Gender Differences in
Executives’ Compensation
ABSTRACT:
In response to the financial crisis one has come to debate about the cause and the role of the gender differences in high-paid positions. This research is intended to contribute to an increasingly informed debate on this topic. For this purpose, I have tested whether there is a gender difference in risk aversion in the executive board by looking at the percentage of variable compensation. Contrary to the majority of former research, I have found no evidence that gender has an impact on the compensation of executives in the period 2007-2012.
Name: Thijmen de Groot
Student number: 10013873 Supervisor: Joeri Sol
1. Introduction
After a prolonged period of economic boom the world economy entered one of the biggest financial crises of modern history: there have been several major bankruptcies in the banking sector and governments have had to intervene to prevent further escalations. The financial crisis has provoked major public discussion on its roots.
Part of the discussion focused on the possible impact of gender and some have argued that the relative absence of females in high-paid positions plays a part in explaining the exceptional size of the crisis (NTR, 2010). The rationale behind this statement is that women have different personal characteristics than men are: men are more competitive (Gneezy, Niederle and Rustichini, 2003), more confident (Barber and Odeon, 2001), and less risk averse (Eckel and Grossman, 2002). The gender difference in risk preferences is discussed in many papers over the years. An overview can be found in the paper of Croson and Gneezy (2009). Most of these papers argue, like Eckel and Grossman, that men are less risk averse compared to women. This conclusion could explain parts of major gender questions that politicians face these days, like the overrepresentation of men in high-paid positions and the wage gap in the executive board: if women shy away for risk, it is less likely that they enter the many competitions that forego a seat in the executive board. In addition to the gender difference in risk-aversion I will examine the difference in risk preferences between men and women in the executive board. I will look at the percentage of variable payments compared to the total compensation and run an OLS regression on the dummy variable female and some control variables. Please note that my research is merely concerned with relative payments and leaves absolute payments out of consideration.
Over the total of 229,043 observations I found a small but very significant influence of gender: male executives earn on average a higher percentage of their income in variable payments. However, from only 115,265 observations the variable age was known and it does not seem to be a realistic subsample so I was not able to include the variable age in these first regressions. Consequently I have decided to make two sub-samples, one for the period 2007-2012 whereof 99% of the age is known and a subsample for the period 1993-2006. By including age and age� among the control variable in the subsample 2007-2012, the gender difference were almost completely gone and the results turn insignificant. For the period 1993-2006 there is still a significant difference, which can be explained in two ways: in this period there was a gender difference in variable payments that disappeared from 2007 onwards or the subsample is not a realistic representation of the complete sample.
I will start in paragraph 2 with a review of the literature on gender differences. In paragraph 3 I will explain something about the data collection, and I will present the
descriptive data. In the subsequent paragraph the results will be shown. Lastly in paragraph 5 I will conclude this paper, and I will discuss its weaknesses.
2. Related literature
Up until the research by Gneezy, Niederle and Rustichini (2003) there were two basic ideas that explain the gender difference on the workplace; the first idea focuses on the gender difference in ability and preferences (Polacheck, 1981), the second concerns a difference in selection and discrimination. Additional to these two explanations Gneezy et al. (2003) find an third explanation for the gender differences: men are more effective in a competitive environment. Gneezy et al. (2003) set up a lab experiment in which men and women were asked to solve a maze. When the participants were paid at the piece rate the results of men and women were equivalent, but when a form of competition is included whereby only the best performing person was paid, men outperform women. The difference is bigger in a mixed group than in a group in which male and female are separated. Gneezy et al. (2003)
demonstrated that there are differences between the intrinsic motivation of men and women, and this is a possible cause of the difference in executive compensation.
An article that examines the gender difference in the financial world is the article from Barber and Odeon (2001). In their empirical research on real life data they found that men on average trade more than women, even if the payoff of trade is negative. The explanation of this difference, according to the researchers, is overconfidence on the part of men. Niederle and Vesterlund (2007) present similar findings in the lab. In their experiment they let participants compete in a lab experiment whereby they had to solve simple mathematical problems. At the end of the four-step experiment they were asked what their relative performance had been. In a group of four persons, 75 percent of the men think they are the best of their group, as opposed top 43 percent of the women. The difference is significant at a 0.016 p-value level. Moreover, it partially explains the gender differences in competition.
