Thursday, April 04, 2019

The Equal Pay Day 2019. Where Echidne Dons Her Economist's Hat And Fixes Mistakes in Beliefs

 The Equal Pay Day 2019 was last Tuesday, the 2nd of April.  It's a day created to remind us how much longer women, on average, must work to earn the same amount men, on average, earned during the previous year.  So last Tuesday was roughly the day when women, on average,  would have caught up to men, on average, in total pay for 2018.

The coverage of the day in social media reveals a great lack of information about the economics of the gender gap in wages.  I saw many false interpretations of what it might mean, and I didn't really see economists chiming in to correct false interpretations.  So I'm going to pick two examples, one from roughly each side of the political aisle, to highlight a few of the central problems.

First example: Everything is Discrimination

This example is picked from what I'd call the left side of the political aisle.  Let's begin by looking at one tweet, picked not for being especially off (many were much more so), but because it highlights a couple of central misunderstandings:

The picture in that tweet is fine.  It says nothing that isn't true, on average, though keep in mind that it compares the earnings of various groups of women to the earnings of white men, not to the earnings of, say,  Asian-American men or African-American men.  Had we used some other comparison group, the percentage deficits in the earnings would have changed (1).

But the written information in that tweet has two errors (or one error and one unrelated comment, depending on one's interpretation).

First, April 2 is NOT the day when white, non-Hispanic women had earned what white, non-Hispanic men earned during the year 2018.  The selection of that date is based on calculations which take into account the earnings of all women.  White, non-Hispanic women's earnings do have a greater influence on where that date is located, simply because white, non-Hispanic women are still the majority among women in the US.

The National Committee on Pay Equity provides the correct information on the date when white, non-Hispanic women will reach, on average, the earnings of white, non-Hispanic men earned in 2018.  That date will be April 19, 2019.

  Here are the dates by which various groups of women will have earned what white men, on average, earned in 2018:

Second,  the end of that tweet, stating that "all women deserve equal pay for equal work," is of course correct in itself and something I fervently support.  But if it's assumed to be linked to the rest of the tweet, then it is more problematic.

This is because the earnings calculations that are used to arrive at the date of the Equal Pay Day do not hold the occupational distributions of the different groups constant, and neither do they control for average group differences in education, experience, geographic location and so on (2).  In other words, it's not the case that we can use the data in the tweet to simply assume that various groups of women are not paid equally for equal work with white men.  This is partly because women and men, on average, tend to work in different occupations, but also for other reasons.

Put differently, I do know, from other sources, that labor market discrimination, in the form of sexism, can affect the earnings of all women, and that labor market discrimination, in the form of racism, can both affect the earnings of women of color directly and also interact with the sexism they face to create more complicated effects on them.  I also know, from other sources, of several cases where women, indeed, have been paid less than men for equal work.

But I also know that direct labor market discrimination in, say, earnings, hiring, firing and promotions is not the only cause of earnings differences.  For us to be able to judge the above figures from that angle we should first standardize for all the other reasons (3) that people earn different amounts.

I am going to use education as one example (4) in this post, because the educational attainment levels are not equal in all the groups the above picture compares, and education affects earnings.

To get an idea of the relevant group differences, let's take as one imprecise measure the percentages of individuals over twenty-five who had a bachelor's degree in 2017:

Asian-American women:  53.8%
White women: 38.3%
African-American women:  25.7%
Native American women:  22.9%
Hispanic Women: 18.6%

The same percentages for men are also informative:

Asian-American men:  57.2%
White men:  37.8%
African-American men:  22.6%
Native American men: 17.7%
Hispanic men: 15.8%

Note how the average educational attainments of the various groups of women correlate with the average earnings ranking in the tweeted picture?

Thinking about that correlation suggests, for example, that part of the reason why Asian-American women earn more than women in any other racial or ethnic group is that more than half in that group hold college degrees. Likewise, that Hispanic women earn the least of all the groups in the tweeted picture is partly due to the fact that only 18.6% of that group hold a college degree.  Jobs which require a college degree tend to pay better than jobs which do not require one, especially for women.

But the education stories and earnings stories are not always conveniently correlated.  Though Asian-American men are more likely to hold a college degree than Asian-American women (and also earn more), the same does not apply to the other racial and ethnic categories which are included in both the tweeted picture and the above lists.  Rather, in all those other categories more women than men hold a college degree, yet earn less than the men in their ethnic or racial group.  Also, note that Asian-American women still earn less than white men, despite having a clear average educational advantage. 

The education example is not intended to imply that direct discrimination based on sex and race wouldn't matter in creating the final earnings outcomes in the tweeted picture.

