3 ms·

> 'Hiring practices which can effectively lower the bar for “diversity” candidates by decreasing the false negative rate' Imagine you have 100 people applying

by imron 9y ago

> 'Hiring practices which can effectively lower the bar for “diversity” candidates by decreasing the false negative rate' Imagine you have 100 people applying for 10 positions, and the demographic breakdown of applicants (based on current trends in tech) is 20 women and 80 men. Let's also say that being male or female doesn't give you any special aptitude, and that the aptitude of all candidates regardless of gender follows a normal distribution along a bell-curve. Because of that and because of Google's standards, let's say that only 25% of all candidates have the skills and qualifications necessary to work at Google, which breaks down to 5 women and 20 men. Once again assuming a normal distribution with neither men nor women over or underrepresented in skill level, then without any sort of correcting for diversity, candidates would need to be in the top 10% of all candidates to receive a job offer (10 / 100). , this is 2 women and 8 men, leaving us with a false negative rate (i.e. people skilled enough to work at Google but who get declined) of 3 women and 12 men. But with an eye for diversity, Google decides to hire all the qualified women (5), with the remaining positions taken by men, so 5 and 5, a 50/50 split. Each of the candidates is qualified and capable of working for Google, but the false negative rate for women has been reduced from 3 to 0. This is what he meant by an effectively lowered bar by decreasing the false negative rate. The women now only need to be in the top 25% of all women candidates, whereas the men need to be in the top 6.25% of male candidates. There's nothing false or wrong about that statement, it's just maths that happens as a result of the diverse hiring practices skewing the distribution of candidates, with the bar being 'effectively' lowered for the diverse candidates from top 10% to top 25%. This has the unfortunate side effect that over time, if you are taking the top 6.25% from the non-diverse group (men in this case), but the top 25% from the other group then the non-diverse group will eventually have on average more high-performers (which affects things like salaries, promotions and more). It's not to say that men are better, just that the hiring practices have been structured to select men from a higher performing percentage of the population. You would get the same results in favor of women if you were only selecting the top 6.25% of women vs the top 25% of men.

