Y Combinator probably makes up the 6% on the confidence boost associated with being selected, alone. Makes it hard to compare unselected/selected companies, I imagine.

http://en.wikipedia.org/wiki/Pygmalion_effect

Probably much higher. I'd bet that it's 50% at least. Aside from health insurance, one of the biggest hurdles to startup generation is the uncertain career effect. Being YC-funded essentially means that Paul Graham is vouching for your time.

Think about college. Aside from the independently wealthy, no one would dedicate 4 years of life to study, given what would be the career costs of doing so-- an empty resume, a younger boss-- except for the fact that a prestigious institution is willing to sign for the time, making it a career-advancing endeavor. In this case, people are actually willing to pay for it. While there are some glaring inefficiencies in the college model, most would argue that it's a good thing for people to be better educated.

The effect of Y-Combinator is similar. YC is the Harvard of angel investors, and no one who has heard of YC would discredit a founder for time at a YC startup, even if it failed. The result of it is that more people are willing to do startups due to the reduced career risk-- a very good thing.

Actually this has started to become a problem for us. When we first started and had no reputation, the only people who applied for funding were people who really wanted to start startups. Now we get people who are attracted by the idea of "getting into" YC. And it can be tempting to accept them, because people who've been using this m.o. tend to have accumulated impressive resumes. But when we do accept them they often do badly, partly because they didn't really want to be starting a startup, and partly because real world users are a different type of test than they're used to.