The observed pattern seems to be that women make more risk-averse career choices, which founding a startup definitely isn't. The pipeline theory applies much better to blacks than to women.

Further, pg somewhere argued that there's a multiplicative factor once you factor in cofounders--cofounding a startup implies an extremely close platonic relationship, people are more likely to form extremely close platonic relationships with members of the same sex, and hence men have a wider range of possible cofounders than women. I think this argument is at least partially flawed--many businesses are started by husband-and-wife teams and there's no clear reason why opposite-sex platonic cofounder relationships couldn't work either.