Focusing on workforce diversity ratios at companies is a terrible idea. Better to focus on process.

TL;DR – Many advocates of ‘more diversity’ in the workplace and in high-paying jobs like technology, are vague, misleading, and misguided in their calls to focus on the demographic ‘representation’ of a given workforce as evidence of bias. They ignore statistical ratios of qualified workers (and underestimate how tricky and complex it would be to estimate what the ‘bias-free’ workforce makeup for a given company “should be”, let alone the random factors that would cause actual ratios to differ.) Those who are concerned with potential bias in personnel decisions should instead focus on the processes of hiring/salary/promotion, and if they believe they see bias, they should investigate more thoroughly than simply arguing that “percentage female” or “percentage minority” are useful metrics to indicate bias.

What’s wrong with diversity ratios?

A number of companies have embraced the nebulous idea of ‘Diversity’, publishing statistics on how ‘diverse’ their workforce is, and publicizing their efforts to increase that diversity. It’s often unclear whether this is done out of genuine belief that having a more diverse workforce will be good for the company, or whether it’s a cynical attempt to foster positive public opinion for its products and employment opportunities, but I suspect it’s a combination of both.

What is also unclear, or in my opinion is often intentionally obscured, is what is meant by diversity, and why ratios of characteristics used to define diversity should be focused on (generally gender and race/ethnicity, but often sexual orientation or family income are mentioned as well.) Focusing on altering the racial, sexual, or [insert demographic variable here] makeup of a workforce is incredibly misguided, and ultimately harmful if actually pursued as a goal in itself.

Instead, companies and activists should focus on making sure that the process of finding, hiring, and promoting qualified candidates is free of bias. Ratios can be a good way to check whether processes are working well (unbiased) or not, but they should not be the goals themselves.

Defining diversity

First, we must determine what is meant by diversity. I would assume that the type of diversity beneficial to an organization would be diversity of opinion, experience, and thought process, but that’s not what most diversity conversations are about (and besides, it would be difficult to measure these things in a meaningful way, since they largely exist inside the heads of workers rather than in their skin pigment, parent’s W-2 forms, and genitals…) Diversity advocates usually look at gender & race (which I’ll combine here with ethnicity), and state that their goal is to ‘increase representation among underrepresented groups’ of people at a given company or industry. ‘Representation’ presumably means ‘% of the workforce’ and ‘underrepresented’ means ‘those who make up a minority as a % of the workforce’. In the US, this seems to mean anyone who is not a white male, or perhaps a straight white male. Outside of tech, white might be sufficient to categorize those who are ‘not diverse’, but since men are overrepresented in tech, it must also mean white men. (One sign that ‘diversity’ has simply become a euphemism for not being a white male is that candidates in the singular are now described as ‘diverse’, which makes no linguistic sense.) Strangely, Asians (which includes those from India for our purposes), are generally assumed to be increase ‘diversity’, despite the fact that they are hugely overrepresented in tech (and other high-paying jobs) in the US.

Computing diversity

One complication here is that under- and overrepresentation is only meaningful if we know what the expected or ‘unbiased’ level of representative should be. If I tell you that the workforce for a tech company in San Francisco is 5% black, does that mean that blacks are under- or overrepresented relative to other groups? Well, if San Francisco’s (SF) population is 10% black, you might say ‘yes, blacks are underrepresented compared to the general population of SF’, but if SF is 2.5% black, then 5% makes blacks way overrepresented relative to the general population.

However, why should the racial makeup of SF be the ‘correct’ level of representation? Tech companies hire from many parts of the US as well as abroad, and their workforces often don’t reflect the racial makeup of the cities they inhabit (sometimes to the annoyance of city officials and their residents who would like to see the locals hired in preference to others, or to the anti-immigrant crowd that believe people randomly born into the US should have more of a right to feed their families than people randomly born elsewhere.)

Furthermore, tech companies likely hire disproportionately from immigrants of certain countries (China & India, say) who move to SF for tech jobs, but less so for other jobs. So maybe we need to look at the hiring pool of qualified candidates both in the US and outside, and then compute the racial makeup of those candidates as a whole so that we know just how many of each group we ‘should’ have in our workforce. (Gender being 50-50 in general makes this much easier for the male-female divide.)

