Friday, August 26, 2011

A Re-Post: There Is No Gender Gap in Wages, Silly Women. Part II

Originally from here. See yesterday's post for the first part.

Let's continue with Carrie Lukas' argument that there is no gender gap in wages. After chatting about unemployment and such she states:
Feminist hand-wringing about the wage gap relies on the assumption that the differences in average earnings stem from discrimination. Thus the mantra that women make only 77% of what men earn for equal work. But even a cursory review of the data proves this assumption false.
I am going to agree with Lukas on one thing: Far too often people (not just feminist people) do talk about the gross gap as if it was completely based on direct labor discrimination, such as paying women less for the same work. The reality is more complicated, though there are enough cases to suggest that women sometimes are paid less for the same job or even for superior job.

But Lukas is indeed correct in arguing that just finding a difference in the average earnings of men and women is not the same as finding that whole difference due to discrimination. Earnings between individuals vary for many reasons (some potentially non-discriminatory ones are education, experience, local labor market conditions and occupation*) and only some of those can be interpreted as caused by direct sex discrimination.

Economists do know this, however, and take it into account when studying gender differences in earnings. Where Lukas goes wrong is in steering to the other extreme: If direct discrimination does not account for the whole gender gap in earnings then it doesn't account for any of it! That is what she argues.

And the next paragraph in Lukas' article is off the topic:
The Department of Labor's Time Use survey shows that full-time working women spend an average of 8.01 hours per day on the job, compared to 8.75 hours for full-time working men. One would expect that someone who works 9% more would also earn more. This one fact alone accounts for more than a third of the wage gap.
Why is it off the topic? Because actual studies of earnings difference between men and women take the differences in working hours into account (see my series on the gender gap in wages)! What is studied is not the gross earnings differences but the differences in pay per time unit, such as an hour, a week or a month of work. The research tries to control for obvious reasons why earnings differ and looks at the differences that remain, and hours spent working per week are one of those controlled variables.

Put another way, Lukas is wrong, because she is talking about the total pay packet, not about the earnings per hour, say. It is the latter which economist study when they analyze the gender wage gap. ( For some statistics on the differences between men and women in hourly wages, check Table 9 in these statistics.)

She is also wrong in a somewhat subtler way: She begins at this point to list various non-discriminatory factors which just might (might!) affect the gender differences in average earnings! She then IMPLIES that they do. But the former does not equal the latter. We need to actually study the factors in those lists to test whether they can account all of the gender gap in wages. Just insisting that they do is meaningless.

One example of Lukas' list is this:
Choice of occupation also plays an important role in earnings. While feminists suggest that women are coerced into lower-paying job sectors, most women know that something else is often at work. Women gravitate toward jobs with fewer risks, more comfortable conditions, regular hours, more personal fulfillment and greater flexibility. Simply put, many women—not all, but enough to have a big impact on the statistics—are willing to trade higher pay for other desirable job characteristics.
Men, by contrast, often take on jobs that involve physical labor, outdoor work, overnight shifts and dangerous conditions (which is also why men suffer the overwhelming majority of injuries and deaths at the workplace). They put up with these unpleasant factors so that they can earn more.
She argues that women freely and gladly "choose" certain occupations because they are comfy, safe and interesting, even though those jobs pay less. Men, on the other hand, daringly choose to work in dangerous jobs which pay better.

But note that even if all this were true Lukas has NOT shown that these differences explain the gender gap in earnings. For one thing, the number of men working in truly dangerous jobs is quite small, too small to account for why women earn on average only 80% of what men earn**. For another thing, men who have "chosen" the same industries which are popular with women still earn more on average, at least per week:
Of the 45 million women who worked full time in wage and salary jobs, 17 million were employed in education and health services, and 5 million were employed in wholesale and retail trade. Financial activities and professional and business services each employed about 4 million women.

Median weekly earnings of women employed in education and health services were $717, which was 77 percent of men's median weekly earnings in that industry. In wholesale and retail trade, women's median weekly earnings were $523 (76 percent of men's earnings)
Thus, picking "jobs with fewer risks, more comfortable conditions, regular hours, more personal fulfillment and greater flexibility" doesn't seem to explain the earnings difference within those industries. If you find the industry classification too wide, similar data is available for individual occupations.

Finally, here is where Lukas makes a serious interpretative mistake:
Recent studies have shown that the wage gap shrinks—or even reverses—when relevant factors are taken into account and comparisons are made between men and women in similar circumstances. In a 2010 study of single, childless urban workers between the ages of 22 and 30, the research firm Reach Advisors found that women earned an average of 8% more than their male counterparts. Given that women are outpacing men in educational attainment, and that our economy is increasingly geared toward knowledge-based jobs, it makes sense that women's earnings are going up compared to men's.
The mistake is this: When economists analyze the gender gap in wages they ideally wish to know the answer to this question: If we could clone a person with only one exception: the person's gender, would the two clones earn the same amount? This is the goal, to compare a randomly selected male worker to a randomly selected female worker in such a way that we are standardizing for all other variables which affect earnings except for gender.

The Reach Advisors study Lukas refers to does NOT standardize for education levels. It's as if the female clone in my example was allowed to have a college degree while the male clone was not. Comparing the earnings of the two do not tell us anything about the pure effect of gender because we forgot to control for education. Indeed, it is even possible that women still earn less than men in those metropolitan areas, if the comparison is done properly, i.e., by comparing women and men with the same education levels. The Reach Advisors study does not tell us anything about the gender gap when it is properly defined, because it still compares male apples to female oranges.

The studies about young workers have an additional problem: Most discriminatory effects will not be seen early in a person's career, because discrimination takes time to operate***. So do some non-discriminatory reasons for earnings differences, such as labor market interruptions by women for childbearing purposes. But the point is that we cannot use data from only the young to make conclusions about all working men and women.

Lukas ends her piece with a nasty statement:
Should we celebrate the closing of the wage gap? Certainly it's good news that women are increasingly productive workers, but women whose husbands and sons are out of work or under-employed are likely to have a different perspective. After all, many American women wish they could work less, and that they weren't the primary earners for their families.
What on earth does the closing of the wage gap have to do with the problems of husbands and sons? Can we swap a greater gender gap in wages for more jobs for men?

And what on earth does the presumed desire of American women to work less have to do with the closing of the wage gap? If anything, higher wages would enable someone to work less for the same total income.

I don't get this paragraph in Lukas' piece at all. It's not logically linked to anything else she makes stories about, and I don't see what its function is supposed to be, unless she argues that we shouldn't care about women's earnings at all because the earnings of men are more important and all women would prefer to stay at home. But if that's her thesis, why didn't she write about it?
*I write "potentially non-discriminatory" because some of these can be affected by past discrimination. For instance, if a woman gets laid off first because she is a woman, her work experience might be reduced over time. Likewise, "occupation" is a tricky concept to the extent that two essentially identical jobs might be given different job titles with different pay, to hide discrimination. Also, occupational steering may affect which occupations are open to women or men.
**As I have written earlier, prostitution may well be the occupation with the highest risk of early death, but it is not included in the statistics because it is an illegal one.
***The Equal Pay Act of 1963 makes it hard to pay women and men in the same job different wages (though it happens), and the Title VII of the Civil Rights Act makes discrimination in hiring illegal (though it, too, happens). If an employer wishes to discriminate against female workers, he or she will have to wait to do it through differential raises, promotions and firings. This is the main reason why data on workers just beginning their working lives cannot prove the absence of labor market discrimination in general.