No really, what’s the expected value of a startup?

In the effective altruism community, there’s a lot of interest in tech startups, because they look like an especially promising avenue for people who are earning to give. This is because if your reason for wanting money is to donate it to charity, the marginal value of your money declines slower, so it makes sense to take greater risks. But no one seems to really know what the expected monetary value of doing a startup is. I’ve been thinking a lot about this question lately, and I want to share my thoughts.

First, let me emphasize that I’m asking a question that no sane person outside the effective altruism movement cares about. If you’re interested in making money for selfish reasons, expected value is a misleading metric for judging startups, because most of your expected value comes from your tiny chance of becoming a billionaire, and when you just want to spend money on yourself, the difference between $0 and $20 million probably matters a lot more than the difference between $20 million and $1 billion.

But hypothetically, if you did want to be the next Bill Gates, not just in money he made but in the donating it to charity part, how good is a startup for that? Carl Shulman’s blog post cites an expected of about $2 million per year for companies that get venture capital funding. As I’ve learned more and more about startups, though, that number has come to seem less and less meaningful.

“Venture capital” generally refers to money from companies whose entire reason for existence is to invest in fast-growing companies. Startups that are just starting out tend not to get venture capital. Rather, they’ll get a small amount of money, known as seed funding from a starup incubator like Y Combinator or “angel investors.” “Angel investors” being random ridiculously rich people who want to invest in a startup for whatever reason. Y Combinator has a reputation for being selective, but my guess is that it’s significantly easier to get money from an angel investor than from a venture capital firm (not that getting money from an angel is easy).

So how hard is it to get venture capital funding? Ryan Carey tried to answer this question and came up with an estimate that about 1% of companies that aspire to venture capital funding eventually receive it. That may be roughly accurate, but I doubt it’s very relevant for tech startups in the Bay Area. Being in the Bay Area seems to be a huge advantage for startups, and they tend to be founded by people who could be making six figures(ish) at a regular job by having technical skills few people have.

Okay, any better ways to estimate the expected value of a startup? Ryan has found data that founders from the first five years of Y Combinator now have net worths averaging $18 million, after an average of 7 years of work. But that might not be predictive of future success, because of regression to the mean. Also, Y Combinator accepts 2.5% of applications.

Okay, what does that mean in practice? Forget about regression to the mean for a second. On that assumption, getting into Y Combinator looks a lot like winning $18 million. Sure, it takes 5-9 years of work, but even if you worked at Google, lived frugally (as startup founders are often forced to do), and bought index funds, you’d be unlikely to accumulate even $1 million in assets. So maybe getting into Y Combinator is like winning $17 million, at least.

Your chances of getting in are only 2.5%, if you don’t know where you stand among Y Combinator applicants. But you can apply every six months. Let’s make believe the probabilities on each round of applications are independent. That’s an expected value of $850,000 per year. Pretty good, if you don’t care about risk, which you should if you’re a sane person who isn’t building their life around trying to imitate Bill Gates. But we threw out the “normal, sane person” assumption long ago. On with expected value.

Is it remotely sane to pretend we don’t know where we are among Y Combinator applicants, and the chances of getting in cycle to cycle are independent? Obviously not… except it might give us as close an estimate to reality as anything will. At least, conditional on you having no obvious disqualifiers from getting into Y Combinator (for example you have the right technical skills, or at are a really smart business type with a technical cofounder). Let me explain.

You might think getting rejected from Y Combinator would make you much less likely to get in on future rounds. But I think that’s actually not true. I’ve talked to a lot of people in the Bay Area startup scene, including two Y Combinator alumni, who tell me plenty of startups get in after being initially rejected. Some get in on like the 5th try. Getting rejected probably does give you evidence that you’re not cut out for startups, but only weak evidence, and it’s probably counterbalanced by learning from experience as long as you keep trying.

But what about that regression to the mean thing? Let’s start by setting a lower bound. Ryan puts the lower bound at $100,000/year, based on how Y Combinator values a startup upon giving them seed funding. But it has to be a lot more than that, because Paul Graham and the other Y Combinator partners wouldn’t invest in such risky companies if they weren’t getting a good risk premium.

I mean, okay, maybe Y Combinator’s early successes, like Dropbox and AirBnB were flukes, and those flukes lead the YC partners to irrationally overestimate their expected return. Still, it seems unlikely that the YC partners are being completely stupid. So what’s their risk premium? One random article I found online by Googling said venture capital firms often want a risk premium of 70%+, and 40% is the low end. Let’s go with that.

Those numbers are per year. Multiply them out over the average of 7 years, and if you use the 70% figure (starting with the initial equity value of $700,000), and you end up around the original average net worth Ryan found of $18 million. Use the 40% figure, and you end up with a smaller but still high figure, around $7 million, or $1 million per year. Even if the Y Combinator risk premium is worse than that, though, it seems unlikely that the partners are being so stupid that the expected value to founders ends up less than just taking a salary at Google.

The “at least it’s better than just taking a salary at Google” conclusion is also where we end up if we just throw out Y Combinator’s three biggest successes, which Ryan says account for 7/8ths of Y Combinator founder earnings. Then you end up with an expected value of a little over $300,000 per year, instead of $2.5 million.

I could go on in this vein for some time, finding various ways to adjust Ryan’s numbers on Y Combinator based on various assumptions. But three remaining points seem worth emphasizing:

  • Y Combinator, by all accounts, does provide significant value to startups they accept. But it would be wrong to think that the value of startups that just barely fail to get into Y Combinator is 0.
  • Conditional on having some of the relevant skills (e.g. programming ability), there seems to be wide agreement among the people I’ve talked to that if you don’t know enough to build a successful startup, one way to learn is do a startup and learn by trying.
  • Economist Noam Wasserman, in his book Founder’s Dilemmas, presents evidence that startup founders face tradeoffs between wealth and control. This suggests founders whose goal is money can increase their expected value by being more willing to give up control (e.g. to VCs) when necessary.

In general, conditional on your being the kind of person who could get a job at Google, my guess is that the expected value of doing a startup is less than $1 million per year, unless you’re something really special. But it’s probably significantly more than your hypothetical salary at Google.

Again, that’s not saying much, because I’m ignoring risk. Have I mentioned that normal, sane people should not take on the risk of starting a startup?

On top of this, there is one final consideration that might tip the balance in favor of not doing a startup, and that’s if you can make a salary of $200,000 per year. That gets you accredited investor status. Why does that matter? Well, the entire argument for effective altruists doing startups is so that you can get the risk premium from doing a startup. But an accredited investor should be able to get that by investing money, rather than time, in startups.

Historical thinkers on human misery
Self-interested activism and gay rights
How bad is life at subsistence?
Effective Altruism and feminism