It does matter, and so does the traditional sexual division of labor inside families which results in women taking on the larger burden in childcare.  As the above table from the National Committee on Pay Equity reminds us, the traditional sexual division of labor inside homes lowers the earnings of women who have children.

My point is, rather, that we need to study discrimination within the overall economic theories about how earnings levels, in general, are determined, and we need to isolate the impact of labor market discrimination from all other causes of earnings differences.

If we fail to do that, we are playing straight into the hands of those who believe that the gender gap in earnings doesn't exist and that talking about the gap is just pure politics.  More about that below.

Second Example:  Nothing Is Discrimination

The Time magazine, in honor of the Equal Pay Day, I guess, posted an article discussing the findings of a new online survey which asked respondents if they believed in the gender gap in wages and so on:

And a new poll from SurveyMonkey, revealed first by TIME, shows that a significant belief gap exists when it comes to the issue of equal pay.
According to an online poll of 8,566 American adults conducted in March, nearly half of men (46%) believe that the pay gap “is made up to serve a political purpose,” rather than being a “legitimate issue.” And about a quarter of men ages 18 to 34 (24%) say that media reports of men and women being paid unequally are “fake news,” one option provided in the poll. Overall, 62% of Americans believe that men make more money than women for similar types of work, with men and younger Americans most likely to incorrectly say that there is no gap in pay.

Bolds are mine.

Before we dive into those survey results, it's pretty important to stress that SurveyMonkey uses a sampling frame which consists of two million people who are  willing to take online surveys.

This is not a good way to create a sample which would be representative of the opinions of all American adults on the question of equal pay for men and women, because the frame excludes all people who are not online and because it is quite likely to consist of greater than average numbers of people who just like taking surveys.

Thus, the sample in this survey is not based on probability sampling, as SurveyMonkey itself acknowledges:

Surveys that use probability-based designs can calculate and report a margin of error estimate for each statistic they produce. You’ll often see language such as “this poll has a margin of error of +/-3.5 percentage points,” which means that if the difference between two estimates is within the margin of error, we can’t tell with confidence which one is greater.
SurveyMonkey research surveys do not have a probability-based design, because there is no well-defined sampling frame of respondents to SurveyMonkey surveys. Therefore, to avoid confusion, we do not report a margin of error term.

This is something to keep in mind when we look at the results.  Strictly speaking, they apply to the two million individuals in SurveyMonkey's sampling frame.  Whether they apply to all adult Americans' opinions is unclear.

Now that I have that off my chest, let's dive into the most interesting part of those results (6):

That men are less likely to believe in the existence of a gender gap in wages, and that it looks like young people are also less likely to believe in such a gap.

What's really great about the survey results is that we can put those two things together in just one table!  Here is the table (click on it to see it bigger):

The first row inside the table shows us  what percentage of the respondents in various sex-age categories agreed with the statement "The pay gap is made up to serve a political purpose."  The second row of the table tells us what percentage of the respondents in various sex-age categories agreed with the alternative statement "The pay gap is a legitimate issue."

A quick glance at the table tells us two things:  First, men are more likely to think that the pay gap is a made up political issue than women, in all age groups.

Second, younger respondents, both men and women, are more likely to agree with that statement than older respondents.

Thus, it won't come as a surprise that the age-sex group which expresses the greatest disbelief in the legitimacy of the gender pay gap consists of young men between the ages 18 and 34.  Almost half of them think that the pay gap is not a legitimate issue.

So what's going on here?  How  would we account for these differing findings?

Here are my speculations about why men, and especially young men are the most likely demographic group to think that the gender gap in pay is fake news:

1.  Men have always been less likely to "get" the impact of, say, sex-based discrimination in the labor force, because they are less likely to witness it and, obviously, less likely to directly experience it.  This, and the desire to defend one's own position in the gender hierarchy, is probably the reason why the female and male answers to this question have historically differed, on average.

2.  Young people will not see as much sex-based pay discrimination among their peers, because starting salaries cannot differ very much between men and women, due to federal laws such as the Equal Pay Act of 1963 which made paying different salaries for the same job illegal, and because of Title VII of the Civil Rights Act which made sex discrimination in hiring illegal.   If those laws are obeyed, men and women starting in the same job category should be paid the same.

It is only over time that any effects of direct labor market discrimination against women would appear, through discriminatory differences in pay raises, promotions and firings  (7).  One needs to grow older to truly see and experience these effects.

3.  Because the average education attainment of young women is now higher than that of young men, the higher earnings of certain groups of young women (single, childless, well-educated women in major urban centers) might be seen as proof that the gender gap in pay no longer exists, or even as proof that a reverse gender gap in wages now exists.

This interpretation is wrong.  It does not compare like with like.  The proper question, in this context,  is not if women with, say,  MBAs earn more than men with, say, high school diplomas, but if women and men with MBAs earn the same and if women and men with high school diplomas earn the same.

