experiment to be effective, parents need to know that their children will be able to continue in their new schools, those contemplating opening new private schools must know that the vouchers they depend on will continue, parents will need time to evaluate the new options available to their children, public schools must have the opportunity to respond to this change in their environment, and private schools must have time to learn from their inevitable mistakes at the outset of the experiment. A 10-year time frame seems sensible. Since we would not expect all efforts to extend choice to be suspended pending the outcome of this experiment, states or individual school districts moving forward with choice on their own could benefit from examining both the design of the experiment and any early findings from it.
Central to the experimental design should be features designed to ensure that low-income families will have an expanded set of schools from which to choose. This goal could be achieved in a number of different ways, and it might make sense to design the experiment so that the relative merits of these alternatives could be evaluated. These alternatives include scaling the size of the voucher to income so that the poorest families would face small, possibly zero, out-of-pocket costs to send their children to private schools. Alternatively, participating schools could be forced to set aside a fraction of the slots in their school for children from low-income families before they can cash vouchers from any children. In some circumstances, it might be essential to provide children from poor families with subsidized transportation. While the experiment could be structured in a number of sensible ways, to the extent possible it should avoid subsidizing upper-income families who would have sent their children to private schools even without a voucher program. At-risk children, including children in special education, should be given larger vouchers so that they can also effectively participate in the experiment.
Some difficult issues emerge related to admissions policy and tuition. If choice is to be effective, private schools should be given a great deal of flexibility; it would make no sense to initiate a large-scale choice experiment and then place such a broad range of constraints on private schools that they will, as a consequence, fail. This principle suggests that private schools be given flexibility, for example, in designing their curriculum and in hiring and promoting teachers. This principle also suggests that private schools should be given a good deal of freedom in setting admissions policy and tuition. But the dangers here are obvious. If they can pick and choose students, the advantages of a randomized experiment are reduced, and if they can set tuition so high that low-income children cannot afford to attend, the experiment will not reach the children who now are often poorly served by the public schools. It might be possible to mitigate some of these problems with means-tested vouchers. Nonetheless, these tough issues will require a great deal of careful thought.
With respect to accountability, at a minimum it would seem sensible to require schools receiving vouchers to provide information to the public on cur-