If people just put away what's left at the end of the month, that's a recipe for failure.

When an economist says the evidence is "mixed," he or she means that theory says one thing and data says the opposite.

The money has to be deferred with what they call "clawback," which means they can get it back if I lose it all. So that guy making ten million a year selling credit default swaps, if we're going to keep five million of it in escrow for ten years, and with the right to go back and get it, if he starts losing money, then we're going to give people the right incentives not too take so much risk.

I'm all for empowerment and education, but the empirical evidence is that it doesn't work. That's why I say make it easy.

LTCM lost money when Russia defaulted on a certain class of bonds, and then they had other investments like on the spread between two different kinds of shares of Royal Dutch Shell Oil Company. Now that seems completely unrelated to Russian bonds. But they were related because other hedge funds saw similar discrepancies and they were all making similar bets.

In a typical 401k plan, when you first become eligible you get a big pile of forms and you're told, fill out these forms if you want to join. Tell us how much amount you've saved and how you want to invest the money. In, under automatic enrollment you get that same pile of forms but the top page says, if you don't fill out these forms, we're going to enroll you anyway and we're going to enroll you at this saving rate and in these investments.

So, what's a nudge? A nudge is some small feature of the environment that attracts our attention and alters our behavior.