A long time ago, I was once a Microsoft intern attending a party at Bill Gates' house. At that party, a high-ranking Microsoft HR employee told me that it was Microsoft's goal to have twice the percentage of female interns in their program than female CS majors.
For example, if women only made up 10% of college CS majors, Microsoft wanted to have 20% female interns.
I replied that, if that were their goal, they would most likely have to lower the bar as compared to a male intern, or else pay the female interns more, give them more perks, or purposefully interview fewer qualified male interns.
My argument was that if Microsoft's hiring bar was the top 1%, most likely only 10% of that candidate pool is female. So, one must either drop the bar for females, interview more females in that pool than males, or somehow double the chance that a female in that 10% of 1% accepted your internship offer. However, in those days almost nobody rejected Microsoft offers, so that last route seemed difficult.
The only way to maintain equality of pay and skill without purposefully rejecting male applicants is to spend a huge amount of effort finding more female applicants than male applicants in that 1% and persuading them to apply. But that's still not really fair, as that really implies that recuiters pay less attention to males, e.g. spending less time and money finding them and recruiting them.
The HR representative got very angry, but couldn't articulate why.
Not really - there's a really simple solution for hiring more women without dropping the bar for them. Make your company the most desirable place for anyone to work, so that you have vastly more applicants than positions, and then you can pick whoever you want as the incoming cohort. Math:
Say that your goal is to hire from the top 1% of the field, and the top 1% is indistinguishable from each other. There are 1 million people in the field, and you will be hiring 1,000 this year. Also say that women make up 10% of the field. In this scenario, there are 10,000 people that you would be happy to work for you, and 1,000 of them are women.
If only 10% of prospective employees would even consider working for you (which is the case for many startups, and probably for present-day Microsoft), then you're trying to fill a class of 1000 from a universe of 1,000 candidates, and only 100 of the women both apply and meet your hiring threshold. The best you can do is 10% female interns.
If, however, everyone in the field wants to work for your company, you have a universe of 10,000 candidates, 1,000 are women, and you're trying to fill a class of 1,000. You can have a female proportion anywhere from 0-100% with no loss of quality.
The strategy you describe drops the bar for all employees in order to achieve a gender target.
This is not how companies operate, for the most part. If a company wants to hire 1,000 people and 10,000 applicants are in the top 1%, they will move the bar up to hire the top 0.1% instead.
Given a normal distirbution of applicants, there is a huge difference in talent (10x?) between top 1% and top 0.1%. The bar always automatically adjusts higher; otherwise, a competitor will hire the fraction of the 0.1% that you've passed over. Now the competitor has a 1,000 workers, and you have 1,000 workers, but the competitors are 10x more talented for the same pay (most of the 0.1% didn't get an offer from you, so there's no bidding war for their talents).
Actually, if a company spends more money recruiting each equivalent female employee than male employee, they do effectively drop their total hiring bar if spending more money lets the company climb the bell curve, because it's effectively a reduction in spending efficency, but that effect is small, and skill parity is still achieved.
You're assuming you can continuously & linearly rank every single applicant. The labor market doesn't work that way. Typically, it's organized into tiers - you have your superstars, and then you have a pool of developers that are good enough, and then you have a bunch of clueless n00bs. Within a tier, it's rare to find significant, measurable performance differences. The studies showed a 10x difference in productivity between the best teams and the worst teams - that does not mean it applies to individuals, or that it means the best developer is 2x as good as the second best developer, at least on an industry-wide level.
(How would you stack-rank John Resig against Rob Pike? The two of them against Zed Shaw? The three of them against Guido van Rossum? Note also that even if you can stack rank their accomplishments, that won't necessarily reflect in their day-to-day performance. Guido van Rossum wrote Python, but he also wrote a bunch of AppEngine code that isn't all that much beloved.)
Even if the population is divided into discrete tiers, the process is still rife with unfairness no matter how you dice it.
If you have 1,000 slots, 10,000 candidates in your tier, and 1,000 of them are women, you can hire any ratio of women to men that you like and all will be equally talented. Great, right? It's great for those that are hired; not so much for everyone else.
Say you make the gender ratio 50%. You hire 500 women and 500 men from the top tier. Every first-class company like Google or Facebook adopts this strategy. This means that the odds of being hired at a first-class company is 50% for women in the top tier, and only 5% for men in the top tier. For every interview a woman does, a man must do ten. Eventually all the slots in all first-class companies are filled up, leaving some top-tier men working for second class companies--but no top tier women are working for second class companies.
