Hacker Read top | best | new | newcomments | leaders | about | bookmarklet login
YC Demographics (blog.ycombinator.com) similar stories update story
77.0 points by austenallred | karma 22490 | avg karma 7.52 2015-01-14 20:56:06+00:00 | hide | past | favorite | 71 comments



view as:

How large is the random sample? Since that you're asserting that YC does not discriminate because the actual proportion of female entrepreneurs funded is more than the number that applied, the conclusion is open to variance.

EDIT: Originally incorrectly said "sampling bias."


for gender, it's the entire applicant pool (we ask for gender so that's easy to compute.)

for race, it's a bit harder. we don't ask for it, but we looked through hundreds of videos and made our best guess (it's also why we left the number as approximate, as of course there are cases where we got it wrong.)


Ah, that clears up the sampling bias.

I might have asked the wrong question. What I meant was "how many people are 5% of YC applicants" that were used to calculate a representative proportion? Let's say 2000 companies apply each with 2-3 founders, so 4000-6000 founders total. 5% of that is 200-300, which can result in a reasonable margin of error for the conclusion proportion but may not be perfect.


more than 10,000. still, definitely small enough for random variance.

Any sample size this small is open to sampling bias.

Small nitpick: What you are talking about is not sampling bias, it is something else.

You are alluding to how confidence in the presented results may be significantly lowered due to a high variance - associated with small sample sizes (it is a statistically significant possibility that you obtained the outcome through chance and not because it is representative)

Sampling bias is another thing altogether, when the method of taking measurements will exclude some of the population. For example, estimating ratio by taking note of the presumed gender of people leaving a particular room - not realising that it is in fact the female toilet.

Nonetheless, it is always a good idea to query the sample size when results are presented in such a manner :-)


You're correct. I double-checked my terminology now. :)

Great to see this data, thanks for releasing it.

I wonder what the age distribution is? (Not just gender and race...)

Yes. Age is a fundamental figure for any demographics. Gender, ethnic group, and age. The fact that age is totally ignored is a huge blind spot.

Exactly. They ask for age in their application and not gender or race.

Not a YC partner, but I help review apps... when I look at age, it's not so that I can discriminate based on it, it's so that I can see how impressive a person's accomplishments are in light of how much time they've had to achieve them.

The problem is that this can be very skewed, especially for younger people by their economic background.

What do you recommend?

For example, I know of someone that changed careers into the tech industry in his mid-30s. He should not be disadvantaged because he entered into the industry later than peers.

Additionally, I can think of countless examples of "slow starters" that didn't really get good at something until there was something that caught that person's interest.

At best, this is a flawed metric with many nuances to consider. So many that you could probably make the same determination by removing age from the application.*

* If you want to collect demographic data, you can just ask for it in the submission form but not show it in reviews.


Interesting... I'd be curious to hear which successful startup founders fit this pattern?

I'm not a founder [but that doesn't mean I wouldn't potentially found a company someday] - I worked in hospitality management until my early 30s. If I were to apply to YC I wouldn't expect you to be interested in my previous career so I wouldn't include that. If I understand you right you're saying you'd look at what you see I've done [in the ~3 years I've been a developer] and be less impressed because it's coming from someone in their mid-30s instead of their early 20s.

Sorry to break it to you, but factoring that into your decision is the very definition of [age] discrimination.


Sorry, I didn't mean to give the impression that I only look at technical accomplishments. I like to look at whatever accomplishments that person has, regardless of what field they're in. So if you've worked a lot in hospitality management, then great, let's hear about your accomplishments there.

>> I only look at technical accomplishments

Without age discrimination

>>accomplishments are in light of how much time they've had to achieve them

Age discrimination by definition


Small correction: The W15 applications did ask for gender.

we've talked about this before. we fund founders of all ages above 18 (though i think the oldest founder in the current batch is 66). we haven't calculated it yet but the median age for this batch is probably in the late 20s, with most founders in their 20s, 30s, and 40s. the last time i looked, we'd funded many more founders in their 40s than in their teens.

there are a lot of promising 16 and 17 year olds that apply that we can't fund, though i think it's probably a good thing for them that we can't.

as we have also discussed before, the majority of the most valuable companies we have funded have been started by founders in their 20s (with exceptions like stripe on the lower side and zenefits on higher side).


In some fields, especially science, you probably need to be 10-15 years in to start something, so 30's and up. The question: what is the age stats if you do not calculate science oriented startups.