This already seems like a difficult enough task, but a further complication is that certain racial groups, genders, and countries may not be equally represented in having the skills required for a given job. This is easiest to see in a field where explicit credentials are required, such as nursing. Nursing in the US is very female-dominated, with only about 10% of Registered Nurses (RNs) being men. If we assume that 10% figure holds both for number of male nurses employed and number of male nurses with a RN certification who are looking for work, then we should not be surprised if a given hospital employs 90% women and 10% men as RNs. If one was to look at the 90-10 split in isolation, you can imagine unsophisticated diversity pundits exclaiming over the “discrimination” responsible for only 10% of RN jobs filled by men: “Male nurses are just as competent & well-educated as women, and yet for every male nurse, there are NINE female ones despite men making up 50% of the population. Hospitals are surely discriminating against men because it’s assumed men can’t be as caring or as conscientious as women in these jobs!” etc, etc.

More thoughtful people might start asking why men are so underrepresented as RNs. They might start professional societies for male nurses, non-profits to encourage men to acquire the education and interest in becoming a RN, or groups supporting men in the nursing profession at their hospital. (“ManlyNursing, a group for male nurses and those who support them!”) Others might misguidedly start calling for the creation of hospital ‘diversity boards’ to ‘encourage’ (read: browbeat) hospital staff to hire more men until the ratio of male nurses increases relative to female, or suggest a #deletehospitalXYZ boycott social media campaign.

I have no problem with, and in fact generally encourage, the efforts to make certain sectors of society more aware of and qualified for opportunities to improve their career position. Looking at employment sectors where certain groups are underrepresented might be a great way to decide where to focus these efforts. However, the fact remains that there’s nothing apparently discriminatory about a hospital employing 90-10 women to men if that ratio represents the talent pool available to that hospital.

The same logic goes for race, etc. Because of the fact that Hispanics and Blacks generally grow up poorer & less educated than Whites and Asians in America, and that family income & education are highly correlated with individual income and employment, we shouldn’t be surprised when we see Blacks & Hispanics underemployed and underrepresented in high-paying fields with respect to whites & Asians. This doesn’t mean that companies are discriminating (or that they’re not), and further analysis is needed to try to determine that*. None of this should dissuade us as a society from improving the lot of marginalized groups by addressing the root causes of poverty or lack of education, it should simply dissuade us from assuming there’s an ‘ideal’ ratio, or even an ‘improved’ ratio of gender/race mix to shoot for, or that imbalances in the workforce are equivalent with discrimination.

To come back to tech, 79% of Computer Science (CS) bachelor’s degrees awarded in 2016 at the top 5 institutions giving those degrees out were to men. Presumably that number was even higher in the past, so the workforce population with CS degrees is likely > 80% men to < 20% women. If CS degrees are a good proxy for “qualified tech workers”, then we’d expect to see about 80-20 men to women in tech, which is in fact what we often see.

(This split is similar for engineering as a whole: in 2015, 18.5% of engineering bachelor’s degrees went to women. In contrast, women obtained over 75% of of psychology bachelors, and graduate with every kind of degree, from associate’s to doctoral degrees, at a higher rate than men overall, and have since 2009.)

So, the real question becomes “why are women [or minorities] underrepresented in acquiring certain technical skills vs men [or whites/asians]?” Not “why aren’t tech companies hiring more women/minorities (hint: bias)?” This former question is a more complex one, and out of scope for this article, but it is the one that I believe companies and diversity advocates should get together on, rather than trying to shame companies in to hiring more of a certain group, which is a zero sum game at best.

Enforced ratio are clearly discriminatory

Another reason to oppose the simplistic idea of trying to enforce an ideal or better diversity ratio is that this practice in itself is discriminatory because companies would necessarily have to take someone’s race or gender into account when making hiring decisions, which is illegal and precisely the practice that diversity advocates claim they are trying to eliminate. Cynically, one could argue that many diversity advocates are not actually anti-bias in hiring processes, but rather anti-white male, and though they seem unaware of this, anti-Asian, and pro-black/Hispanic/female. Arguing that companies should hire a candidate from an ‘underrepresented’ group that they’ve deemed less qualified than a candidate from an ‘overrepresented’ group is anti-meritocratic, discriminatory, and generally outlawed by Federal law.

