Some findings from an independent investigation of the Tennessee STAR experiment and from other investigations of class size effects

While random-assignment experiments have considerable conceptual appeal, th e validity and reliability of results depends crucially on a number of desi gn and implementation issues. This paper reviews the major experiment in cl ass size reduction-Tennessee's Project STAR-and puts the results in the con text of existing nonexperimental evidence about class size. The nonexperime ntal evidence uniformly indicates no consistent improvement in achievement with class size reductions. This evidence comes from very different sources and methodologies, making the consistency of results quire striking. The e xperimental evidence from the STAR experiment is typically cited as providi ng strong support of current policy proposals to reduce class size. Deraile d review of the evidence, however uncovers a number of important design and implementation issues that suggest considerable uncertainty about the magn itude of any treatment effects. Moreover; there is reason to believe that t he commonly cited results are biased upwards. Ignoring consideration of the uncertainties and possible biases in the experiment, the results show effe cts that are limited to very large (and expensive) reductions in kindergart en or possibly first grade class sizes. No support for smaller reductions i n class size (i.e., reductions resulting in class sizes greater than 13-17 students) or for reductions in later grades is found in the STAR results.