Some choice quotes from his Google+ post: "The lottery is not a tax on people who are bad at math. It's trading a little bit of expected-value for a lot of variance."
"There are many other cases in real life where mean/variance tradeoffs arise, and there are perfectly plausible objective functions under which it makes sense to follow a policy that doesn't maximize expected value.
The obvious example is insurance, where you pay a small bit of money each year to reduce the chance that you end up suddenly needing to pay a lot of money. Same with forming or joining a startup; on average you'll make less money than if you take a "normal" job, but there's some small chance that you'll have just joined the next Apple or Google and you'll end up rich.... "
One common assumption is that utility is logarithmic in money, i.e. that you become "one unit happier" not by adding $100K but by doubling your income. This matches actual human behavior fairly well, and implies that you should buy insurance (losing everything is really bad) but not lottery tickets (winning $BIGNUM makes you less happy than you'd think).
Actually, according to this article, certain kinds of insurance are less effective at optimizing happiness, because we are psychologically equipped to adapt to new (bad) circumstances:
Obviously, certain kinds of disaster pose a greater threat to happiness than others. For example, health insurance may be a worthwhile investment for some, while property insurance may not.
People in Colin's circles are probably not the desperate, poor, uneducated lottery players that opponents of the lottery are concerned about. Colin's right not to judge his friends who choose to entertain themselves with the lottery fantasy every so often. They can afford it.
Yeah, when you think about it the lottery and insurance are surprisingly similar ideas, though a bit in reverse. Getting in an accident and winning the lottery are both unlikely. By buying insurance you remove the small chance of a large loss. By playing the lottery you create the small chance of a large win. Both reduce your expected value, and doing both even more so. If you buy both, your variance is still high but in the positive direction.
>Yeah, when you think about it the lottery and insurance are surprisingly similar ideas, though a bit in reverse. Getting in an accident and winning the lottery are both unlikely.
Getting into an accident is far more likely. And everyone has health issues sooner or later, so the probability of that is 1.
The probability of being hit by lightning is massively diluted by the fact that most people don't live places where lightning is a serious threat. I've heard it's actually a serious risk some places.
> on average you'll make less money [in a startup] than if you take a "normal" job
I disagree. If he's using "average" in the sense of "mean" (not median), then he's saying that starting a startup has a lower expected return and is higher risk than working a normal job, meaning it's inferior in both regards (assuming diminishing marginal utility). Entrepreneurship is a classic "high risk, high return" venture. It has a higher expected return than a conventional job; that's what makes it financially attractive to founders and VCs in the first place. The astronomical success of the minority offsets the lack of success of the majority.
In a similar vein I disagree with his assertion that joining a startup (taking a risk) is like buying insurance (avoiding a risk). They are opposites.
Do you have any proof that, on average, startup founders make more money than people employed in regular jobs, once you take into account all of the ones that make no money and then collapse?
The mean YC startup about a year ago [1] $22.4 million, let us assume 3 founds on average and 3 years of work, and that they hold 1/3 of the value, and you get an average return well north of 500k a year per founder.
And most of that money doesn't actually go to the founders - YC owns a chunk of them, and so do other investors (as PG says, 94% of them get funding from elsewhere too).
That's a good question. Would be interesting to see if someone can dig up any studies on this. e.g what % of the country's wealth is owned by founders (and their heirs), divided by the % of people who are current or former founders.
My argument is founded not on experimental data but rather on the axioms of utility theory. If startups had higher risk AND lower expected payoff, it would make no financial sense to found or invest in one, since that would be a textbook example of a bad investment:
There is a massive difference between founding a startup and investing in startups.
Professional investors (VCs, Angels, Incubators) choose the best startups (most efficient choices) and do not fund the rest which eliminates a huge number of founders who simply fail and get zero return for their work. On top of that, there are a number of VCs and angels who are completely unsuccessful picking startups and end up out of the business unable to raise additional institutional money or who lose their additional investment capital.
The expected return of investing in startups is very likely negative if you add all the smart and non-smart investors together. (I don't have data, and can't even imagine where to get it, so I'm saying "likely".)