Diversity advocates might counter by stating that they’re merely ‘correcting’ for some underlying bias that has actually suppressed more qualified minority/female candidates from getting certain jobs. This gets into ‘hard to prove’ territory that needs much more careful analysis than is usually given. I think these advocates are at least overestimating the amount of the underrepresentation that is the result of biased hiring/promotion in favor of white (or Asian) men, and are overlooking the root causes like educational attainment, incarceration rates, and the factors that lead to these proximate causes (poverty, single parenthood, etc.)

One round-about benefit of pressuring companies to hire less-qualified candidates might be that companies would respond by trying to improve the talent level of these groups, and also that those candidates would gain the resources to improve the future talents of their offspring (at the expense of the more qualified candidate’s offspring…) The latter ‘benefit’ is of dubious value to society, but the former could be good, and can be seen when tech companies partner with non-profits like Code like a Girl. (Whether it’s the best use of resources to coerce a company to spend its resources in this way when it might not have otherwise is up for debate.)

Even if advocates were correct in their assumption of wide-spread bias against certain groups in corporate hiring decisions, the correction would be to remove the bias, not enforce discriminatory ratios that will never be tuned enough to ensure that the most qualified people are hired for a given job in a given place within a given company. One simple, non-intrusive example of how to do this is to remove names, which often give away gender & sometimes race/ethnicity, from resumes before they’re given to recruiters & hiring managers. Another might be to put more weight into phone screens + resume qualifications vs in-person interviews. (This one would be a tougher sell. Despite criticisms, interviewers, including myself, like to talk to someone face-to-face and put a lot of weight on those encounters when hiring.) Using a panel discussion format for hiring decisions made up of diverse interviewers is probably also a good practice to reduce bias (and more importantly, really helps to make better hiring decisions, in my experiences at two large tech firms.)

If you are for more equal diversity, you are against Asians in tech

Interestingly, while much has been made of ‘white male privilege’ in America, I’ve heard little talk of ‘Asian male privilege’ in tech, despite the fact that Asians are hugely overrepresented in tech. (Which I attribute to their education, skills, and ambition, not to heavy bias in their favor!) Uber recently released its diversity stats, so let’s use them as a hopefully-representative example. In 2018, Uber’s tech jobs were filled by 46% white people and 45% asian (with the remaining 9% about equal parts black, hispanic, and ‘multiracial’.)

 In 2015, about 63% of the US population was non-Hispanic white, 5% Asian, 13% black, and 17% hispanic. Compared to the US as a whole, blacks & Hispanics in Uber’s US workforce are way underrepresented (about 20-25% of their US-population representation), whites are also somewhat underrepresented (about 73% of their US representation), and Asians are way overrepresented by about 900%!

One might argue that Uber is an SF-headquartered company, and the Bay Area has a larger-than-average Asian population. The Bay Area is 42% non-hispanic white, 23% Asian, 24% Hispanic, and 7% African American. Compared to the Bay Area instead of the US, whites look roughly equally represented at Uber, Hispanics look even more underrepresented (13%), and Asians much less overrepresented (but still 200% of the local average), and blacks a bit less underrepresented (about 35%.)

Either way, it seems that Asians are vastly overrepresented, whites about equally, and blacks & Hispanics way underrepresented. Strangely, even such upstanding (but left-leaning) news sources as the Washington Post want to cling so badly to the idea that Uber has poor diversity that it falsely claims that “Uber employs a relative dearth of women and racial minorities, particularly in technical roles” [emphasis mine]. Uber’s US numbers for 2018 show us plainly that whites make up 49% of Uber’s total (tech + non-tech) workforce, even though non-Hispanic whites make up 63% of the total US population. The Post goes on to admit that “Uber employees identifying as Asian make up 30.9 percent of the ride-hailing company’s U.S. workforce and hold 47.9 percent of technical roles”, but the author is incapable of seeing how this single fact invalidates its initial conclusion.

Corporations already have a vested interest in unbiased hiring

Profit-loving owners/shareholders of large companies want to hire the best talent regardless of demographics. It is in their economic self-interest to do so, and any company that doesn’t will lose out somewhat on the talent battle to a more unbiased company.

