Evidence, Politics, and the Class Size Debate

Abstract

While the debate over class size reduction is largely a political one, some recent discussion has focused on the evidentiary base. Krueger (2000) provides two re-analyses that tend to support reduction policies: a re-interpretation of existing econometric evidence and a proposed demonstration that the small expected effects of class size reduction policies are worthwhile. Unfortunately, as shown here, neither is very convincing. First, a review of past experience from large-scale class size reduction shows no discernible effects on student achievement. Second, while Krueger’s re-interpretation of the econometric evidence suggests more positive outcomes than prior reviews, his results come from placing much larger weight on low quality statistical estimates. Third, reliance on results from Project STAR does not change the policy conclusions. Finally, his benefit-cost calculations presume that more efficacious policies such as those operating on teacher quality are not feasible and that the only alternative is the current system. His demonstration that the small effects suggested by Project STAR might still be good policy relies on a series of heroic assumptions and entirely neglects any of the uncertainty in his illustrative calculations. 1 Stanford University, University of Texas at Dallas, and National Bureau of Economic Research. Helpful comments were provided by John Kain, Alan Krueger, Steve Landsburg, Ed Lazear, Terry Moe, Paul Peterson, Macke Raymond, and Steve Rivkin. Support was provided by the Smith Richardson Foundation and the Packard Humanities Institute. Evidence, Politics, and the Class Size Debate Eric A. Hanushek With the suddenness of a summer storm, politics thrust the issue of class size policy onto the national agenda. Before the political popularity to voters of reductions in class size became known, most educational researchers and policy makers had discarded such policies as both too expensive and generally ineffective, leaving only teachers unions and other with clear vested interests in the policies to support such ideas. When the political appeal of class size reductions became known – largely through the reactions to the 1996 California policies – there was a scramble to backfill evidence supporting such policies. In this current environment, the evidence about the effectiveness of class size reduction has been thoroughly spun in the political debate in order to match the preconceived policy proposals, making it difficult to conclude that the debate has been guided very much by the evidence. This political backdrop is necessary to understand the significance of Alan Krueger’s reanalysis of the existing evidence on class size (Krueger 2000). He focuses attention directly on the scientific evidence and its implications for policy, thus attempting to move the policy debate away from pure politics and toward a better basis for decision-making. While he offers no new evidence on the effects of class size on student performance, he contributes two different analyses that point toward a more aggressive policy of class size reduction: a massaging of the econometric evidence on effectiveness of class size reduction and of overall spending and a proposed demonstration that small outcome effects are still worthwhile. Upon careful inspection, however, neither is convincing. Nonetheless, policy makers should not ignore the emphasis on the importance of a solid evidentiary base. Because supporters of class size reductions are likely to be attracted to his defense of such policies, it is important to understand the nature and substance of his analysis. First, his discussion omits mention of the long history and dismal results of class size policies. Second, his analysis of the existing econometric evidence derives its results from giving excessive weight to

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 106 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.