The promising results from the new Harvard/Mathematica study of New York City’s private voucher program quickly came under fire from school choice opponents, who tried to discredit the analysis by questioning the motives of lead researcher Paul E. Peterson and dismissing the “small” academic gains as a “predictable” consequence of “smaller classes, smaller schools, and parents with higher levels of income and education.”
The only surprise, sniffed American Federation of Teachers president Sandra Feldman, “is that they didn’t do even better.”
Significantly, Peterson and his colleagues did in fact address the issue of the size of the impacts of the New York City program in their report. They compared their results with those of an evaluation of another randomized experiment heavily promoted by teacher unions and used by Congress to justify the spending of more than $1 billion to reduce elementary class sizes: a Tennessee experiment to reduce average class size from 25 to 15 students.
The Tennessee experiment showed one-year gains of .15 to .30 standard deviations only in first grade, while the New York City experiment shows one-year gains of .18 to .23 standard deviations — effects deemed “small” by the AFT–in grades four and five.
“The effect sizes observed in this evaluation of the New York scholarship program do not differ materially from those observed in Tennessee,” note Peterson and his colleagues, suggesting that the New York City results would also be called “quite substantial” if the Tennessee terminology were applied to their work.
But are the gains of voucher students a “predictable” result of smaller classes, smaller schools, and parents with higher levels of income and education, as the AFT’s Feldman claims?
Few researchers would make such a statement based on the minute differences, shown in the accompanying table, between the eligible population and actual scholarship users. Scholarship parents are uniformly low-income, have limited education, and their children benefit from class sizes averaging 24 students rather than 26–a far cry from the Tennessee reduction of 25 to 15.
|% Mothers College Grads||3.7%||7.8%||+4.1%|
|% Fathers College Grads||8.0%||10.4%||+2.4%|
|* Control Group
** Scholarship Group