Betsy DeVos Wants to Spend Millions on School Vouchers. Two New Studies Say They Don’t Actually Work

President Donald Trump’s proposed 2018 budget includes $250 million for voucher initiatives, state-funded programs that pay for students to go to private school. Another $1 billion is set aside for encouraging schools to adopt school choice-friendly policies. Betsy DeVos, Trump’s education secretary, is an outspoken advocate for such programs, which she has argued would lead to better educational outcomes for students.

But a pair of new studies released Monday provide evidence that voucher programs don’t improve students’ performance on standardized tests, one quantifiable method of testing educational outcomes.

“There may be some places where vouchers are working for some students, but generally we need to understand a lot more about the conditions under which vouchers are effective or not,” Mark Berends, co-author of the Indiana study, told TIME.

The first study, a joint project from Tulane University’s Education Research Alliance and the University of Arkansas’ School Choice Demonstration Project, found that voucher programs did not produce improvement on students’ test scores.

The researchers used standardized test scores spanning from 2012 to 2015 to analyze students’ outcomes from the Louisiana Scholarship Program, an initiative that paid for 7,000 children from low- and moderate-income families to attend private schools. They found that students in the Louisiana voucher program initially had lower test scores, but after three years, their scores matched those of students who stayed in public schools.

A second study, meanwhile, examined the statewide school voucher program in Indiana, one of the largest initiatives of its kind in the U.S. The unpublished study from the University of Notre Dame and the University of Kentucky, which is pending peer review, found that Indiana’s 34,000-student program had a negligible effect on educational performance for children in third grade through eighth grade from 2011 to 2015.

The researchers found that the Indiana program resulted in an initial drop in math test performance for low-income students. They made up the losses only if they stayed in the private school for three or four years, though many students who performed the worst left and went back to public school.

The study found no significant difference in students’ performance in English.

But nothing is “going to be a fix-all for everything,” said R. Joseph Waddington, who co-authored the Indiana study.