This is largely off-topic: Speaking as a guy who used to work for a local technology incubator, I really really hate the "most startups fail" factoid because it unintentionally distorts the statistic.
The statistic is that most new businesses will not be going concerns X years after inception. This is true, but it does not indicate failure. I find this to be clarifying: if someone asked you for advice, saying (with wavering voice) "I'm considering taking a... job... where this company, like, pays me money every other week... is this safe?!", would you say "Pfft, most jobs fail"? Because most twentysomethings who start a job today will not be employed at the same job, by the same company, in 5 years.
My second "real job" was with the aforementioned technology incubator. I worked there three years, made decent money and learned a lot, and had a better opportunity when it came to contract renegotiation, so I took it. If I had been a startup at the same incubator, that would have put me in the "fail" column. Is there any sanity in that definition of fail? (Give me another few months and I'll be failing my way forward out of the better opportunity, hopefully.)
Can you clarify your argument? It sounds like you're saying, "sure, most businesses will cease functioning within 2 years, but you get so much out of starting your own company that it doesn't matter".
I'll tell you a story. I started a business, worked about 50 hours/week on average for about a year, and made $12,000. That's 2500 hours at a rate of... $4.80 per hour[1]. When I could've made $15/hour right away at a boring office manager job doing estimates, phone support, and yelling at the men when they leave their tools a mess. Or could've gotten an entry level job at BigCorp and been making... I don't know, whatever people who work hard and get promoted fast make within a couple years.
So when I start a new company that pays me more the next year, my first business "closed" - that means it's not going. You could say it "failed", but really, I learned the lessons, made a tiny bit of coin, and then promoted myself into the next business. I know lots of entrepreneurs who leapfrog themselves the same way - and similarly, lots of small companies get bought and integrated or stripmined by a bigger company. This is actually a hard part of being a businessman - that's your baby, y'know? - but if you're ready to move on, it's certainly not a failure.
Paul Graham sold Viaweb. Viaweb company is no longer a going concern. It didn't fail by any stretch of imagination, but will be included as a "failure" in these silly "most small businesses don't exist (fail) within five years" statistics.
[1] Numbers might be off a bit, my memory's kind of hazy. I remember calculating at one point that I was making something like $2.20 an hour, maybe after six months or so. $5,500 net split with my partner after six months? I'm smiling now, but it's kind of ugly to remember those days...
I think it's reasonable to consider a business that does not make the investment + minimum wage a failure. IMO, it's more a question of how large the failure was. If you made 2/3 what you could have been making at a normal job and then quit it's a failure but not a major one. If you end up with 30k of debt then that's a major failure.
I think that business continuity is a poor proxy for "success" and that lack of it is a poor proxy for "failure".
For example, let's say you hang out your shingle as a self-employed software consultant today. Two years and 3 successful projects from now, one of your clients offers you enough money to convince you to go full time for them. You, accordingly, close your business. While that is counted as a "failure" by most metrics, that doesn't look like a failure to me. (Losing all your money and being reduced to bankruptcy, now I'll agree that certainly looks like a failure, but nobody tracks that.)
I'd like to see that data. Specifically, I'd like to see the data that starting the types of businesses he mentions are less likely to succeed than the majority of the types of businesses started by the people on this site. Almost all businesses fail.
