Gaming the system could be avoided by using “mystery shoppers”.

Indeed – but therein lies the problem – there is a constant need to refine and re-refine the contract and KPI’s to stay one step ahead of people who would gain the system. Given that there are literally thousands of behaviour changes you can make to save costs that might impact quality or public safety it is an uphill battle and at some point you have to wonder whether the overhead of the endlessly growing private sector incentive system costs you any less than the overhead of supposed public sector innefficiency.

I wouldn’t worry about reduction in repeat business – I’m sure an innovative private sector provider can think of ways to maintain their customer numbers:
http://www.ccpoa.org/news/entry/judge_guilty_of_taking_bribes_from_private_prison_sending_juveniles_to_for-/

You’d probably also want some KPI that punishes the prison if inmates are killed, so that the prison doesn’t have incentive to put hits on folks with high likelihood of recidivism. 

Shleifer is mostly worried about inefficient cost chiselling in industries with little scope for beneficial technological innovation and where there are margins for chiselling that can’t be contracted. I was 100% with Shleifer, and posted as much, saying that prisons are about the last thing I’d privatize in NZ. Then I saw the reports on the private prison chiselling on costs by treating prisoners like human beings to save on guard costs. And I updated my priors.

So long as they’re paid more for overachieving the recidivism target, all’s fine though, no? Though I suppose we eventually hit some laffer curve max where the reduction in revenue from repeat business overwhelms the incentive payments.

The other thing worth noting is that by definition the involvement of private providers introduces and additional ‘performance requirement’ that doesn’t exist in the public sector – the requirement to make a profit.

It is important to consider this in relation to those unmeasurable behaviours. There are a massive number of ways an organisation can save costs (many of them with negative consequences either for inmates, staff or the public) and this particular performance objective has no cap and often runs counter to other objectives.

Thought experiment:
I am a prison company and I have a KPI to reduce recidivism by 10%. Suppose I innovate and find that I can run an effective but reasonably expensive ‘peace n love’ workshop with inmates and the effect is a 20% reduction in recidivism. Arguably I would be acting irresponsibly towards my shareholders if I put all inmates through this expensive course because I could reduce spend, increase profits and still achieve my contracted 10% recidivism KPI by only running the course for half of the inmates. In fact reducing recidivism amongst more inmates than contractually necessary would be detrimental to the company long term because it would decrease my future supply of ‘customers’.
Note these same incentives don’t necessarily apply to prisons run by central government. For them its worth the extra cost to run the course for everyone because the reduction in reoffending delivers savings in the police budget so the government’s ‘business’ is better off. 

I guess you don’t need a stick if you have a monopoly over carrots! I wonder if that is consistent with Shleifer’s recommendation against private provision to violence-prone inmates. It surely could be, but I don’t know the details of who is housed in which NZ jails.

http://www.offsettingbehaviour.blogspot.co.nz/2012/03/prison-economic-illiteracy.html