...with numbers like these — why would women want to work in games?
April 4, 2013 7:35 AM Subscribe
"The next time someone tells me that men and women get paid equally for their talents in the game industry, I wanted something to link to them." [via]
Also, what is the average number of years' experience in the industry for women as compared to men? The statistics show a massive numerical gender imbalance - if that is correcting itself over time, it may be that the women have significantly less experience overall. If you look at the pie-charts, experience seems to have a huge impact on salary, far more so than gender.
Not saying that there isn't a big problem here, but this article is just pontificating without all the facts.
posted by YAMWAK at 7:56 AM on April 4, 2013 [3 favorites]
Not saying that there isn't a big problem here, but this article is just pontificating without all the facts.
posted by YAMWAK at 7:56 AM on April 4, 2013 [3 favorites]
See also this horrendous story and this hot bullshit. Makes me want to stop playing games altogether sometimes.
posted by kmz at 7:57 AM on April 4, 2013 [5 favorites]
posted by kmz at 7:57 AM on April 4, 2013 [5 favorites]
Statistics and polling are extremely powerful ways to gather a lot of fairly reliable information. They are also rather cheap and susceptible to any number of biases. Most people who know how to read and analyze this type of data know what is meaningful and what is not. Most people who, in this case, refuse to believe the data will point out to any number of small flaws or weaknesses to claim that the whole enterprise is bunk.
Which is a long winded way of saying that someone needs to fund some audit experiments where people (males and females) pose as prospective employees and see how their gender affects the salary being offered. There are much less variables in that type of data. You can also control for experience, skills, etc.
posted by MisantropicPainforest at 8:09 AM on April 4, 2013 [3 favorites]
Which is a long winded way of saying that someone needs to fund some audit experiments where people (males and females) pose as prospective employees and see how their gender affects the salary being offered. There are much less variables in that type of data. You can also control for experience, skills, etc.
posted by MisantropicPainforest at 8:09 AM on April 4, 2013 [3 favorites]
kmz - she mentions in her update that the website the journalist was writing for outed themselves. Does anyone happen to know which website so I can blacklist them?
posted by royalsong at 8:14 AM on April 4, 2013
posted by royalsong at 8:14 AM on April 4, 2013
Are these numbers worse than national averages for all industries? This is a deep rooted cultural issue that spans all industries and should be attacked on many fronts. I encourage women who want to work in games to work in games and demand the salary they deserve. I feel like the industry is becoming more self-aware of these issues than most.
posted by jeffamaphone at 8:32 AM on April 4, 2013
posted by jeffamaphone at 8:32 AM on April 4, 2013
On first read of YAMWAK's comment, I have to admit it leaves a bad taste in my mouth regarding "it may be that the women have significantly less experience overall". I'm a woman in programming - not GAME programming, but still. Going on ten years in the industry. I have a female coworker who is also a coder who just got her 20 year certificate. My CompSci advisor in college over ten years ago was a woman who'd been in the industry for decades herself. Wasn't there a recent FPP about a woman who was a programmer for IBM back in the vacuum tube and card era?
I may be reading too much into that comment, though. It just reads to me that YAMWAK thinks women haven't been in the industry for very long. And I just don't think that's true at all.
posted by jillithd at 8:34 AM on April 4, 2013 [14 favorites]
I may be reading too much into that comment, though. It just reads to me that YAMWAK thinks women haven't been in the industry for very long. And I just don't think that's true at all.
posted by jillithd at 8:34 AM on April 4, 2013 [14 favorites]
Not saying that there isn't a big problem here, but this article is just pontificating without all the facts.
The article addresses that exact point:
posted by Catseye at 8:37 AM on April 4, 2013 [3 favorites]
The article addresses that exact point:
I’m sure there are more details that might make these numbers less damning. For example, we all know that games have been long dominated by men and the industry is taking small steps to change that. As a result, many of the women who answered the survey might be new to the game industry, might not be in as senior of roles as the men who responded. However, I don’t think this changes the fact that we need to recruit and encourage more women at all levels of every organization — and we’re failing to do so.I would like to believe it's all down to levels of experience, but when gender pay gaps exist across so many sectors where they don't seem to have a direct relationship to experience (like this chart for HR professionals, broken down by level of seniority), I'm sceptical that the games industry is going to be so much better.
posted by Catseye at 8:37 AM on April 4, 2013 [3 favorites]
From the blog comments:
I’m male, and I understand that 50% of the population is having what is probably a very different experience of life to me, one which I will never fully understand. However, the best opportunity for me to do so is by playing those experiences in games, which under the current balance of genders I’m unable to do.
...this so perfectly encapsulates my cranky old fart notions of how a heavy duty gamer might view the world that I have to ask, is he joking?
posted by bonobothegreat at 8:37 AM on April 4, 2013
I’m male, and I understand that 50% of the population is having what is probably a very different experience of life to me, one which I will never fully understand. However, the best opportunity for me to do so is by playing those experiences in games, which under the current balance of genders I’m unable to do.
...this so perfectly encapsulates my cranky old fart notions of how a heavy duty gamer might view the world that I have to ask, is he joking?
posted by bonobothegreat at 8:37 AM on April 4, 2013
I'm sure there's an actual gender gap in salaries, but hasn't anyone considered that the real problem is that women are fictional and therefore don't deserve to get paid as much, and certainly not in real physical dollars?
posted by shakespeherian at 8:37 AM on April 4, 2013 [17 favorites]
posted by shakespeherian at 8:37 AM on April 4, 2013 [17 favorites]
If you want to nitpick this, why not pick the WHY THE HELL DID YOU USE PIE CHARTS FOR THIS nit!?!?! Pie charts are great if you want to say X% of a set is A, Y% of a set is B and Z% of the set is C, over a single set of variables where A, B and C add up to 100%. So they gave us years in the industry data on the pie chart (which does add up to 100% but isn't really the statistic we care about) and then shoehorned in wage data and then realized that they didn't include gender data so they shoved that in somewhere else in a two member table. There is probably meaningful data here, but armed with those graphs I mostly have to take their world for it.
A much better way to have done this would be with a group bar graph with six bars - <3 years (male), <3 years (female), 3-6 years (male), etc. Then I could at least know if I was looking at something that was apples to apples, or not.
Also, would it kill them to tell me the number of respondents in each class?
posted by Kid Charlemagne at 8:37 AM on April 4, 2013 [8 favorites]
A much better way to have done this would be with a group bar graph with six bars - <3 years (male), <3 years (female), 3-6 years (male), etc. Then I could at least know if I was looking at something that was apples to apples, or not.
Also, would it kill them to tell me the number of respondents in each class?
posted by Kid Charlemagne at 8:37 AM on April 4, 2013 [8 favorites]
I may be reading too much into that comment, though. It just reads to me that YAMWAK thinks women haven't been in the industry for very long. And I just don't think that's true at all.
