Wednesday, February 05, 2014

Christina Hoff Sommers: Nope, Women Don't Earn Less Than Men. Part 1 Of 3

Meet Christina Hoff Sommers.  It It's Bad For Women, Women Chose It.  If It's Bad For Men, Women Chose It.

Are you familiar with the work of Christina Hoff Sommers?  She is an anti-feminist* American writer, employed by the conservative think tank the American Enterprise Institute (AEI), where her field to plow is anything having to do with feminism.  In 1995 she published a book fetchingly entitled Who Stole Feminism? How Women Have Betrayed Women, but she is most famous for her book about the educational problems of boys and how those problems are the fault of feminists.  That 2001 book is titled The War Against Boys. How Misguided Feminism is Harming Our Young Men.

In other contexts she has argued that masculinity is viewed as politically incorrect in the United States, that "fair" wages are not equal wages because women simply happen to prefer jobs which just happen to  pay less or because women voluntarily choose to drop out of the labor force to have children and those gaps naturally mean that the free-market-gods will have to pay women less.  She has recently stated that so few women are CEOs because women don't want to be CEOs.  The LeanIn movement must fail because women don't want to Lean In; women want to lean against the strong shoulders of their male partners, or to Lean Out because pay isn't that important for those people-centered women interested in caring occupations.

None of that is surprising.  Conservatives usually assume that any differences between men and women are a) biology-based and unchanging or  b)  based on utterly free and unconstrained choices by both men and women but especially women.

These rules are sometimes applied in the weirdest of manners.  For example, you might think that reproductive differences, such as who it is who gives birth, would be viewed as biology-based.  But the general conservative interpretation of what might be going on when women take time off from work to have children is that this is just a free choice by the ladiez!  It's  not something that the labor markets need to take into account, not at all the same as, say, the human need for sleep or rest. 

But if reproductive gender differences were truly seen as innate and not just a choice by women, we'd have to address the kinds of questions we do about the need for sleep:  Firms are not allowed to keep workers working 24 hours a day without rest.  Human reproduction requires that those giving birth get to have some time off, but conservatives have fought tooth-and-nail against any kind of maternity leaves for those workers.  What conservatives regard as innate is women's lesser skill or interest in STEM-careers, not the way humans are reproduced.  The latter is one of those choice items.

There is one exception to these rules:  When it comes to boys' problems at school or the lower male participation rate in college, conservatives do not look for a choice-based or innate explanation. Rather, Hoff Sommers' book blames feminism for the "war against boys" in the United States.  That the same patterns are observed in essentially all countries which don't outright exclude women from higher education, including such countries as Saudi Arabia, cannot be explained by the blame-feminism theory.  Indeed, in other contexts conservatives would probably argue that something so widespread as the greater percentage of women among college students must be based on innate differences, not on environmental or economic or cultural incentives.

I'm not buying that, of course.  But it's important to note that the conservatives (and quite a few liberals) believe that  "voluntary choice" because of [sic] those "biological differences" is the usual explanation for gender differences when they appear not to favor women.

Hoff Sommers:  No, Women Don't Make Less Money Than Men

So why am I going on about this particular anti-feminist?  Because she has written a commentary on president Obama's State of the Union speech with that inane headline (No, Women Don't Make Less Money Than Men) and because it's worth taking out that flea comb and combing through her arguments, partly to see How It Is Done and partly to see if there's anything in her opinions with which I can agree.

Here's Hoff Sommers:

It’s the bogus statistic that won’t die—and president deployed it during the State of the Union—but women do not make 77 cents to every dollar a man earns.
President Obama repeated the spurious gender wage gap statistic in his State of the Union address. “Today,” he said, “women make up about half our workforce. But they still make 77 cents for every dollar a man earns. That is wrong, and in 2014, it’s an embarrassment.”
OK.  Actually, women do make 77 cents to every dollar a man earns, on average.  Despite the "bogus statistic" reference and the headline-in-search-for-clicks Hoff Sommers doesn't disagree with any of that.  What she disagrees with is what she thinks the president thinks by that statement!  Or

