Saturday, June 01, 2013

Fun With Economics: The Politics of The Gender Gap in Wages

Some of you may be familiar with my three-part series on the gender gap in wages.  If not, click on the site given at the top of this page and read it.  The study I use in it is a bit old by now but all the theory should be fresh as dew.

Politics doesn't handle the earnings gender gap at all well.  Some lefties rush into regarding the gross (unadjusted) gap as all discrimination, most (in my experience) righties and every single anti-feminist view it as women's private choices.

Neither is correct.  But in many ways the wingnut view is less correct, for one very simple reason:  Studies can sometimes prove that sex discrimination exists in the labor markets, but studies cannot really prove that the absence of concrete evidence means that the differences are just choice.  You pick chocolate ice-cream, because you like it!  I pick a dead-end job because I happen to be regarded as responsible for children!

The same thing, in the wingnut minds.  But more importantly, we cannot conclude free choice as the explanation for wage discrepancies from the sort of studies used that way by the conservatives, and that is the purpose of this post:  To explain why that is the case in some detail.

The most recent study which "proves" that the little ladies choose their lower earnings is this one:

Salary tracking website PayScale released a report Thursday pushing back on the idea of a gender pay gap.
The report found that although women earn an average 81 cents on the dollar to when compared to men, it's because women choose lower paying jobs.
"Unequal pay for equal work? Not really," wrote Katie Bardaro, lead economist at PayScale.
The site found that the salary difference between men and women with the same types of jobs was negligible. The reason for the wage gap is that females tend to gravitate toward jobs that are societally beneficial, where as [sic] men choose more lucrative careers, according to the report.
Salary differences for the same types of jobs were negligible after controlling for occupation, experience, education and so on? 

I am unable to find a writeup of the study at the PayScale site, though I have now asked them for one.  The lack of that makes interpreting their results difficult.  But let's try.

If you go to the site, you can see the raw gender comparisons in a graph and then choose to see the adjusted comparisons.  It's not that there are no differences in the second graph.  There are, but they are much smaller.  I want to see the actual numbers, because essentially all studies of the gender gap find that not all of the gap is unexplained by education, experience and so on, and it is only the unexplained part of the gap which could be discriminatory.

Let's inject a political point here:  To focus so much on a study which isn't a proper academic study (or at least isn't presented that way) is usually driven by politics.

Now onwards and upwards:  I did find a methodology section at the PayScale site.  From that we learn which variables the research took into account when it moved from the gross earnings gap to the adjusted earnings gap:

Using our unique database and compensation algorithm, we estimate the controlled median pay by adjusting for outside compensable factors across genders. These factors include years of experience, education, company size, management responsibilities, skills and more.  In order to provide an apples-to-apples comparison, we determine the characteristics of the typical man within a job and then adjust the characteristics of the typical woman in the same job to match those of the average man. The result is the median pay calculated for the average woman if they had the exact same breakdown of compensable factors as the average man.
The last two sentences sound like they used the Oaxaca decomposition method but only in one direction.  It would be good to see what the average man would have earned if he had had the exact same breakdown of compensable factors as the average woman.  These two figures may not result in the same net residual if the labor market rewards men and women differently for education, experience and so on.

But  that's not important.  What is important has to do with that list of  characteristics I have bolded in the quote.  They are controlled for because they are regarded as factors which naturally explain why someone would earn more, in the absence of any discrimination.  That there is something called "more" is pretty important, because I think the list over-controls and thus may be wrong about the lack of discrimination against women.

Note the term "managerial responsibilities."  That one is controlled for as just one of the innocent outside reasons for higher earnings. 

But you get managerial responsibilities at least in part by being promoted.  If I hate green-eyed people* and want to pay them less at my snake company, the easiest way to do so and not to get caught is not to promote them at all!  Alternatives, such as paying them less for the same job, can get me into trouble if people find out.  But I can probably invent good reasons why the people I promote don't have green eyes.

The lesson:  It's dangerous to control for variables which can hide discrimination in studies like this one.  That the list of controlled variables is not complete is also worrisome.

Alternative explanations for the "managerial responsibilities" variables exist.  Perhaps women don't want them and choose not to apply for jobs which have them.  But you can't assume that this is the case because the data does NOT tell us the relevant reasons.

Then to the wider question about women holding lower-paying jobs than men and the reasons for that:  It is possible that women and men have "freely" chosen the types of jobs they tend to congregate in.  It is possible that women choose lower-paying jobs because they are more concerned about flexibility of a job than its pay, given the societal expectation that they are going to be responsible for child-care.  It is possible that men choose higher-paying jobs and women lower-paying jobs because both expect the man to support a full family one day.

But none of this is proven by finding that men and women tend to be found in different kinds of jobs and that the jobs women are found in pay less, on average.  That's an important point, because the above quote dives straight into assuming that what we have here is free choice.  Have another chocolate Sundae!

To return to my hatred of green eyes (I have those, by the way), suppose that I want to keep green-eyed people earning less (for whatever reasons) and that I do this by approving their applications to jobs which pay less but not approving their applications to jobs which pay more.

If I'm careful I can get away with that and nobody needs to be the wiser.  Or I can rename jobs which are roughly the same by moving some stuff between them and giving one of the jobs a new name.  I can then make sure that the green-eyed person is in the job that is going to pay less.  Or I can have annual increases vary in size between green-eyed people and the rest. 

In short, finding men and women in different job categories does NOT disprove the existence of discrimination and it does NOT prove that the job choice was freely done by the workers.

At the same time, it is clear that there are more young women than young men who choose to study for lower-paying occupations.  This could be because of societal views about the kinds of jobs which are appropriate for women and men or it could be because of the gendered division of labor about child-rearing, assuming that lower-paying jobs are more flexible (not necessarily the case).

Here's the final puzzle for you:  Are women choosing lower-paying jobs or do jobs become lower-paid when many women choose them?  There's some evidence for the latter, too.

I hope I haven't bored you.  The purpose of this post was to introduce some of the complexity that goes into studying discrimination and alternatives as explanations for the gender gap in earnings, and to point out how simplistic the uses of studies are when they are employed as political weapons.

If I had a dollar for every time some MRA* or anti-feminist has told me on the net that "everybody knows the gender gap has been shown not to exist" I'd be a goddess with a yacht.  I get that those people go from their desired conclusions to the search of supporting data, not from a general study of the data to whatever conclusions that leads to.  But at least here is an alternative story.

*This example is not meant to be taken seriously.  Hatred is probably not a common reason for discrimination.  It stands for the real reasons here, such as the belief that women will have children and drop out so they are not worth promoting, or the belief that men are more suited for leadership roles or the belief that women don't care as much about money and promotions etc.