What he's trying to say is that most people crimes under the assumption that they won't get caught, not that, if they get caught, they can withstand the punishment.
In the 'purely rational' crimes, it's going to be some sort of function of both right? The expected benefit, the risk that you get caught, and the severity of the punishment.
While I agree that people conducting corporate fraud think they will get away with it - I don't agree that the long sentences won't act as a deterrence. If you set the sentence for these sorts of crimes to 1 year rather than 15+, that completely changes the risk profile for people who think there is a 90% chance they will get away with it.
It's simpler than that. You just don't think about getting caught, it's not part of the plan.
Sentence length has a small effect on crime rates, but what really matters is enforcement levels. If you have a 99% chance of getting caught & punished, you don't bother.
Surely this can’t be true - as a trivial example I would be surprised if removing parking fines wouldn’t increase parking violations, or if Singapore stopped punishing littering that it wouldn’t affect the amount of littering etc
Maybe the difference between a 10 year or 20 year sentence for murder doesn’t make much difference, but if murder had no punishment at all I would be very surprised if that wouldn’t raise the murder rate!