Comments Eric Alan Hanushek and Mariano Tommasi Eric A. Hanushek: Much of education is driven by fads. Most frequently, these fads involve issues of curricula or programs. Some, however, involve the organization and management of schools, such as the degree to which decisions are made by central authorities as opposed to local school personnel. A common thread among all of these fads is that they are seldom systematically evaluated in terms of their impacts on student performance. That characteristic solidifies their position as a fad, because, lacking information about success or failure, they are subject to being replaced by the next fad. In this light, Galiani and Schargrodsky's analysis of decentralization is refreshing. They address the impact of educational decentralization in Argentina in a serious and thoughtful way, and they provide some preliminary support for policies that would move more decisionmaking out to the local provinces and schools. Decentralization has been advocated by a variety of people and organizations, including the World Bank. The strength of support for this idea has not, however, been accompanied by any substantial evidence about the success of such policies. As Galiani and Schargrodsky point out, even though there is clear support for such policies on the grounds of information about local conditions and needs, there are other arguments about factors that could mitigate, if not reverse, the advantages of decentralization. They identify issues of local political economy and local bargaining, of spillovers and externalities, and of decisionmaking capacity. This list could easily be expanded. Thus they quite correctly point out that the advantages and disadvantages of decentralization are empirical questions. Their work capitalizes on a policy change in Argentina during the 1990s that called for the decentralization of federal secondary schools to the provincial level. Because different provinces had historically different levels of federal school provision and because the decentralization occurred at different times across the provinces, there is variation in the decentralization process that can be exploited to try to identify its impacts. [End Page 303] Identifying the causal impact of decentralization is nonetheless difficult. Many factors that affect achievement are likely to be correlated with the amount of decentralization—either casually or simply by association. It is clearly very difficult to identify and measure all of these other things, making the problem of omitted variables bias very real. Galiani and Schargrodsky pursue a clever idea: many of the factors that might differ by province could be accounted for by considering the difference in performance between public and private schools. Factors such as economic conditions, parental education, language differences, or general governmental programs outside of schools might reasonably be thought of as affecting both public and private school students. By defining their performance measure as the difference between public and private school achievement, they can thus eliminate any common elements. What this approach does not deal with is any factor that has a differential effect on the two sectors. For example, private school attendance is not randomly determined, and the factors involved in this might naturally intrude on their identification strategy. If the public schools in a province are more attractive than private schools and thus are able to bid away both teachers and students from the private sector, then there are obvious selectivity differences that do not fall out with their strategy of contrasting public and private performance. More seriously, if decentralization in fact made public schools more attractive over time to teachers and parents, the change in selectivity could be directly related to the amount of decentralization. Similarly, if private schools tended to be more urban than public schools, a variety of factors including economic conditions could operate differentially on the two sectors. The authors provide a useful discussion of potential advantages and disadvantages of their strategy, but there is a limit to how far they can go without undertaking more detailed analyses of the differences between the sectors. On net, their strategy is undoubtedly useful. Nonetheless, some concerns necessarily remain because it relies on strong assumptions about the operation of the two sectors. I therefore interpret their results as suggestive, but surely not definitive. With regard to the details of the analysis, one real concern involves the measurement of...