Of course, this theoretical idea doesn’t prove anything about how companies and individuals at those companies actually hire. Just because a company loses out from not hiring the best across all demographics doesn’t mean that its agents (employees in charge of hiring) are doing what they should be. Individuals, and even private owners, have other motivations besides maximizing company value, and might indeed be acting on their own prejudices. That said, it should suggest that the default systematic behavior of most large private companies, especially corporations since they are typically owned by diverse shareholder groups, will be to make relatively unbiased hiring, salary, and promotional decisions.

Individual bias

While individual bias may play a role in seeing less women or minorities in tech, this needs to be studied explicitly. In one randomized simulation on hiring an academic, for example, significant bias was found… in favor of hiring women.

Conclusion

Ratios (or worse, quotas) of diversity should not be the goal of diversity programs. Instead, maintaining unbiased personnel processes to cast the widest net possible for talent should be the goal. One might start with a ratio and look at how it compares to the talent pool from which a company can choose from to get clues about bias, but that should lead to a study of whether bias is actually occurring, and the ratio itself should not be confused for the goal, or as a way to measure bias in hiring.

Diversity advocates should focus more on the much harder problem of addressing the (non-discriminatory) causes of workforce underrepresentation such as poverty and educational attainment, both of which aren’t neatly segmented into race and gender categories. Fair employment and well-being should be a constructive discussion about improving the lot of all humans, not one of zero sum social politics where each group’s default assumption is that the ‘others’ are trying to keep them down or hold them back.

 

Preemptory clarifications

Since this is a controversial subject, here a few clarifications of what I’m trying to say above:

  • I’m talking here specifically about personnel decisions including hiring, firing, promotions, and compensation. I’m NOT talking about workplace environments, or other forms of ‘discrimination’ like hostile work environments for women or minorities. Undoubtedly these would lead to lower-than-expected ratios of the negatively-impacted demographic group in those environments. That kind of discrimination is potentially separate from what I’m talking about (although surely biased hiring might be correlated with an unfriendly environment towards the group experiencing the bias.)
  • Hiring bias is likely not the primary driver of most gender/racial disparity among jobs in the US. Instead, educational attainment, family income, physique, cultural values (male deaths in the military), and prior workforce ratios (some of which may have occurred due to past discrimination or demographic-influenced choices, like female teachers and nurses, say) are probably responsible for more of the variation we see in employment. This is NOT to say that there is NO bias in hiring decisions as a whole, and especially not in individual hiring decisions.
  • I’ve steered clear of the debate over why women get paid slightly less than men (about 95 cents to a man’s $1 after controlling for various factors) and are less represented, along with minorities, among management ranks. (Arguing that the cause is anything other than sexism is a good way to lose a career.) Frankly, I find most arguments made by both sides of the female wage gap unconvincing. These arguments seem more like narrative fallacies to me than carefully constructed causal explanations based on clear evidence. It has been argued that some of the unexplained difference between male & female salaries and positions is due everything from biology, sexism, to women prioritizing their family lives over their careers (which again might be due to social expectations that are the result of bias/traditional values, etc.) I can see how those arguments might be true, but I can see equally how they might be entirely fictitious ‘common sense’ which will evaporate in a few decades as women continue to become more equal players in the workforce (and as men in turn take a more equal role in child-rearing and domestic duties.) In general, many arguments that have ‘explained’ why women (or minorities) historically have done, or not done, this or that have turned out to be self-serving falsehoods by the, typically, white men who’ve espoused them. (See Stephen J. Gould’s The Mismeasure of Man for a good overview of the faulty biology and anthropology that produced erroneous conclusions as a result of racial bias.)
  • From looking at education & income statistics, as well as my personal experience, I think women are going to be just fine career-wise, and that America should focus more on the impoverished, which contain a disproportionate amount of black, Hispanic, and Native Americans.  This confidence with regard to female career prospects is not meant to take away from other challenges women face such as sexual violence and unwanted advances from men. These social problems (still) haven’t received enough mitigation in society, especially outside the US.

Leave a Reply

Your email address will not be published. Required fields are marked *