The only cheap/practical way to do this is to sample a subgroup chosen because they can be measured. Say, YC S13 founders today, or Stanford alumns in a specif

The only cheap/practical way to do this is to sample a subgroup chosen because they can be measured. Say, YC S13 founders today, or Stanford alumns in a specific group on campus (START?) 5 years later who are in close communication. The group will not be representative of the total population of entrepreneurs/founders, but the group will give internally high response rate and thus significance for that group. And an ACCURATE survey of Stanford alum founders will actually be more representative of overall founders than a voluntary-response set would be. You then ALSO ask better questions. That's also bad in this survey, but not the worst part -- the voluntary response bias is. (I want to be positive overall, though -- it is awesome that you are surveying this. I just want you to do the best job possible.)