In 2009, Gneezy, Leonard, and List conducted a real life experiment in order to test to what extent the difference in the willingness to compete can be explained by reference to nurture differences between men and women. Gneezy et al. (2009) compared the Maasai, a male dominated culture, to the Khasi, a female dominated culture. They tested their
competitive payout scheme that has higher expected payoff. The researchers found, in line with the former literature, that the Maasai men are more competitive than their women. Surprisingly, the women of the Khasi turned out to be more competitive than the men of their tribe. One can conclude from this research that the assumption that women compete less than men is not universally valid. It also shows that the difference in competitiveness is largely learned behaviour.
A paper of Buser (2009) examined the impact of nature on the competiveness of women. Buser (2009) found a relation between the willingness to compete and the menstrual cycle of women. In his lab he conducted an experiment like Niederle and Vesterlund (2007) and he let the participant solve simple mathematical problems. The likelihood of choosing the competing wage scheme is influenced significantly by the menstruation cycle. This research shows that also nature has an influence in the gender difference in competition.
A research that focused only on the impact of risk aversion is the paper of Eckel and Grossman (2002). The researchers conducted a lab experiment where the payoff was totally random. Students were asked to participate in one of five lotteries. In lottery 1 the expected payoff is $10 (100% chance of winning $10), and in the fifth the expected payoff is $18 (50% chance of losing $6 and 50% chance of winning $42). Lotteries 2-4 had a payoff and risk profile in between the first and fifth lottery. The results were significant and showed that men choose the fifth lottery more often; women choose the first and less risky lottery more often.
In contradiction to the results found by Eckel and Grossman (2002), Schubert et al. (1999) found that there is a situation where men are more risk averse than women: the
situation of a possible loss. The researchers also argue that the difference between risk taking behaviour of men and women strongly depends on the way in which the question is
formulated. Datta Gutpa, Paulson and Villeval (2005) concluded from their lab research that the difference in risk-aversion between men and women does not play a role in explaining competitiveness’ differences. Instead they argue that it is the influence of risk aversion that matters. The researchers asked the participants to place their risk-aversion on a one to five scale, and they let them solve mazes. The participants could choose their payment scheme, either at a piece rate or a tournament payment scheme. The results indicated that the decision to compete by men is not influenced by their risk aversion whereas women are: more risk-averse women avoid competition.
All papers considered thus far are concerned with risk-aversion as one of the possible explanations for gender differences. However Barber and Odeon (2001) argue that it is not a suitable explanation for the difference and Niederle and Vesterlund (2007) find that risk
aversion gives only a partial explanation of the differences that occur. The idea that gender differences have an influence on risk-taking behaviour is supported by Eckel and Grossman (2002), and Datta Gupta et al. (2005).
As opposed to many of their colleagues, Graham, Harvey and Puri (2012) did their research in a real life context by gathering data from active executives. Graham et al. (2012) studied the question of executive compensation and the gender difference and conclude that the wage schemes of women are indeed less performance driven compared to the wage schemes of men. They took a questionnaire among the readers of the Chief Executive Magazine and asked them for their preferred compensation scheme. Their results show that there is a significant gender difference: men prefer a high variable payment whereas the women prefer a high base salary. In contradiction to the study of Graham et al. (2012), I will focus on data provided by databases whereas the previous study was done. An overview of articles on gender differences can be found in the article by Croson and Gneezy (2009).
Having considered former literature, I expect my research to show that women earn lower variable payment relative to total payment compared to men because of the higher risk of a more bonus driven compensation. I expect that women will negotiate more certain compensation whereas men have a more risky compensation. This hypothesis is line with the findings of the paper from Graham et al. (2012).
