Are real estate agents encouraging lowball offers, because their take doesn't change much and they just want to make a deal? He answered this by looking at how agents sell their own homes (those houses stay on the market longer, and go for more).
How much does money matter in elections? Hard to say, because the challengers don't get a lot of money unless there's a good chance they'll win, and incumbents don't spend a lot unless there's a good chance they'll lose. So Levitt looked at repeat matchups -- same people, different races. It's kind of like how scientists trying to tease nature vs. nurture look at separated twins. Levitt found that money mattered a lot less than expected.
Levitt's big take home message: the data is out there, the hunches are out there, you just have to find an interesting way to prove your point, and publish it.