If (as I think the case is) YC doesn't discriminate on age at all, one might still expect a lower acceptance rate for higher ages. Allow me to explain why.

As much as we hate to admit it, our brains (like the rest of our organs) start to decline in our 20's.[1] We're not quite as quick as we used to be. We can't work as hard. We take a little longer to learn new things. In almost every discipline, crystallized knowledge more than offsets this loss of fluid intelligence. Eventually though, age catches up with us. Combine that with YC's very high bar (accepting the top 3% of applicants), and even a slight dulling due to aging would cause a discrepancy.

1. http://lesswrong.com/lw/4gi/age_fluid_intelligence_and_intel...


> even a slight dulling due to aging would cause a discrepancy

Only if you completely discount the value of experience.


Not at all. I addressed that:

> In almost every discipline, crystallized knowledge more than offsets this loss of fluid intelligence. Eventually though, age catches up with us.

At some point, gains from knowledge can't make up for decline in cognitive ability. For some disciplines, it usually happens sooner (mathematics). For others, it's typically later (biology). For startups, I bet it depends heavily on what the company is doing. But at some point, age makes one less likely to be in the top few percent.

Also, while experience is valuable, the value of experience is probably sub-linear. When learning a new skill, most important knowledge is acquired early on. After that, returns diminish.


Some of this discussion might have to be revised if the startup is active in a health or scientific field. For example, people entering the Chemistry field with a PhD tend to be 30+, particularly if they post-doc. I would imagine if fully-trained (i.e., post resident) MD's are required, you might see people approaching 35+. So, while they're on the wrong end of the power curve, they're what you have to deal with.

Fortunately, for the startup incubators, there are few scientific endeavors that lead to a useful exit strategy. Unless, of course, you're trying to start a biotech or pharma startup. And at that point, you're 1-2 orders of magnitude more expensive as physical plant and even regulatory approval start to enter the picture. Medical endeavors are probably information based (deal with privacy and patient access issues, hence MD's) which is easier to handle. If they're medical devices, you'll face approval processes, I think, similar to pharma. Possibly faster in that there aren't so many steps in the clinical trial process - devices tend not to have a tox problem, for example.


Uh, (citation needed)? As far as I'm aware, the scientific community is pretty convinced that the adult brain has a large degree of neuroplasticity [1] and can retain a significant amount as we age [2]. It's also been shown what we refer to as "fluid intelligence" can actually be trained and optimized for [3]. I would argue that until real, systemic cognitive decline begins occurring (let's say ~65), any differences in an individual's ability to learn new things and retain facts based on age can probably be more strongly attributed to their internal biases from their pre-acquired knowledge than an inability to learn.

Maybe in some part I'm arguing semantics ("either way, what appears to be a decline still occurs"), but to say that cognitive decline happens sooner when related to aptitude for maths as opposed to aptitude for biology is a little silly without a citation. Would love to read these articles if you have resources.

It's probably a mistake to write off an older adult's ability to learn. I would lean to heavily favour Elon's anecdote [4], "I do kinda feel like my head is full! My context switching penalty is high and my process isolation is not what it used to be. Frankly, though, I think most people can learn a lot more than they think they can. They sell themselves short without trying." --- more knowledge means you have the ability to be more discerning which can be more computationally expensive (and take longer), but it doesn't necessarily offset ability to learn. In fact, you could probably make a convincing argument that with proper pattern recognition you can learn the right things more quickly.

1. Plasticity in the human cerebral cortex: lessons from the normal brain and from stroke. http://www.ncbi.nlm.nih.gov/pubmed/15070490

2. Plasticity of the human brain - “We never use the same brain twice” https://www.mpg.de/38899/hm09_PlasticityHumanBrain_basetext....

3. Improving fluid intelligence with training on working memory http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2383929/

4. https://www.reddit.com/r/IAmA/comments/2rgsan/i_am_elon_musk...


If you want something more substantial than the diagram in my earlier comment, I recommend When does age-related cognitive decline begin?.[1] The figures on pages 11 and 12 are depressing, but the study seems pretty sound.

Many of the replies to my comments are misconstruing me. I'm not saying that older people are useless, or that they can't learn, or that they should be discriminated against. I'm simply saying that if you select the very best and blind yourself to age, your sample is likely to skew younger. This is due to the effects of aging, including the aging brain. Nobody contests this when it comes to sports, including mental ones such as chess. Why should it not also be true for startups?

1. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2683339/pdf/nihm...


That's blatant ageism (and a very common misconception, too).

