Why you maybe shouldn’t start (or invest in) a startup

The Effective Altruism community is quite enthusiastic about startups. 80,000 Hours lists “tech entrepreneurship” as one of their top four careers, notably beating out working as a software engineer for a big company–which requires a similar skill set. Benjamin Todd of 80k has also advocated startup investing as good use of money for people who want to make the world a better place. But lately, I’ve come to suspect startups are being oversold. Here’s why.

About a year ago, a friend of mine was raising seed investments for a startup. I chipped in a small amount of money, he applied to Y Combinator (the startup incubator responsible for AirBnB and Dropbox), got rejected, and the startup promptly went bust within a year. This in itself was not surprising–I knew going that most startups fail, and the reason to found or invest in a startup is for the small chance of a mega-success.

What shocked me was this: while I’d heard that in previous rounds, Y Combinator would interview 10% of applicants and accept 2 or 3%, in the round my friend applied to they only interviewed 3% of applicants. This is in spite of the fact that number of people accepted to Y Combinator grows every round–apparently, the applicant pool has grown even faster. I’ve also heard independently that Y Combinator thinks its quality of applicants has gone up over the years.

In a way, this isn’t surprising. Once everyone knows a career path is lucrative, talent tends to pour in, making the competition much stiffer, and the career path less attractive. Incidentally, 80,000 Hours has written about this too recently. It’s a problem that’s infamously afflicted doctors, lawyers, and college professors (the last of these, while not financially lucrative, offers a pretty nice life to those who can get tenure–yet getting tenure has gotten much harder for precisely that reason).

One thing that makes the situation especially grim for would-be startup founders is that not only has the chance of getting into Y Combinator gone down, the chance of being successful without Y Combinator’s backing may have also gone down. YC partner Paul Buchheit has claimed that YC rejectees rarely go on to become successful, and he thinks its not because YC is so good at picking successful startups, but because Y Combinator’s resources give startups it incubates such an overwhelming advantage.

The key factor may be that once everyone knows all the good startups come from Y Combinator, it becomes much harder for non-YC startups to attract investors and talent. Economist Robin Hanson has, in fact, argued that this “elite evaluator” role may be the main way that Y Combinator, top later-stage VC firms like Andreessen Horowitz, and schools like Harvard make their money. This model should lead us to predict that the startup biz will continue to be good for YC and Andreessen Horowitz for the foreseeable future–but less of a good deal for founders and random individuals who happen to have a lot of money to invest.

I think it’s worth being explicit about how this situation looks from an investor’s perspective. In a world where Y Combinator dominates the startup ecosystem, and getting in is insanely competitive, the would-be angel investor faces the following dilemma: On the one hand, you can invest in a company that hasn’t gotten in to Y Combinator yet, and then have terrible expected return because getting in to Y Combinator is very difficult and the startup won’t be successful otherwise.

On the other hand, if you try to invest on YC’s Demo Day, other investors will be tripping over themselves to invest in the same startups you want to invest in, so you’ll be forced to invest at inflated valuations that also leave you with a terrible expected return. Indeed, I’ve heard Silicon Valley investors complain that YC’s Demo Days are feeding frenzies where it’s impossible to get in on the good deals.

Again, I’ve heard people say that it seems like these days in Silicon Valley, everyone is talking about their angel investments. But if everyone is doing it, it’s probably not an exceptional investment opportunity, at least anymore. As Matt Yglesias said in another context:

By the time the word is out that stocks are a great investment, or that hedge funds can get you 8 percent returns, or whatever else it is it’s probably already too late. Hedge funds had their best moment before they got famous. Once everyone’s rushed in, it’s too late.

I expect many people to dismiss what I’m saying, as it conflicts with Silicon Valley conventional wisdom. Perhaps they’ll cite an old Paul Graham article, or even hard data on the amazing returns to startup investors. The problem is that I’m not claiming startups were never a good deal–just that now that everyone “knows” they’re a good deal, that fact has made them stop being a good deal. The most rigorously-collected data in the world doesn’t matter if the data is out of date.