I had a similar thought to YAMWAK, not because I think women haven't been in the industry that long, but because comparing men and women's overall average salaries just seems like a sloppy way to analyze the data. I suspect that the data, properly adjusted to account for experience, would reveal similar patterns, but adjusting the data for the single factor that most strongly determines salary within an industry seems like an obvious thing to do.
posted by Bulgaroktonos at 8:38 AM on April 4, 2013 [1 favorite]
I had a similar thought to YAMWAK, not because I think women haven't been in the industry that long, but because comparing men and women's overall average salaries just seems like a sloppy way to analyze the data. I suspect that the data, properly adjusted to account for experience, would reveal similar patterns, but adjusting the data for the single factor that most strongly determines salary within an industry seems like an obvious thing to do.
posted by Bulgaroktonos at 8:38 AM on April 4, 2013 [1 favorite]
Wasn't there a recent FPP about a woman who was a programmer for IBM back in the vacuum tube and card era?
There's also this FPP about the women who programmed ENIAC, and this great article (which I'm sure was linked to on here but can't now find, dammit) on "Computer Girls" in the 60s.
posted by Catseye at 8:51 AM on April 4, 2013 [3 favorites]
There's also this FPP about the women who programmed ENIAC, and this great article (which I'm sure was linked to on here but can't now find, dammit) on "Computer Girls" in the 60s.
posted by Catseye at 8:51 AM on April 4, 2013 [3 favorites]
It's worth noting that saying "this dataset is insufficient" is not the same thing as saying "there is obviously no problem with disparate treatment of men and women in this industry."
posted by yoink at 8:54 AM on April 4, 2013 [3 favorites]
posted by yoink at 8:54 AM on April 4, 2013 [3 favorites]
The AAUW released a report last fall that looked at the gender gap in pay. The report uses descriptive statistics and regression analysis to describe pay differences between women and men one year after college graduation so length of experience in a field was not at issue. From the pdf of the executive summary:
posted by rtha at 8:57 AM on April 4, 2013 [32 favorites]
Both discrimination and cultural gender norms can play a role in the “explained” portion of the pay gap. With that in mind, we find that college major is an important factor driving pay differences. Men are more likely than women to major in fields like engineering and computer science, which typically lead to higher-paying jobs. Women are more likely than men to major in fields like education and the social sciences, which typically lead to lower-paying jobs. But college major is not the full story. One year after graduation, a pay gap exists between women and men who majored in the sameFull report (.pdf)
field. Among business majors, for example, women were paid just over $38,000, while men were paid just over $45,000. Gender differences in college major only partially explain the pay gap.
[...]
Yet, when we control for each of these factors [number of hours worked, sector of the economy], women still tended to be paid less than their male peers were paid. Within a number of occupations, women were already paid less than men were paid just one year out of college. Among teachers, for example, women were paid 89 percent of what men were paid. In business and management occupations, women were paid 86 percent of what men were paid; similarly, in sales occupations, women were paid just 77 percent of what their male peers were paid.
When we compare the earnings of men and women who reported working the same number of hours, men were paid more than women were paid. For example, among those who reported working 40 hours per week, women were paid 84 percent of what men were paid. Among those who reported working 45 hours per week, women’s earnings were 82 percent of men’s.
[...]
Regression analysis allows us to analyze the effect of multiple factors on earnings at the same time. Controlling for occupation, college major, hours worked per week, and many other factors all at once, we found that college-educated women working full time were paid an unexplained 7 percent less than their male peers were paid one year out of college.
posted by rtha at 8:57 AM on April 4, 2013 [32 favorites]
rtha,
that is pretty compelling. Thanks for posting.
posted by MisantropicPainforest at 9:03 AM on April 4, 2013
that is pretty compelling. Thanks for posting.
posted by MisantropicPainforest at 9:03 AM on April 4, 2013
Among teachers, for example, women were paid 89 percent of what men were paid.
How does this happen? Aren't most teacher contracts based solely on experience and education?
posted by nolnacs at 9:07 AM on April 4, 2013
How does this happen? Aren't most teacher contracts based solely on experience and education?
posted by nolnacs at 9:07 AM on April 4, 2013
This is a very weak statistical presentation of what is certainly a real and significant problem, which is frustrating because presenting statistical "evidence" that is as full of gaps as this just gives ammunition to people who would try to deny the existence of systemic gender bias in the workplace.
posted by Scientist at 9:20 AM on April 4, 2013 [2 favorites]
posted by Scientist at 9:20 AM on April 4, 2013 [2 favorites]
Scientist, in what way are the statistics weak? Genuine question, this is a subject I don't know very much about.
posted by KathrynT at 9:26 AM on April 4, 2013
posted by KathrynT at 9:26 AM on April 4, 2013
How does this happen? Aren't most teacher contracts based solely on experience and education?
There could be a variety of reasons. One, with teaching, is that women are often encouraged, rather individually or on the whole, to teach younger kids--you rarely see men as kindergarten teachers--which pays less (while men are encouraged to teach in high schools or colleges). For some reason, teaching age groups where one is assumed to be more nurturing pays less. I'm sure it has nothing to do with how feminine that career is perceived to be.
posted by PhoBWanKenobi at 9:34 AM on April 4, 2013 [5 favorites]
There could be a variety of reasons. One, with teaching, is that women are often encouraged, rather individually or on the whole, to teach younger kids--you rarely see men as kindergarten teachers--which pays less (while men are encouraged to teach in high schools or colleges). For some reason, teaching age groups where one is assumed to be more nurturing pays less. I'm sure it has nothing to do with how feminine that career is perceived to be.
posted by PhoBWanKenobi at 9:34 AM on April 4, 2013 [5 favorites]
I guess not enough women were present in every category of experience to make a clearer comparison. But, it would have been helpful to also mention the average experience of men vs women when mentioning the average salary of men vs women. These breakdowns were available to the author, right?
posted by asra at 9:39 AM on April 4, 2013
PhoBWanKenobi: For some reason, teaching age groups where one is assumed to be more nurturing pays less. I'm sure it has nothing to do with how feminine that career is perceived to be.
For whatever it's worth, high school teachers often have more education than those who teach younger students, since it is considered beneficial for them to have advanced training in the subjects they teach. This is particularly true of instructors who teach advanced science and math classes.
posted by Mitrovarr at 9:48 AM on April 4, 2013 [1 favorite]
For whatever it's worth, high school teachers often have more education than those who teach younger students, since it is considered beneficial for them to have advanced training in the subjects they teach. This is particularly true of instructors who teach advanced science and math classes.
posted by Mitrovarr at 9:48 AM on April 4, 2013 [1 favorite]
gilrain: "Maybe clarify why you find it funny? In any case, I doubt that poster was joking. For instance, aside from being friends with and talking to the women in my life, an important way I've increased my understanding of the experience of being a woman is by reading novels from their perspective.
