A Simple, Legal Way to Help Stop Employment Discrimination

Originally posted at TheAtlantic.com.

Women and racial minorities are no longer making progress toward equal representation in the workplace. Here’s a way to maybe fix that.

Jacquelyn Martin

Progress toward gender and racial equality in the workplace has basically stalled. One reason for that is the government’s lack of antidiscrimination enforcement. As Donald Tomaskovic-Devey and Kevin Stainback show in their book Documenting Desegregation, ever since the reign of Clarence Thomas as head of the EEOC in the 1980s, the Equal Employment Opportunity Commission (EEOC) has been underfunded, understaffed, and largely ineffective at doing its job. To help get things moving again, under the existing law (more or less), we could use the power of social media and the principle of government transparency to allow workers and consumers themselves to apply pressure on discriminating employers. Would it work? It couldn’t hurt. First a little background.

Anti-discrimination today
Here is the occupational segregation trend from 1966 to 2005, fromDocumenting Desegregation, just comparing white and black men and women. The index of dissimilarity shows what percentage of a group would have to change jobs to have the same representation as white men.


The figure shows white women made a lot of progress in the 1970s and 1980s, but less since. Black women have a similar pattern but much slower progress. And black men haven’t budged since 1980. The same pattern holds for representation in managerial jobs.

The burden to fight discrimination today is mostly on workers who have been discriminated against to first discover this fact and second file a complaint and/or lawsuit themselves. The courts have tightened their definition of discrimination to include only deliberate acts proven to have been motivated by discriminatory intent – a very steep burden. And they have reduced workers’ capacity to bring class actions, most notably in the Wal-Mart decision, which makes it hard to get good legal teams. As a result, few cases make it to court, and virtually no one wins. A study of 1,672 employment discrimination cases from 1988 to 2003 found that about half resulted in settlements (with a median value of $30,000), 6 percent went to trial, and one-third of those were victorious (with a median award of $110,000). Although more than 100,000 people file discrimination complaints with the EEOC, most workers lack basic information not only about the law and their options, but about their own employers’ practices (as was painfully revealed when Lilly Ledbetter discovered she had been discriminated against by Goodyear for many years). And people who aren’t hired in the first place have an even smaller chance with the law.

In the 1964, Congress passed the Civil Rights Act, which included in Title VII a mandate to collect information about employment in the private sector. Since 1966, all large employers are required to submit a simple accounting: the number of workers, by race and sex, in each of nine occupational categories. This has produced a treasure-trove of data, which Tomaskovic-Devey and Stainback used to document the trends. But this information could be used more proactively by the government itself, if stopping discrimination were a higher priority.

Anti-discrimination tomorrow
Defining and proving discrimination is difficult. Many employers have no outward motivation to discriminate—they just don’t do enough to stop discrimination by individual supervisors, recruiting practices that produce narrow applicant pools, and malicious co-workers. So not every workplace with an underrepresentation of women or minorities is a case of willful discrimination. But when a workplace has significant underrepresentation in either its management or its overall employee pool, it’s at least worth taking a look to see what’s going on.

Here’s my suggestion, inspired to by Documenting Desegregation. Underrepresentation is very widespread, and easy to detect. Why not label it?

Using the same EEOC data, my colleague Matt Huffman and I identified workplaces in which there were fewer African-American managers than would be expected by chance, using a test common in employment litigation. With a wide statistical margin—95 percent confidence—we found, for example, that 7 percent of black private-sector workers in the D.C. metropolitan area worked for employers with easily identified underrepresentation of black managers. That is, they had fewer black managers, compared with other firms in their same industry in their same town, than would have occurred by chance. Maybe they aren’t discriminating on purpose, but they’re probably doing something wrong. As a customer, client, business partner or job applicant at that firm, wouldn’t you like to know that? (Of course, as researchers we are prohibited from revealing information about individual employers.)

So why doesn’t the EEOC generate a simple certificate, like the one I have mocked up here, to notify the employer, the employees, and the public, about such cases? (This would only apply to those with 50 workers or more.)

This hypothetical firm has an overrepresentation of white men in management compared with the local industry (for example, a department store with 60 percent white male managers when the local industry average is 30 percent). They have underrepresentation of black women across the board, and Latina women compared with the rest of the industry locally. Representation of the other groups isn’t outside the range of the 95 percent test, or there aren’t enough cases to judge. The test accounts for sample size—if you only have two managers at your business, and one is a white man, you’re not going to fail.