What is wrong and embarrassing is the President of the United States reciting a massively discredited factoid. The 23-cent gender pay gap is simply the difference between the average earnings of all men and women working full-time. It does not account for differences in occupations, positions, education, job tenure, or hours worked per week. When all these relevant factors are taken into consideration, the wage gap narrows to about five cents. And no one knows if the five cents is a result of discrimination or some other subtle, hard-to-measure difference between male and female workers. In its fact-checking column on the State of the Union, the Washington Post included the president’s mention of the wage gap in its list of dubious claims. “There is clearly a wage gap, but differences in the life choices of men and women… make it difficult to make simple comparisons.”

Emphasis is mine.  More about that bolded assertion below and later in this post. For the time being, let's just assume that what Hoff Sommers means is this:  The gross gender gap, measured as the 23 cent difference above, does not take into account such average differences between men and women which are not necessarily evidence for gender discrimination in the labor markets.  If that's what she means here (and I'm being kind), she has a valid point.  The rest of her piece presents a second partially valid point and then adds a lot of weird stuff and interpretation. 

Let's define that 23 cent gap in earnings as the gross gender gap.  It is arrived at by comparing average male and female earnings over some time period.  Those earnings depend on both how much a particular person works during that time period and how much she or he is paid for that work per some time unit.  Because men, on average, work more hours than women in the labor market (though not necessarily more overall when household work is also included), some part of the gross gender gap reflects simply that fact of working longer hours.

It is also possible that the average male worker differs from the average female worker in other ways which might not be a consequence of gender discrimination.  For instance, in the past the average male worker had more education and work experience than the average female worker, and such differences should be taken into account if we are looking for the impact of discrimination in the labor market.  The education and experience differences are no longer as important because women's educational achievements have risen over time and because women are now less likely to drop out of labor market for long periods of time, which reduces the experience gap.  But such differences should still be taken into account in the analyses of the gender gap in wages.

What Hoff Sommers would like us to use in such arguments is the net gender gap, not the gross gender gap.  Ideally, the net gender gap would measure the average pay differences between full-time workers who are identical in everything except their gender.   It's like identical twins, except that one is male and one is female!  Real studies cannot quite create such an ideal measure, but they can get pretty close by taking into account all the possible non-discriminatory reasons which cause differences in the average earnings of individuals.  Once all those are taken into account, what still remains could be due to labor market discrimination about gender.

To see how the net gender gap in earnings might be calculated, have a look at my gender gap series, available at  Although the study I analyze in that series is now outdated when it comes to the data, the methods the series discuss are still valid.  Thus, what we really wish to compare are the earnings of two full-time workers, equal in all relevant characteristics but sex.  The relevant characteristics include years of education, years of experience, occupation, location, age, marital status, number of minor children, ethnicity and race.**

Now, Hoff Sommers is correct that the gross gender gap cannot be interpreted as a direct measure of nothing but labor market discrimination.  On the other hand, what she fails to point out is this:  Not all the relevant characteristics which she wants to see controlled for ("differences in occupations, positions, education, job tenure") can be determined to always be non-discriminatory in nature:

As an example, consider that "positions" variable she wants to see held constant in the evaluations. If positions here refer to organizational positions and if workers are promoted into them, the following scenario is possible:  Joe and Jane have the same experience, education and job skills.  They begin in the same job and earn the same amount.  When a promotion slot opens up on a higher level, both Joe and Jane apply.  Suppose now, that the people picking whom to promote believe that men should be the leaders in the society.  Even if Joe and Jane are equally competent and hard-working, even if Jane works harder and is more competent, Joe might be promoted because of his gender.

IF this happens, what is the outcome of controlling for positions in studies of labor market discrimination?  It is to downplay actual discrimination, because once Joe has been promoted, his higher earnings will be attributed to the "non-discriminatory" reason that he is in a different position than Jane.