For example, consider this piece of conventional wisdom: angel investors shouldn’t worry about getting a cheap valuation, they should worry about making sure they don’t miss out on the one company that will give them ridiculous returns. Well, it is true that if only 1 in 10 of the companies you invest in will be a hit, the difference between 20x and 30x returns doesn’t matter. But if competition for angel deals raises valuations to the point where you’re only getting 10x returns, you may as well go play some high-stakes dice games with your rich friends.

This is not to say that no one should invest in startups. I expect Y Combinator will continue to make money for its partners for some time. Nor is it to say no one should found a startup–maybe you have some exceptional advantage over the great mass of would-be startup founders. Of course, you’d better be careful to ask yourself if your “exceptional advantage” is actually something that many thousands of Google employees and recent Stanford grads also have.

I also think there’s a general lesson here: the EA movement needs to be much more careful about giving career advice that is likely to go out of date rapidly. If something sounds too good to be true–well, maybe it was true five years ago, but it might no longer be true to day. Especially if “everyone knows” it’s a great deal.

Correction: An earlier version of this post misidentified a claim by Paul Buchheit as having been made by Paul Graham.

Update 7/21/2015: Current YC head Sam Altman calls angel investments the biggest status symbol in Silicon Valley, bigger than first-name name drops. This should be a warning sign that if you try to do angel investing to make money, you’ll be competing with people who do it for the status symbol.


14 thoughts on “Why you maybe shouldn’t start (or invest in) a startup

  1. Cool article! Minor correction, though – the HN comment you link to regarding YC rejectees that go on to be successful wasn’t written by Paul Graham, but rather by Paul Buchheit.


  2. If investing in YC startups no longer yields significantly positive returns, one would expect risk-seeking investors to strike much more favorable deals with the startups that are YC rejects – especially if the primary reason those startups are failing compared to YC ones is a lack of investors. If they’re not doing that, there’s money lying on the ground and you should go pick it up.

    From the founder side, this dynamic seems like it’d work in your favor. You have a clear and early indicator of whether your startup will succeed – does it make YC? – and apparently investors so desperate to invest in you that they’re madly undercutting each other. If founders are failing in that environment, it’s by dragging along a low-expected-return project after it’s been rejected from YC or by sinking six months into a bad idea that’ll predictably never get into YC – but at least it’s only six months, whereas five years ago you could sink several years into your bad idea.

    Liked by 1 person

    • Re your first paragraph: unless you’re wealthy enough to keep giving a startup money until its profitable (which may take years), the problem is that without a well-respected investor backing the startup, it may have trouble raising the next round it needs. And it doesn’t solve the problem of lack of prestigious signals -> trouble hiring good employees.

      From a founder’s perspective, yeah you can get a pretty strong signal after six months, but is the expected value you get from those six months of work higher than the $75-$100k or more you could have made working at Google? (And yeah, Google engineers make that much in six months.)


  3. There’s some reason to believe that startups will continue earning high returns even if they’re well-known. Startups are much riskier than most investments or career paths, so even if everyone has perfect knowledge about their expected returns, lots of people are still not willing to invest because they’re not willing to take on so much risk. This is (probably) why stocks have pretty consistently outperformed bonds, even though it’s public knowledge that stocks outperform bonds.

    It’s conceivable that startups are fun or sexy, so people over-invest to the point where they are taking on too much risk for the return, like people did with publicly-traded tech companies in 1999. With the dot-com bubble, there were clear indicators that companies were overvalued (they had unprecedentedly high price/earnings and price/book ratios); but it’s much harder to value a startup that is growing rapidly and making no profits, so there’s no similar safety mechanism.


  4. Most of the newer Y Combinator startups seem pretty uninteresting to me. I think the space of web services you can build an impressive prototype for in 3 months is pretty well mined out. Romeo and I have generated lots of startup ideas that seem interesting to us that aren’t the sort of thing Y Combinator funds.


