Benefiting from new data provided by experimental economists, behavioral economics is now moving beyond empirical tests of standard behavioral assumptions to the problem of designing improved institutions that are tuned to fit real-world behavior. It is therefore worthwhile to consider the potential for new experiments to advance school choice debates and the daunting task of institutional reform. Deciding which new institutions to introduce and test is an important scientific question. New institutional alternatives are, however, by definition, untested, and empirical arguments to guide this process are difficult to construct without adding more layers of theoretical assumptions. This article analyzes normative criteria that can be used when introducing new institutions for delivering school services. Because of the long time-frames for testing the effects of new institutions and the extraordinarily high stakes for children and families who participate in such experiments, performance metrics that are normally uncontroversial, such as cost minimization, will likely be insufficient for justifying the introduction of new institutions. (Contains 1 note.)