That example should be viewed as a pedagogical one.  Many of the variables economists regard as relevant and non-discriminatory determinants of earnings differences can also hide discriminatory outcomes (promotion chances, probabilities of being fired, amount of on-the-job training one receives).

Hoff Sommers points out the possible problem of omitted variables:  That we don't really know what the net gender gap in earnings measures if we don't have data on all important variables, such as the detailed skill differences between individuals.  But what I describe here is an equally important problem:  The danger of over-controlling for all sorts of variables, in an attempt to nail down pure discrimination, when those variables themselves can have discriminatory components.

What more can we say about that last Hoff Sommers quote?  She glides over all the studies which actually have found labor market discrimination of various kinds.  I also cover some examples in my gender gap series. 

And while I agree with Hoff Sommers that the gross gender gap in wages is not, in itself, a measure of labor market discrimination, I'm considerably less impressed by her firm statement that the final "net" gender gap in earnings shrinks from 23 cents to five cents when all the other characteristics are taken into account.  She doesn't give a link to that argument, but my guess is that it comes from this 2009 study (pdf), because the creator of the mega-analysis in that study makes similar arguments.  I wrote about that study earlier:

Surprisingly, the conclusion that "the aggregate wage gap "may be almost entirely the result of the individual choices being made by both male and female workers" doesn't come from the surveyed studies but from the survey-maker's own study.

This study begins with a classification of variables which might explain differences in men's and women's earnings levels. These are
-work experience
-career interruptions

Note that the author of the survey assumes that all these are based on workers' choices alone. That rules out any alternative explanations even before we are beginning to look at the data. Yet promotions, for example, can be used to discriminate between the genders. The promoted person ends up in a different occupational category (say, management) from the non-promoted one (say, sales personnel). Likewise, it's possible that women don't work in certain industries because of workplace harassment which creates a barrier to entry, and it's also possible that employers discriminate against women who have children by giving them less lucrative tasks or fewer promotions and so on. Work experience itself could be based on not only the workers' own choices if employers discriminate in firing.

My intention is not to argue for those alternative explanations, but to point out that we cannot really regard all those factors as unambiguous outcomes of workers' choices in the sense the survey treats them.

The actual analyses in the study are of two types: The first one is a conventional cross-sectional analysis of the gender gap in wages. It uses variables commonly included in such studies (though without work experience measures) and its conclusions are also fairly standard: The conventional analysis accounts for between 28.8% and 44.9% of the raw gender gap in wages, leaving the rest unexplained.

The alternative analysis is one I find very troubling. It attempts to capture something that a cross-sectional study cannot do very well: patterns of labor market attachment over time. The survey-maker does this by adding market data: The percentages of workers with the same gender, age, and number of children who either are not in the labor force for reasons other than retirement or disability or who are working part-time.

These are supposed to be proxies for the worker's own potential career interruptions. Perhaps Hoff Sommer's argument that women are more likely to leave work to care for children or aged parents?

But note that such variables would also capture any statistical discrimination employers might practice against women of child-bearing age, where statistical discrimination is defined as treating all women of given characteristics as if they were equally likely to leave their job or start working part-time. IF statistical discrimination is practiced, women in the relevant age and family composition classes might not be given on-the-job training or promotions and might earn less for those reasons.

The alternative analysis 'explains' between 74.4% and 130% of the raw gender gap in wages, depending on the specification. Oops!

The survey-maker hastens to point out that the 130% figure is an anomaly caused by a high percentage of young men being out of the labor force. Other specifications for the alternative analysis 'explain' between 65.1% and 76.4% of the raw gender gap in earnings. But remember that including variables about the general behavior of the cohort a worker belongs to would also capture any statistical discrimination. Or to put it more bluntly: They also measure being a woman in a fertile age category, because taking time off and working part-time are correlated with being female. If labor markets discriminate on the basis of sex this variable would pick up some of that effect.

Note that the net gender gap in earnings from one particular study with certain problems shouldn't be assumed to equal the net gender gap other studies might find (I usually avoid stating a particular number for that reason), and note, also,  that even the study Hoff Sommers favors couldn't explain away 22% of the gross gap, despite essentially throwing into the analyses every single thing they could think of, including the kitchen sink.