Right, but you're going to get this unfairness no matter what criteria you use, gender or otherwise.
Say you leave the gender ratio unspecified and instead decide based upon the interviewer's gut feeling. Then you'll bias the hiring process toward schmoozers with good social skills.
Or you decide based on which college the applicant went to. Then you bias it towards people who were willing to shell out for a prestigious piece of paper.
Or you decide based on whoever responds to your offer first. Then you bias it against people who have lives and better things to do with their time than refreshing their e-mail waiting for a callback.
Really, the only solution is to acknowledge that life's not fair, and people sometimes get things for completely arbitrary reasons. Which is really hard for a lot of people to do - it was hard for me - but you end up being a lot more successful when you don't think too hard about all the folks who get undeserved job offers and promotions and think more about how you can tilt the odds toward being one of the lucky ones instead.
I completely agree. I make this point, though, because the parent article begins,
“You only got that internship because you’re a woman.”
Note that this statement does not imply she is unqualified. She could be absolutely qualified (and probably is). However, in the ficticious tiering example above, the female applicant has 10x higher odds than a male applicant of getting a sought after job at a first-lass company, even though both are equally qualified. For this example, at least, the above statement is explainable (minus the "only" part, which is just mean) by the huge difference in probability between her and her friend. Her friend would have to apply to ten times more internships in order to land an equivalent gig.
It largely is in an employment situation. There are a number of people that are all well-qualified for the job, and any one of them can do it as well as any other.
It's not necessarily true when you compare between organizations, where different groups may make subtly different decisions with wildly different outcomes (although even then, a lot of the difference comes from different circumstances and not innate talent of the people involved). But in an organization, the need for teamwork and to "row in a common direction" tends to flatten out individual differences, and there're generally two possible classes: "valued contributor" and "holding back the group".
(This is also why large organizations like Google tweak their hiring processes to avoid false positives more than false negatives. One bad hire forces the team down to his level, as they always have to stop and explain things to him, or he'll block them from implementing a cleverer solution that he wouldn't understand. One good hire, however, very rarely raises the level of the team - he needs a lot of patience and very good empathetic & people skills to do so.)
> It largely is in an employment situation. There are a number of people that are all well-qualified for the job, and any one of them can do it as well as any other.
No, that is simply not true. There might be a cutoff over which everyone is "reasonably competent", and perform "acceptably", but there are still huge differences between people.
You immediately say "no" to people that don't reach the "acceptable" bar. But above that, you have hiring quotas, and you try to maximize the value of the people you hire. Hiring someone just above the bar now means you get one less hire later. Someone much better might show up, and often does. And on the other hand if the current candidate is much, much above the bar - you fight to get that person hired, you work on convincing your peers in the hiring process. Because they are worth it. Such people have a much higher chance to be hired.
I've been on both sides of the hiring process many times and worked for many years in tech. To say that above some skill level everyone is the same, "can do it as well as any other" - not in all of my experience.
How large an organization? There are big differences between what a 10-person startup needs and what a 20,000 person mega-corp needs.
(FWIW, I would pay much more attention to individual performance differences if I were hiring for my own startup than I would when interviewing for Google. But I thought the context of this discussion was organizations large enough for gender quotas to matter, i.e. the Microsofts and Googles of this world. If it were my own startup, I'd try to hire from the population I've personally worked with, avoiding this whole discussion anyway. And I have never worked in an organization with a hiring quota - the companies I've worked for will all take you if you meet the hiring bar, and hold cash in reserve so they can scoop up a suitably-qualified employee if one presents herself. For that matter, I've been given offers at several places that were "not hiring", so I'm guessing quotas are just guidelines in many other places as well.)
> And I have never worked in an organization with a hiring quota - the companies I've worked for will all take you if you meet the hiring bar
You must have worked only at places where the amount of acceptable candidates is greatly constrained. Either because the hiring bar was extremely high, or there simply were extremely few candidates out there with the right skills.
In practice I've seen quotas everywhere I've worked. At small startups, at the beginning you often have no money to pay salaries, so you give out equity, and you don't want to be diluted into nothing from day one. It's also crucial to find great talent for the very beginnings of your company and codebase.