Why sample as opposed to looking at the whole population? The typical reason is, "because it's difficult to survey the entire population". But in this case I'd think that it'd be easy because I assume you have all the data.

They likely don't record the team composition, so this would require hand-labeling of any data points.

Do we know total applicants? I'd like to know what n produced a 5% sample.


It would be interesting to see the proportion of foreign entrepreneurs (applied from abroad).

Sam answers this in a different point in this thread -- https://news.ycombinator.com/item?id=8889333 (40% are born abroad)

Socioeconomic diversity is the elephant in the room.

How could YC even be able to calculate that?

It's pretty easy to figure out what class people are in if you actually care to know.

> Competition is extremely high to enter Stanford, with an admit rate of 7.2 percent. And those who make it in are, as a group, financially better off than the rest of the American population. According to Director of Financial Aid Karen Cooper, the median family income at Stanford is approximately $125,000; by contrast, the median family income in the United States in 2008, the last year for which data are available, is $61,521.

http://www.stanforddaily.com/2010/08/12/diversity-remains-on...

http://en.wikipedia.org/wiki/Household_income_in_the_United_...

Now try to guess what class these people were born into: http://www.ycombinator.com/people/

I'm still a huge fan of YC (very much not a hater) but on this rather large issue I think they're on the wrong side of history. A truly good actor compared to most, but far short being egalitarian.


I very much doubt that YC is in a position to ask every applicant for proof of their family's income and assets , another factor to note is that many applicants are from overseas.

I guess you're right, it's simply not possible and they should just not worry about it.

How about the non-hispanic, non-black minorities?

Or even more interesting to me, how about the distribution of education level and background?


Yeah, I find a lot of software organizations it's becoming more and more a disadvantage of being white. Not sure I can complain, it's worked out for me, but I worry about my kids.

All evidence points to the contrary. 34% of Facebook's employees are Asian, but only 19% of managers are.

NYT article: http://bits.blogs.nytimes.com/2014/06/25/facebook-mirrors-te...

Facebook's "Equal Employment Opportunity" report: https://fbnewsroomus.files.wordpress.com/2014/06/facebook-ee...


Data on citizenship would also be interesting. How many founders are foreigners and how many are US citizens?

for last batch, more than 40% of the founders were born outside the US.

we didn't look at this here because we fund a lot of asian and indian founders, and we just don't have enough of a sample to look at e.g. native american founders.

on things like education level, country of birth, age, etc we've previously looked at and discussed the data and it seems very diverse.


What about someone who's mixed?

Michael, I think you should really ask yourself: what good can come of this? It doesn't really demonstrate anything about YC's fair-opportunity goals, and it opens the topic up for significant scrutiny--both by rights activists, and now (even worse?) by stats majors who are always looking for an opportunity to tear apart someone's oversimplified models.

>Michael, I think you should really ask yourself: what good can come of this?

I got this one!

The good that comes is you solve the current status quo, where on a thread like "Why silicon valley works" people make comments and genuinely feel like this: https://news.ycombinator.com/item?id=8552487

>If you are a white male, in your early 20's and went to Stanford, go for it. If you're not, don't even try to play the game; it's rigged to use you, not help you.

or up-thread someone else

> Complete and utter bull. Maybe in [YC's] world, sure! White? 20-something? No-kids? Harvard/Stanford? B2C? Let’s chat! > Early-40's/Late-30's? Not-white? Married w/ kids? Not-Harvard/Stanford? B2B? Nope. (Don't bother to apply to YC either).

Theirs is a strong argument when 0.0% of founders are black. When it's 4%, man, 1 in 25 is better odds than 80% of the things you're doing to build a $100M exit anyway. You'll go for it and apply.

I gave up as I'm in a hurry, but search 'Hackernews search' (hn.algolia.com) for white/asian, "bother to apply" and similar words - there are a ton of comments by people who think they shouldn't because it's 0%. They should!!!

This blog post is totally great, and a bit surprising. Kudos.


The question is, would this post still be made publicly, in the same manner, if the data actually found serious discrimination happening?

Who knows, maybe YC is actually divinely meritocratic and there are no biases. Maybe they are transparent enough to make their biases public if they were discovered, and I absolutely admire that YC takes the initiative to check.

But if we were to be scientific about it, having YC take a look at discrimination within YC wouldn't be considered a reliable experiment that you could draw any conclusions from.


"The good news is that there is no disadvantage to applying to YC as a female or minority founder."

I don't understand why that's good news.

