Two new studies cast doubt on benefits of class-size reduction
Reducing class size may not improve student achievement, two new studies conclude. [1]Conducted separately by economists Eric A. Hanushek of the University of Rochester and Caroline M. Hoxby of Harvard University, the studies question the wisdom of proposals for reducing class size.
The studies are especially important in light of Tennessee’s Project STAR, which claimed to show improvement in student achievement following class-size reduction. [2] From 1985 to 1989, Project STAR attempted to track student achievement in both big and small classes from kindergarten through third grade. [3]
STAR’s exaggerations
According to Hanushek, STAR's poor methodology exaggerated benefits from class-size reduction. For instance:
- between 20 and 30 percent of the students quit the project each year, with less than half the original number remaining at the end; [4]
- the students who quit tended to be below-average achievers, giving the smaller classes a perceived boost in achievement; [5]
- no pretests were conducted on any students upon enrollment, which provided no benchmark to assess their level of achievement; [6]
- neither the teachers nor the schools chosen for the project were selected randomly. [7]
Harvard’s Hoxby found that reductions in class size within the range of 15 to 30 students had no effect on achievement. Hoxby criticized Project STAR, stating that its participants were "aware of being evaluated" and "mindful of the rewards being contingent upon the outcome." [8]
$500 million price tag in Ohio
The Buckeye Institute found similar results when it analyzed school building-level data in Cleveland, Columbus, and Cincinnati: smaller class sizes did not help explain student achievement. [9]
A similar Project STAR class-size reduction proposal in Ohio would cost taxpayers $499.2 million. [10]
Notes
[1] Eric A. Hanushek, "Some Findings from an Independent Investigation of the Tennessee STAR Experiment and from Other Experiments of Class Size Effects," working paper, March 1999 (Educational Evaluation and Policy Analysis, Vol. 21, No. 2, forthcoming Summer 1999); and Caroline M. Hoxby, "The Effects of Class Size and Composition on Student Achievement: New Evidence from Natural Population Variation," NBER Working Paper 6869 (Cambridge, Mass.: National Bureau of Economic Research, December 1998).
[2] STAR is an acronym for Student-Teacher Achievement Ratio. The State of Tennessee Department of Education funded the original study, while additional research on Project STAR students has been financed by grants from the National Education Association (NEA) and the American Federation of Teachers (AFT), labor unions which represent bus drivers, janitors, food service workers, and public school teachers. See "Small Classes Hold Long-Term Benefits," NEA Today, Vol. 16, No. 7 (March 1998), p. 17.
[3] Project STAR "randomly" assigned 6,341 students to "small" classes (13 to 17 students), "regular" classes (22-25 students), and regular classes with a teacher’s aide. Hanushek, p. 18.
[4] Hanushek, pp. 18 and 22. Only 48 percent of the original number remained at the end of the experiment.
[5] Hanushek, pp. 22-23. Also see Harvey Goldstein and Peter Blatchford, "Class Size and Educational Achievement: A Review of the Methodology with Particular Reference to Study Design," British Educational Research Journal, Vol. 24, No. 3 (1998), pp. 255-268. During the experiment, a significant number of students crossed over from large classes to small classes, strongly suggesting biased results (Hanushek, pp. 23-26). Additionally, the students who quit the project were replaced with others midway through the experiment, adding to the methodological problems (Hanushek, p. 19).
[6] Hanushek, p. 19. Moreover, a large percentage – between 3 and 12 percent – of the students failed to take the required proficiency exams each year, thereby suggesting some bias in the results (p. 24).
[7] Hanushek, pp. 20-21 and 26. Also, since kindergarten attendance was not compulsory in Tennessee at the time the experiment was conducted, it cannot be known whether any positive effects of class-size reduction shown by the experiment are simply due to the fact that some students had attended kindergarten while others had not, especially since the researchers did not control for this fact (pp. 29-31). Hanushek states that, if anything, there is less support for lowering student-teacher ratios at the elementary level (p. 12). Hanushek’s findings confirm other research on class size. A summary of 277 econometric studies analyzing the effect of teacher-pupil ratios is summarized in Eric A. Hanushek, Steven G. Rivkin, and Lori L. Taylor, "Aggregation and the Estimated Effects of School Resources," Review of Economics and Statistics, Vol. 78, No. 4 (November 1996), pp. 611-627. While Project STAR defenders criticize Hanushek’s work by stating that pupil-teacher ratios and class sizes are not the same thing, the best available evidence is that they do tend to move together in the aggregate. See Eugene M. Lewit and Linda Schuurmann Baker, "Class Size," The Future of Children, Vol. 7, No. 3 (Winter 1997), pp. 112-121.
[8] Hoxby, p. 1. Hoxby used natural population variation rather than random assignment (like Project STAR) in her study of 11 years of Connecticut school data.
[9] Public Choices, Private Costs (Dayton, Ohio: The Buckeye Institute for Public Policy Solutions, September 1998).
[10] Estimated by reducing the average regular instruction K-4 class size from 22.6 to 15. The number of K-4 classroom teachers would increase from 25,398 to 38,267, at an average salary of $38,794. Data are from Ohio Department of Education, "State Profile – FY96 vs. FY97," Education Management Information System (EMIS). Reducing the average class size would also involve extensive capital costs for new facilities which are not calculated here.