Where both this 2009 study and Hoff Sommers go wrong, in my opinion,  is the ready willingness to accept "choice" as the reason for women's lower earnings, without a much clearer definition of what "choice" consists of.  In reality, we make choices under constraints, and those constraints matter.

Among the most common constraints are financial ones, but societal constraints also operate.  For instance, if childcare is expensive to obtain and if the parent responsible for hands-on child-rearing is assumed to be the mother, the average woman's labor market "choices" will be made under different constraints than those of the average man.  In the traditional setup he is expected to focus on finding an occupation which pays enough to cover most of the costs of a future family.  The average woman, on the other hand, may also need to consider how flexible that occupation will be for her gendered chores.  If more flexible jobs pay less, the "choices" of men and women under these constraints will result on different earnings patterns, and not necessarily because women (or men) like it that way.***

Such choices are not the same in meaning as the choices between chocolate or vanilla ice-cream, and they certainly depend on the constraints which are linked to gender, whether biologically or through social expectations and sanctions.  Finally, the labor markets themselves in the United States are still largely designed to regard the worker as someone who has full backup at home, who never needs to take time off to care for children or a sick relative and who is expected to focus on nothing but the job.  These assumptions are particularly harsh on single parents, but they also affect all labor market participants who don't fit the traditional model.

And, as I noted at the beginning of this post, the job of human reproduction requires giving birth.  The conservative view (eminently present in the comments to the Hoff Sommers piece) seems to be that all this is not that different from picking chocolate ice-cream instead of vanilla, that it's something women "voluntarily" do, so why compensate them for it?  On the other hand, we are also told (even by the same commenter) that  markets are never wrong, that markets are always free and that markets will never pay women the same as men because women take time off for childbirth and after they return they should get paid less as they now have less job experience.  Anything else is discriminatory.  It's like a merry-go-around.

To conclude this first post out of the two on Hoff Sommers' arguments, I should make it very clear here that the various measures of gender gaps in earnings are getting smaller in the US.  Part of the reason for that is the higher education levels of women now when compared to those in the past, and, the fact that women have somewhat shifted their occupational "choices" (either actual choices or as a consequence of social norms changing) towards better-paying fields.  But much of the reduction, I'd guess,  is due to something sadder:  The decline of well-paying blue-collar jobs in traditionally male industries.  That, my friends, is due to outsourcing and globalization.

The second post addresses occupational segregation as the cause of some part of the gross gender gap in earnings, while the third post asks whether it is true that the gendered choice over college majors really can account for much of the gross gender gap in earnings.  That's what Hoff Sommers argues in her opinion piece.
*The term "feminism" in this post refers to a focus on providing equal opportunity to all individuals, irrespective of their gender. That seems to be the definition of feminism most conservative anti-feminists use and the one they attack.

**Ethnicity and race are included in that list not because they are  OK determinants of earnings differences but because of the attempt to isolate gender discrimination as the cause of earnings differences while holding all other causes for differences constant, including any ethnicity or race based discrimination or differences in nondiscriminatory variables which correlate with ethnicity or race.

The gross gender gaps in earnings can  differ between different racial or ethnic groups, and there are also sizable gross race/ethnic gaps in earnings.  I have not been able to find a good study which analyzes the combined effect of race and gender but I am still looking because such a study is important.  In the meantime, have a look at one of my older posts where I look at the gross gender gaps in earnings by race and ethnicity and also the gross race/ethnicity gap in earnings by gender.

Keep in mind that those figures are not the net gaps but the gross gaps, and that they apply to 2009.

***They can also affect wider societal patterns and even fertility rates if women are allowed to keep control of their own fertility.  This is because in the conservative worldview women are supposed to bear the brunt of the financial losses caused by childbearing, and one possible response to that is to have fewer children.  Thinking about how all this links to divorce, who gets custody of the children and recent proposals to get rid of alimony altogether is also worthwhile.