Thursday, January 19, 2006

The Gender Gap. Part I: Theories



This is the beginning of my series on the gender gap in earnings. It turns out that I can't write about this topic except in the academic style. My apologies for that.

The first post in this series gives you a concentrated summary of the economic explanations for why men earn more than women, on average, all over the world. The next post will look at actual evidence about the gender gap and how that evidence fits with the theories. The last part of the series explains why the right-wing's explanation for the gender gap in earnings is misleading and mostly incorrect.

The theories I describe here are just that: theories, and that they are discussed does not mean that they have proven to be true. The question of their value in explaining real world gender differences in earnings will be thoroughly investigated in the next part of the series. It's unfortunate that the theories need to be talked about before the fun stuff, but that's life for you.

A good place to start is to define the concept of an earnings gender gap. In most statistical sources this takes the average earnings of women who work full-time and compares it to the average earnings of men who work full-time. The gender gap is the ratio of women's average earnings to men's earnings, both defined for full-time workers. If men and women on average had the same earnings this ratio would equal one. The lower women's earnings are as a proportion of men's earnings, the lower the figure we arrive at. In the United States the recent gender gap, defined for full-time workers only, has hovered around 0.75 or 75% in percentage terms. (The term "gender gap" is a clumsy name for this ratio. It would have been better to find the actual gap (say, 25% in the case of the U.S.) and to talk about that directly. But what is done is done and I will apply the common clumsy concept in this series.)

Most studies don't include part-time workers' earnings in the calculation of the earnings gender gap. When these are included the gender gap increases dramatically, because most part-time workers are women and because part-time work tends to pay less well than the otherwise equivalent but full-time positions. The existing numbers on the gender gap underestimate the actual earnings differences between the sexes.

Fringe benefits, things like retirement benefits and health insurance, are usually also omitted from the gender gap calculations. These benefits tend to go up with earnings, though, so it's fairly safe to argue that if they affect the size of the gender gap it will be in the direction of making the gap bigger.

The gender gap is found all over the world but its size varies greatly from the smallest gap of 0.9 (meaning that women earn 90% of what men earn, on average) in countries such as Sweden and Australia to as large a gap as 0.4 in Russia. That the gender gap is not the same size in all countries means that it is at least partly affected by the laws and labor market customs of the countries. It is not some sort of a Biblical constant that cannot be altered, in other words. This is also reflected in the fact that the gender gap changes over time. In the United States, for example, the gap narrowed from the 1960s to the 1980s and then slightly widened during the 1990s. The gap can narrow not only when women's earnings rise but also when men's earnings fall, and at least some of the narrowing in the U.S. has been due to the worsening of men's earnings in blue-collar occupations which have been especially hit by newer production methods using less workers and the outsourcing of the labor the developing countries.

How does this read so far? Academic and technical and not fun, but I can't really write it much better. I know too much, sadly.

Before I tackle the theories themselves, a few more words on the definition of the gender gap: Note that the earnings concept that is used is the gross pay of workers. This means that it depends not only on how much the worker is paid per hour or per week or per month but also on how many hours, weeks or months the worker is toiling at the job. The reason why the gender gap is usually measured only for full-time workers is that doing this controls for differences in how much people work as a reason for income differences. The control is not complete, as some full-time workers work more hours than others. But the studies I discuss in the next part take this into account by adjusting the gender gap measure to reflect actual reported working hours. The rest of this first part assumes that we have taken into account the variation in hours worked so that the gender gap can be viewed purely as a measure of higher wages or salaries for men than for women.

Note also that the gender gap is calculated across all jobs. Men and women often work very different types of jobs. This is called labor market segregation, and it may be voluntary (chosen by the workers) or it may be involuntary (chosen by the firms through steering workers to certain jobs). What this segregation means is that we may be comparing the earnings of men and women in quite different occupations. Luckily, it turns out that the empirical methods we use can control for at least some of these differences.

