To briefly respond to that, what I would initially say is, when you think about that notion like a movement to a private space that is less inclusive, that ultimately that would likely result, in fact, in worse outcomes for at least some subset of the population, perhaps for those who didn't participate in this sort of programming.
So what tools like social impact bonds can do, if well designed—and that's a caveat, an “if”—is ensure that the measurement architecture, which is how things are evaluated for outcomes, take note of those factors. As an example, when we think about things like crime reduction, it is important to note the difference between high-frequency offenders—people who offend quite frequently—and those who may just have had their first offence. If something is done just on the basis of a binary measure—whether they committed crimes or not—you might just choose to work with those easiest-to-work-with offenders. That is the sort of cherry-picking concern. But if you can design it such that we actually address the frequency of offence, then we might actually have an incentive to drive outcomes toward those who are the hardest-to-treat populations.
Equally, as we think about issues of cultural sensitivity, that may be a driving factor in the lack of performance of these new funded programs. So having that orientation toward outcomes, performance, and effectiveness and actually learning frankly whether or not that is true—perhaps cultural sensitivity matters an incredible amount to your committee in particular—might be a tool, and social impact bonds might actually be a tool to develop that knowledge base in a very rigorous, credible way.