"
I think the "joking" bit comes from the idea that gaming can actually provide the fullness of the experience. Gaming can't provide an experience outside of certain artificial boundaries: there's never going to be, ex, an achievement for dealing with microagressions or any other situation that happens to women on the daily.
We have things like Tomb Raider's reboot, which has handled complex situations with a woman protagonist( not without controversy), but how do the setpieces in that game help me as a man experience "woman-ness" in any meaningful way? Is that even a possibility? I'd like to think so, but the state of the industry doesn't seemed geared to it.
Talking to women and reading seem like a better way to understand the experience than playing video games, in my estimation.
posted by boo_radley at 9:50 AM on April 4, 2013 [2 favorites]
"
I think the "joking" bit comes from the idea that gaming can actually provide the fullness of the experience. Gaming can't provide an experience outside of certain artificial boundaries: there's never going to be, ex, an achievement for dealing with microagressions or any other situation that happens to women on the daily.
We have things like Tomb Raider's reboot, which has handled complex situations with a woman protagonist( not without controversy), but how do the setpieces in that game help me as a man experience "woman-ness" in any meaningful way? Is that even a possibility? I'd like to think so, but the state of the industry doesn't seemed geared to it.
Talking to women and reading seem like a better way to understand the experience than playing video games, in my estimation.
posted by boo_radley at 9:50 AM on April 4, 2013 [2 favorites]
Well, the first thing that jumped out at me was what YAMWAK pointed out up above, that they didn't control for years of experience. If women on the whole have fewer years in the industry than men (as you might expect, in a field which has historically been and still is so heavily male-dominated but which recently has been showing signs of starting to be colonized by women) then the pay gaps could in theory be explained by that alone.
It also relies on self-submitted data, and doesn't offer any evidence of an attempt to validate that data or to control for the self-selected nature of the people who were surveyed. It's not taking a random sample of the industry -- it's using the sample of the industry that felt moved for whatever reason to enter their job statistics into the Game Developer Magazine salary survey. Perhaps men who are making above-average salaries are more likely to fill out this survey than others -- we don't have any way of knowing, based on the presentation in this article.
The article also makes no mention of sample sizes, confidence intervals for the data, or of the significance values. Let's say that 100 audio developers responded to the survey, for instance (which for all I know might be a realistic number -- we have no idea how many people responded to this survey, but it's a niche magazine and audio developers are only a subset of respondants so it's plausible). That would mean that only four female audio developers responded to the survey, which is far too few to draw any conclusions about what the average salaries are for women in that sub-field as a whole. As a baseline, you'd typically want an absolute minimum of 30 respondants in a survey of this type (assuming that your sample set was really high quality, which this one isn't since it's a self-reported survey) if you were trying to publish a peer-reviewed paper and wanted to avoid getting laughed at by your reviewers.
So basically we have some obvious possible confounding variables that haven't been accounted for, a poor methodology, an unknown (and possibly too small) sample size, and no indication of whether the results are actually statistically significant or whether they might be equally well explained by random noise in the data.
I don't want to say systemic bias against women in the game-developing industry isn't a problem, though! I'm sure it is, I'm certain of it. If nothing else, we have numerous reports from women describing their own lived experiences in the industry, and they are not to be discounted. Bias against women is absolutely a problem in the games industry. My issue is just that this article is full of holes and opens up far too many lines of attack for those who would like to deny that the problem exists. Anyone with a working knowledge of statistics (which is not everyone of course, but there are plenty of us out there and we tend to be pretty vocal on the internet) could poke this study full of holes very easily, and an unscrupulous rhetorician (also a pretty common and vocal creature on the internet) could use those holes to make an at least superficially-persuasive argument that evidence for bias against women in games developing is lacking and therefore presumed to be nonexistent. You can see people using techniques like that every day, from the seedy underbelly of Reddit all the way up to the international news media and the floor of national legislatures. It's poisonous and despicable and also very hard to defend against.
And that's why it's important to use good statistics to support your case -- statistics are a weapon, and if you don't know how to use them then your enemy can turn them against you.
posted by Scientist at 9:51 AM on April 4, 2013 [18 favorites]
It also relies on self-submitted data, and doesn't offer any evidence of an attempt to validate that data or to control for the self-selected nature of the people who were surveyed. It's not taking a random sample of the industry -- it's using the sample of the industry that felt moved for whatever reason to enter their job statistics into the Game Developer Magazine salary survey. Perhaps men who are making above-average salaries are more likely to fill out this survey than others -- we don't have any way of knowing, based on the presentation in this article.
The article also makes no mention of sample sizes, confidence intervals for the data, or of the significance values. Let's say that 100 audio developers responded to the survey, for instance (which for all I know might be a realistic number -- we have no idea how many people responded to this survey, but it's a niche magazine and audio developers are only a subset of respondants so it's plausible). That would mean that only four female audio developers responded to the survey, which is far too few to draw any conclusions about what the average salaries are for women in that sub-field as a whole. As a baseline, you'd typically want an absolute minimum of 30 respondants in a survey of this type (assuming that your sample set was really high quality, which this one isn't since it's a self-reported survey) if you were trying to publish a peer-reviewed paper and wanted to avoid getting laughed at by your reviewers.
So basically we have some obvious possible confounding variables that haven't been accounted for, a poor methodology, an unknown (and possibly too small) sample size, and no indication of whether the results are actually statistically significant or whether they might be equally well explained by random noise in the data.
I don't want to say systemic bias against women in the game-developing industry isn't a problem, though! I'm sure it is, I'm certain of it. If nothing else, we have numerous reports from women describing their own lived experiences in the industry, and they are not to be discounted. Bias against women is absolutely a problem in the games industry. My issue is just that this article is full of holes and opens up far too many lines of attack for those who would like to deny that the problem exists. Anyone with a working knowledge of statistics (which is not everyone of course, but there are plenty of us out there and we tend to be pretty vocal on the internet) could poke this study full of holes very easily, and an unscrupulous rhetorician (also a pretty common and vocal creature on the internet) could use those holes to make an at least superficially-persuasive argument that evidence for bias against women in games developing is lacking and therefore presumed to be nonexistent. You can see people using techniques like that every day, from the seedy underbelly of Reddit all the way up to the international news media and the floor of national legislatures. It's poisonous and despicable and also very hard to defend against.
And that's why it's important to use good statistics to support your case -- statistics are a weapon, and if you don't know how to use them then your enemy can turn them against you.
posted by Scientist at 9:51 AM on April 4, 2013 [18 favorites]
gilrain: "That's why the original post we're discussing was about wishing the gender balance in the games industry was better, so that we'd have a chance of cultivating more games that could."
