One essential feature common to all demographic phenomena is variability across units of analysis. Individuals differ greatly not only in attribute and outcomes of interest to social and behavioral scientists, but also in how they respond to a common treatment, intervention, or stimulation. We call the second type of variability heterogeneous treatment effects. The proposed research assembles an interdisciplinary team, encompassing such diverse fields as sociology, economics, statistics, and demography, to investigate the consequences of and methodological approaches to heterogeneous treatment effects. The proposed research has five specific aims: 1. It will demonstrate, with combined observational and experimental data, heterogeneous treatment effects in demographic research. 2. It will develop new statistical methods for estimating heterogeneous treatment effects using the instrumental variable approach with quasi-experimental data. 3. It will further demonstrate the usefulness of estimating heterogeneous treatment effects in observational data with propensity score methods, especially with applications to studies of the impact of family-level shocks on children's psychosocial skills. 4. It will demonstrate, through micro-level simulations, that in the presence of heterogeneous treatment effects, the use of standard statistical methods may give rise to treatment effect estimates with compositional biases. 5. It will develop a set of diagnostic and analytical tools that will help researchers and practitioners o analyze heterogeneous treatment effects in demographic research. The proposed research will provide new knowledge and practical solutions for dealing with heterogeneous treatment effects in demographic research. We will make computer programs produced by this research publically available. We expect several articles and a book to result from the project.
The proposed research will provide new knowledge and practical solutions for dealing with heterogeneous treatment effects in demographic research. While the methodological work will be applied only to six selected studies in the proposed research, this work has general implications for almost all demographic and medical research dealing with causal inferences for human populations in such areas as socioeconomic status and health, family and children's wellbeing, environment and health, immigrants' assimilation into the U.S., prevention education and health outcomes, and social networks and AIDS.
