A study that influenced President Bush to recommend a big cut in spending for a federally financed after-school program has serious flaws, according to a statement by seven members of the study鈥檚 technical working group.
The first-year report on the 21st Century Community Learning Center program, conducted by Mathematica Policy Research, has 鈥渕ethodological problems that call into question its findings and that violate basic principles governing how evaluation should be used to guide policy and affect program budgets,鈥� the statement contends.
Mathematica Policy Research posts a to the statement issued by some members of the study鈥檚 technical working group. (Requires .)
Officials of the U.S. Department of Education, which paid for the Mathematica study and has championed the value of scientifically based research in determining the effectiveness of education programs, said they 鈥渟tand behind the study.鈥�
The principal investigators for the study, Mark Dynarski and Mary Moore, said in their own response last week that the technical advisers had had sufficient time before the study was released to raise their concerns, and that the study had never suggested that funding for the $1 billion program be cut.
President Bush is requesting a 40 percent reduction in the program for fiscal 2004.
The study鈥檚 only policy recommendations, Mr. Dynarski and Ms. Moore said, were that programs look for ways to increase student participation and improve academic content.
Mr. Dynarski added in an interview that one of the members who signed the statement鈥攚hich was e-mailed to the media last week鈥攈ad hardly participated in the meetings, and that two other members did not sign the statement.
鈥淚t feels political to me,鈥� he said. 鈥淚 think they got caught up in the moment of the budget cuts.鈥�
Christopher T. Cross, a member of the nine-person working group and a senior fellow at the Washington-based Center on Education Policy, responded that the group never saw the final version. Mr. Cross, one of the signers of the statement, served as an assistant U.S. secretary of education for research during the first Bush administration.
The debate over the Mathematica findings indicates that more work needs to be done to determine the best way to design and evaluate after-school programs, said Megan Beckett, a social scientist working in this field at the RAND Corp., a think tank based in Santa Monica, Calif.
鈥淚t鈥檚 a rough field, and it鈥檚 just now starting to be held accountable,鈥� she said. She added that the Mathematica study is 鈥渢he most rigorous so far.鈥�
Deputy Secretary of Education William D. Hansen, meanwhile, testified before a Senate Appropriations subcommittee last week to further explain the administration鈥檚 proposal to cut funding for the initiative by $400 million. One reason, he said, is that the program grew 鈥渞apidly with little consideration of its effectiveness.鈥�
鈥楰ey Measures鈥� Questioned
Released in February, the study by Princeton, N.J.- based Mathematica concluded that the program had had few positive effects on student achievement and had not led to improvements in students鈥� behavior or feelings of safety. (鈥淪tudy Critiques Federal After-School Program,鈥� Feb. 12, 2003.)
The seven working-group members who released the statement say some of those conclusions were not justified, however.
For example, they say, the study did not reflect the fact that middle school students in the control group started out scoring higher than those who were being served by the program.
鈥淚n our judgment, the first-year analyses and any subsequent analyses are un-interpretable given these substantial baseline differences on key measures,鈥� the statement says.
But the Mathematica researchers countered that while the treatment and control groups were different in some respects, the study used 鈥渟tatistical methods specifically designed to adjust for the differences.鈥�
In the elementary school evaluation, the seven working-group members say, the size of the study sample was so small that they do not have much confidence in the results.
Data were collected on 403 students in the treatment group and 226 students in the control group from 18 elementary sites. But in the 2001-02 school year, there were more than 6,800 21st Century Community Learning Center sites operated by more than 1,400 grantees, offering a variety of activities such as homework assistance, recreation, and intellectual enrichment.
The statement notes that the researchers later added 1,600 more students to the sample, but still reported the findings on the smaller sample.
Mr. Dynarksi and Ms. Moore argue, however, that while a larger sample would have been preferable, their study is still the first one on after-school programs to use random assignment. And they noted in the report that the findings were preliminary.
Analyzing Outcomes
Whether academic improvements should have been expected from all of the sites is also a point of contention between the seven working-group members and the researchers. Four of the elementary sites in the study offered programs aimed at adults, such as efforts to improve job skills, and the children attended only with their parents or grandparents.
鈥淚t is not clear why or how these adult-focused programs would be expected to impact child outcomes,鈥� the group鈥檚 statement says.
But the researchers argue that the approach used at those sites was 鈥渂ased on the hypothesis that by supporting parents, programs can achieve better academic outcomes for students,鈥� and understanding whether that was accomplished was an important objective.
The working-group members also say they found inconsistencies between the full report and the executive summary, a contention that the researchers dispute. And the members say in their statement that positive outcomes might not have been found because students didn鈥檛 participate regularly enough鈥攁 factor that other after-school studies have concluded is necessary for results to be achieved.
