NEW ORLEANS — In some school districts, the implementation of charter schools has been the catalyst for significant improvements in student outcomes. But in other places, charter schools have had negative effects on math and English test scores. A new national study from REACH, Tulane University’s National Center for Research on Education Access and Choice, looks at factors that cause charter schools to succeed in some places and fall short in others.
“On average, the lower the initial achievement in traditional public schools, the more likely charter schools are to be effective,” said Harris, who is also a professor, chair of the Department of Economics, and the Schlieder Foundation Chair in Public Education at the School of Liberal Arts.
The story of New Orleans’ educational transformation following Hurricane Katrina was a key motivation for the study. After the storm, control of the city’s school system shifted into the hands of a state agency, the Recovery School District. Harris and colleagues found that test scores, high school graduation rates and college attendance all rose — sometimes by double digits — after the city shifted almost all public schools to independently operated charter schools governed by performance contracts.
“New Orleans has probably improved more, faster, and across a wider variety of outcomes than any district has ever improved in the United States,” Harris said. “The question was, why? Why did it work so well here and not as well elsewhere? This study examined all the theories we could think of. The idea that it was because outcomes were so poor to start with turned out to be the key one.”
The study also helps explain a long-standing pattern in prior research: charter schools appear especially effective at raising test scores for Black, Hispanic and low-income students. The findings suggest this pattern has less to do with students’ demographics than with the low-performing schools they are more likely to attend.
“Another way to interpret the same finding is that charter schools reduce achievement gaps,” Harris said.
Low achievement alone isn’t the only factor that may predict the success of charter schools, the study found. Charter schools were also more effective in places where schools had more money to spend.
“Money matters,” Harris said. “States that authorize charter schools to save money may be undercutting the results they are hoping for.”
Notably, total spending across charter and traditional public schools mattered more than charter school funding alone. That’s because charter schools affect other schools, too, Harris explained. They can help the students who enroll in them, and they can push nearby public schools to improve so they don’t lose students, but those other schools need resources as well, Harris said.
By contrast, state charter laws appeared to make almost no difference to the success of charter schools. The researchers examined five policies rated as important by the National Alliance for Public Charter Schools, a charter advocacy group, including caps on charter growth, performance-based contracts and exemptions from rules that apply to traditional public schools. None of the factors ranked among the top predictors of charter success, according to the study.
“Other than funding, state policy doesn’t seem to matter very much,” Harris said.
The study is one of very few national studies on school choice. The findings arrive as states continue to expand school choice, and as some policymakers pitch choice as a way to improve schools, sometimes without spending more.
“Some people view school choice as a way of improving education on the cheap,” Harris said. “Using choice as a substitute for spending is probably not going to work.”
Instead, Harris said, the findings point toward targeting charter schools where they are most likely to help.
“We should focus on low-performing districts,” Harris said. “Those are the ones most likely to benefit from this. A large expansion of school choice everywhere probably isn’t going to improve outcomes.”
Contributing authors on the study include Nicholas Lacoste of the University of Hawaii at Manoa, who developed a novel machine learning method for the analysis, and Feng Chen of Liaoning University.