For large companies, there are always quotas because otherwise they would grow until they quickly become unprofitable, and of course there is a limit to how fast you can integrate new people into an existing structure. Each division and team has a target size for the next year, and they hire up to that limit.
I have never been in a company, big or small, where we said "hire as many good people as you find! no matter how many! we'll take 'em all!"
I work at Google. The hiring bar is definitely high, but the quotas are also definitely just guidelines. (I was hired during a hiring freeze, for example.)
I've previously worked at 2 startups, founded 1, and also interned at a large and a mid-size company. None of them had quotas. The two startups were very constrained in the number of acceptable candidates, but the internships had a fairly large pool of candidates to choose from, and would make a position for a suitably-qualified candidate if one did not exist.
I think we are saying basically the same thing in other words.
If the hiring bar is very high, it can be set high enough so as to limit the number of people you hire. So you end up hiring only (ones you think are) the very best.
The fact remains that even in such a "quotaless" situation, if you saw 1,000 amazing people you would not hire them all for your 10 person startup. That would be lunacy. In fact you would stop interviewing after hiring a tiny fraction.
The point here is that, contrary to the discussion before, it isn't that there is a "near-infinite" amount of candidates of equal talent. All companies want to hire the best, and the right amount of them within some reasonable range.
So it isn't that you can pick a criterion like "we will hire only left-handed people" without that having an effect, since left-handedness is only about 10% of the population (of all coder skill levels). If you start looking at far fewer candidates, you will miss some of the very best that otherwise you would want to hire.
The startups I've worked at hire everyone who's good. Quotas and headcounts exist at companies like Intel, but VC funded startups can't possibly hire quickly enough to spend all their investment.
Really? I always see funded startups limit their hiring to their investment and because it just isn't practical.
If you are a 10 person startup, there are likely thousands of people in silicon valley who could work for you, and hundreds of thousands who could work remotely. You're not going to hire them all - it would take too long, you would run out of money, and you would have your equity diluted into nothing.
Do you have evidence that talent can be linearized in such a manner that it is possible to tell objectively within 1% how well an intern will fit at the company? Of course you can linearize on test scores or number of publications or something, but is there evidence that these values will translate precisely to job performance? In my experience, such evidence does not exist, so the zero-sum reasoning is fallacious from the start.
In my experience, when looking for people to fill a position, that we end up with some quantity who exceed our requirements, all of whom we'd be happy to hire. The correlation between that quantity and the number of available slots is not strong. Once the candidates cross the "happy to hire" threshold, the decision comes down to what, in retrospect, seem like random factors (if we vaguely think the company might need a certain type of skill later, we might bump up candidates who have that skill, if someone on the team particularly got along well with some candidate that's a plus, etc etc). Evaluating people is not very precise at all.
The line about spending a "huge" amount of effort finding more females seems a bit misogynist on its face. There doesn't seem to be any evidence that it would be a particularly huge expense. If the goal is to have an incoming ratio with 10% more females, it seems that the additional expense could reasonably be only 10%. It also is fallacious to assume that this is a zero-sum equation; that additional 10% (or 50%, or whatever) could be provided solely for that purpose, and would be unavailable otherwise.
Congratulations on successfully trolling the HR representative, though.
I'm interested in your position. I mainly don't agree about the non-zero sum or definite ordering parts (the BigTechCos that I've worked out explicitly stack rank applicants, so that means they do think there is some definite ordering that can at least be approximated), but I don't think that is a really interesting debate to have.
I think more interestingly, it seems like what I've seen is looking to hire interns where we can't even fill the spots we want with people who appear that they are likely to be actually qualified, they hire interns who are unlikely to be qualified because its worth it to filter down to the few that are. It seems that your suggestion that you get more qualified applicants than you have positions for, which really flies in the face of my experiences. Is it really true or were you exaggerating somewhat to make a point?
Not exaggerating at all. I've never been in a position where I felt I had to lower my standards below "the best in the world (for our team)", and I've always been able to turn down a candidate who doesn't meet that criteria without having to worry about having an empty seat. That's how tech hiring should be; you're building an effective team rather than laying bricks.
Maybe I've been lucky with the organizations I've been associated with, but none of them has thought of hiring as filling slots, or of internships as anything other than nurturing the most promising young people to become great hires later.
Are people really hiring interns to get work done? No wonder we have so much unimaginative cookie-cutter software! I mean, there's no doubt that tons of people use absolutely abysmal hiring practices where making certain decisions would make them even worse, but that's not really interesting to talk about.