Why would men on average receive higher wages (or salaries, but I will use wages for simplicity from now on)? At the risk of tremendous oversimplification, I offer three basic explanations:

1. Men are better workers than women, on average.

2. Women want to work in jobs which pay less.

3. Women are treated unfairly at work.

The first of these reflects the simple idea that more productive workers will be paid more. Are men more productive workers? How would you try to find out about this for, say, systems analysts or physicians? Productivity is really hard to measure except in the simplest of cases, especially if we want something very objective and easily countable as the measure.

Because of this lack of good productivity measures, economists usually replace them with what might make someone a better worker: education and work experience. The guess is that less educated and less experienced workers would earn less. If there was a big difference in the average education and experience levels between the sexes then we might have explained the wage gap. Too bad that this isn't what works in explaining the current gender gap in the United States, because if it did I could stop writing this post now. But women have at least as much education as men do, these days, and although the average work experience of men is a little higher than that of women the difference is too small to account for much of the gender gap in earnings.

The second theory is the absolute most favorite of the wingnuts. For them the idea that women "choose" jobs which pay less would be great, because it would mean that we need to do nothing about the gender gap. Everybody is happy! Everybody is doing their own thing!

Why would women want to have work which pays less? The answer lies in the idea that jobs offer not only income but a whole bundle of job characteristics, such as safety/riskiness, dirtiness/cleanliness, flexibility/rigidity and so on. If women, on average, value things like safe and clean working conditions and lots of flexibility they might be willing to accept a lower wage rate in exchange for all these goodies. The same would be naturally be true for men, too, which means that this theory only explains the gender gap if we assume that the sexes are, on average, different in what they want from a job.

This theory is mostly used to argue that women "choose" jobs which allow them to care for their children and to do household chores more efficiently. The idea is that it is the women who are responsible for childcare and household management, and that these extra tasks take away from the energy that women can give to their jobs. Thus, women, on average, might choose jobs which offer less stress and thinking and more flexibility in terms of telecommuting or flexible working hours. Because of all these desirable goodies the jobs are acceptable to women even when they don't pay very much. But they are not acceptable for men, on average, as men are expected to provide for the family.

To apply this theory, one doesn't have to decide whether women "choose" lower earnings because of biological differences between the sexes or because of social indoctrination that makes women feel responsible for all the nonmarket work. The conservatives tend to jump to the biological argument right away. Because they'd like that one, I guess. But in practise quite a lot of social indoctrination goes on and at least some of it is likely to have an impact over and above any biological differences.

I have noticed this explanation misused a lot recently, by combining what I have written here with the idea of women who quit working as an explanation of the gender gap. But that is nonsensical. Women who are no longer in the labor force are not affecting the gender gap as they have no earnings. Neither are the women who work part-time, because the gender gap is measured for full-time workers. Indeed, as women who feel the burden of family responsibilities strongly are quite likely not to work full-time the power of this general explanation is considerably weakened.

The way to test this theory is by finding out if the jobs that are predominantly female indeed have flexibility, and all the other things that the theory assumes that women find desirable but that men don't. Then one would also somehow test that women actually "choose" these jobs rather than being steered into them by career counselors, parents, teachers and the clergy. That is pretty hard to do.

This is a good place to say a few words about "choice". The wingnuts use "choice" as a codeword for things which require no meddling by the government. If something is "chosen" and turns out to be a mess, the chooser should suffer the consequences unaided. This trivializes choice. It makes the choices a mother makes no different from the choices I make when I decide which chocolate bonbon I will eat next. Yet the choices women make about working or not have repercussions elsewhere in the society and on themselves. A woman who takes a job that pays less in order to care for her children or her partner will also end up with less retirement benefits when she is old, and this is because of the way we have decided to determine retirement benefits.

More generally, when we make choices we make them under constraints. If I am sentenced to death and offered the choice between hanging and being shot you could argue that I'm choosing to die in a certain manner, but I'm not choosing the death part. Likewise, the choices women make about their working circumstances are not totally free choices; they are made under constraints: a husband who earns more and will not adjust his hours for this reason, in-laws that scold you for being a bad mother if you take that job with traveling and so on.

