Addendum to the Game Audio Industry Survey 2016:
An Analysis of Income Differences Between Men and Women in Game Audio
The 2016 Game Audio Industry Survey results found a significant discrepancy in both median and average incomes between male and female respondents. The average male income reported as consistently higher than the average female income. We felt these results warranted a deeper analysis and discussion.This follow-on to the Game Audio Industry Survey attempts to dig deeper into this discrepancy.
High-level Summary of Results:
A greater percentage of women reported income from game audio (10.4%) than prior years (7% in 2015, 3.5% in 2014)
Even after accounting for the lower average number of years of experience in the game audio industry, women make less than men
The difference in total income is equivalent to approximately 2.1 years of experience; That is, the ‘cost’ of being female in game audio is approximately the equivalent of having 2.1 fewer years of experience in the game audio industry.
First the good news: This year saw a continued increase in the percentage of female respondents to the Game Audio Industry Survey. 61 of the 587 usable responses were by women (10.4%); this is up from 7% in 2015 and 3.5% in 2014.
However, the 2016 Survey suggested a gender-based discrepancy in incomes and salary, showing women earning less than men in game audio. We wanted to see if the discrepancy was indeed a factor of gender, or was a byproduct of some other factor. As we noted in the main survey, although the reported incomes of women were lower than those of men, women also had fewer years of experience in the game industry. And since years experience in game audio was shown to have a significant correlation with salary and income, perhaps the fact that women in game audio reported fewer years of experience in game audio might be among the reasons for the difference.
To determine if this was the case, we hired professional statistician, Dr. Mary Siegrist, PhD from IMSA Consulting to analyze the raw data. We specifically commissioned a statistical analysis to determine what amount of the difference in income, if any, was attributable to gender and what might be due to other issues such as education or number of years’ experience in game audio, sampling error or other factors. We asked to consider three different types of respondents:
All respondents, considering their total income (freelance or salaried income, including any 'on the side' income of employees of game companies or audio contracting houses)
Salaried employees only (not including any ‘on the side’ income they might earn)
First, we asked for a more thorough statistical analysis to verify our prior finding that "years in the game audio industry" has a statistically valid impact on total income.
The analysis verified that for both salaried and freelancers:
“Total Income has a positive correlation with Years in Game Industry, which means that the longer you work in the gaming industry the more likely your income will also increase.”
Given the verified relationship between income and years of experience in the industry, variance analyses were performed to determine if the income difference between men and women could be attributed to the differences in experience. The analyses concluded:
"There is sufficient evidence to determine that gender has a correlation with total [game audio] income earned. Females earn less than males regardless of the length of time working in the game industry."
The difference in income was then quantified, with the result:
“The cost of being female is 2.15 years of experience. However, females never do catch up males.”
Analysis of Salaried Employees and Freelancers
Note that for all analyses, we removed respondents who reported that more than half of their income was from non-audio sources (aka their ‘day job’) and also filtered out any total annual compensation < $1,000. This was to remove any hobbyists or ‘aspiring’ respondents from the analysis, or for whom 'game audio' is not their primary income.
1) All Respondents (Full-time Freelance and Salaried, including ‘on the side’ income by salaried employees)
We looked at “total merged income” which is the sum total of any salaries paid to an employee plus any freelance income these employees had on the side, or the total amount of freelance income (if the person was a pure freelancer). The analyses showed a statistically significant correlation between gender and income earned; women make less than men, even accounting for differences in experience.
“Total Merged Income has a positive correlation with Years in Game Industry, which means that the longer you work in the gaming industry the more likely your income will also increases. Correlation was significant, r = .513, p < .0001.
Gender has a negative correlation with Total Income, which means that gender plays a role in total income regardless of the number of years in game industry. In this study females earn less than males. Correlation was significant”
There is a demonstrated statistically significant difference between genders and total income earned, F(1,427)=9.553, p=.002. Females (n=48) average total income was $45,469.27 and Males (n=381) average total income was $75,615.68.
2) Salaried Employees; Salary Only
We also looked specifically at Salaried employees only, removing any full-time freelancers and also any “on the side” income reported by the salaried employee. The analyses showed a statistically significant correlation between gender and income earned; women make less salary income than men, even accounting for differences in experience.
Total Salaried Employee Only Income has a positive correlation with Years in Game Industry, which means that the longer you work in the gaming industry your income also increases. Correlation was significant, r = .580, p < .0001.
Gender has a weak negative correlation with Total Salaried Employee Only Income, which means that gender plays a role in total income, but that there may be variables that have a stronger impact. The females earn less than males in relationship to this study. Correlation was significant, r = -.151, p -= .009.
