I think the point is that for each wallet stolen we would send a guy to jail for, say, two days. If we think that the guy has stolen approximately 100 wallets we prevent the next 900 from suffering by imprisoning him for the most part of a year, and that's it.

What I haven't seen yet is a comparison of cash, not to the cost of imprisonment, but to the cost of getting a job. That is, if someone is in prison for 200 days, that costs a few tens of thousands of dollars to the US taxpayer, so the arguments being advanced in these threads say "this should be a response only to someone who is stealing $10,000 from us, otherwise the government is taking much more from us in taxes than the guy was taking from us in crime." That makes some amount of sense. It ignores certain problems (like "who do the taxes come from?" and psychological damage from getting mugged and so on), but it does have some core "thrust" to it.

On the other hand, to be an effective deterrent to crime, you'd figure that we'd want to make the crime net-unprofitable; that is, you'd want the 9-to-5 job at $7/hour to be a more profitable way of living than stealing $100 wallets. So this would suggest that for an expected number of 100 wallets stolen you should really incarcerate for, say, 2 years or so, so that the original crime "really doesn't pay", in the sense that you lose job-access for a total sum of more money than you gained. (That punishment might also have to be increased if the chance of catching someone who steals 100 wallets is not 100%.)

I don't agree that prevention of future nonviolent crimes is a mandatory goal. Think about a routine drug user. It is a crime, but I am not willing to imprison someone just for repeating that crime. Some crimes just don't have enough social impact. It is unjust to viciously enforce compliance simply because some behavior is illegal.