Note also that a woman might "choose" a job that pays less because at an earlier time in her life she was excluded from getting the right kind of education for better-paying jobs by her family or by the school system. This may not be common today in the United States but continues to be a severe problem in many other parts of the world.

The third and final theory has to do with the idea that women are treated unfairly by the labor market or by the wider societal institutions, and that the combined effect of these unfair treatments is to make women earn less. I'm talking discrimination here, but in a slightly different sense than the word is commonly employed, and I need to define the term in a little more detail. I'm going to start with direct wage discrimination.

Direct wage discrimination is what we usually think of as sex discrimination in employment: a man and a woman do the same job and the woman gets paid less for it. Doing this is illegal since the 1963 Equal Pay Act (which Alito might destroy), but in reality it is very difficult to find out if somebody else is getting paid more for the same job. This is because earnings are often kept secret. It is also possible to adjust one of the jobs ever so slightly and then call it a different job, though the Equal Pay Act is supposed ot ignore such fine-tuned differences.

But direct wage discrimination is not the only way in which women, on average, might end up being paid less than men. Consider hiring and promotions. If women are not hired or promoted into the best-paying occupations they cannot earn the high wages, even without any direct wage discrimination. Title VII of the Civil Rights Act therefore bans sex discrimination in hiring and promotions (and Alito might get rid of that one, too). A worker needs to at least suspect that she has been the victim of discrimination to use this Act, though, so the number of actual cases filed does not measure the whole extent of possible hiring and promotion discrimination. I talk about some alternative ways of measuring this in Part Two.

Labor market segregation, the pattern of finding certain jobs mainly male and other jobs mainly female, may also be caused by discrimination, not just by free "choices" by the workers based on what they want from a job. A famous discrimination case against the AT&T company in the 1970s showed that the company used different hiring literature for men and women, and different jobs were described in the leaflets, and once it used to be common to advertize a job for either men or women. If such steering happens, women may be steered into the female jobs in such large numbers that the wage rate in these jobs drops. Usually a dropping wage rate would make job-seekers turn to other occupations. But discrimination against women might make this unfeasible. There is lots of anecdotal evidence that men in some blue-collar occupations use sexual harassment as a way to keep female competition for the same jobs down.

Why would sex discrimination exist in the first place? Economists have three major groups of explanation for its existence, and all of these would also apply to, say, race discrimination. These groups are

1. Bigotry or hatred as held by either employers, coworkers or consumers in a society

2. The fact that discriminating can actually be income-increasing to the discriminator

and

3. Explanations based on lack of information or biased information, such as prejudice.



The oldest of these explanations is the first one, the idea that some people just are sexist. Remember the idea that workers might pick jobs not just on the basis of the wage rate but also on the basis of other job characteristics, such as risk or dirtiness? Well, one job characteristic that might matter to a bigot would be whether the other workers are women or men. A misogynist would refuse to work with (or under) a woman, unless he or she was paid a lot more than in alternative all-male jobs. This will not happen in a well-functioning market place, but if there were enough misogynists (so that the firms needed to hire them to get enough workers) then firms would start segregating women from men at the workplace, to keep things peaceful. This is one example of the way that sex-segregation might not be by female choice.

If employers dislike working near women they would treat female job applicants in a similar manner to the previous explanation: The woman would have to be either a lot better to be paid the same wage rate as a comparable man or she would have to agree to work for less. This so that the mental suffering being near women would cause the owner or manager gets compensated.

Consumers can also discriminate. Think of the patrons of a restaurant who refuse to be waited on by a woman, say. If such patrons are many, women might find it hard to get jobs at this restaurant, unless they are willing to work for less than otherwise comparable men. Because hiring them means losing some business. This example also shows how hard productivity might be to measure if consumers are bigots themselves.