Will salary parity put women in a place where they can direct the types of games that are being created? If you're a woman working as an artist for Shootmans 5: Manshooter Down, can you start up a more thoughtful and meaningful game at the same time simply because you're paid equitably? Will your employer back that project? Or do you just have the personal financial stability to kickstart something with likeminded people and take the risk yourself? I don't know the answer to these questions -- I'm trying to draw the line from "pay equity" to "creative agency", and I'm not sure how to get there.
gilrain: "For instance, in terms of microaggressions: play an MMO or other online game as a female character, and if you do so mindfully you'll notice how very differently you're treated. "
Hadn't considered this. I wind up playing a female toon for most online games, but -- and this may be where I'm missing the experience -- I tend not to game outside my personal social circle. Mefight Club is the exception, and I think they tend to be respectful. Outside of that, I find most MMOs to be pretty boring.
Your PA article was a good read. It's inspirational to see how there are mechanisms for trans people to express themselves more genuinely online.
posted by boo_radley at 10:43 AM on April 4, 2013
Will salary parity put women in a place where they can direct the types of games that are being created? If you're a woman working as an artist for Shootmans 5: Manshooter Down, can you start up a more thoughtful and meaningful game at the same time simply because you're paid equitably? Will your employer back that project? Or do you just have the personal financial stability to kickstart something with likeminded people and take the risk yourself? I don't know the answer to these questions -- I'm trying to draw the line from "pay equity" to "creative agency", and I'm not sure how to get there.
gilrain: "For instance, in terms of microaggressions: play an MMO or other online game as a female character, and if you do so mindfully you'll notice how very differently you're treated. "
Hadn't considered this. I wind up playing a female toon for most online games, but -- and this may be where I'm missing the experience -- I tend not to game outside my personal social circle. Mefight Club is the exception, and I think they tend to be respectful. Outside of that, I find most MMOs to be pretty boring.
Your PA article was a good read. It's inspirational to see how there are mechanisms for trans people to express themselves more genuinely online.
posted by boo_radley at 10:43 AM on April 4, 2013
...we found that college-educated women working full time were paid an unexplained 7 percent less than their male peers were paid one year out of college.
I like how this is phrased without jumping to conclusions about causes, because there are a lot of subtle and sometimes hard-to-measure things* that can affect pay, and jumping to conclusions helps no one.
* Anecdote time: When my then girlfriend and I were both ~1 year out of college, working in the same technical industry, with similar skill levels and ability, she made ~18% more than I did. If I had to explain it, I'd attribute it to personality differences that led her/allowed her to job hop more aggressively.
posted by cosmic.osmo at 11:06 AM on April 4, 2013
I like how this is phrased without jumping to conclusions about causes, because there are a lot of subtle and sometimes hard-to-measure things* that can affect pay, and jumping to conclusions helps no one.
* Anecdote time: When my then girlfriend and I were both ~1 year out of college, working in the same technical industry, with similar skill levels and ability, she made ~18% more than I did. If I had to explain it, I'd attribute it to personality differences that led her/allowed her to job hop more aggressively.
posted by cosmic.osmo at 11:06 AM on April 4, 2013
Hold on... male QA testers are making $50K/year on average?!
posted by Mister Fabulous at 11:26 AM on April 4, 2013 [1 favorite]
posted by Mister Fabulous at 11:26 AM on April 4, 2013 [1 favorite]
The article also makes no mention of sample sizes, confidence intervals for the data, or of the significance values. Let's say that 100 audio developers responded to the survey, for instance (which for all I know might be a realistic number -- we have no idea how many people responded to this survey, but it's a niche magazine and audio developers are only a subset of respondants so it's plausible). That would mean that only four female audio developers responded to the survey, which is far too few to draw any conclusions about what the average salaries are for women in that sub-field as a whole.
I suspect given the very exact number of $50,000 (not $49,999) for average salary, which is the sort of thing that people tend to give when replying to open-ended surveys, that there were 24 men and 1 woman surveyed in this category. (At most, two women, one earning $45K and one earning $55K or something like that.) Which doesn't mean that women aren't underpaid in the industry, but rather that we need better statistics about that.
posted by Homeboy Trouble at 11:39 AM on April 4, 2013
I suspect given the very exact number of $50,000 (not $49,999) for average salary, which is the sort of thing that people tend to give when replying to open-ended surveys, that there were 24 men and 1 woman surveyed in this category. (At most, two women, one earning $45K and one earning $55K or something like that.) Which doesn't mean that women aren't underpaid in the industry, but rather that we need better statistics about that.
posted by Homeboy Trouble at 11:39 AM on April 4, 2013
This is a really unfortunate blog post. I don't doubt that there's a wage gap in the game industry -- there is everywhere else, after all -- but there's just not enough in these so-called statistics to even begin to support that conclusion.
It'd be much more useful to weigh the modes (i.e. what most men earn vs. what most women earn) rather than the means (which I assume "average" refers to in these charts), as the latter can be very easily skewed by a small handful of ridiculously overpaid male outliers. As so many more men than women responded to the survey, that scenario is much more likely than if the sexes were split 50-50; ditto for a small sample size.
Is there any raw data in the original GD article all these charts came from? Do they even say how many respondents there were?
posted by Sys Rq at 12:48 PM on April 4, 2013
It'd be much more useful to weigh the modes (i.e. what most men earn vs. what most women earn) rather than the means (which I assume "average" refers to in these charts), as the latter can be very easily skewed by a small handful of ridiculously overpaid male outliers. As so many more men than women responded to the survey, that scenario is much more likely than if the sexes were split 50-50; ditto for a small sample size.
Is there any raw data in the original GD article all these charts came from? Do they even say how many respondents there were?
posted by Sys Rq at 12:48 PM on April 4, 2013
Not as far as I can tell, Sys Rq. Which does render the statistics meaningless. Hell, we don't even know if they're statistics. If the sample size for women is 1, that's an anecdote.
posted by Justinian at 12:55 PM on April 4, 2013
posted by Justinian at 12:55 PM on April 4, 2013
And that's why it's important to use good statistics to support your case -- statistics are a weapon, and if you don't know how to use them then your enemy can turn them against you.
This.
Ideally when you display data you want to pick a display method which elucidates all of the relevant information and its interrelationships. The question that one might be asking if all the data was available is "Why are so many women leaving the games industry after fewer than three years?"* That's still a type of bias, but it wouldn't be fixed no mater how level the pay scale was per years of service.
*Restless_Nomad made a comment once that explained this, though, in case you're wondering.
posted by Kid Charlemagne at 12:59 PM on April 4, 2013
This.
Ideally when you display data you want to pick a display method which elucidates all of the relevant information and its interrelationships. The question that one might be asking if all the data was available is "Why are so many women leaving the games industry after fewer than three years?"* That's still a type of bias, but it wouldn't be fixed no mater how level the pay scale was per years of service.
