Workarounds for hiring discrimination in tech

2022-09-04

Dan Luu has written very convincingly about why discrimination in tech hiring exists despite "efficient markets" arguments for why it doesn't. This was famously argued by venture capitalist Marc Andreessen to New York Magazine, who asserted that systemic discrimination in Silicon Valley can't exist because of how competitive hiring in tech is:

Let's launch right into it. I think the critique that Silicon Valley companies are deliberately, systematically discriminatory is incorrect, and there are two reasons to believe that that's the case. ... No. 2, our companies are desperate for talent. Desperate. Our companies are dying for talent. They're like lying on the beach gasping because they can't get enough talented people in for these jobs. The motivation to go find talent wherever it is unbelievably high.

The efficient markets hypothesis comes from the field of financial economics, and roughly states that market forces cause all available information to be priced into assets. The irony is that it doesn't even hold up in its original field, never mind labor markets: there continues to be an entire industry of people who reliably make profitable trades in financial markets based on publicly available information. Likewise, hiring discrimination persists, despite US laws for protected categories like race and sex. In addition, the literature reveals persistent biases that generally have no legal protection whatsoever, which we'll get into in a bit.

To be clear, discrimination shows up in the data. And it persists despite decades of evidence and efforts to address it. For example, annual Stack Overflow surveys consistently show fewer than 10% of respondents identifying as female, and they are systematically paid less and receive fewer interview callbacks than similarly qualified male applicants in job postings. Eliminating these biases would be not only a "feel-good" measure (e.g., because "it's the right thing" for tech companies to do), it is also unambiguously a pure win by conferring a huge competitive advantage to firms where it counts: recruiting the best people. It's even good from a macro perspective, because it results in better allocation of talent and resources in the economy.

Let's dive into discrete biases for which we have evidence, and some obvious workarounds that come to mind:

Names

Names that encode or imply race

Black sounding names get fewer interviews than identical resumes with whites-sounding names [1].

Names that suggest gender

Tech recruiters in one study were 6.47% less likely to reach out to women with comparable qualifications than men, even though female programmers who self-report programming language experience have more experience on average than males. This appears to be partly because women are 11% less likely to self-report knowing a programming language despite working in it. This causes large disparities in outcomes, since candidates with self-reported programming language experience are ~30% more likely to be recruited [2].

A study of resume reviews of college seniors of all majors suggested no discrimination (on average) on account of race or gender, but for STEM fields employers displayed "large, statistically significant preference" for white males, equivalent to a penalty of 0.27 GPA for female or nonwhite candidates [3].

Alphabetical ordering

A new research paper suggests alphabetical sorting of names significantly affects early careers [4], which disappears by mid-adulthood, "presumably because it is superseded by observable characteristics that are more directly expressive of ability."

"Surnames with initials farther from the beginning of the alphabet were associated with less distinction and satisfaction in high school, lower educational attainment, more military service and less attractive first jobs."

Hard-to-pronounce names

A study of PhD students and job outcomes found that hard-to-pronounce names were less likely to be hired in academic or tenure-track positions, and when they were hired, took less prestigious posts overall [5].

A longitudinal study of immigrants from the early 20th century shows they experienced larger "occupational upgrading" and higher earnings if they adopted Americanized names [6].

Takeaway: if you have a "weird" foreign name, a "black" name, or anything that doesn't code as white anglo-saxon male, go ahead and pick an alias that works around these issues. No one should discriminate because they subconsciously think your name is weird, but why give them the opportunity?

Social class

"High class" hobbies

Resumes from selective law schools to elite law firms showed a higher callback rate when they contained multiple signals of "high class." This included a "high class" last name, and references to sailing, polo, and classical music. This effect only helped men, not women. High class women, lower class women, and lower class men with identical qualifications had a lower callback rate [7].

In free-form interviews, almost all attorneys reviewing resumes worried about the "commitment" of "high class" women, worrying they might "leave paid employment entirely." I also laughed a little at a reviewer verbalizing their real-time rationalization for why those hobbies are good signals:

Mark said of the higher-class man: "If you look at the interests, it's classic cultural capital. It would help with being around people who [he pauses] work hard."

Ah yes, sailing, the hobby for the hardest-working of Americans!

This study is worth a read and presents more shades of gray than I am letting on here - but the general takeaway is that "cultural fit" (as in having the same prevailing hobbies and interests at a firm) is almost certainly a huge leg up.

Takeaway: Don't name interests or hobbies that are associated with a "low class" status (e.g., "hoops"); consider picking up and listing hobbies associated with "high class" status (classical music, etc). Actually, you should probably just aim to try everything once and then list the "right" ones on your resume depending on the culture of the firm you are applying to [8].

Speech and accent

Hiring managers who watched a brief "tell me about yourself" video devoid of qualification information picked up on class cues, and judged higher social-class applicants as more likely to be competent for the job, a better fit, and more likely to be hired [9]. Some other interesting findings in the research paper, including pronunciation patterns as cues of social class. Listeners could discern from 7 out-of-context words whether a speaker was a college graduate or not, with better-than-random probability (albeit only 55% accuracy).

