The expected value of doing a startup is high -- I think a little higher than being able to guarantee myself $80K/year. If I work on a startup with 1 cofounder for 5 years and it's got a 10% chance of exiting for $10M, my expectation is roughly $100K/year. So if I can afford to take the (rather high) risk of getting nothing, then I should do the startup.

If I value my time differently depending on what I'm working on -- if the startup is lots more fun -- then it's an easy choice.