*Restless_Nomad made a comment once that explained this, though, in case you're wondering.
posted by Kid Charlemagne at 12:59 PM on April 4, 2013
Hadn't considered this. I wind up playing a female toon for most online games, but -- and this may be where I'm missing the experience -- I tend not to game outside my personal social circle. Mefight Club is the exception, and I think they tend to be respectful. Outside of that, I find most MMOs to be pretty boring.
I've played MMOs for a while - usually my characters are split roughly even genderwise. I can only think of a couple of occasions where someone has said something sexually inappropriate - but I'm guessing most people assume a given person in an MMO is a guy, anyway. One thing I have noticed, and I'm not sure if its just me, is that people assume you are less competent. Like if there's a group event, people are more likely to try to "help" me at the expense of doing something useful - even if I'm clearly the most competent person on the team.
Then there was that one time I created a black character in WoW - I seriously had someone saying racist crap to me before I left the starting zone.
posted by Zalzidrax at 1:44 PM on April 4, 2013 [1 favorite]
I've played MMOs for a while - usually my characters are split roughly even genderwise. I can only think of a couple of occasions where someone has said something sexually inappropriate - but I'm guessing most people assume a given person in an MMO is a guy, anyway. One thing I have noticed, and I'm not sure if its just me, is that people assume you are less competent. Like if there's a group event, people are more likely to try to "help" me at the expense of doing something useful - even if I'm clearly the most competent person on the team.
Then there was that one time I created a black character in WoW - I seriously had someone saying racist crap to me before I left the starting zone.
posted by Zalzidrax at 1:44 PM on April 4, 2013 [1 favorite]
I'm sorry both that my first comment wasn't clear and secondly that I got distracted today and didn't come back to the discussion until now.
Scientist's excellent post clearly explains the weaknesses of the statistics given - far more clearly than I ever could.
I'll try and put my view forward more clearly.
I absolutely believe that women are underpaid and under-represented in the gaming industry. I believe that the under-representation is at least partially due to the misogyny that at times appears ubiquitous.
If a portion of women join the industry and are then driven away, they will not work their way up to the better paying jobs, skewing the data. It's their game creation experience that I believe would be statistically lower, not their skills or overall professional experience. Others have made this point better than I.
I know that there have been women programmers since the beginning, from Ada Lovelace onwards. Thing is, why would a female programmer work in a blatantly hostile environment when they don't need to? If the skills can carry over to another job, where they will get more respect and possibly more money too, why hang around? Same applies for the other disciplines.
The paragraph at the end of the article identified by Catseye fails to address the second paragraph of my post. The article's author loudly bemoans the disparity between incomes for the different genders and makes that the focus of the article. The paragraph cited basically says 'Well, I could be wrong, but look at this different problem instead'. These are two separate but related issues. The data we've been given in the article does not give us enough information to comment reliably on either one (although the gender disparity is sufficiently blatant that you'd probably be pretty safe in saying things are bad).
rtha's links are far more useful in this regard, but lack the precision to see the gaming specific issues that are almost certainly there. I believe that the gaming industry, by the nature of their customer base and the nature of their product, has a bigger problem with this than other related industry sectors.
There is a huge problem here. This article doesn't help.
Sorry about the miscommunication in my first post, and for any in this one. I don't have a problem with women engineers, programmers or anything like that. I do have a problem with bad statistics.
posted by YAMWAK at 2:10 PM on April 4, 2013
Scientist's excellent post clearly explains the weaknesses of the statistics given - far more clearly than I ever could.
I'll try and put my view forward more clearly.
I absolutely believe that women are underpaid and under-represented in the gaming industry. I believe that the under-representation is at least partially due to the misogyny that at times appears ubiquitous.
If a portion of women join the industry and are then driven away, they will not work their way up to the better paying jobs, skewing the data. It's their game creation experience that I believe would be statistically lower, not their skills or overall professional experience. Others have made this point better than I.
I know that there have been women programmers since the beginning, from Ada Lovelace onwards. Thing is, why would a female programmer work in a blatantly hostile environment when they don't need to? If the skills can carry over to another job, where they will get more respect and possibly more money too, why hang around? Same applies for the other disciplines.
The paragraph at the end of the article identified by Catseye fails to address the second paragraph of my post. The article's author loudly bemoans the disparity between incomes for the different genders and makes that the focus of the article. The paragraph cited basically says 'Well, I could be wrong, but look at this different problem instead'. These are two separate but related issues. The data we've been given in the article does not give us enough information to comment reliably on either one (although the gender disparity is sufficiently blatant that you'd probably be pretty safe in saying things are bad).
rtha's links are far more useful in this regard, but lack the precision to see the gaming specific issues that are almost certainly there. I believe that the gaming industry, by the nature of their customer base and the nature of their product, has a bigger problem with this than other related industry sectors.
There is a huge problem here. This article doesn't help.
Sorry about the miscommunication in my first post, and for any in this one. I don't have a problem with women engineers, programmers or anything like that. I do have a problem with bad statistics.
posted by YAMWAK at 2:10 PM on April 4, 2013
The paragraph at the end of the article identified by Catseye fails to address the second paragraph of my post.
I identified it to point out that the author (of the linked post) wasn't actually ignoring the issue, not to claim that the authors (of the graphs) actually deal with it. They don't, and I agree that it's a puzzling shame they don't, especially when that information obviously was gathered. (The 2011 version is also available but doesn't give that info either, although the gap in pay looks much the same.)
posted by Catseye at 2:26 PM on April 4, 2013
I identified it to point out that the author (of the linked post) wasn't actually ignoring the issue, not to claim that the authors (of the graphs) actually deal with it. They don't, and I agree that it's a puzzling shame they don't, especially when that information obviously was gathered. (The 2011 version is also available but doesn't give that info either, although the gap in pay looks much the same.)
posted by Catseye at 2:26 PM on April 4, 2013
And that's why it's important to use good statistics to support your case -- statistics are a weapon, and if you don't know how to use them then your enemy can turn them against you.
So in the spirit of Scientist's quote, and in the spirit of "everybody complains about the weather, but nobody does anything about it", I did something about it. The next two paragraphs are methodology, you can skip to the table if that's not your thing.
Data source: I used the 2006-2010 American Community Survey Public Use Microdata Sample (ACS PUMS), which is a random sample of US citizens from the Census Bureau, which gets rid of the sample size and self-selection biases. I found the people in the database who worked 30+ hours a week, and who worked 50+ weeks in the previous year. I then found only the workers for companies in NAICS 51-12xx, which is the software publishing industry. (I'm using the 2002 NAICS and SOC definitions). Of these workers for software companies, I used those who were in SOC 15-1xxx occupations, which are technical computer science occupations (i.e. not high level managers, nor general support staff). Finally, I looked at 30-49 year old workers with either a Bachelor's or Master's degree, a compromise between a large enough group and a homogenous enough group. There are 801 observations in this dataset (including 173 women), and the regression is done against the total personal income, which for Census definitions notably excludes capital gains (which could be significant, as stock options are relatively common in software).
