Excellent work and a fascinating result. Seems to me that both kinds of questions are fair game for any entrepreneur. Maybe if men were asked more "prevention" questions we'd get fewer big flameouts (Color Labs [0] comes to mind).
(Though I wonder about their use of the term "Turing test" -- do they mean "A/B test"?)
The original Turing test had the interlocutors guessing whether the subject was a man or a woman, I think it's an oblique reference to that to suggest disguising the entrepreneur's gender would make things fairer.
This research illustrates beautifully how tech's pervasive and durable gender disparity problems are not solely attributable to a handful of deviant "assholes being assholes" and egregious actions such as obvious sexual harassment.
Many people who think of themselves as well-meaning and who are perfectly comfortable acknowledging that problems exist also play a role in perpetuating the negative status quo. These individuals are an opportunity for constructive engagement and real progress.
Yes. The unconscious assumption -- evidently made by female as well as male VCs! -- seems to be that a female entrepreneur is much less likely to found a company that will become huge. If an opportunity has less upside than others, it is natural to look more closely at the downside risk.
I wonder whether YC has managed to free themselves to some extent from the effects of this bias by standardizing their interview process.
I would be interested to know if there were any variance in marketing materials/information provided by male vs female led firms that might seed the discrepancy.
Men and women are different. It makes sense to ask different questions. If I'm choosing which companies to invest millions of dollars in, you can bet I'll take woman specific issues, such as pregnancy, into consideration. The idea of a universal, gender blind rule set is absurd to rational decision making.
Something you should keep in mind is that a man is also affected if he has a child. Available time, sleep quality, stress levels are all disrupted, and it affects our cognitive abilities to extreme degrees often times. So just be sure to discriminate against men who plan to have kids to some reasonable factor as you do women.
But I got just as little sleep, if not less, than my wife did, because I was the one getting up and changing my baby in the middle of the night. Please don't generalize.
It does not adjust for sample quality. Perhaps the women engaged radiated their insecurities, and potentially their insecurity about their gender.
If there's anything a VC is wary of, than it's a wary rather than optimistic attitude.
I have not personally met a competent computer scientist woman, I have met competent logicians, lawyers, and businesswomen. I cannot avoid being skeptic without positive evidence.
How one dresses makes a big difference (see "Dress For Success" by Molloy), as well as height and general level of physical attractiveness (for both men and women). Scott Adams notes in "How To Fail" that even altering the timbre of your voice makes a difference, as well as your word choices.
I suspect that everywhere you look for a difference, you'll find one.
It seems like the researchers missed a golden opportunity to also slice the data on objective criteria like height, age, etc., and look for other correlations.
This is precisely why we conducted an experiment to remove the influence of variation in speech patterns, vocal intonation, nonverbal gestures and physical appearance. Note that the co-author on the male attractiveness in venture funding paper is a co-author on this paper, so we are well aware of these factors. I would point you in the direction of that paper as well since male not female attractiveness was found to improve funding outcomes.
That's good. (I see the article did mention "Controlling for factors that may influence funding outcomes — like measures of startups’ capital needs, quality, and age, as well as entrepreneurs’ past experience" but not the other factors.) How about height? I recall a study years ago that found that salaries for men increased by $1,000 per inch of height. CEOs tend to be tall.
It seems "prevention" questions would imply skepticism on the part of the questioner. You lightly touched upon this, but were there equally drastic discrepancies between the types of questions posed to male versus female founders in the presence of equal growth/success metrics? I would certainly believe it -- maybe even expect it, but would love to hear more as this point really hammers home the extent of the bias
Yes, these results hold when controlling for such factors; this study would never have been published otherwise Obibring! I urge you to read the actual paper since it appears you are interested.
Success of your project isn't entirely objective or logical. Why do we expect the process to funding your project would be? You know, I heard that sometimes wealthy parents will invest millions of dollars in their children's projects. How incredibly biased of them. They don't even ask tough questions in the due diligence process.
Given a problem statement, the questions that emerge should objectively be largely based on the problem itself: the industry, competition, business models and risks. Regardless of whether it is being presented by a man, woman, monkey, or dog.
The only way the questions could change is if you already knew about the person, and if the objective of the questions was to test only weaknesses in the whole system including the people driving it, assuming the strengths.
This would then imply that on average the men should being tested on what the men are weaker at! And correspondingly for women.
This study does not adjust for age, experience, height, or anything really.
It's enormously suceptible to a hidden variable effect. E.g. suppose males and females has different experience levels. The VCs questions might be the same after controlling for that, and so we should be researching why males and females have different experience levels. Instead, we're looking in all the wrong places.
I'm not saying this is the case, but it reflects the quality of the study that even the basic stuff wasn't considered.
Paul,
If you read the paper itself, not just the highlights, you will find that we have controlled for every factor that has been found to impact funding, including past experience of the entrepreneur. The 2nd study (the experiment) then controls for all these factors and only manipulates prevention/promotion levels, finding that the allocation differences are significant.
And please note that it is really damaging to science in general when people make comments like "the basic stuff wasn't considered" without actually reading academic studies.
Hope you will keep this in mind in the future before posting comments!
27 comments
[ 4.9 ms ] story [ 49.8 ms ] thread(Though I wonder about their use of the term "Turing test" -- do they mean "A/B test"?)
[0] https://en.wikipedia.org/wiki/Color_Labs
Many people who think of themselves as well-meaning and who are perfectly comfortable acknowledging that problems exist also play a role in perpetuating the negative status quo. These individuals are an opportunity for constructive engagement and real progress.
I wonder whether YC has managed to free themselves to some extent from the effects of this bias by standardizing their interview process.
If there's anything a VC is wary of, than it's a wary rather than optimistic attitude.
I have not personally met a competent computer scientist woman, I have met competent logicians, lawyers, and businesswomen. I cannot avoid being skeptic without positive evidence.
I suspect that everywhere you look for a difference, you'll find one.
Given a problem statement, the questions that emerge should objectively be largely based on the problem itself: the industry, competition, business models and risks. Regardless of whether it is being presented by a man, woman, monkey, or dog.
The only way the questions could change is if you already knew about the person, and if the objective of the questions was to test only weaknesses in the whole system including the people driving it, assuming the strengths.
This would then imply that on average the men should being tested on what the men are weaker at! And correspondingly for women.
I do not see this in the result set.
Welcome to the stone age all over again :)
It's enormously suceptible to a hidden variable effect. E.g. suppose males and females has different experience levels. The VCs questions might be the same after controlling for that, and so we should be researching why males and females have different experience levels. Instead, we're looking in all the wrong places.
I'm not saying this is the case, but it reflects the quality of the study that even the basic stuff wasn't considered.