But, here's what I tell myself to rationalize those leanings away:
The wild card that Marx didn't account for was that employees in startups generally own a small piece of the company that they work for. That's the reason that employees in startups are willing to work 80+ hours a week.
Generally, those employees would be much better off financially working at a Java shop that paid more. But, startup employees tend to spin off and start their own startups. Having a front row seat in a startup, and learning from the founder's fails and wins is also worth something. Also chances are, you're probably not going to get to built stuff in Node, Haskell, Erlang, Python, etc... at BigCo. That's probably also worth a price differential to account for the risk involved in choosing "unproven" technologies.
Finally, there is a substantial risk that the startup you're working for is going to go belly up. And, the founders of the startup are generally making less per hour than you are. And, despite rumors to the contrary, the average startup in Silicon Valley works really, really hard to get funded. Many, many great companies are under capitalized right now. Everybody from the founders to the employees are working below market salaries with the hopes that some day the exit will pay off.