Model structure: I estimated a log-linear hedonic model, in the form ln(Income) = intercept + Σ αz, which has the nice property of ultimately producing a model where e^intercept is the average income, and the α are the elasticities to the characteristics z; that is, an estimated α of 0.1 corresponds to a 10% increase in income for every 1 of that thing. Since the model uses entirely binary (1/0) variables, the α represents the change in income for the property being present. Due to Excel's limitations, I did a piecewise estimation, where the wage elasticities of working in the five largest markets for software were estimated (with parameters controlling for specific occupation, age range and degree status); these five parameters were then combined into a single MajorMarket property that had an estimated α of 0.971. (This allowed me to free up space in the 16-parameter limitation.) I'm presenting the individual markets as if they were individually estimated. I'm also transforming the intercept and the α for specific occupations into average incomes for each occupation, for clarity's sake.
Model results:
Average incomes:
Discussion:
Women are, like, totally underpaid in the tech world.
Obviously, this is using proxy measures like age and education rather than experience and skill, but it's still a pretty strong statement. After controlling for age, education, specific occupation and even the labour market, there is still a very strong effect showing that women are paid 80 cents on the dollar.
I tried interacting the female parameter with various characteristics, and found (while not statistically significant) that women got less benefit from a Master's degree, and did worse in the "core" occupations of programmers / software engineers than in the other, more supporting roles. On the other hand, younger women did better, and women in the Bay area did better -- although never as well as their base-case male counterparts.
posted by Homeboy Trouble at 3:25 PM on April 4, 2013 [232 favorites]
So in the spirit of Scientist's quote, and in the spirit of "everybody complains about the weather, but nobody does anything about it", I did something about it. The next two paragraphs are methodology, you can skip to the table if that's not your thing.
Data source: I used the 2006-2010 American Community Survey Public Use Microdata Sample (ACS PUMS), which is a random sample of US citizens from the Census Bureau, which gets rid of the sample size and self-selection biases. I found the people in the database who worked 30+ hours a week, and who worked 50+ weeks in the previous year. I then found only the workers for companies in NAICS 51-12xx, which is the software publishing industry. (I'm using the 2002 NAICS and SOC definitions). Of these workers for software companies, I used those who were in SOC 15-1xxx occupations, which are technical computer science occupations (i.e. not high level managers, nor general support staff). Finally, I looked at 30-49 year old workers with either a Bachelor's or Master's degree, a compromise between a large enough group and a homogenous enough group. There are 801 observations in this dataset (including 173 women), and the regression is done against the total personal income, which for Census definitions notably excludes capital gains (which could be significant, as stock options are relatively common in software).
Model structure: I estimated a log-linear hedonic model, in the form ln(Income) = intercept + Σ αz, which has the nice property of ultimately producing a model where e^intercept is the average income, and the α are the elasticities to the characteristics z; that is, an estimated α of 0.1 corresponds to a 10% increase in income for every 1 of that thing. Since the model uses entirely binary (1/0) variables, the α represents the change in income for the property being present. Due to Excel's limitations, I did a piecewise estimation, where the wage elasticities of working in the five largest markets for software were estimated (with parameters controlling for specific occupation, age range and degree status); these five parameters were then combined into a single MajorMarket property that had an estimated α of 0.971. (This allowed me to free up space in the 16-parameter limitation.) I'm presenting the individual markets as if they were individually estimated. I'm also transforming the intercept and the α for specific occupations into average incomes for each occupation, for clarity's sake.
Model results:
Average incomes:
Computer software engineers $106,485 Computer programmers $99,435 Computer scientists & systems analysts $103,003 Computer support specialists $80,170 Network systems & data comm. analysts $89,009 Network & computer systems admins $137,234 Database administrators $81,386Effects on income:
Denver: 18.7% ** San Jose (Sta Clara): 28.6% *** Seattle (King county): 8.7% * San Francisco / San Mateo: 16.8% *** Raleigh/Durham: -15.6% ** Not in major metro (1M pop): -5.5% Age 30-32: -29.5% *** Age 33-35: -22.0% *** Age 36-39: -10.0% ** Age 40-44: -2.8% Has Master's degree: 4.8% Female: -19.5% ***The R-square is 0.173; one star indicates significance at the 0.9 level, two at the 0.95 level; and 3 at the 0.99 level.
Discussion:
Women are, like, totally underpaid in the tech world.
Obviously, this is using proxy measures like age and education rather than experience and skill, but it's still a pretty strong statement. After controlling for age, education, specific occupation and even the labour market, there is still a very strong effect showing that women are paid 80 cents on the dollar.
I tried interacting the female parameter with various characteristics, and found (while not statistically significant) that women got less benefit from a Master's degree, and did worse in the "core" occupations of programmers / software engineers than in the other, more supporting roles. On the other hand, younger women did better, and women in the Bay area did better -- although never as well as their base-case male counterparts.
posted by Homeboy Trouble at 3:25 PM on April 4, 2013 [232 favorites]
A bunch of people (here and elsewhere) keep asking about the breakdowns. The editor posted an update on Gamasutra with some numbers. Here's some of the relevant details:
Across all disciplines, the men we surveyed are more likely to have more experience. 623 male devs have over six years of industry experience, 426 devs have 3-6 years, and 284 have less than three years. Women were mostly in the 3-6 year range (77), then >6 years (50), then less than three years (46). So, yes, a higher proportion of highly-experienced male devs means we'd expect higher salaries for men than women overall. To me, this seems to reflect what we as an industry already know; it's not a particularly hospitable industry for women (as indicated by the meager gender ratio), meaning they're less likely to stick around than men are.
The "less than three years" response for women is very worrisome to me, as I expected an opposite trend. There are either fewer women entering the game industry now than in the past... or there's way fewer women aware of the survey, despite the fact that everyone talks about it every year. I can't see why more men, proportionally, would know about and respond to it than women.
posted by subject_verb_remainder at 5:03 PM on April 4, 2013 [3 favorites]
Across all disciplines, the men we surveyed are more likely to have more experience. 623 male devs have over six years of industry experience, 426 devs have 3-6 years, and 284 have less than three years. Women were mostly in the 3-6 year range (77), then >6 years (50), then less than three years (46). So, yes, a higher proportion of highly-experienced male devs means we'd expect higher salaries for men than women overall. To me, this seems to reflect what we as an industry already know; it's not a particularly hospitable industry for women (as indicated by the meager gender ratio), meaning they're less likely to stick around than men are.