But that's still not really fair, as that really implies that recuiters pay less attention to males, e.g. spending less time and money finding them and recruiting them.
You're wrong. It is fair. "Normal" recruiting promotion budgets are obviously paying more attention to males. For example, if you market primarily to CS departments which are 90% male, then that is money largely spent on recruiting males. Spending money specifically on recruiting females brings the budget to parity.
> You're wrong. It is fair. "Normal" recruiting promotion budgets are obviously paying more attention to right handed people. For example, if you market primarily to CS departments which are 90% right handed, then that is money largely spent on recruiting right handed people. Spending money specifically on recruiting left handed people brings the budget to parity.
Dividing up the population into two groups using your favorite method and then demanding that the amount spent on each of the groups is equal is ridiculous. Not only that, but mathematically you almost surely can't even satisfy two people that have that philosophy simultaneously.
What you want here is that the amount spent on a person is conditionally independent [1] of characteristics that are irrelevant to that persons performance given the characteristics that are relevant to that persons performance.
What you want here is that the amount spent on a person is conditionally independent [1] of characteristics that are irrelevant to that persons performance given the characteristics that are relevant to that persons performance.
Why? Just on topic that leaves out two very, very important things:
1. The process is already inherently unfairly biased towards men and it will take a nontrivial investment of resources to counter that.
2. Although it wouldn't affect an individual's performance, diversity (and other things that don't affect individual performance) can affect team performance and thus need to be considered.
"The process is already inherently unfairly biased towards men"
This is untrue. There are probably all sorts of societal issues that affect whether one enters the tech industry based on gender, but those aren't inherent to the process. It's also not clear that hiring practices will help the problem. A desirable company can improve its gender ratio, but only by hurting the ratio at other companies. (I don't think this is a bad thing.) Something has to cause more women to become programmers, and I doubt hiring practices are the solution.
Parent meant "fairness" to apply on the individual level, rather than on a collective level, I believe. That is, resources expended per candidate would be comparable, rather than having the total sum of resources expended on female candidates be equal to those spent on male candidates (mutatis mutandis: hires, members of the public, etc.). Which is presumably what you mean.
But even resources per candidate isn't a good measure of fairness. Just consider the trivial illustrative possibility of a normal job board posting that every potential candidate sees. In today's world it is likely that more men would respond than women by a nontrivial factor. Everyone who responds receives equal treatment and equal devotion of resources and, in the end, significantly more men are hired than women. Now, imagine that we have the exact same scenario but somebody is hired to stand next to the job board and tell every woman who comes by "you should apply for this!" More resources are now being expended per woman than per man but the proportion of people responding might be different without being unfair to any particular person.
Granted, there are a lot of trivialities in this example, but it's pretty representative of how such programs are supposed to work.
Actually, this is unfair to particular people. Two students (a man and a woman) walk by the job board for company X. Both of them are ideal for company X, and they would love to work there, but they don't notice the board, so they both take a job at Y for less money and are unhappy.
Now, rewind. The situation is the same, but nearby is a recruiter. The man walks by again, not noticing the job board. The recuiter does not flag him down. Now the woman walks by and the recruiter walks up to her and interests her in company X. She takes her ideal job and is happy. The man takes a job at company Y, for less pay, and is unhappy.
Ah, but remember that in this example everyone sees the posting (or, if you prefer, the recruiter only reaches out to women who see the posting) so your scenario is impossible. The recruiter's role isn't to increase visibility but rather to be a signal to women that they wouldn't be wasting their time with this company.
When programs like the one briefly mentioned in the article are done well their net effect is really signalling more than anything. As trivial as it sounds, such signalling can actually have a large impact.
Both of the students are CS majors. They presumably will both eventually work for similar companies. Why would the woman who is equally interested in computer science pass over the job posting, requiring an actual recruiter to be present, when the man does not? Do women get CS degrees and then leave to work in some other field post-graduation, not writing code? If so, that is a huge problem. Do they lose interest in computer science? Do they get sick of the male dominated environment in CS but just stick it out for their degree?
I would hope that the recruiter isn't, in fact, signalling something along the lines of, "your odds of getting hired at this company are greater than your equally talented male friend." If I were receiving that much extra recruiting attention because of my gender, that is the impression I would get.