kybernetikos 9y ago
If the fact that the applicant population is slanted towards a particular gender means (or is caused by the fact) that it is harder for the other gender in that field, then the smaller population will in fact be better on average than the larger, because it will have taken them more motivation, determination, work and ability to win over others just to get them to where they are. At least, this seems like the most likely explanation for the fact that individuals in the much smaller group of female startup founders outperform the male startup founders that has been reported in a number of places. e.g. > For growing start-ups, those with female founders almost universally outperformed their male-only counterparts. The fastest-growing companies at 200%+ growth, are 75% more likely to have a female founder. https://www.tinypulse.com/startup-culture-report https://www.tinypulse.com/startup-culture-report > Once again assuming a normal distribution with neither men nor women over or underrepresented in skill level This is probably wrong, and if your hiring practices assume it in a scenario where it is harder for one gender to even make it to the applicant stage then they are unfairly discriminatory.
- imron 9y ago
  > means that it is harder for the other gender in that field, then the smaller population will in fact be better on average than the larger, because it will have taken them more motivation, determination, work and ability to win over others just to get them to where they are. Do you have any studies or statistics to back you up or is this just what you believe? Also, how does it account for males that come from a poor/under-privileged backgrounds that also requires them to have more motivation, determination, work and ability to win over others and get where they are? > The fastest-growing companies at 200%+ growth, are 75% more likely to have a female founder. Revenue growth is meaningless without any sort of revenue numbers to compare it against - compare growing revenues from $5,000/year to $15,000/year (200% growth) with growing from $10,000,000 to $20,000,000 (100% growth). As a corollary to your point though, how many women founders are there in unicorns? Unfortunately not that many: http://swaay.com/todays-unicorn-companies-where-are-the-women/ http://swaay.com/todays-unicorn-companies-where-are-the-wome... Would such companies be more or less successful with female founders, I don't know, but I suspect gender doesn't really play much of a role, in much the same way I suspect that gender of founders doesn't play much of a role in revenue growth. > This is probably wrong, So you think that there is difference in ability based on gender?
  - kybernetikos 9y ago
    > Do you have any studies or statistics to back you up or is this just what you believe? I wasn't clear enough. There are some situations where smaller subgroups are overrepresented at the top of the field compared to their starting position, startups being one. This is something that can be observed as an empirical fact. The reason for it is perhaps a matter for debate as working out reasons is harder to study and I don't know of any that do in this case, but the most likely seeming reason to me is what I said - that in a field where the predominate culture has a preference, those who don't meet that preference will have to be better just to pass the bar for entry. Here's a similar article on female CEO performance: http://fortune.com/2015/03/03/women-led-companies-perform-three-times-better-than-the-sp-500/ http://fortune.com/2015/03/03/women-led-companies-perform-th... guessing at the same underlying reason as I did. > As a corollary to your point though, how many women founders are there in unicorns? I actually don't know, since it's difficult to find stats and there is a relatively small population of unicorns (meaning that any conclusion would have to be caveated), but I know of these unicorns with women on their founding teams: Ant Financial, Alibaba, GrabTaxi, Houzz, SunRun, Gilt Groupe, renaissance learning, lynda.com, cloudflare, kabam, 23andMe, fanduel, nextdoor, medallia, eventbrite, and there are a few unicorns that were spun out of other companies with women taking a lead role in that process. FirstRound.com has an interesting article: > companies with a female founder performed 63% better than our investments with all-male founding teams. And, if you look at First Round's top 10 investments of all time based on value created for investors, three of those teams have at least one female founder — far outpacing the percentage of female tech founders in general. http://10years.firstround.com/#one http://10years.firstround.com/#one > So you think that there is difference in ability based on gender? I didn't mean to make any statement on that topic. I was trying to say that if you assume that your applicants (not the population in general) have ability normally distributed regardless of gender, then you're probably wrong.
tomp 9y ago
Your analysis assumes that it's always the best candidate that's hired. Given how much tech companies are afraid of false positives (i.e. hiring candidates that turnout to be bad), I'm not sure that assumption is warranted. It's more likely that there's a bar, and everyone that clears that bar could be hired, and then a (random) sample of those people are. So in the above example, it's still top 25% of women and men that are hired, it's just that a top-25% woman has a higher chance of being hired (because of "fast-tracking" or other affirmative action policies) than a top-25% man. Personally I find it hard to believe that the companiesdoing affirmative action manage to do it without lowering the bar, but it's definitely logically and statistically possible.
- imron 9y ago
  > Your analysis assumes that it's always the best candidate that's hired. Which is something Google has gone on record as saying it does, see for example here: http://www.businessinsider.com/how-google-hires-exceptional-employees-2016-2/#1-set-an-uncompromisable-high-standard-1 http://www.businessinsider.com/how-google-hires-exceptional-... But you are correct in that my analysis makes several assumptions that might not hold up in real life, it was meant merely to show that the statement from the original memo doesn't mean (as many people incorrectly thought) that Damore was saying Google was lowering its standards for diversity hires. > Personally I find it hard to believe that the companiesdoing affirmative action manage to do it without lowering the bar Ok but even assuming as you do above that a random sample is hired such that "it's still top 25% of women and men that are hired," then you have lowered the bar for everyone from 10% to 25%. Yes that still meets their own internal requirements, but is what the author meant when he said 'effectively' lower the bar by decreasing the false negative rate.

by imron 9y ago

kybernetikos 9y ago

If the fact that the applicant population is slanted towards a particular gender means (or is caused by the fact) that it is harder for the other gender in that field, then the smaller population will in fact be better on average than the larger, because it will have taken them more motivation, determination, work and ability to win over others just to get them to where they are. At least, this seems like the most likely explanation for the fact that individuals in the much smaller group of female startup founders outperform the male startup founders that has been reported in a number of places. e.g. > For growing start-ups, those with female founders almost universally outperformed their male-only counterparts. The fastest-growing companies at 200%+ growth, are 75% more likely to have a female founder. https://www.tinypulse.com/startup-culture-report https://www.tinypulse.com/startup-culture-report > Once again assuming a normal distribution with neither men nor women over or underrepresented in skill level This is probably wrong, and if your hiring practices assume it in a scenario where it is harder for one gender to even make it to the applicant stage then they are unfairly discriminatory.