    • They mostly fund webservices, but haven’t they also funded biotech and nuclear fusion startups? Though from what I can tell the nuclear fusion startup is almost certainly nonsense.


      • There’s an entire economy outside of web services & biotech. Has Y Combinator funded a single startup that opens brick and mortar locations, for instance? Where’s my TDCS parlor where I can rent a brain zapping session?

        I think Peter Thiel is right. Most people are too conformist in their thinking to see zero to one startup ideas.


      • I’m curious to hear your startup ideas that you don’t think Y Combinator would fund (you’re welcome to e-mail me them if you don’t want to talk about them publicly). Also, have you looked up what it costs to rent the space you’d need for a TDCS parlor in, say, Berkeley? I suspect you couldn’t do it with YC’s ~$100k, but I might be wrong.


      • We’ve thought about sharing the startup ideas. Some reasons we haven’t yet: (a) many are in the health/wellness space, meaning MealSquares could conceivably expand to do them, (b) I’m skeptical that people will work hard on startup ideas they didn’t come up with themselves, (c) most take some time to explain (either there is no one-sentence version or the one-sentence version fails to communicate why the idea seems exciting, and either way there’s a lot of additional info I’d want a potential founder to know), (d) most are fairly capital-intensive, and since they aren’t the kind of thing Silicon Valley typically funds, it’s not clear where a potential founder would get the capital.

        I haven’t looked in to the cost to rent a TDCS parlor, no. I did buy a high-end TDCS device and have the subjective experience of semi-permanent mood improvement. But I also experienced some side effects (temporary sleep disturbances, likely due to using the device before bed, which I later learned is a bad idea because it’s stimulatory in nature). I’m happy to lend/sell the device to anyone who’s interested.

        I think the legal hurdles would be a bigger issue for TDCS. Doctors are notoriously jealous of anything that looks like someone who’s not a doctor helping people with their health. You’d probably exist in a legal grey area like MetaMed, and you’d have to advertise carefully… “zap your brain” as opposed to “cure your depression”. Since it’s an experimental technology, there are issues of liability insurance etc. Still might be a good fit for an EA with neuroscience expertise. (I guess maybe that’s you?) Also, I’m told that the latest meta-analyses show TDCS doesn’t actually work that well.


      • Maybe a good solution to these problems would be for people who think they can work on someone else’s idea to come talk to me and I can suggest a couple of ideas that might be good fits for them personally. Any EAs who really want a personalized startup idea or two are welcome to email me at dreamalgebra on gmail with info on their interests, skills, resume, EA involvement, ability to get capital, and thoughts on how motivated they’d be working on someone else’s idea and I can take a quick look through the idea list (maybe 40 ideas long?) for them.


  5. I think investing in a startup is very different than founding a startup. I would agree that investing in startups typically doesn’t make much sense, but think that’s a very different question than founding a startup, which is what 80k discusses.

    One strong benefit of YC is that they may be able to help founders by rejecting ideas that would have failed anyway. If not getting into YC is a strong indicator of later success (and this appears to be true), then the information value of the choice of your acceptance can be quite significant. I really don’t buy an argument here that YC is bad for startups as a whole, but it does seem to be doing a good job at selecting ones that will be successful (which is very useful).


  6. > in the round my friend applied to they only interviewed 3% of applicants

    As an analogy, top technology companies hire less than 1% of job applicants. This doesn’t mean that software development is hopelessly competitive, it just means that every jackass with a resume and some free time spams recruiters at Google.

    Of course, with software development you can create a reference class and then ask e.g. what fraction of CS grads from top colleges get offers from Google? With startups there isn’t an easy way to create a reference class of “possibly good startups”, so it’s harder to figure out what fraction of possibly good startups actually get into YC. I would still wager that your odds are much better than the <1% you imply.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s