Still, as I have written before, the anti-feminist right uses the above data to argue that there no longer exists a meaningful gender gap in pay, and they have been very successful in seeding those ideas all over the social media.

4.  The general form the anti-feminist right-wing arguments takes is not to negate the gross earnings gap between men and women, the one described in the first picture in this post.  Rather, it is to argue that pretty much all of it can be explained away by women's voluntary choices.

Thus, women earn less because they are assumed to choose occupations which pay less (avoiding, say, STEM careers while plunging for feminist art therapy) and because they are assumed to choose to work fewer hours in the labor force and more hours at home (8).

These types of arguments are common on the more conservative sites where they are given without any rebuttals. Because the numerical majority of men in the US, white men, are more likely to be Republicans, it's possible that men have been more exposed to the choice-based arguments as an explanation for the gender pay gap than women (9).

In Summary

If you waded through this enormously long post, congratulations!  But if you did not, the short form of my argument is this:

It's wrong to interpret the gross gender pay gap as fully caused by labor market discrimination.  It's even worse, however,  to assume that there is no real discriminatory gender gap in pay by simply stipulating that the reason why women earn less is based on women's "voluntary choices."

The actual story is more complicated, of course.  But to appreciate that you need to read this long post.  Sorry.

(1) Because Asian-American men earn, on average, more than white men while African-American men earn, on average, less than white men.

For an earlier take by me about these kinds of gross pay gap comparisons and what's wrong with them, see this post.

(2) In fact, the National Committee on Pay Equity states, explicitly, that the gender wage gap they calculate an aggregate. It does not show men and women doing the same work or in the same jobs. But it does show changes over time, with progress in narrowing the gap in the 1990s and little change in this century.

(3) If you are interested in how that might be done, in general,  read my Gender Gap Series at the website listed at the top of this page. 

(4) It's crucial to remember that a) education is not by far the only variable which matters in determining someone's earnings, b) the educational attainment variable can itself be affected by discrimination if schools, for example, treat most students of color more harshly than white students and/or have lower expectations for them or c) apply particular sexist expectations to female students (possibly leading them to careers which pay less but are deemed more suitable for women) and, perhaps, apply particular sexist-cum-racist expectations to African-American girls.

It's also crucial to remember that educational access depends partly on the wealth of the student's parents.  Because the racial distribution of wealth in the United States is extremely unequal, some earlier racist acts and practices which caused that racial wealth inequality (as well as institutional racism) can still have an impact on the ability of African-American parents to pay for their children's higher education.  When this ability is weakened, fewer African-Americans will have college degrees for purely financial reasons.

Finally, recent immigration status can affect educational attainment.  Hispanics, for example, are a group with many more recent immigrants than the other groups except for Asian-Americans.

Recent immigrants, in general, may have fewer resources to educate their children and may also be less likely to already hold college degrees from their source country.

That this is not the case for Asian-Americans is because the immigration authorities appear to have favored applicants from Asian countries who possess high educational qualifications.

(5)  See my Statistics series, available at the website posted on the top of this page.  It's a very simple introduction to, among things, the concept of sampling for surveys.

(6)  You can see the answers to the survey questions broken down by the respondents' gender here.  You can also choose, on that same page, to see the results broken down in various alternative ways, including by age, by race and ethnicity, or by both gender and race/ethnicity, or by both age and gender.

(7)  It's true that the gross gender gap in wages also increases over time because motherhood, and the traditional gendered division of labor often associated with it,  increases with time.  But it's possible to analyze the various causes for earnings divergence separately if the studies are done well and have access to good data.
(8)  I have written so much on all this in the past that my fingers should have worn to nubs by now.  Here are a few earlier posts which cover the field: 

This one talks about the right-wing use of "choice" in general, this one talks about why female-dominated occupations pay less, and this series discusses extensively the way occupational sex segregation can be or cannot be explained by women's "voluntary choices."

That series is my response to Christina Hoff Sommers' arguments that there is no actual gender gap between men and women, once we allow for women's "choices." I still love the third post in that series a lot.

(9)  Young men may also be the most likely demographic group to visit the more extreme MRA sites which argue that men earn more than women because men deserve to earn more.  Men work longer hours per week and men are more likely to be employed in the most dangerous occupations.  Besides, men killed the mammoths to protect the women who were reclining in the home caves eating bonbons.

Sorry.  I got a little carried away there.  The more relevant response is that a) the gender gap in pay still exists once hours worked per week is controlled for and b) physically dangerous occupations have too few workers to account for the existence of the overall gender gap.  Besides, the higher male earnings are not caused by the large earnings of fishermen or fire fighters, but by the large earnings of stockbrokers and so on.