imron 9y ago
> means that it is harder for the other gender in that field, then the smaller population will in fact be better on average than the larger, because it will have taken them more motivation, determination, work and ability to win over others just to get them to where they are. Do you have any studies or statistics to back you up or is this just what you believe? Also, how does it account for males that come from a poor/under-privileged backgrounds that also requires them to have more motivation, determination, work and ability to win over others and get where they are? > The fastest-growing companies at 200%+ growth, are 75% more likely to have a female founder. Revenue growth is meaningless without any sort of revenue numbers to compare it against - compare growing revenues from $5,000/year to $15,000/year (200% growth) with growing from $10,000,000 to $20,000,000 (100% growth). As a corollary to your point though, how many women founders are there in unicorns? Unfortunately not that many: http://swaay.com/todays-unicorn-companies-where-are-the-women/ http://swaay.com/todays-unicorn-companies-where-are-the-wome... Would such companies be more or less successful with female founders, I don't know, but I suspect gender doesn't really play much of a role, in much the same way I suspect that gender of founders doesn't play much of a role in revenue growth. > This is probably wrong, So you think that there is difference in ability based on gender?
- kybernetikos 9y ago
  > Do you have any studies or statistics to back you up or is this just what you believe? I wasn't clear enough. There are some situations where smaller subgroups are overrepresented at the top of the field compared to their starting position, startups being one. This is something that can be observed as an empirical fact. The reason for it is perhaps a matter for debate as working out reasons is harder to study and I don't know of any that do in this case, but the most likely seeming reason to me is what I said - that in a field where the predominate culture has a preference, those who don't meet that preference will have to be better just to pass the bar for entry. Here's a similar article on female CEO performance: http://fortune.com/2015/03/03/women-led-companies-perform-three-times-better-than-the-sp-500/ http://fortune.com/2015/03/03/women-led-companies-perform-th... guessing at the same underlying reason as I did. > As a corollary to your point though, how many women founders are there in unicorns? I actually don't know, since it's difficult to find stats and there is a relatively small population of unicorns (meaning that any conclusion would have to be caveated), but I know of these unicorns with women on their founding teams: Ant Financial, Alibaba, GrabTaxi, Houzz, SunRun, Gilt Groupe, renaissance learning, lynda.com, cloudflare, kabam, 23andMe, fanduel, nextdoor, medallia, eventbrite, and there are a few unicorns that were spun out of other companies with women taking a lead role in that process. FirstRound.com has an interesting article: > companies with a female founder performed 63% better than our investments with all-male founding teams. And, if you look at First Round's top 10 investments of all time based on value created for investors, three of those teams have at least one female founder — far outpacing the percentage of female tech founders in general. http://10years.firstround.com/#one http://10years.firstround.com/#one > So you think that there is difference in ability based on gender? I didn't mean to make any statement on that topic. I was trying to say that if you assume that your applicants (not the population in general) have ability normally distributed regardless of gender, then you're probably wrong.

tomp 9y ago

Your analysis assumes that it's always the best candidate that's hired. Given how much tech companies are afraid of false positives (i.e. hiring candidates that turnout to be bad), I'm not sure that assumption is warranted. It's more likely that there's a bar, and everyone that clears that bar could be hired, and then a (random) sample of those people are. So in the above example, it's still top 25% of women and men that are hired, it's just that a top-25% woman has a higher chance of being hired (because of "fast-tracking" or other affirmative action policies) than a top-25% man. Personally I find it hard to believe that the companiesdoing affirmative action manage to do it without lowering the bar, but it's definitely logically and statistically possible.

imron 9y ago
> Your analysis assumes that it's always the best candidate that's hired. Which is something Google has gone on record as saying it does, see for example here: http://www.businessinsider.com/how-google-hires-exceptional-employees-2016-2/#1-set-an-uncompromisable-high-standard-1 http://www.businessinsider.com/how-google-hires-exceptional-... But you are correct in that my analysis makes several assumptions that might not hold up in real life, it was meant merely to show that the statement from the original memo doesn't mean (as many people incorrectly thought) that Damore was saying Google was lowering its standards for diversity hires. > Personally I find it hard to believe that the companiesdoing affirmative action manage to do it without lowering the bar Ok but even assuming as you do above that a random sample is hired such that "it's still top 25% of women and men that are hired," then you have lowered the bar for everyone from 10% to 25%. Yes that still meets their own internal requirements, but is what the author meant when he said 'effectively' lower the bar by decreasing the false negative rate.