But for both are the findings not significant. Thus the results of both can be spurious.

There are only 7 cases of downgrading of the credit rating to the lower edge of the investment category, that is a very low amount for a logit regression with 345 observations. This mean that the result can give spurious results (Flack and Chang, 1987). Table 7 shows that DOWNBBB- is dropped out of the

regression. Because no downgrading of the credit rating to the lower edge of the investment category predicts perfectly that there is a transition from female to male CEO/CFO in the logit regression. Thus the findings of this logit regression are unusable, because DOWNBBB- dropped out of the model.

Table 7

Robustness check of logit regression model (2) with DOWN and with DOWNBBB- (2) with DOWN (2) with DOWNBBB-

TRANSITION Coefficient z-statistic Coefficient z-statistic DOWN 0.367 0.75 DOWNBBB- Omitted ROA -0.698 -0.33 -0.588 -0.28 LEV 1.658 0.59 2.298 0.82 COVER 0.002 0.91 0.002 0.91 LOSS -0.351 -0.75 -0.309 -0.66 INTAN 8.642 0.43 6.201 0.31 (EQ) -0.651 -1.3 -0.608 -1.23 RET -0.135 -0.64 -0.133 -0.64 BTM 0.231 1.59 0.227 1.58 MV -0.266 -0.64 -0.412 -1.06 SDRET 0.357 0.48 0.499 0.69 SDNI -1.304 -1.05 -1.530 -1.21 Constant -1.774 -9.13 -1.686 -9.49 Observations 272 272 Log Likelihood -113.09 -111.85

(34)

6. Conclusion

This research examined if credit rating agencies impound the gender of a CEO/CFO in their risk assessment. The effect of gender of the CEO/CFO is measured with pre- and post-transition period of the CEO/CFO (POST). The credit rating is measured on a credit rating score on the scale of 1 to 20 (RATING). It is examined for a sample of 1050 firm-year observations with 84 cases of male to female CEO/CFO transitions. The findings provide evidence that credit rating agencies impound the gender of a CEO/CFO in their risk

assessment. But the model with a sample of transition from female to male CEO/CFO could not confirm that those findings are true, because the findings of this model were not significant.

Further there is examined if companies that have credit rating concerns, because they recently experienced a downgrade of their credit rating, are triggered to switch from male to female CEO/CFO. This is examined with a regression of a downgrade of the credit rating on transition from male to female CEO/CFO. Transition of male to female CEO is measured as when there is a female CEO/CFO in year t and a male CEO/CFO in year t-1. Downgrade is

measured on two ways. First way is measured as if there is a downgrade in year t compared with year t-1. Second way is measured as if there is a downgrade of the credit rating to the lower edge of the investment grade in year t compare with year t-1. The findings give no evidence if companies with credit rating concerns are triggered to switch from male to female CEO.

The result of this research can have an effect on the decision of companies to hire a female or male CEO/CFO, because the findings suggest that credit rating agencies impound the gender of a CEO/CFO in their risk assessment. For instead there are 23 female CEO’s out of the S&P 500 companies in 2015. This means that 4.6 percent of the CEO’s are female. This is percentage is very low. This research can help companies to consider if it is a better to take a female or male CEO/CFO. If there are more female CEO/CFO in the future, then the result of further research on the effect of female CEO/CFO will be more robust.

The findings provide evidence that credit rating agencies impound the gender of the CEO in their risk assessment. However there are several limitations

(35)

for this research. First, the model omitted the following variables; a company is from the financial industry, the level of financial transparency, absolute value of abnormal accruals, market beta and managerial risk taking incentives. While prior literature (e.g. Kuang and Qin, 2013; Cheng and Subramanyam, 2008) found that those variables have influence on the credit rating. Second, the sample consists of CEO and CFO male to female transitions, because otherwise the

sample is too small. However for instead the influence of CEO on earnings management is different than the influence of a CFO (Jiang et al., 2010). Third, the size of the sample of male to female CEO/CFO transition is still not large, which can lead to spurious result (Flack and Chang, 1987). But the distributional characteristics of variables of this sample are almost the same as those found in prior literature (e.g. Kuang and Qin, 2013; Cheng and Subramanyam, 2008). Fourth, the Francis et al. (2014) used for transition for a CFO that male and female CFO must be in office consecutively for at least three years excluding the transition year, but in this study the CEO/CFO must be in office consecutively for only one year excluding the transition year. A CEO/CFO who works only one year for a company can may have not enough influence on companies credit risk. However for filter of must be in office consecutively for at least three years excluding the transition year, the sample for this study will be to small and can give spurious results (Flack and Chang, 1987).