It could be like the health department certificate posted on a restaurant wall (and online). Then, maybe someone who worked there would get up the courage to file a complaint. Maybe customers wouldn’t shop there. Maybe politicians running for office would promise to improve the local statistics. Maybe concerned managers would honestly consider their hiring practices to look for ways to do better.

This doesn’t reveal any trade secrets. It doesn’t increase the reporting burden on employers, since they’re already required to submit the forms. It wouldn’t cost much. But it gives the public a little more leverage and increases the accountability for employers. It wouldn’t solve everything either. But if equal opportunity employment were a major priority, a small step like this would seem pretty reasonable.

17 thoughts on “A Simple, Legal Way to Help Stop Employment Discrimination

  1. In Norway, perhaps the most egalitarian country in the world, 90 percent of engineers are men and 90 percent of nurses are women. Different groups may simply have different interests and abilities.


  2. Staffan, exactly. Given the IQ distribution (much higher variability of IQ in males, resulting in much higher proportion of idiots and geniuses) it’s inevitable that at most employers in the lowest-ability demanding jobs (janitors etc) and highest-ability demanding jobs (higher management) must be males. If there is 50% females and males in high management it means that males are discriminated against.
    This whole post assumes that ability is distributed equally amongst different groups, which is completely unfounded assumption.


    1. szopeno, do you really think that higher variability in IQ is great enough to explain why 90% of engineers are men? Engineering at the bachelor’s degree level is not so mentally challenging that only the top 0.5% of the human population can handle it. Higher variability in IQ might result in a totally equal society awarding more men with Nobel prizes in science than women, but it doesn’t make sense as an explanation for a heavily skewed gender ratio in reasonably attainable science careers.


  3. velella sure not. I saw the claim that at 140 IQ level only 60% are males. However in addition to general IQ, you have for example spatial skills which may result in males (on average) being more interested in subjects where they can use their widely reported higher ability.

    This is a question of abilities AND interests. Just to illustrate what I mean (ant not to prove anything, this is just an illustrative example):
    My daughter is highly intelligent and has no problem in math. I also urge her to learn math as it is crucial in getting good job (almost ANY job), at least in Poland. Nevertheless, she says math is boring. OTOH my son is fascinated with numbers, though he is 6. He was constantly coming and boring us with questions about what is after a million? What is the results of such and such operation? Even if he has lower math ability than my daughter, I find it more likely that HE will be engineer, and SHE will not.

    And IF (but this is a reasonable if) males are more interested in being engineers than females, THEN obviously there cannot be equal number of female and male engineers. Moreover, demanding equal number of males and females will discriminate males.


  4. “Sex differences in the right tail of cognitive abilities: A 30 year examination”, Wai, Cacchio, Putallaz, Makel, 2010.
    In short, in the last 20 years the gap has stopped to narrow. In the top 6% of SAT-Math scores (>700 points) there are at least 1.75 males for each female, and in top 3% there are 2 males for each female. It seems unlikely, as vellela rightly remarked, that this difference alone explains the fact that 90% of engineers are males: According to http://collegeapps.about.com/od/sat/a/undergraduate-engineering-sat-scores.htm, 25% of students enrolling to engineering courses has at least 630, but in majority of it seems it’s more than 700 (sometimes much more). Only 25% has score lower than 500, but in some colleges only 25% has score lower than 700.

    So, while it is reasonable to expect more than 50% of male engineers on ability alone, vellela is right that this is not enough. The rest may be explained simply by the different interests. Now, where is the data of declared interests of girl and boys in engineering related topics?


      1. Thanks for the link. It’s surprising that the ration of females in computer science is so high; it’s only something like 1/3 lower than what I would expect based on upper bound guesstimated on SAT-Math scores.

        And thanks for the patience, too; especially since it is undeserved, since indeed I have missed the fact that the companies were compared to other companies. I must say I am puzzled by your answer, because, if I understand you correctly a sentence “a firm consisting of all genius engineering managers in a town [compared?] with other similar firms, would have the same distribution of men in top positions” is not necessarily true and it would be true only if it assumed equal distribution of “genius” managering skills amongst all analysed groups, which is just an assumption, isn’t it?

        And I stand by opinion that since a) all studies I know found higher variance in many abilities amongst males than amongst females, then b) in professions in which this kind of ability is important, in both lowest- and highest- ability-demanding, there must be more males than females, and the difference must be the higher, the more (or the less) demanding is the job.


  5. BTW, something bothers me about SAT scores and percentiles published by Cacchi. I don’t know US system, but it seems to me from the quick google that 700 is not top 6%, as Cacchi claims?