The "less than three years" response for women is very worrisome to me, as I expected an opposite trend. There are either fewer women entering the game industry now than in the past... or there's way fewer women aware of the survey, despite the fact that everyone talks about it every year. I can't see why more men, proportionally, would know about and respond to it than women.
posted by subject_verb_remainder at 5:03 PM on April 4, 2013 [3 favorites]
There are either fewer women entering the game industry now than in the past... or there's way fewer women aware of the survey
Or women are entering the industry and being rapidly chased off.
posted by KathrynT at 5:05 PM on April 4, 2013 [7 favorites]
Or women are entering the industry and being rapidly chased off.
posted by KathrynT at 5:05 PM on April 4, 2013 [7 favorites]
Well, to rephrase, it implies women are being much more rapidly chased off now than in the past.
The implication is the same: there are even fewer women in entry level positions than before, even with more scholarships, diversity summits, outreaches to college students, game development-focused academic programs, heightened awareness and discussion of gender and sexism in games, and more diversity in the types of games the industry makes.
Like I said, it's the opposite of what I would expect.
posted by subject_verb_remainder at 5:20 PM on April 4, 2013
The implication is the same: there are even fewer women in entry level positions than before, even with more scholarships, diversity summits, outreaches to college students, game development-focused academic programs, heightened awareness and discussion of gender and sexism in games, and more diversity in the types of games the industry makes.
Like I said, it's the opposite of what I would expect.
posted by subject_verb_remainder at 5:20 PM on April 4, 2013
The statistics show a massive numerical gender imbalance - if that is correcting itself over time, it may be that the women have significantly less experience overall.
There are either fewer women entering the game industry now than in the past... or there's way fewer women aware of the survey.
A bachelor's degree in computer science is not the only possible credential for entering this industry, but a partial answer to both of these wonderings is that the proportion of people obtaining a bachelor's in computer science in the US and/or Canada who are female has been declining ever so slightly in the past decade or two.
posted by eviemath at 5:40 PM on April 4, 2013
There are either fewer women entering the game industry now than in the past... or there's way fewer women aware of the survey.
A bachelor's degree in computer science is not the only possible credential for entering this industry, but a partial answer to both of these wonderings is that the proportion of people obtaining a bachelor's in computer science in the US and/or Canada who are female has been declining ever so slightly in the past decade or two.
posted by eviemath at 5:40 PM on April 4, 2013
Wow Homeboy Trouble, nice. I could hug you right now.
posted by Scientist at 8:32 PM on April 4, 2013 [3 favorites]
posted by Scientist at 8:32 PM on April 4, 2013 [3 favorites]
Also, note to bloggers: that is what proper statistics looks like. To someone who knows what they're doing, it is not that hard. It's not magic and it doesn't even require that much work -- though Homeboy Trouble definitely went above and beyond the call of duty here. Generating decent statistics is actually fairly accessible a lot of the time to people who have the expertise, so let's not settle for shitty statistics instead.
posted by Scientist at 12:59 PM on April 5, 2013 [10 favorites]
posted by Scientist at 12:59 PM on April 5, 2013 [10 favorites]
"Also, note to bloggers: that is what proper statistics looks like."
Seriously, that is like the proper use of a lamppost - for illumination rather than drunken support. Thanks for the effort Homeboy.
posted by Blasdelb at 4:10 AM on April 6, 2013 [6 favorites]
Seriously, that is like the proper use of a lamppost - for illumination rather than drunken support. Thanks for the effort Homeboy.
posted by Blasdelb at 4:10 AM on April 6, 2013 [6 favorites]
After controlling for age, education, specific occupation and even the labour market, there is still a very strong effect showing that women are paid 80 cents on the dollar.
The control for occupation isn't quite specific enough.
Gender ratios vary a lot between computer science specializations. I'm told by two female friends of mine with significant HCI experience (SIGCHI papers, etc.) that HCI/user experience has far more women than most other CS areas. The job market (and correspondingly, salaries) also varies a lot between specializations and research areas.
I have no idea how much this skews your results (or even in which direction), but I wish the occupation and income data was granular enough to control for CS subspecializations.
posted by ripley_ at 10:31 AM on April 6, 2013
The control for occupation isn't quite specific enough.
Gender ratios vary a lot between computer science specializations. I'm told by two female friends of mine with significant HCI experience (SIGCHI papers, etc.) that HCI/user experience has far more women than most other CS areas. The job market (and correspondingly, salaries) also varies a lot between specializations and research areas.
I have no idea how much this skews your results (or even in which direction), but I wish the occupation and income data was granular enough to control for CS subspecializations.
posted by ripley_ at 10:31 AM on April 6, 2013
Homeboy, I appreciate that you did the labor, but.... Did you really break down by age, rather than years in the industry? It seems like the latter is a lot more likely to be relevant to one's pay grade than the former.
posted by ThatFuzzyBastard at 11:19 PM on April 6, 2013
posted by ThatFuzzyBastard at 11:19 PM on April 6, 2013
Hey, cool, I made Best Of! Thanks for your kind words, folks!
The "less than three years" response for women is very worrisome to me, as I expected an opposite trend.
The way they worded it is confusing, but women are the same proportion of <3 year and 3-6 year respondents, so this doesn't show a trend. I wouldn't be surprised if this was people unfamiliar with the survey in their first years in the industry or something. (There are fewer women in the 6+ years, but given the small sample / self-reporting, I would hesitate to draw any conclusions.)
The job market (and correspondingly, salaries) also varies a lot between specializations and research areas. [...] I wish the occupation and income data was granular enough to control for CS subspecializations.
You and me both. Occupation classifications shift over time as the workforce shifts, and the newer classification standard (the 2010 SOC) has broken computer science into somewhat finer categories (but only, like, from 7 to 11 or so). The dataset I'm using is transitioning into the newer codes, since the American Community Survey is a continuous survey. But right now, newer editions of the dataset have a mix of new and old codes and that's a knot of vipers I wasn't interested in untangling for something I was sort of banging out on my lunch hour. It's tough, because the more specific things get then the data gets spread thinner (and the less anonymous it is, for the public release stuff I work with).
Sadly, I suspect that areas with more women like HCI tend to get paid less, but that's cynicism more than anything else talking.
Homeboy, I appreciate that you did the labor, but.... Did you really break down by age, rather than years in the industry? It seems like the latter is a lot more likely to be relevant to one's pay grade than the former.
If I had years of experience, believe you me, I would have used it. (Years of experience as a field in the dataset, that is; I do have years of experience in statistical-type analysis, I suppose.) Age is mostly a proxy for that, and not a great one. But the PUMS dataset I used (because it was already loaded into a SQLite database on my computer) is a general-purpose survey -- the old long form census in new clothes; I use it for transportation planning purposes primarily. So years of experience isn't in the data set. They've just started asking about field of degree, which will be interesting for labour analysis purposes going forward.
posted by Homeboy Trouble at 12:09 AM on April 7, 2013 [6 favorites]
The "less than three years" response for women is very worrisome to me, as I expected an opposite trend.