All these types of bigotry could exist at the same time, which means that the overall effect would be to lower the wages women can get and to enforce sex-segregation at work. But this theory is almost always applied in the form where the only bigots are the owners or managers of the firm. This particular type of discrimination, it is said in Wingnuttia, can't survive in the long-run (meaning a time period long enough for new firms to enter the market and for existing firms to leave) if competition works, and therefore discrimination doesn't exist. The argument goes like this: Suppose neutral and nice people start new firms in the market. They immediately spot that the bigots are paying women a lot less, and these women are every bit as productive workers as the more expensive men. So these new firms go out and hire all women. Their total wage bill is less than that of the bigot firms and their profits higher. Over time the bigot firms will be outcompeted and must leave the industry, and both men and women end up earning the same wages. End of story.

Well, they are wrong, even within the narrow submodel they have chosen to apply. The reason is that the eradication of discrimination in the long-run would only take place in very competititive markets which are quite rare in the real world, and only if large existing firms had no specific advantages from being large (such advantages could keep the good guys entering with small new firms from ever getting established). More generally, they are wrong because both coworkers and consumers can discriminate against women, and because the models we have looked at so far ignore other reasons why discrimination might exist. Most importantly, they assume that information on workers' productivity is perfectly known to everybody.

The second group of explanations argues that discrimination may be the right thing for firms to do if they want to make as much profits as possible and if they have at least some market power in hiring. This means that there are few alternative places where the workers could find jobs. Mining towns are sometimes used as an extreme example of this situation: if you don't work in the mine you don't work, but university towns might also offer a good example of something similar.

Men and women might have different labor supply elasticities. What this rigmarole means is that women might be willing to work for less money than men for reasons that have to do with family arrangements. Think about dual career academic couples. If such a couple decides to accept a job offer for one of them in a small college town the other one is likely to be very restricted in the kinds of jobs he or she (but usually she) can find. The college knows this and may be able to get the following spouse for very little money.

This model is an example where the firm makes more money by discriminating than by not discriminating. Its application may not be very wide as the college town type examples are not too common, but it's a good one to present, just to remind the wingnuts that markets don't necessarily self-correct away from discrimination. There are other models in this group which concentrate on not how income can be increased by discriminating but on how punishments can reduce the income of a non-discriminating firm. To get a flavor of these, think about what might have happened to a firm which employed women in the Taliban-era Afghanistan.

The third group of explanations lumps together various theories which apply the idea that information often has large holes in it and that we may think we have truthful information about something when we do not. This esoteric stuff turns out to explain labor market discrimination, too.

One example of it is plain old prejudice. Prejudice here would mean holding incorrect (perhaps outdated) or no information about the true skills of some group of people, say, women. An employer with prejudice would not want to hire women, because he or she would fear that they turn out to be bad workers. It's irrelevant that this might be untrue; as long as the employer dislikes taking risks the women will not be hired, unless something forces the employer to do so. One such force is wartime. Women broke into many occupations during the World Wars because the firms had not alternative but to hire them. In the case of pure prejudice just seeing that the women could do the jobs would change the employer's attitudes and stop the discrimination against women.

Another example of the information theories is what is called statistical discrimination. This consists of using the average group characteristic as a proxy for any one individual in that group. Thus, all young men might be assumed to drive with the same care as the average young man, for insurance purposes. Because on average young men have a poor accident record the premia that all must pay are high. This discriminates against the careful drivers in the group of young men and favors the most reckless ones.

The insurers do this because they don't know how to get good information on any one individual's driving habits, but it's discrimination nevertheless.

The same principle may apply to women in the labor market. People who are hiring or promoting workers in a large firm may treat all women as if they are going to be just like the average woman. Add to this something like the belief that women are more likely to quit a job than men (which is not necessarily true, by the way), and you can see why firms would choose to train and promote men rather than women. But they would be discriminating.

This is not a thorough overview of the relevant economic theories, just a quick walk-through. The amount of words I give per theory is not an indication of the importance of that theory but dictated by how much background I believe my readers have. I stress the discrimination theories not because I would find them to be of overwhelming importance but because the recent onslaught of conservative arguments that there is no discrimination, that there can be no discrimination.

The next post will look at one study in greater detail, to see how the various theories perform and what proportion each explanation can account for out of the total gender gap in earnings in the United States.