3. Data collection and description
I will use the dataset of ExecuCom to test for gender differences in executives’
compensation. ExecuCom provides information about the compensation and characteristics of executives of US firms. The information on compensation is used to construct a measure that expresses the percentage of a pay that is variable, henceforth called RelBonus. I will run an OLS regression with gender, industry, size of the firm, fiscal year and exchange market as independent variables and RelBonus as the dependent variable. The variable ‘size of the firm’ is constructed by looking at market capitalization: observations are divided in small-, midcap, and S&P 500. The variable ‘exchange market’ controls for different stock exchanges. Like size and exchange market, the variables gender, industry, and fiscal year are used as dummy variables. I choose to use dummy variables for the fiscal years because I cannot assume a linear relationship over time.
Variable Description
RelBonus Percentage of payment that is variable
Fiscal Year Dummy variable for the fiscal year
Size of the firm Dummy variable that divides the observations in small-, midcap, and S&P 500
Exchange Market Dummy variable for the stock exchange. (NYSE, ASE, and NASDAQ)
Industry Dummy variable for the different industries
Crisis Dummy variable that is equal to 0 before 2008 and 1 afterwards
Table 1.
To filter the influence of the financial crisis I construct another variable next to the normal dummy variable for time. This dummy is equal to 0 before 2008 and 1 afterwards. The year 2008 has been chosen because of the bankruptcy of Lehman Brothers and the fact that the financial crisis and governments have acknowledged the need for action.
The data cover the period 1992 to 2012 and include a total of 229.043 observations. Table 2 presents the descriptive statistics. 94.07 % of the data are from male executives. The average age of an executive is slightly higher for male executives (52.85) than for female executives (49.6). The base salary is, just like the variable payment, higher for men than for women. Men receive higher variable payment than women: 25.4% of their total current compensation consists of bonuses whereas this is only 21.6% for women.
GENDER EXECUTIVE
Male Female
Mean Std Mean Std #Observations
Salary 381,556 278,015 334,167 220,886 229,041
Bonus 263,468 929,865 161,971 430,250 229,041
RelBonus 0.254 0.239 0.215581 0.225 228,325 #observations 215,453 (94.07%) 13,590 (5.93%) 229,043
GENDER EXECUTIVE IF AGE KNOWN
Male Female
Mean Std Mean Std #Observations
Age 52.85 7.958 49.6 6.466 122,415 Salary 464,458 315,419 395,153 250,333 122,415 Bonus 304,606 1.157.110 137,665 391.054 122,415 RelBonus 0.214688 0.252 0.152910 0.221 114,743 #Observations 115,265 (94.16%) 7,150 (5.84%) 122,415 Table 3.
The choice of independent variables is for a large part determined by the availability of data. The variable age can only be used after 2006 because of only 53% of the observations the age is known. By running an OLS regression on both the whole sample and the sub-sample, I found that the 53% of which the age is known is not a realistic sub-sample. This due to the fact that in 1992 from 3.19% of the data the age is known whereas in 2012 this
percentage is 99.3. See figure 1.
Other differences can be found in table 2 and 3. Except for the Bonus payments for women, the absolute numbers are higher in the sub-sample. Yet, the Relbonus decreased. I will look further into these differences in paragraph 4.2.
Another remarkable statistic is the increase of women in executive positions. In 1992 1.6% of the executives in this dataset was female, whereas in 2012 this percentage was 7.9. See figure 2 for a graphical expression of the increase.
Figure 1. 0 20 40 60 80 100 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 percentage of age known
Figure 2.
4. Results
Table 4 shows the results of the OLS regression. In the first regression I regress the variable RelBonus on the dummy variable female without controlling for any other variable. This regression shows that women on average receive 3.8%-point less variable payment than their male colleagues. This result is significant at a 1% level, but its size is very small: 16% of the standard deviation. Furthermore, by controlling for time, the result becomes smaller and is only 6% of the standard deviation.
In the third regression I also control for the size of the firm, the market whereupon it is traded, and the industry. I found a significant gender difference. Women, on average, receive 0.8% points less variable payment compared to men. However, this difference is again a lot smaller than it was in regression 1.
