riculum, admissions policies, staff, and student test scores. Schools that receive vouchers also should be required to have their students take specified standardized tests as part of the evaluation of the experiment. It would also make sense, however, to be somewhat cautious about establishing these accountability measures. The problem here again is one of providing sufficient flexibility to private schools. If private schools were required to meet a very long and detailed list of mandates, they would lose much of their freedom and the result of the experiment would be preordained.
The design will also need to specify which types of schools will be allowed to receive vouchers. Religious schools are an obvious issue. A recent court decision allowed religious schools in Milwaukee to participate in that city's choice experiment, but clearly this question raises some difficult issues in the separation of church and state. It might make good sense to allow students to use their vouchers to attend public schools outside their district.
Evaluation should be built into any such experiment from the start. It should focus largely on academic outcomes, but should look at other outcomes, such as school safety, as well. For a number of reasons, that evaluation will be far from straightforward. Suppose vouchers are offered to some of the children in a school district, with the intent of comparing the education outcomes of those who receive vouchers with those who do not. But now suppose that the public schools respond to expanded choice by offering their students a better education. In this case we might find that the control and experimental group have the same outcomes, but clearly it would be wrong to conclude that the choice experiment had been a failure. Suppose further that vouchers were offered randomly, but only to children whose parents agreed to participate in the experiment. This design would provide information about the children from families in which the parents care enough about their children to learn about the choice experiment and would be willing to have them be a part of the experiment, but very little about the rest of the students in the district. These evaluation challenges mean that setting up such a project will require significant planning by a team of experts in the field of research design.