I hope you don't mind if I'm not interested in spending a few hours tracking down all the sources I've referenced before but to give you a short answer to your first paragraph: for the past decade or so women have held 20-30% of the CS degrees with recent times being somewhat higher. Women also hold less than 20% of the software positions (http://www.pbs.org/newshour/rundown/2012/04/science-engineer... ctrl-F for "19 percent"; it cites the Bureau of Labor Statistics but I'm not sure what part). Obviously there's not a perfect correlation between degrees and jobs for a number of obvious reasons but it's evidence against equal chance of CS majors staying jobs. And I've seen separate sources more clearly showing that there is larger fall-off for women than men. Unfortunately, I don't know what the cause is exactly.
However, your second paragraph sort of has things backwards. The current scheme of things strongly signals to women (pretty accurately) "your odds of getting hired are lesser than your equally talented male friend." Such signalling as I described before isn't about saying that their odds are better than their male friend but rather that the company is actively trying to not disadvantage women.
The current scheme of things strongly signals to women (pretty accurately) "your odds of getting hired are lesser than your equally talented male friend."
Do you have some sort of evidence for this? Anecdotally, talented female programmers stay on the market far shorter than males do. Most companies desire a gender balance, which makes the limited stock of female programmers more vlauable.
"Fairness" as you use it should have no part in a business decision. The Microsoft HR team identified that for them to maximize their competitiveness (i.e. profit), they need more female staff members. That means that an "underqualified" woman suddenly becomes more qualified because she brings a business advantage to the table that a man does not. I have no problem with this.
1. Microsoft does not make decisions (mostly) based on fairness to prospective employees. It DOES make decisions based on having a good public image, part of which is being seen as diverse and inclusive.
2. The reason being diverse and inclusive is seen as good is because of a ton of research showing that being diverse is good for companies, good for employees, and good for entire underrepresented communities.
So it's not fine, but the fineness or not does not exist in a vacuum. The two situations are not arbitrary, and therefore, not analogous.
Okay, first, can we drop the overuse of female/male? It makes us sound like we're talking about specimens in a bio lab rather than people and introduces issues wrt the nonequivalence of female and woman or male and man.
That said, to a large extent you are correct but you are framing it poorly. One of the major steps of breaking the gender gap is to spend more resources on, not just finding women, but also encouraging women. The net result may be in paying disproportionate amount of attention to the group of women compared to the group of men but it should not imply either of the more obvious ways to interpret "pay less attention to males". That is, it shouldn't involve ignoring men or spending less time with any individual nor should it mean reducing any focus on finding good people regardless of gender. It should involve adding focus on the women, which is substantively different despite the ability to describe it with the same words. It's also not necessarily feasible for everyone but it certainly is for Microsoft, especially with the sort of focus they put on trying to bridge the gender gap.
You have a budget of one hundred identical recruiters and need to get 1000 female applicants and 500 male applicants to apply for up to 100 internship slots to meet your goal of hiring twice the percentage of women in CS programs. In your country, all of the colleges are segregated; there are 100 all-male and 100 all-female colleges. All of them are having job fairs next week. For each job fair you attend, you know you will get 10 applicants. How do you assign your recruiters?
Okay, sure, in your highly-fictional, strangely-exact, not-at-all-representative-of-reality example more mens' colleges will be left out than womens' colleges. In reality, not a single part of your scenario actually holds and even the parts that are close aren't that close. I'd do just as well to assume in your scenario that all candidates are uniform, frictionless spheres in a vaccum where gravity is reversed.
For example, if women only made up 10% of college CS majors, Microsoft wanted to have 20% female interns.
I replied that, if that were their goal, they would most likely have to lower the bar as compared to a male intern, or else pay the female interns more, give them more perks, or purposefully interview fewer qualified male interns.
My argument was that if Microsoft's hiring bar was the top 1%, most likely only 10% of that candidate pool is female. So, one must either drop the bar for females, interview more females in that pool than males, or somehow double the chance that a female in that 10% of 1% accepted your internship offer. However, in those days almost nobody rejected Microsoft offers, so that last route seemed difficult.
The only way to maintain equality of pay and skill without purposefully rejecting male applicants is to spend a huge amount of effort finding more female applicants than male applicants in that 1% and persuading them to apply. But that's still not really fair, as that really implies that recuiters pay less attention to males, e.g. spending less time and money finding them and recruiting them.
The HR representative got very angry, but couldn't articulate why.