Using the security market line to prove your point isn’t really fair since not much research has been done on applying it to startups. One of the major issues with applying the SML is the fact that it is very hard to calculate beta of startup investments since they have very long term time horizons. SML has more to do with portfolio construction and how you can get a portfolio which doesn’t deviate from the benchmark too far while still producing Alpha. (Alpha is active return). The Tryenor ratio is more about making sure public equity managers are not doing really crazy stuff.
Overall my point is that there is no evidence that suggests the expected return from investing is startups in net positive. Whether founding or working for a startup produces positive expected returns or not is also very hard to know, but I would estimate that working for early stage startups for lower than market pay has a negative expected value even after considering the possibility of huge exits (a lot of companies fail that you have never heard of).
Where I think startups have a much higher expected return comes with the experience and growing up you can do while working in a small company. When you calculate total earnings over a lifetime for people who join even failed startups early on, I would not be surprised if they made more and had more enjoyable lives, even if they ended up at big stable companies later. When you add the possibility of a big exit to the higher future earnings I think the expected value goes net positive.
I've rarely heard people try and sell startups as a way to make oodles of cash. It's more frequently sold as a way to be your own boss. As such I'd expect it to attract people of a particular personality type, who want to drive their own ideas forward and do things their own way, without anyone telling them what they can or cannot do. As such I'm not convinced that it is always a rational decision (or, at least, not an entirely rational one - there's definitely emotional input there).
Looking at all of the people who applied to YCombinator and seeing how many of them have successful businesses a couple of years later would be fascinating.
>Entrepreneurship is a classic "high risk, high return" venture.
Sure, for the ones investing capital. For the workers, I can't imagine startup's mean pay being anywhere near as high as a "normal" job. A mean will cut out the Zuckerbergs, Serg, Brin, etc. Also, remember you can't compare mean startup salary to mean "normal" job salary since the bulk of startups are in the valley. To get a closer estimate you'd need to give the same weight to "normal" jobs in a given area as number of starups they have there (e.g. lots in SV, some in NYC, etc.).
>It has a higher expected return than a conventional job; that's what makes it financially attractive to founders and VCs in the first place.
In the exact same sense than a lottery has a "higher expected return that a conventional job", just with slightly better odds.
>The astronomical success of the minority offsets the lack of success of the majority.
That doesn't make sense. Except in the light of something lke the Dunning-Krueger effect, where most people involved in startups expect they will be that "minority".
The point of a lottery is to generate income for the government/company running it, which would not be possible if players collectively won more than they spent.
tl;dr there are circumstances where the expected value is greater than 1, due to the accumulation of prize money when nobody wins several times in a row.
In aggregate, all lottery runs put together take in more than they pay out. But an individual lottery run might be -EV for the state if the jackpot has grown enough; however, that only happens after no one wins the lottery for weeks on end, at which point the state has run even bigger profits than before. Once someone wins that big jackpot, the state may pay out more than it took in that particular week, but still less than it took in for the several weeks leading up to the jackpot finally being won.
"Some choice quotes from his Google+ post: 'The lottery is not a tax on people who are bad at math. It's trading a little bit of expected-value for a lot of variance.'"
If you want high variance, wouldn't you then want to make high-risk investments in the stock market? I'm not very knowledgeable on the subject, but my understanding is that the expected value rises with increased variance on the stock market, rather than decreasing.
https://plus.google.com/107814003053721091970/posts/6sPpxMo6...
Some choice quotes from his Google+ post: "The lottery is not a tax on people who are bad at math. It's trading a little bit of expected-value for a lot of variance."
"There are many other cases in real life where mean/variance tradeoffs arise, and there are perfectly plausible objective functions under which it makes sense to follow a policy that doesn't maximize expected value.
The obvious example is insurance, where you pay a small bit of money each year to reduce the chance that you end up suddenly needing to pay a lot of money. Same with forming or joining a startup; on average you'll make less money than if you take a "normal" job, but there's some small chance that you'll have just joined the next Apple or Google and you'll end up rich.... "