Matt Yglesias points out that prison privatization has not been the success its proponents suggested. Costs seem to be higher in many places, and the prison privatization has not eroded the power of the prison lobby, but merely replaced it with another group of lobbyists:
There are important general lessons here. The genius of the real private economy is that firms that are really poorly run go out of business. It's not that some magic private sector fairy dust makes the firms all be runs soundly. Lots of bad businesses are out there. But they tend to lose money and close. Meanwhile, well-run firms tend to earn profits and expand. The public sector doesn't have this feature. Just because a public agency is inept is no guarantee that it will go out of business. Resources are allocating according to political clout rather than any criteria of merit. It's a problem. But it's not a problem that "privatizing" public services actually solves. There's no magic private sector fairy dust.
That's not to say government services should never be contracted out. As an extreme example, public agencies don't manufacture their own printer toner. DC contracts out its Circulator buses to First Transit and it's very successful. There are some very good charter schools. But there's no fairy dust. Not in prisons and not elsewhere.
One should be cautious about this: the study in question was done by the Department of Corrections, which may not be a neutral arbiter, and it doesn't really address quality, only costs. Nonetheless, I wouldn't be surprised by this result. Privatizing government services doesn't necessarily solve the problems inherent in providing those services, and it may create new ones as you add another layer of conflicting interests--the managers and shareholders of the private company--into the services.
Privatizing welfare services, for example, doesn't seem to have done much good, in part because the metrics used to measure performance emphasized getting people off welfare--full stop--rather than, say, ensuring that they got and kept jobs. The incentive, obviously, is to mess with people's checks, making it very difficult for them to enter the system, and finding arbitrary reasons to cut them off. Maybe you're okay with that because you don't think we should have a welfare system--but in that case, the proper thing to do is to gather a political constituency to abolish it, not to give some corporation money to end-welfare-as-we-know-it by randomly throwing people out of the system until they hit their quota.
I'd be much more sanguine about the prospects for an agency in either the prison or welfare system that got paid only if their "clients" stayed employed, out of jail, and drug free for a year after leaving the system. But I doubt you could find many companies willing to take on such a contract, because who wants to bet on your ability to change the behavior of legal adults who have, as a group, a higher-than-normal propensity to be out of work and take drugs?
If you can't create a good contract that aligns the incentives of the contractor with the incentives of society, then you shouldn't just go ahead and privatize anyway because the free market is awesome. The free market is awesome. But it is not actually a cure-all.
So when should we privatize services? I think the number one rule is that we should be extremely cautious about privatizing services with a captive audience, especially when the people they are supposed to serve have little political voice. So no to privatized courts, prisons, or welfare agencies, though it may well work to contract out specific functions within those agencies that have easily agreed-upon objective measures which are hard to abuse.
We should also be careful about outsourcing monitoring functions such as building inspection. It's very hard to come up with any sort of performance metric that won't tempt the company to either issue too many violations, or contrariwise, pass substandard work (especially if you expect them to collect fees for their services.) Since it's hard to monitor their work, you don't gain much from privatization, and indeed, you may lose, since unlike the government, they need a profit margin.
On the other hand, services where the "clients" are primarily customers seem ripe for privatization. I'd say that most of the rest of the DMV's services should be outsourced, buses are an obvious candidate, and even community centers could probably be operated on some sort of capitated rate.
And what about day care and schools? Obviously, I think these things can be outsourced. Unlike prisoners and welfare clients, this adds monitoring and exit rights to the system, actually improving the alignment of incentives with what parents want. Which is, largely, what taxpayers want: happy, educated kids.
Will this be perfect? Of course not: parents will pull their kids for stupid reasons, they will (like the rest of us) have difficulty ascertaining the quality of their school, and principals and teachers will try to game the system. But it's actually hard to game the metric of "Parents want to send their kid there." Even grade inflation, the most obvious way which principals and teachers might pander, is actually pretty uncommon at private schools (or was, when I was there); my 2.7 GPA would have been an absolute bar to admission to an Ivy League school if I'd gone to a public school, but as it was, I got in (barely) because the classes were harder, and A's were not so generously given out. The number of kids with a 4.0 in my class of 95 was either one, or none*. And of course, you can also condition receipt of state money with meeting some minimum standard, using whatever metrics you think are now doing a good job of helping the government provide excellent school performance.
Even food stamps, which are pretty damn simple and objective, are not perfect: there is fraud, and people long managed to use the stamps to buy things other than food. The metric for outsourcing is not whether the program will be perfect; it's whether there's some compelling reason that government provision of services (rather than simply funding them) actually allows you to provide them at lower cost or higher quality.
There are lots of services like this--I'm generally in favor of the government being the single-source provider of police and military, for example. On the other hand, I don't want to get them into the business of manufacturing police cars or fighter planes.
* The only possible candidate was sort of a moody loner, so I can't say for sure.