ABSTRACT
Further, in most cases, the studies are not randomized experiments. Some studies are better controlled than others; some studies are more externally valid than others. Does the choice of method affect the conclusion? Should we dispense with a study just because it is not well con trolled, even though it is nationally representative? Should we ignore a study that is well con trolled, but based on data from a single site? Meta-analysis can provide answers to such questions. Further, meta-analysis can help program managers understand why some programs appear to work better than others. For example, it may be able to help managers evaluate whether central ized or decentralized management of statewide programs makes programs more “effective,” as suming that effective has a specific definition or set of definitions.