Punitive Damages and Statistical Error
In all that has been written about punitive damages (see here and here for two recent papers), one important rationale for allowing, even encouraging, punitive damages has been ignored. Punitive damages, at least in some instances, allow us to adjust for type II errors brought about by the high standard of proof required in criminal law. Here's what I mean:
Suppose we think a defendent has commited a crime; in fact, we're pretty sure the defendant committed the crime, but we doubt if we can prove it beyond a reasonable doubt. This standard of proof (beyond a reasonable doubt) is in place because we don't want to convict innocent people -- i.e. we want to minimize what statisticians refer to as Type I errors. But reducing Type I errors by requiring such a high standard of proof means that there will be more guilty defendants on the loose - Type II errors.
One way to punish some of the guilty defendants who would otherwise be let off (i.e., one way to reduce the incidence of Type II errors) in some instances is to allow plaintiffs to sue them in civil court for both compensatory and punitive damages. The standard of proof in civil court is a "balancing of the probabilities", a much lower standard of proof than "beyond a reasonable doubt". This lower standard means more Type I errors will be committed, but fewer Type II errors will be committed. It appears that by allowing plaintiffs to seek punitive damages, society is saying that we are willing to commit more Type I errors (punish the innocent) in this way, with less punishment than criminal punishment. But we're doing it to cut down on the Type II errors and create some more deterrence and punishment of some types of activities.
Note that it doesn't matter who receives the punitive damages for them to have this deterrent and punishment effect. And so my suggestion is (somewhat) consistent with the State of California's recent legislation, attempting to claim for itself 75% of all punitive damages awarded in cases in that state.