Paul E. Peterson and Matthew M. Chingos
Harvard University
November 2007
Harvard political scientist Paul Peterson has devoted a good chunk of his time this year (see here and here) to correcting what he sees as methodological flaws in RAND Corporation's recent evaluation of Philadelphia's privately-managed public schools. (In 2001, the state enlisted several for-profit and non-profit organizations to run 45 failing district schools.) In his latest report, Peterson and colleague Matthew Chingos use student-level data (the same that RAND used) to compare gains in student performance among for-profit, non-profit, and district-run schools. Their analysis differs from RAND's in a couple of key ways. First, the Peterson-Chingos model compares student performance gains, while the RAND model compares student performance levels. Second, Peterson and Chingos compare the privately-managed schools to comparable (i.e., low-performing) district schools, while RAND measured them against a statewide (and, in Peterson's view, not-so-comparable) sample. Third, Peterson and Chingos break out the results of the for-profit and non-profit schools separately; RAND lumped them all together. Peterson and Chingos conclude that for-profit schools achieved greater gains than district schools by a statistically significant margin in math, but not in reading. Schools under non-profit management showed a negative, but not statistically significant impact on student gains in both subjects. In other words, the for-profit schools got the best results, then the district, then the not-for-profit ones. This raises interesting questions about the enthusiasm in many quarters (including our own) for non-profit charter school chains. Still, it's just one program--and a far-from-ideal one at that. RAND's study, for example, noted that "continued district involvement in provider schools... constrained provider autonomy," and Peterson wrote earlier this year in the Wall Street Journal of "concessions to unions that allowed many of the old regulations to remain in effect." One wonders how different the data would look were the providers given true autonomy. Read the report here (and read the Peterson/Chingos op-ed in yesterday's Journal here). We interview Peterson here.
