You still have the problem where crime X (let's say stealing a car) and crime Y (maybe shooting someone in the leg) result in the same prison term, when normal people would say that one of the crimes is a lot worse. Prison plus fine is barely multidimensional, and converting prison to money reduces it to one dimension again.

In a study like the one you propose, they should also see what effect the crime causing the prison term has on the valuations - will time in prison for rape cost more or less than time in prison for murder?