(36)

7. Reference list

Arch, E. (1993). "Risk-Taking: A Motivational Basis for Sex Differences." Psychological Reports, 73(3), 6-11

Ashbaugh-Skaife, H., Collins, D. W. and LaFond, R. (2006). “The effects of corporate governance on firms’ credit ratings.” Journal of Accounting and Economics, 42(1-2), 203–243

Bandiera, Oriana, Luigi Guiso, Andrea Prat, and Raffaella Sadun, 2014, Matching firms, managers and incentives, Journal of Labor Economics, forthcoming.

Bhojraj, S. and Sengupta, P. (2003). “Effect of corporate governance on bond ratings and yields: The role of institutional investors and outside directors.” Journal of Business, 76(3), 455–75

Bradley, M. and Chen, D. (2011). “Corporate governance and the cost of debt: Evidence from director limited liability and indemnification provisions.” Journal of Corporate Finance, 17(1), 83–107

Bullard, J., Neely, C.J. and Wheelock, D.C. (2009) “Systematic Risk and the

Financial Crisis: A Primer” Federal Reserve Bank of St. Louis Review, 91(5-1), 403-417

Cheng, M. and Subramanyam, K.R. (2008). “Analyst following and credit ratings.” Contemporary Accounting Research, 25(4), 1007–1043

Croson, R. and Gneezy, U., (2009). “Gender difference in preferences.” Journal of Economic Literature, 47(2), 448-474

DeFond, M.L. and Jiambalvo, J. (1994) “Debt convenant violation and

manipulation of accruals.” Journal of Accounting and Economics, 17 (1-2), 145-176

Denis, D. J. and Denis, D.K. (1995) “Performance Changes Following Top Management Dismissals.” Journal of Finance, 50(4), 1029-1057

Dwyer, P. D., Gilkeson, J. H., and List, J. A. (2002). “Gender differences in revealed risk taking: Evidence from mutual fund investors.” Economics Letters, 76(2), 151–158

.

Eckel, C. and Grossman, P., (2008). “Forecastingrisk attitudes: An experimental actual and forecast gamble choices.” Journal of Economic Behavior &

Organization, 68(1), 1-17

Ederington, L.H., Yawita, J.B., and Roberts, B.E. (1987). “The informational content of bond ratings.” Journal of Financial Research, 10(3), 211-226

(37)

Eisenhardt, K.M. (1989). “Agency theory: An assessment and review.” The Academy of Management Review, 14(1), 57-74

Elsaid, E. and Ursel, N. (2011). “CEO succession, gender and risk taking.” Gender in Management: An International Journal, forthcoming.

Estes, R. and Hosseini, J. (1988). "The Gender Gap on Wall Street: An Empirical Analysis of Confidence in Investment Decision Making." Journal of Psychology, 122(6), 577-90

Faccio, M., Marchica, M.T. and Mura, R. (2014). “CEO gender, corporate risk-taking, and the efficiency of capital allocation.” Working paper, Purdue University and Manchester Business School.

Federal Register (2009a), “Securities and Exchange commission”, Federal Register, 74(25), 6456-6484

Federal Register (2009b), “Securities and Exchange commission”, Federal Register, 74(232), 63832-63865

Forbes, D. (2005). “Are some entrepreneurs more overconfident than others?” Journal of Business Venturing, 20, 623-640.

Flack, F. and Chang, C. (1987). “Frequency of Selecting Noise Variables in Subset Regression Analysis: A Simulation Study.” The American Statistician, 41(1), 84-86 Francis, B., Hasan, I., Park, J.C. and Wu, Q. (2014) “Gender Differences in Financial Reporting Decision Making: Evidence From Accounting Conservatism.”

Contemporary Accounting Research

Graham, J.R. and Harvey, C.R. (2001) “The theory and practice of corporate financial: Evidence from the field.” Journal of Financial Economics, 60(2-3), 187-243

Grossman, M. and Wood, W. (1993). "Sex Differences in Intensity of Emotional Experience: A Social Role Interpretation." Journal of Personality and Social

Psychology, 65(5), 1010-22

Gu, Z. (2007). “Cross-sample incomparability of R2s and additional evidence on value relevance changes over time.” Journal of Business Finance and Accounting, 34(78), 1073–98

Harshman, R. A., and Paivio, A. (1987). "'Paradoxical' Sex Differences in Self-Reported Imagery." Canadian Journal of Psychology, 41, 287-302

Jensen, M. and Meckling, W., (1976). “Theory of the firm: Managerial Behavior, Agency costs and Ownership structure.” Journal of Financial Economics, 3(4), 305-360

Influence of the gender of CEO/CFO on credit rating