N.Y.C. Study Finds Gains for Charters
Research Shows Schools Closing City-Suburb Gap
New York City’s charter schools are making strides in closing achievement gaps between disadvantaged inner-city students and their better-off suburban counterparts, a new study concludes.
The study, conducted by Stanford University researcher Caroline M. Hoxby and her co-authors Sonali Mararka and Jenny Kang, is based on eight years of data for students applying to the city’s growing number of charter schools. It finds, for instance, that attending a charter school from kindergarten to 8th grade can close the achievement gap with a similar student in the affluent suburb of Scarsdale, N.Y., by 86 percent in mathematics and 66 percent in reading.
By comparison, the “Harlem-Scarsdale” gap only widens over the same span of grades for students who remain in regular public schools, according to the study.
For high school students, the study also found, attending a charter school increased the likelihood that a student would earn a state regents diploma by age 20 by 7 percent for each year spent in that school.
Charter schools, which are public schools allowed to operate largely independent of many traditional school district rules, are a major focus of the Obama administration’s education plans. Federal education officials have made it clear, in fact, that states deemed unfriendly to charter schools will be at the back of the line for economic-stimulus funds from the $4 billion Race to the Top program. ("Obama Team's Advocacy Boosts Charter Momentum," June 17, 2009.)
Studies comparing student achievement in charter schools with that in regular public schools are difficult to do credibly, however, because applicants and their families are presumably more academically motivated than students who remain in regular schools.
Ms. Hoxby was able to overcome that problem in New York City because its charter schools are oversubscribed, requiring administrators to use lotteries to allocate seats. That allowed researchers to compare students randomly assigned to charters with students who applied but did not win a seat in a charter school, thus providing for a more apples-to-apples comparison.
The promising findings contrast sharply with those from a widely reported national study published this past summer by the Center for Research on Education Outcomes, which is also based at Stanford. In that study, CREDO Director Margaret E. Raymond examined data for 2,403 charter schools in 15 states and the District of Columbia. She found that students in more than 80 percent of the charter schools performed the same as, or worse than, students in regular public schools on mathematics tests. ("Study Casts Doubt on Charter School Results," June 15, 2009.)
How Much to Generalize?
“As with all studies of charter schools, you have to look at what you’re comparing,” said Ellen B. Goldring, a principal investigator for the National Center on School Choice, at Vanderbilt University in Nashville, Tenn. Ms. Hoxby’s study was financed by the federal Institute of Education Sciences through the center.
“Clearly, the lottery method is the gold standard,” Ms. Goldring added, “but it is one locale and it is generalizable only to schools with oversubscribed lotteries.”
“The CREDO study is looking across states and using a matching method, so they’re just asking different questions,” she said.
Jeffrey Henig, a professor of political science and education at Teachers College, Columbia University, said the explanation for the differing results could be even simpler. “It could just be that New York’s charter schools are better,” Mr. Henig said. “This is a pretty strong study design-wise,” he said of Ms. Hoxby’s study, “and it shows effects that, if aggregated over time, are substantial.”
But another possible explanation for the sharply contrasting results, said Ms. Hoxby, may be that the CREDO findings suffer from a “serious mathematical mistake.”The problem, she said in a separate analysis, is that Ms. Raymond compared the achievement of individual charter students with that of groups of students from nearby public schools without making the statistical adjustments necessary to account for the natural downward biases that result from that sort of calculation.
In an e-mail message, Ms. Raymond declined to comment last week on Ms. Hoxby’s criticism. “Her comments deserve similar careful attention,” she wrote, “and we will reserve comment until we have had a full opportunity to consider her memorandum.”
In her study, Ms. Hoxby found that by the 3rd grade, the average charter student was 5.3 points ahead of lottery “losers” on state exams in English and 5.8 points ahead in math. (That’s on test scales of 475 to 800 points.) After that, the charter school students gained an additional 2.4 to 3.6 points a year beyond the regular public school students who failed to win a charter spot in the lottery.
Ms. Hoxby extrapolated from those year-to-year gains to determine the overall impact of attending a charter school from kindergarten through 8th grade,she said, but the results were in line with those for students who had actually been in charters from kindergarten through 5th grades. The study draws on data for 4,200 charter students per grade and an equal number of public school students.
Experts said Ms. Hoxby’s study is also important because it “gets inside the black box” by exploring what successful charter schools might be doing differently with their added autonomy. Schools that held classes 10 or more days longer each year tended to produce better achievement, for example.
Other school characteristics associated with better student achievement included: more time on English instruction; teacher pay plans based on teachers’ effectiveness at improving student achievement, principals’ evaluations, or whether teachers took on additional duties, rather than traditional pay scales; an emphasis on academics in schools’ mission statements; and a classroom policy of punishing or rewarding the smallest of student infractions. While those factors don’t necessarily cause higher achievement, Ms. Hoxby said, they offer clues for further study.
Vol. 29, Issue 05, Pages 1,11