Skip Navigation
Evaluation of the DC Opportunity Scholarship Program: Impacts After Two Years

NCES 2009-4023
June 2008

Executive Summary

The District of Columbia School Choice Incentive Act of 2003, passed by the Congress in January 2004, established the first federally funded, private school voucher program in the United States. As part of this legislation, the Congress mandated a rigorous evaluation of the impacts of the Program, now called the DC Opportunity Scholarship Program (OSP). This report presents findings from the evaluation on the impacts 2 years after families who applied were given the option to move from a public school to a participating private school of their choice.

The evaluation is based on a randomized controlled trial design that compares the outcomes of eligible applicants randomly assigned to receive (treatment group) or not receive (control group) a scholarship through a series of lotteries. The main findings of the evaluation so far include:

  • After 2 years, there was no statistically significant difference in test scores in general between students who were offered an OSP scholarship and students who were not offered a scholarship. Overall, those in the treatment and control groups were performing at comparable levels in mathematics and reading (table 3).
  • The Program had a positive impact on overall parent satisfaction and parent perceptions of school safety, but not on students’ reports of satisfaction and safety (tables 4 and 5). Parents were more satisfied with their child’s school and viewed the school as less dangerous if the child was offered a scholarship. Students had a different view of their schools than did their parents. Reports of dangerous incidents in school were comparable for students in the treatment and control groups. Overall, student satisfaction was unaffected by the Program.
  • This same pattern of findings holds when the analysis is conducted to determine the impact of using a scholarship rather than being offered a scholarship. Twentysix percent of students who were randomly assigned by lottery to receive a scholarship chose not to use it in either the first or second year. We use a common statistical technique to take those “never users” into account; it assumes that the students had zero impact from the OSP, but it does not change the statistical significance of the original impact estimates. Therefore, the positive impacts on parent views of school safety and satisfaction all increase in size, and there remains no impact on academic achievement and no overall impact on students’ perceptions of school safety or satisfaction from using an OSP scholarship.
  • There were some impacts on subgroups of students, but adjustments for multiple comparisons indicate that these findings may be due to chance. There were no statistically significant impacts on the test scores of the high-priority subgroup of students who had previously attended schools designated as in need of improvement (SINI). However, being offered or using a scholarship may have improved reading test scores among three subgroups of students: those who had not attended a SINI school when they applied to the OSP, those who had relatively higher pre-Program academic performance, and those who applied in the first year of Program implementation. The Program may also have had a positive impact on school satisfaction for students who had previously attended SINI schools. However, these findings were no longer statistically significant when subjected to a reliability test to adjust for the multiple comparisons of treatment and control group students across 10 subgroups; the results may be “false discoveries” and should therefore be interpreted and used with caution.
  • The second year impacts are generally consistent with those from the first year.1 The main difference is that after 1 year, the non-SINI and higher performing groups of students appeared to experience statistically significant positive impacts on math achievement, while in the second year the impacts were on reading achievement. Adjustments for multiple comparisons suggest that both sets of results may be false discoveries.


1 See Wolf, Gutmann, Puma, Rizzo, Eissa, and Silverberg 2007.