It'd be good news if YC were actively preferring female and minority founders. Everyone acknowledges the glaring diversity problem in tech, but apparently YC is happy standing on the sidelines.


From the OP:

> "We want to be sure we’re not discriminating against founders in our funding decisions at YC"

This is not an affirmative action program. YC is an accelerator, not a charity. It primarily seeks long term positive returns, and to do that you have to evaluate people and their ideas based on merit, not their gender. Having more female founders would probably be better for the startup ecosystem as a whole, but "actively preferring" a class of candidates based on attributes poorly correlated with their probability of success would be bad for YC.


This is not an affirmative action program. YC is an accelerator, not a charity.

Let's not inject FUD about affirmative action into this conversation.

MIT, for instance, has a robust affirmative action program and I can assure you the Institute isn't in the business of picking unqualified applicants. One of the most famous (and successful) affirmative action programs in the world is implemented by the NFL — hardly a charity.

http://en.wikipedia.org/wiki/Rooney_Rule

Institutions that implement affirmative action do so selfishly to ensure that they're fielding the best proverbial team possible.


Look, whether we like it or not, VCs prefer white or asian founders over black or hispanic founders. That's fact, not conjecture [1]. I understand that YC is an accelerator. It's also a cultural institution. Blacks and hispanics are at a disadvantage when raising because VCs DON'T evaluate strictly on merit; they evaluate on a whole host of characteristics, many of which are subject to bias.

The YC stamp of approval is a valuable badge when fundraising. I'd like to see them find a way to use this to combat the inherent bias that women, blacks, and hispanics face when fundraising.

[1] https://www.cbinsights.com/blog/venture-capital-demographics...


The YC article suggests that black and hispanic founders are slightly more likely to be funded by YC than other ethnicities. 3% of applicants to the batch were black or hispanic but 7.7% of funded founders in the batch were black or hispanic.

The article you linked does not seem to adjust for demographics of the applicants, it only discusses those who were funded.


> Look, whether we like it or not, VCs prefer white or asian founders over black or hispanic founders.

Something something correlation, causation...

I mean look at the YC numbers here. When 90+% of your applicants are white or Asian that's going to be reflected in the final numbers unless you actively favor a certain group of applicants over another.

We're talking about greedy VCs. Not a single one of them is listening to a pitch and going "man this is a billion dollar idea, but they're _black_, ugh..."


Why should they prefer different genders or races? What was that thing called... oh yes, equality.

Who gives a shit? YC should fund good ideas with good founders regardless of skin color, gender or whatever else.

> I don't understand why that's good news.

This is because of "implicit bias", wherein people discriminate against women and/or minorities unconsciously and unintentionally.

For example, here is an article demonstrating that orchestra auditions were subject to bias against women:

http://scholar.harvard.edu/files/goldin/files/orchestrating_...

I suspect that everyone (or nearly everyone, anyway) evaluating the auditions wanted the best musicians regardless of gender, but the linked study determined that when orchestras began using blind auditions, more women were offered positions.

In YC it is impractical to conduct "blind interviews", so that it is good practice for the partners to ensure that they aren't accidentally disadvantaging women and minorities.

Whether YC should actively prefer female and minority founders is, I believe, much more subjective.


Good to know that YC doesn't seem to be affected by biased selection process. The low overall percentage of minority or female founders is a reflection of the current state of society, which, alas, isn't where it could be.

The post does not add any value -- "you are not at a disadvantage" means we are evaluating everyone with the same scale. Why should this be a surprise? this is the least we would expect anyway.

It would be interesting to have some more insights, like how many of rhw minority groups ( black or hispanic as the article says ) were funded, also more meaningful like:

Race/Sector of interest Age/Sector of interest Gender/Sector of interest People with Higher Scholar degree/ Sector of interest Startups funded by sector


> this will not be a zero-sum game

Is there some kind of rule at YC that mandates that blog posts must insist that nothing is a zero-sum game?


yes! It is the same rule that says that we must be sad that 50% of all applicants aren't women.

It's not like the number of entire YC 2015 winter applications is super large. Why take a sample?

I really want to see what percent introvert founders take.

I have been reading articles saying introversion is not good for entrepreneurship. I wonder if this is true.

if it is true, will VCs treat introvert founders like they do to minorities?


Some types of discrimination are not "fashionable" enough. If you are not of right color/gender you are out of luck as a minority.

the big question is of course about self selection: how many nonwhite, female and non-Stanford entrepreneurs would like to apply for YC but don't because they think their chances are basically zero?

Legal | privacy