[hansvm]

It's possible for both of those views to coexist. I have no idea if the data fits such a model, but off the top of my head here are a few factors which might matter.

1. White-collar criminals might be more significantly deterred by the threat of any prison time, perhaps because they have more to lose or because such crimes have more premeditation.

2. Prison might work as a deterrent in general, but if 20yrs will already ruin your life then the additional threat of another 80yrs might have little to no impact.

3. Rehabilitation might have a stronger effect than deterrence, which could point to hybrid solutions leveraging both effects, supposing they mix appropriately.