It depends on what you mean by "fail". If it means "drop out before getting a PhD", then you are surely right. But that's not how most people who enter PhD prog

It depends on what you mean by "fail". If it means "drop out before getting a PhD", then you are surely right. But that's not how most people who enter PhD programs in sciences define success. They want a tenure track position afterwards (possibly with a few years of posdoc positions in between), and eventually they want to become tenured professors. And I suspect--I can't back it with any numbers though--the failure rate for them is of the same order of magnitude as for startups.