Hacker News Reader

Exactly. There are other problems with the data as well, but the big problem is that he's talking about averages in an industry defined by outliers.

You should start a startup if you think you have it in you to revolutionize an industry, not because someone's spreadsheet says you will earn an average of $X/year doing it. The median founder should expect to earn a lot of valuable experience, and very few dollars.

Definitely true that the median can't expect to earn much. In fact, looking at the data, we can see that about 70% exit with only their salaries (typically under $100k).

Like we say in the summary, since all the value is in the tail, an individual should be thinking in terms of 'what are my chances of creating a new DropBox or AirBnB' i.e. revolutionising an industry.

The problem is, no-one knows who exactly that's going to be.

This post shows that getting into YC provides some useful information. It shows you're now in a reference class of ppl who can be expected to earn millions of dollars per year (although the median is still zero).

Beyond that, there's lots more to consider when judging whether to become a tech entrepreneur e.g. are you exceptionally technically able, gritty, able to accept risk, in possession of a potentially huge idea etc. If you got into YC but it's clear you'll never be Drew Houston, then your expected outcome is probably still pretty low.

If you're attempting to earn money to donate to effective charities, and you're risk neutral, it might make sense to consider average rather than likely outcomes.

It's not just the donations. It seems likely startups produce lots of economic value by developing new products, and the extent to which you do that is somewhat correlated with how much money you earn. That's another reason to consider the non-risk-adjusted average outcome.

What are the other problems with the data? We're keen to hear.

> Why is the piece talking about averages, when wealth is distributed as a power curve?

80,000 Hours is explicitly focused on maximizing social good, often by advising people on what career will maximize their "earning to give" potential. That is, earning a lot of money to donate to an effective charity (e.g. high rated charities on GiveWell).

Because of this, their analyses tend to look different than it would if it was optimizing towards a risk-adverse individual maximizing personal utility.

The justification for looking at expected value (averages) is expanded on here: http://80000hours.org/blog/12-salary-or-startup-how-do-goode...

The key assumption with this line of thinking is that a model built on a population who don't base their decisions on 80000hours principles will perform well on those who do.

That seems unlikely to true if a passion for fixing a particular problem is both a significant factor in predicting a startup's success and unlikely to be present in one who makes decisions by modeling the value-maximizing way to live one's life.

That's a great point. If you read this article and think "Now I'm going to apply to YC and do good!", you've probably got lower chances of success than the average YC founder.

That's one reason we point out that the expected value mainly depends on your chances of being the next Dropbox or AirBnb, which probably involves having a huge passion for fixing a particular problem.

On the other hand, I don't think the information should have no impact on decisions. For instance, I think someone at college could read this, and set themselves on a path towards becoming the type of person who makes a good entrepreneur. They could learn to program, hang out around startup people, try out their ideas. They could find a problem and become passionate about it.

This is interesting, I think you're saying that this advice could be for utilitarian good, ignoring the individual.

From a purely utilitarian perspective, maybe the startup compensation curve is best for philanthropy.

Hypothesis: 1,000 people making $100,000 will give significantly less in aggregate than 1 person with $100,000,000. Therefore, if more people go into a corporate ladder job for the $100k, there will be less total giving.

There's quite a bit of data on the distribution in the post as well. Quoting from the summary:

Most of the returns have gone to a tiny minority of super-successes. The founders of AirBnB, Dropbox and Stripe are worth about US$7 billion, about 78% of all founders’ equity, although they account for 0.5% of the companies.

Outside of the most successful companies, it was still possible to earn significant returns. 12% of companies from the first five years of Y Combinator are now worth US$40 million or more, and a further 10% have sold for US$5-40 million. The remainder probably earned little more than their (low) salaries.

I agree, I am not sure there is really that much interesting in this article. I'd love to have the author re-write it looking at the distribution, rather than averaging in returns for the top companies which makes up less than 1% of YC companies.

It would be interesting (but unfortunately not practical or possible) to have two groups.

Members of both groups are accepted to YC but then randomly selected as to who actually goes with YC or not. [1]

Of course just knowing that you have been accepted to YC (or telling others) would no doubt skew any outcome. Just like knowing you were accepted to "a Harvard" and/or telling people you were will skew how you are perceived.