Takeaway: Write and speak using "standard english" as might appear on the largest US websites, or like a college graduate.

Physical Attributes

Height

Employers are reluctant to hire short applicants, and perceive taller applicants as more competent. Additionally, short people are promoted at considerably lower rates than taller peers [10].

All else being equal, people seem to prefer to be led by folks who are literally taller.

So, basically: Wear a pair of shoes / hairstyle that increases your perceived height.

Facial dominance

The ranks of cadets entering the Military Academy at West Point are known to be related to "facial dominance" (essentially, how masculine the facial features are) based on ratings of graduation photos at West Point. Further research revealed that officers with more facial dominance were also promoted faster, earned more promotions in total, and were more likely to reach the General ranks (Brigadier General, Major General, Lieutenant General, etc) [11].

The methodology in the linked paper is pretty impressive - it attempts to control for a large variety of factors to isolate the effects of facial dominance. They also adjusted for whether the candidate was smiling in the photo (which is known to lower ratings of facial dominance).

Face with dominance score 6Face with dominance score 3
This cadet had a facial dominance score of 6 (out of 7)
This cadet had a facial dominance score of 3 (out of 7)

We select leaders, in part, by how much they literally "look like a leader." Or more accurately, our preconceived notions of a what leader should look like. So, I dunno, I'm not like a makeup expert or anything, but it can't hurt to emphasize how manly your face is, or something like that.

Weight

Obese and overweight women make less money [12] [13]. The effects are somewhat uneven as the effect appears reversed for black men (emphasis added):

For a two standard deviation increase in weight (64 lbs.), white women are paid 9 percent less. Black men seem to receive higher wages with higher weight, and Hispanic men incur a wage penalty.

So appear tall and don't be overweight. Got it.

Age

Older job candidates are perceived as less desirable [14] and get fewer responses to job applications [15]. And when you adjust for experience and education, older workers make less than younger colleagues [16].

Takeaway: if your resume implies you are 50+ (but probably sooner), start truncating some older experience and drop any graduation years from your education.

Attractiveness

Attractive people make more money [17] [18].

Attractive men and women are more likely to be called when applying for jobs, but the starkest difference seems to be for unattractive women, who are heavily penalized.

A set of studies with seemingly conflicting results seems to boil down to: If you are an attractive person, try to make sure the person judging you is the opposite sex.

Takeaway: Be attractive. Don't be unattractive.

This is weird.

I'm generally against being deceptive or manipulative for personal gain, but is it deceptive if you intentionally mask an attribute that shouldn't be a factor in a hiring decision? It's not like your interviewers would be able to consciously object - "I offered you the SQL developer job on the basis that you were taller" just doesn't come out right?

It should go without saying that the existence of these biases is unfortunate, and the onus shouldn't be on candidates to work around them. But I just wouldn't blame anyone who adopts any of these workarounds (or similar ones) to their personal benefit.



[1] https://eml.berkeley.edu//~crwalters/papers/randres.pdf Involved 123 firms from the Fortune 500 for a total of 84,000 job applications. Methodology was "audit" of job postings involving submission of fictional resumes
[2] https://pubsonline.informs.org/doi/10.1287/mnsc.2021.4077 . Sample size: 170,886 candidates, 7% female
[3] https://faculty.wharton.upenn.edu/wp-content/uploads/2018/09/KesslerLowSullivan_Revision1.pdf
[4] https://www.colorado.edu/faculty/zax/sites/default/files/attached-files/alphabetism_-_cauley_and_zax_-_6-26-17.pdf
[5] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4031991
[6] https://www.journals.uchicago.edu/doi/abs/10.1086/692531
[7] http://www-2.rotman.utoronto.ca/facbios/file/RiveraTilcsik.pdf The researchers specifically picked law because it's a field requiring education and credentials; its unclear if this generalizes to other fields like STEM.
[8] If you try something once, I don't feel like listing it as a hobby on a resume is a lie. And if someone brings it up in an interview, you'll have some experience at hand to talk about.
[9] https://www.pnas.org/doi/epdf/10.1073/pnas.1900500116 [10] https://opencommons.uconn.edu/cgi/viewcontent.cgi?article=1432&context=law_review
[11] https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.878.6872&rep=rep1&type=pdf . Results over 289 career officers who were in service for 20 years or more (e.g., analyzing only those who made a life commitment to the career)
[12] https://www.bls.gov/osmr/research-papers/2007/pdf/ec070130.pdf
[13] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2379575
[14] https://academic.oup.com/esr/article/37/1/49/5900768
[15] https://www.tandfonline.com/doi/full/10.1080/13504851.2011.581199?src=recsys - or if you believe https://www.tandfonline.com/doi/full/10.1080/09585192.2011.540160?src=recsys, age discrimination starts in the late 30s.
[16] https://www.tandfonline.com/doi/abs/10.1080/08874417.2004.11645817 - the abstract is really good, but unfortunately I couldn't find a free version of the paper itself to read.
[17] https://press.princeton.edu/books/paperback/9780691158174/beauty-pays
[18] https://journals.sagepub.com/doi/10.1177/0022185608096808