Several controversial social policy debates involve disagreements about the outcome and inequality implications of reallocating individuals across groups. Examples of reallocating policies in public education alone include school busing and desegregation, school vouchers, ability tracking and single-sex schooling. Location-constrained rental vouchers, as provided in the Moving to Opportunity (MTO) demonstration of the U.S. Department of Housing and Urban Development, change the assignment of public housing recipients to neighborhoods. Affirmative action admission rules alter the assignment of students to universities; as the passing of the passing of Proposition 209 in California has dramatically illustrated. Despite the prevalence and controversial nature of reallocating policies, statistical methods for measuring their effects are virtually nonexistent. This project would develop estimands which characterize the average outcome and inequality effects of reallocations as well as corresponding estimation and inference procedures.
The potential intellectual merits of the proposed activity include increasing our understanding of how to evaluate reallocating policies empirically. Conventional approaches to program evaluation focus on characterizing the average effect of a policy within an affected population. This information is, at best, only indirectly helpful for predicting the impact of a reallocation. Consider the problem of assigning teachers, of varying quality, to classrooms of students, of varying mean ability. The average effect of a unit change in teacher quality, the traditional evaluation estimand in this context, may be zero, even if the effect of reassigning teachers across classrooms is substantial. The evaluation of reallocations involves a complex interplay between the production technology and the constraints imposed by feasibility. Reallocations are distinct from other policies in that they involve no augmentation, only a redistribution, of resources. This is a feature which helps to explain their often controversial nature.
The broader impacts resulting from the proposed activities includes an expansion of the methodological toolkit available to applied researchers interested in the study of reallocating policies, and, more generally, social spillovers or neighborhood/peer effects. STATA and MATLAB software implementing the proposed estimators are developed. The growth in various forms of single-sex public education has been rapid in recent years, and is likely to expand further given recent changes to Title IX regulations. The researchers also apply their techniques to study the effects of the gender composition of a classroom on student achievement and hence contribute to evaluating the case for single-sex schooling.
