Sweden has a school voucher system with universal coverage and full acceptance of corporate providers. Using a value added approach, we find that students at upper-secondary voucher schools on average score 0.06 standard deviations lower on externally graded standardized tests in first year core courses. The negative impact is larger among lower achieving students (but not among immigrant students), the same students who are most prone to attend voucher schools. For high achieving students, the voucher school impact is around zero. Comparing internal and external evaluations of the exact same standardized tests, we find that voucher schools are 0.14 standard deviations more generous than municipal schools in their internal test grading. The greater leniency in test grading is more pronounced among students at academic than at vocational programs. The findings are consistent with voucher schools responding more strongly to differences in educational preferences than municipal schools.