The National Institute for Justice (which is the research, development and evaluation agency of the U.S. Department of Justice) says that prison is an ineffective deterent: http://nij.gov/five-things/pages/deterrence.aspx
Not as much as you'd expect. There are plenty of studies that confirm this (I'm too lazy to google them for you). The problem is, that when crimes are done, one does not usually think rationally or does not expect to be caught (or else he wouldn't do that in first place). Another big category of crimes (especially violent ones) are the ones committed out of passion, where severity of sentence does not play significant role either.
What matters in deterrence is getting caught, more than the actual duration or intensity of the punishment.
(It's also important that people are not punished when they're not guilty, which the US tends to forget. The other side of reinforcement learning. The tendency of US police to administer adhoc punishment up to and including execution on the street is a problem. You can't say "don't commit crimes or you'll be in trouble with the police" when simply driving while black can get you in trouble with the police.)
The evidence supporting it is very weak, especially the assumption the more punishment always leads to more deterrence. The more effective deterrence in many cases is increasing the probability of getting caught, not increasing the punishment for the few that do get caught.
When even The Bloody Code[1] didn't stop people from offending, you have to question the effectiveness of future punishment as a reason to not commit a crime.
I do not think it is, no. I think the difference in deterrence between prison, and prison where you might starve is likely to be very low.
The real question is not whether rehabilitation-based prison systems give better outcomes across factors such as reoffending and reintegration into society (they do, but at a perceived higher cost), but how to transition from a punitive system to a rehabilitative system.