  6. Szopeno: You missed one of the main points: each firm is compared to other firms in their industry in their town. Although I have little patience for your argument, it nevertheless is true that under this scheme a firm consisting of all genius engineering managers in a town with other similar firms, would have the same distribution of men in top positions. If they differed from the rest of the local labor market, this would be an easy thing to explain given the general level of sexism in the population.


  7. Sweden’s high levels of segregation are pretty easy to explain without resorting to gender essentialist / IQ differences mumbo jumbo: Sweden’s workforce and childcare policies encourage women to remain in the labor market (and are “egalitarian” in this sense), but funnel them into pink collar occupations. Read pretty much any of Maria Charles’ pieces on comparative sex segregation.


  8. I am constantly impressed by the work you do here. I often send my SOC students to your blog. I am teaching SOC of Gender this term and discussing Gender and Work this section so this entry will be very helpful!


  9. You know what, I’ve thought again about mr. Cohen idea and thought what would happen, if more categories would be added. For example “Catholics”, “orthodox christian”, “Jewish”. In my opinion, anyone proposing such form would be suspected (at least, I would suspect someone like this) of hidden agenda and maybe even anti-semitism.

    But this is just a sidenote. This form (showing whether some company have less/more than average of some group than the average in their area) would be bad. Hiring is not a random process and a company with different pattern of hiring does not necesarilly discriminates. Hiring involves checking you social networks often, to find first whether your friends can recommend someone trusted. People tend to form their social networks with similar people; Poles usually know other Poles, so they would have a lot more of Polish managers and workers than others, Hiring may involve a lot of otherd decisions; for example, a law company serving mainly Polish catholic community could hire exclusively Polish catholics. Or imagine than you have some construction company found by Poles, whichi would hire people they can trust and with them can they share language – this firm would be almost exclusively Polish, despite serving diverse customers. Why the Polish owners of the company should be punished by hiring their Polish friends, members of their Polish families etc?

    In addition, IF different classes of people (e.g Polish catholics vs German protestants) have different ability (e.g. Polish catholics have lesser rate of good managers), THEN it is easy to construct an example in which companies following perfect meritocratic policies will have radically different rates of Polish catholics; e.g. one could hire only the best of the best, and it would have 9 Germans and 1 Pole; other could not afford to get the best, and would have to get the best of the others left for the hiring, with the result being that some companies would have radically higher ratio of German managers, not because they are doing something bad, but on the contrary, because they are doing something good. The form indicating that some of those companies use discriminating practices would be highly damaging to their business and would be completely unfair. A state which would, therefore, would force such policy, would be immoral and it would be moral to oppose it.

    (I am Polish, and I have used Polish vs German on purpose, so noone in his sane mind could accuse me of some kind of hidden racist agenda).


  10. Just ran across the blog, and enjoy it. I like numbers and arguments. Thanks for writing. All comments intended in good faith.

    “Underrepresentation is very widespread, and easy to detect. Why not label it?”

    Interesting idea. But why, again, is statistically proportionate representation the appropriate yardstick to use for measuring “equality”? And why is antidiscrimination enforcement” the answer?

    It seems as though you have very little faith in either market forces to correct such “inequalities” (e.g. the argument that if–for example–women are cheaper to hire and just as able, we’ll see firms that hire them have a competitive cost advantage; so, forcing companies to hire more women through “antidiscrimination enforcement” is raising the market value of women through imposing penalties, not creating value, which is a variant of the broken-window fallacy) or the rights of firms to do as they please. If a firm wants to discriminate, why not let it and let it also bear the market consequences, since it’s best able to judge its situation?

    Interference by 3rd parties usually occurs without full knowledge of the impacts on the firm and usually has bad and unintended consequences for many parties. In your example, you’re trying to lever companies into “employment equality.”

    Let’s assume there’s no discrimination, and a company’s a statistical hiring outlier, which tautologically you’re bound to have in any statistical sample (for instance, the 7% of employees you cited with the 95% confidence interval.). Your solution unfairly publicly penalizes them for who they hired. Why do you want to do that? Let’s now assume discrimination. Why would a firm discriminate? Maybe there is an operational or economic reason behind it, in which case you’re now forcing suboptimal outcomes on the firm to achieve your numerical parity goal. Maybe it’s purely irrational discrimination, but how would you tell the difference? And how would you know you’re right? And what gives the right to anyone to open these companies up to certain legal action based on the supposition?

    For background, see Thomas Sowell on the nature of mundane knowledge in his book Intellectuals and Society.


Comments welcome (may be moderated)

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s