In Let Markets Help Criminal Defendants, I wrote that “If I were running a public defender service, I’d consider setting up an online prediction market for the conviction of my clients.” I still think this is a good idea, but someone suggested a serious problem that would have to be remedied for the scheme to be possible.
Right now, prediction markets bets on judicial events, like the conviction of Lewis Libby (whose graph is to the right), pay off at 100 for conviction, and 0 for any other ending of this set of charges, including a plea. This creates noise which renders them useless for criminal defendants looking to see if they ought to plea. That is, as I didn’t fully appreciate before, traders must be estimating the probability of conviction, tempered by the likelihood of a plea – prices are lower than the actual market estimate of a guilty verdict independent of a plea. That is, if the current price of Libby’s “stock” is .40, that means that incarceration is not 40% likely. It means that traders think it is 60% likely that Libby will win at trial, receive a mistrial, obtain a dismissal, be granted a pardon, or plea. I imagine that the likelihood of a plea accounts for a large percentage of this figure.
If traders thought that conviction prices affected defendant behavior, then presumably they’d seek to put in sell orders at prices above those where rational defendants would plea. This would put downward pressure on price and make the entire system useless from defense counsel’s perspective.
For my system to work, you’d have to exclude the possibility of a plea (i.e., nullify all bets if there is a plea). Of course, this still would create some dynamic tension, as bettors presumably would become eager to invest time and trade only as pleas become less likely – near trial, or in jurisdictions, like Philadelphia, where the District Attorney has a no-plea policy. But the resulting prices would be more informative than those offered by the current system.