The way they worded it is confusing, but women are the same proportion of <3 year and 3-6 year respondents, so this doesn't show a trend. I wouldn't be surprised if this was people unfamiliar with the survey in their first years in the industry or something. (There are fewer women in the 6+ years, but given the small sample / self-reporting, I would hesitate to draw any conclusions.)
The job market (and correspondingly, salaries) also varies a lot between specializations and research areas. [...] I wish the occupation and income data was granular enough to control for CS subspecializations.
You and me both. Occupation classifications shift over time as the workforce shifts, and the newer classification standard (the 2010 SOC) has broken computer science into somewhat finer categories (but only, like, from 7 to 11 or so). The dataset I'm using is transitioning into the newer codes, since the American Community Survey is a continuous survey. But right now, newer editions of the dataset have a mix of new and old codes and that's a knot of vipers I wasn't interested in untangling for something I was sort of banging out on my lunch hour. It's tough, because the more specific things get then the data gets spread thinner (and the less anonymous it is, for the public release stuff I work with).
Sadly, I suspect that areas with more women like HCI tend to get paid less, but that's cynicism more than anything else talking.
Homeboy, I appreciate that you did the labor, but.... Did you really break down by age, rather than years in the industry? It seems like the latter is a lot more likely to be relevant to one's pay grade than the former.
If I had years of experience, believe you me, I would have used it. (Years of experience as a field in the dataset, that is; I do have years of experience in statistical-type analysis, I suppose.) Age is mostly a proxy for that, and not a great one. But the PUMS dataset I used (because it was already loaded into a SQLite database on my computer) is a general-purpose survey -- the old long form census in new clothes; I use it for transportation planning purposes primarily. So years of experience isn't in the data set. They've just started asking about field of degree, which will be interesting for labour analysis purposes going forward.
posted by Homeboy Trouble at 12:09 AM on April 7, 2013 [6 favorites]
I have a probably-very-ignorant question. You wrote:
...and that the entire dataset includes workers 30-49. Presumably, for workers age 45-49 the income effect is positive? What explains the fact that only around the top age quartile do workers reach the average income? Are workers 45-49 really so much better paid even than 40-44-year-olds, or so much more numerous? I'd have expected a more roughly normal distribution with workers 30-40 seeing a negative income effect and workers 40-49 seeing a positive income effect.
posted by enn at 12:22 PM on April 8, 2013
Age 30-32: -29.5% ***
Age 33-35: -22.0% ***
Age 36-39: -10.0% **
Age 40-44: -2.8%...and that the entire dataset includes workers 30-49. Presumably, for workers age 45-49 the income effect is positive? What explains the fact that only around the top age quartile do workers reach the average income? Are workers 45-49 really so much better paid even than 40-44-year-olds, or so much more numerous? I'd have expected a more roughly normal distribution with workers 30-40 seeing a negative income effect and workers 40-49 seeing a positive income effect.
posted by enn at 12:22 PM on April 8, 2013
...and that the entire dataset includes workers 30-49. Presumably, for workers age 45-49 the income effect is positive? What explains the fact that only around the top age quartile do workers reach the average income?
Since I was doing a linear regression with categorical variables (i.e. each variable was a "yes" or a "no"), not every category can be represented. For instance, if both "male" and "female" had parameters, then any solution is possible - Male + 19.5%, Female 0% is mathematically equivalent, or Male +10% or Female -9.5% or Male +100% Female + 80.5% or so on. The model is "overspecified" and a solution can't be found.
In essence, something needs to be chosen as the reference case, with the parameters reflecting variations with respect to that. So in the case of my regression above, the reference case is men, aged 45-49, with no Master's, in medium-to-large cities outside the five I picked out. The elasticities are then relative to this reference case; workers in Raleigh made less money, workers with Master's made more, and so on. So the things that aren't in this regression are effectively zero, and the average salaries are for that reference case.
So my (arbitrary) choice of 45-49 as a reference case meant that younger workers would have the effect of reduced wages, which is strong for the youngest workers, but pretty well negligible (and not statistically significant) by their 40s). I could have done other structures; picking 30-34 as a reference would produce positive effects for older workers. If I'd picked 40 (or, because there's not so many observations, say 38-42) as a middle ground, then there would have been something of the effect you expected; reduced income for younger workers, and higher income for older workers, although the parameters show a much greater effect for younger than for older.
posted by Homeboy Trouble at 1:16 PM on April 8, 2013 [2 favorites]
Since I was doing a linear regression with categorical variables (i.e. each variable was a "yes" or a "no"), not every category can be represented. For instance, if both "male" and "female" had parameters, then any solution is possible - Male + 19.5%, Female 0% is mathematically equivalent, or Male +10% or Female -9.5% or Male +100% Female + 80.5% or so on. The model is "overspecified" and a solution can't be found.
In essence, something needs to be chosen as the reference case, with the parameters reflecting variations with respect to that. So in the case of my regression above, the reference case is men, aged 45-49, with no Master's, in medium-to-large cities outside the five I picked out. The elasticities are then relative to this reference case; workers in Raleigh made less money, workers with Master's made more, and so on. So the things that aren't in this regression are effectively zero, and the average salaries are for that reference case.
So my (arbitrary) choice of 45-49 as a reference case meant that younger workers would have the effect of reduced wages, which is strong for the youngest workers, but pretty well negligible (and not statistically significant) by their 40s). I could have done other structures; picking 30-34 as a reference would produce positive effects for older workers. If I'd picked 40 (or, because there's not so many observations, say 38-42) as a middle ground, then there would have been something of the effect you expected; reduced income for younger workers, and higher income for older workers, although the parameters show a much greater effect for younger than for older.
posted by Homeboy Trouble at 1:16 PM on April 8, 2013 [2 favorites]
It's a bit late to this thread but this heartbreaking Disney rejection letter from 1938 came through my facebook stream.
"Women do not do any creative work in connection with preparing cartoons for the screen, as that work is performed entirely by young men. For this reason girls are not considered for the training school."
posted by bonobothegreat at 7:48 PM on April 17, 2013 [2 favorites]
"Women do not do any creative work in connection with preparing cartoons for the screen, as that work is performed entirely by young men. For this reason girls are not considered for the training school."
posted by bonobothegreat at 7:48 PM on April 17, 2013 [2 favorites]
« Older AppleSoft BASIC in JavaScript | Tolstoy, the Circassians, and Lincoln Newer »
This thread has been archived and is closed to new comments
So there are two questions that come up: Are the men and women who completed the survey truly representative of the industry? Are men and women equally candid about their salaries?
posted by justkevin at 7:51 AM on April 4, 2013 [2 favorites]