The effects of class size on academic achievement have been studied for decades. Although the results of small-scale, randomized experiments and large-scale, econometric studies point to positive effects of small classes, some scholars see the evidence as ambiguous. Recent analyses from a 4-year, large-scale, randomized experiment on the effects of class size (Project STAR in Tennessee) provide evidence that small classes in the early grades lead to increased academic achievement. A major question is whether these benefits of small classes are uniformly distributed across schools and districts. The analyses reported here suggest that class size effects at the 3rd grade vary across districts and that the variation in effects is linked to average teacher salaries in those districts. The question of how to allocate resources to most effectively further the aims of educational systems is one of the enduring questions facing educational researchers and policymakers. Among the most fundamental allocation decisions is the decision of how to assign instructional staff to classes. This includes the decision of how large classes should be and whether ancillary instructional staff, such as classroom aides, should be used. Manipulating class size is a policy option that is gaining increasing attention throughout the nation. In the last two decades many states have adopted policies that reduce class sizes with the goal of improving achievement, and class size reduction often is included in state and local education initiatives. Although decisions about resource allocation (and class size in particular) are made in every educational system, there is serious debate about whether these allocation decisions influence academic achievement and other desired outcomes of education. This research investigates whether assignment to small classes affects academic achievement and, if so, how those effects are distributed across schools with different social compositions and teacher characteristics. The variable of local district teacher salary and its effects is also investigated. (Contains 2 tables and 13 footnotes.)