If fining them is a problem because the variance of “how much they are fined” is too large, what if uh, had an external organization evaluate the probability of

If fining them is a problem because the variance of “how much they are fined” is too large, what if uh, had an external organization evaluate the probability of each officer getting fined and use that to price insurance that they have to buy? That way the variance is reduced, and they have a direct monetary incentive to have habits which reduce the estimated probability that they get fined (or, rather, the expected value of how much they get fined for) ?