Estimating the Impact of School Vouchers and Private Schooling When the Impact Depends on the Duration of Participation.1 (Job Market Paper)

This paper analyzes the impact on academic achievement of school vouchers and private education when the impact depends on the duration of attending private schools, using data from the New York School Voucher Experiment. This experiment randomly assigned vouchers for attending private schools to families that met the eligibility criteria. However, voucher recipients decided whether or not to send their children to private school, and once in the private sector, decided whether or not to stay in that sector. A substantial fraction of families either did not use the vouchers or stopped using them before the end of the program, and only about half of the voucher recipients stayed in the private schools for the entire program period. The impact of offering vouchers has dominated the empirical literature of school vouchers as the central question of interest. However, this conventional impact parameter by itself does not provide a complete picture of the effectiveness of the voucher program in raising academic achievement due to the dynamic aspect of self-selection into the program, and this piece of information alone is insufficient to answer many important policy questions. For example, if the impact of offering vouchers is found to be low, should the program be terminated? If not, how should it be restructured so as to achieve a higher impact in the future? Should it be targeted to a certain population? To answer these types of questions, it is desirable to know the impact of using vouchers for the entire program period and the impact of staying in private schools for part of the program period. As the decision about how long to use the vouchers is nonrandom, non-experimental estimators are required t