Finally, I ran a regression with a dummy variable for the crisis, which equals 0 before the start of the financial crisis and equals 1 afterwards. Despite of the significant influence of this dummy variable, it has no influence on the gender difference. The gender gap in variable payment stays at the 0.8%-point level.
0 2 4 6 8 10 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 percentage of female executives
Dependent variable: RelBonus Independent variables (1) (2) (3) (4) Female -0.038 (0.002)*** (0.002)*** -0.0153 (0.002)*** -0.0084 (0.002)*** -0.0084 Controlled for
time NO YES YES YES
S&P 500 0.0667 (0.001)*** (0.001)*** 0.0668 MidCap 0.2278 (0.001)*** (0.001)*** 0.2279 SmallCap 0.0020 (0.001) 0.0020 (0.001) Dummy’s for exchange market NO NO YES YES Dummy for crisis (0.003)*** -0.1963 Dummy’s for
industry NO NO YES YES
#Observations 228,325 228,325 228,325 228,325
R� 0.002 0.202 0.298 0.298
Table 3. *significant at a 10% level **significant at a 5% level ***significant at a 1% level
4.2 Robustness analysis
In all regressions in the former paragraph the variable age was not taken into account because of the unrealistic sub-sample. However, there is a lot empirical research that suggests there to be a relation between the compensation of an executive and his age (Ryan and
Wiggins III, 2001). To test for this influence I have made a different regression for the years 1993-2006 and 2007-2012. For these periods respectively 38 and 99 percent of the variable age is known. I excluded 1992 because only 3% of the age variables were known.
Dependent variable: RelBonus 2007-2012 1993-2006 Independent variables (1) (2) (3) (4) Female 0.003 (0.002) (0.002) 0.001 (0.004)*** -0.014 (0.005)*** -0.016 Age -0.008 (0.001)*** (0.001)*** 0.010 Age^2 0.00007 (0.000008)*** (0.0000008)*** -0.0001 Controlled for
time YES YES YES YES
Dummy’s for
size YES YES YES YES
Dummy’s for exchange
market
YES YES YES YES
Dummy’s for
industry YES YES YES YES
#Observations 62,318 62,318 59,308 59,308
R� 0.109 0.111 0.218 0.220
Table 5. *significant at a 10% level **significant at a 5% level ***significant at a 1% level
I tabulated the results in table 5. There is a big difference between the two sub-samples. In the years 2007-2012 one can no longer observe a difference between the relative executives’ compensation schemes. However, if we look at the sub-sample 1993-2006 we find an influence that is bigger than in table 4. The variable age is, in line with former literature, in both regressions significant at a 1% level.
The difference between these two sub-samples can be explained in different ways. Firstly one can imagine there to be a gender difference between the relative payment schemes of executives in 1993-2006 that disappeared in the period 2007-2012. This might be due to the fact that female executives of the last years have different personal characteristics than their colleagues ten years ago. Another explanation would be that the sub-sample, the executives whose age is known in 1993-2006, is not a good sub-sample.
5. Discussion and Conclusion
In this paper I examined the influence of gender on the percentage of variable pay in executives’ compensation. I ran an OLS regression with the variable part of the total
compensation as the dependent variable, and I tested the influence of the dummy variable Female while controlling for other variables. In the first regressions I found a significant, but very small, influence of gender. This difference decreased in the process of introducing control variables like time, industry, exchange market, and size. Nonetheless, it stayed
significant at a 1% level. Subsequently, I split up the observations in a period before and after 2007. In contradiction to the period 1993-2006, the period 2007-2012 (whereby I was able to add the variable age and age�) showed no significant influence of gender.
These findings can be explained in two different ways. One possibility is that there was indeed a gender difference, but that it disappeared due to changes in the social- and/or economic- environment. Besides, there is the possibility that there has never been a gender difference and there are other unobservable variables that explain this difference. This
conclusion is not totally in line with the findings of Graham et al. (2012): among their female respondents they found a strong preference for a low percentage of variable pay whereas I found no gender differences. A possible explanation for these different findings is that the influence of executives on their compensation scheme is negligible. Another explanation would be that their answers in a survey are different than their actual negotiation behaviour.
