Risk vs. Return

Suppose if you win an election you get a payoff of one but if you lose you get a payoff of zero.
Two players are running for election and right now, given the information the median voter has, candidate O is beating candidate M. That is, the posterior of O's expected quality is higher than M's. If candidate M's actions continue to generate signals that leave his posterior to the median voter below O's, he will still lose and get zero. If he manages to generate high signals, he wins and gets one. Basically, the payoff structure generates a convex payoff structure that makes you risk-loving rather than risk averse. (Ironically, it is similar to the payoff structure of a lender who knows he is going to be bailed out from losses but enjoy the upside returns!)

So, if safe actions leave the posteriors roughly the same, the only way to win is to take risky actions that give you some chance of leap-frogging your opponent. This is what McCain is doing with the Palin pick and the "call off the debate and let me solve the bailout problem" strategies. It's kind of like an R+D model where the follower works harder if he is close enough to the leader so he can take leapfrog over him and win the race.

So, if rational choice theory predicts this, I guess it also predicts more of it in the future if the two attempted strategies fail - which they seem very likely to do. I'm trying to think of what McCain might do. Ditch Palin? Seems too drastic. So, I think he'll try to appoint a cabinet in waiting. Maybe try to get Colin Powell on board, maybe get Bloomberg to be Treasury Secretary etc. But will anyone serious want to sink with McCain when he is so obviously a wild card and likely loser? Who will take that risk for so little return?

PS Check out Jonah's great post below about mechanism design.