There is a demonstrated statistically significant difference between genders and total salaried employee only income earned, F(1,297)=6.921, p=.009. Females (n=32) average total income was $52,716.41 and Males (n=266) average total income was $75,131.68. Females average total income was higher than total income (including freelance income) for females. Males average total income was similar with total income (including freelance income).
3) Freelancers Only
We finally considered Freelancers only. Here the results show no significant correlation between gender and income. Freelance income differences between men and women could be due to factors other than gender
Total Freelance Only Income has a positive correlation with Years in Game Industry, which means that the longer you work in the gaming industry your income also increases. Correlation was significant, r = .489, p < .0001.
Gender has a weak negative correlation with Total Freelance Only Income, which means that gender plays a role in total income, but that there are other, factors which could have a stronger impact on the income difference. Using a significance level of .05 (also referred to as alpha), there is no statistically significant difference between gender and total income earned. Correlation was significant, r = -.160, p -= .068.
There is no statistically significant difference between genders and Total Freelance Only Income earned, F(1,130)=3.384, p=.068. Females (n=16) average total income was $26,750.00 and Males (n=115) average total income was $69,667.64. Females average total income is lower than Males average total income, however the difference between genders is not statistically significant using an alpha of 0.05.
All Respondents Data
It is helpful to view compensation of all respondents graphically by gender (Freelancers and Salaried, including ‘on the side’ income, including all respondents). Note that this includes all respondents who reported their income, except those explicitly calling themselves 'hobbyists or aspirational.'
Finally, we looked to see if the difference in gender income was due to differences in education level. There was not a significant relationship between respondents’ gender and their level of education. In addition, somewhat surprisingly, when asked to consider the effect of education on income, the conclusion was:
“There is no statistical significance between Education Levels and Income [Freelance or salaried]”
It should be pointed out (as noted in the Game Audio Industry Survey), however, that 95% of respondents engaged in game audio had at least some college; 72% had graduated with Bachelors degrees or higher.
Although there is certainly good news in the increasing participation of women in game audio, even accounting for number of years experience, there is a clear discrepancy in the compensation rates between men and women; a "Gender Gap" exists in game audio. In addition, the highly compensated and very highly compensated women are significantly underrepresented.
It should be noted that in the survey, any questions on compensation were optional; this was done to maximize survey participation and increase accuracy in other areas of the survey. 483 of the 559 non-hobbyists, non-0-income respondents provided their compensation, including 51 women, salaried and freelancers; 16% of total female respondents (10 of 61) declined to provide compensation information.
It is also important to note that there may be causes beyond those which were analyzed that caused the results in the findings presented. The type and format of the survey render it difficult to make any definitive case whether the discrepancy in compensation is systemic (gender bias) or is an artifact of the survey itself. Among possible causes are:
Women are paid less than men, systemically (Gender bias)
Highly compensated women were less likely to report their income than highly compensated men
Men were more likely to exaggerate their income
Women were more likely to underestimate their income
Some other unknown bias or artifact of the survey itself
However, given the data provided by the respondents, statistical analysis showed a clear correlation between gender and income. The amount that women make less than men in game audio is the equivalent of just over two years of experience in the game audio industry.
Analyses are performed in 3 steps:
First an analysis is done to see if there is a relationship between # of years in the game audio industry and income (First bullet point).
Next, a two-tailed Pearson correlation is performed to determine the strength of the correlation (if any) between gender and income (second bullet point in the analyses).
'Then an ANOVA (Analysis of Variance) is done to determine if the correlation is due to gender and not sampling error or chance (third bullet point in the analyses)
Calculation of “cost of being female in years experience” analysis done using ANOVA with a post-hoc Bonferroni.
Although fewer women than men participated in the survey by a 9:1 ratio, a Levenes Test for Equality of Variances test was performed. The Levene’s test was significant, p = .012 which means equal variance for each group and then the equality of means was significant, p = .002, also evidence that the results can be trusted.
Limitations of the Survey Format
The 2016 Game Audio Survey, like any survey, has inherent limits and biases. Although the survey attempts to obtain an accurate picture of the industry, from those working in the industry, it may not be fully representative of the industry as a whole. Some potential issues include, but are not limited to:
The survey was publicized via social media and email networks and known audio groups, and via some major music industry web sites. This may bias results towards the ‘more connected’ composers and sound designers in the industry.
A small number of very anomalous looking responses were all or in part discarded. This may result in pre-conception bias.
A very small number of responses were not self-consistent. These were analyzed manually to determine intent or discarded. This may result in pre-conception bias.
In order to increase participation, survey questions directly related to compensation were optional.
Some number of participants may have misrepresented their data or misinterpreted survey questions.
Thank you to the Game Audio Network Guild for assisting in the Survey
Executive Director, GameSoundCon