Finally, it is good to notice that women who get into the executive board have selected themselves in their position over a very long period. It is plausible to assume that their
personal characteristics are different from other women and this could explain why the findings of this paper are different from the findings of Eckel and Grossman (2002) in their laboratorial research. Therefore, the conclusion of this paper does not extend to the gender differences in general.
Most of the shortcomings of this paper are due to a lack of information on the executives. I was not able to find reliable information about their tenure and experience whereas it can be assumed that this has an effect on their compensation scheme (Ryan and Wiggins III, 2001). Besides tenure and experience, the variable age plays a significant role in the executives’ compensation scheme. Yet, I was not able to find age data for a large minority of the observations.
Another shortcoming of this paper is the fact that I only looked at the package of current compensation, the base salary and the bonuses. If you want to look at the aspects of
gender differences in risk aversion you have to look at all variable payments, including stock options, pension funds and other variable and fixed payments. On the other side, it also has some advantages: by looking at current compensation only, the difference is better aligned with risk-aversion. If stock options and pension funds were to be included, other variables would start to play a role (like gender difference in intertemporal preferences).
Further research should focus on the missing data in this paper. If I would have more data on the age of executives before 2006, and data about the tenure and experience, the results would be stronger. In a situation wherein these drawbacks are overcome one could further develop the analysis of the relationship between the gender difference and the current financial crisis.
Reference List:
Barber, Brad M. and Terrance Odeon (2001), Boys Will Be Boys: Gender, Overconfidence and Common Stock Investment, Quarterly Journal of Economics.
Buser, Thomas (2009), The Impact of the Menstrual Cycle and Hormonal Contraceptives on Competitiveness, Journal of Economic Behavior & Organization.
Croson, Rachel, and Uri Gneezy (2008), Gender Differences in Preferences, Journal of Economic Literature, 47(2): 1 – 27.
Datta Gupta, Nabanita, Anders Poulsen and Marie-Claire Villeval (2005), Male and Female Competitive Behaviour – experimental evidence.
Eckel, Catherine C. and Philip J. Grossman (2002), Sex Differences and Statistical
Stereotyping in Attitudes Towards Financial Risk, Evolution and Human Behaviour, 23: 281 – 295.
Gneezy, Uri, Kenneth L. Leonard, and John A. List (2009), Gender Differences in
Competition: Evidence from a Matrilineal and a Patriarchal Society, Econometrica, 77(5): 1637 – 1664.
Gneezy, Uri, Muriel Niederle and Aldo Rustichini (2003), Performance in Competitive Environments: Gender Differences, Quarterly Journal of Economics, 118(3): 1049 – 1074. Graham, John R., Campbell R. Harvey and Manju Puri (2012), Managerial Attitudes and Corporate Actions.
Niederle, Muriel, and Lise Vesterlund (2007), Do Women Shy Away from Competition? Do Men Compete Too Much?, Quarterly Journal of Economics, 122(3): 1067 – 1101.
Polachek, Solomon W. (1981), Occupational Self-Selection: A Human Capital Approach to Sex Differences in Occupational Structure, Review of Economics and Statistics, 63(1): 60 – 69.
Ryan, Harley E. jr., Roy A. Wiggens III (2001), The Influence of Firm- and Manager Specific Charasteristics on the Structure of Executive Compensation, Journal of Corporate Finance, 7:101 - 123
Schubert, Renate, Martin Brown, Matthias Gysler and Hans Wolfgang Brachinger (1999), Financial Decision-Making: Are Women Really More Risk-Averse?, American Economic Review, 89(2): 381 – 385.
Smit, Jeroen (presentation) (2010), De Vrouwelijke Leider, NTR Leiders Gezocht, consulted on: 28 June 2013 from: http://www.uitzendinggemist.nl/afleveringen/1005160