That said there certainly could be a study (sans the halo effect) done of startups, as a group that applied but were rejected by YC vs. those that were accepted.

Sounds like something that mattermark could potentially take on. Or maybe priceonomics in a blog post.

[1] Then, years later, compare outcomes.

Obviously we aren't going to intentionally reject good companies in order to perform some kind of study, but we do our best to keep track the ones we reject that later become successful. There are a couple of them, but so far they are an order of magnitude smaller in both size and number than the successful companies that we did accept.

Taking the concept in a different direction:

You currently have people you accept and people that you reject.

You recognize that many of the people that get rejected might actually get accepted if the pool of companies you could manage could be larger. (Similar to how many people are rejected by top schools but the schools recognize that there are people that qualify but miss because of some arbitrary decision process). (After all the class size is pretty much fixed even when the applicant pool rises).

So you create a third class which is, for lack of a better way to put it, "worthy runner up". Like having a silver or bronze medal. (Noting that merely even getting into the Olympics and not getting anything is worth something.)

And then those rejected applicants get some of the halo of YC (which further increases the applicant pool because now it's not solely pass or fail). Plus of course further YC name branding. "While they were rejected by YC they did get a Silver medal and with that..."

Anyway my intuition says that someone branded as "runner up" would do better than someone not given that status.

I think you should seriously consider doing this. It's the best way to measure the true impact of YC on the companies it selects. Otherwise, it's hard to know to what extent you're just picking the most successful teams vs. genuinely making the teams more likely to succeed.

Addition: To avoid having to completely reject people, you could take randomly select half to enter immediately, and let the other half enter in one year. Then you could measure the difference between the two groups after a year.

It's not important know, and certainly not important enough to justify discarding half of our profits. It's also very difficult to get anything statistically significant out of the data when there are so few data points (there's probably only about one DropBox per batch).

Btw, if it were truly just a matter of us picking the right companies, that would be even more impressive given that we reject over 97% of applicants :) That's obviously not the case though, since some of our most successful companies (such as AirBnb) wouldn't exist without YC.

Not saying it's not impressive either way!

Good point that the question is much less valuable from the perspective of YC partners. From that perspective, all that matters is that the companies perform well in the end. It matters more from the perspective of the applicants. Though even then, because the costs of joining YC are low, it's probably not a big issue for many people.

Agree the success rates would be hard to compare. You'd need to have some other proxies for eventual success besides profits.

Excellent point that if you can show you significantly helped the most successful couple of companies (which account for 80%+ of the value), and there were no other large successes among the people you rejected, then the job is done. You've shown YC increases the expected value of the companies you select.

While the statistic is incomplete, for the reasons you say, it's not entirely without value.

For example, if the average return were below the typical developer salary, I'd call it a sucker's bet and say that no one should participate unless it was for the yucks.

Are you sure being accepted into an average YC startup is worse than being accepted into the average non-YC company?

There are more considerations than just money. Quality of life seems like most important. Experience and connections are probably a close second.

In statistics, taking average on any distribution is perfectly valid metric that give you an expected value - even on power curve. For example, let's say you roll a dice and get a million dollar if you get 6 but nothing otherwise. Then your expected return is $1M divided by 6.

I was always interested this number: Sum the personal networth increase of each founder and divide that by number of founder years. That would give expected networth gain per year for being YC founder. Due to outliers, I think it would be significant number and actually makes a case for being YC founder.

One another advantage in calculating this number is that you can compare it with other accelerators (provided you have data for large enough number of founders for each). This would be interesting for rank accelerators from founder perspective (as opposed to investor perspective).

They're doing two averages that lose useful information. You've pointed out one, which I agree with, but the article also averages salary over the life of the company, which I think is losing more useful information.

I'd be much more interested in an average of what the CEOs make in their first year and second year. The Dropbox, AirBnB and Stripe founders wouldn't impact this average all that much since I doubt they were pulling in 7-8 figure salaries from the beginning. And, if someone is thinking of starting a company, they salaries that they're most interested in are those first two years...they want to know just how little they'll have to survive on.

More than two years into a startup, the values might be interesting, but they're not relevant because outcomes vary so wildly.

Exactly. It would be interesting to know the breakdown based on 20 percentile increment.

You can get a good idea of that here, plus the stats in the summary: http://80000hours.org/blog/341-how-much-do-y-combinator-foun...