Three essays in applied econometrics

My dissertation is motivated by the general question of how one can econometrically account for heterogeneity in economic outcomes. Chapter 1 studies issues related to variation in socioeconomic outcomes across different groups in the society and measuring the strength of social interactions within groups. Chapter 2 is on welfare system improvement and the evaluation of new social programs. Explaining cross-country differences in growth patterns is the focus of Chapter 3.; Chapter 1 presents a framework for the estimation of models with social interactions where the composition of groups affects the structure of these interactions. In this chapter I derive identification conditions for a heterogeneous social interactions model and propose an instrumental variable (IV) estimator. The framework is applied to estimate peer effects on students' test outcomes using the data from Project STAR where the heterogeneity of social interactions is caused by varying duration of being classmates. Controlling for this heterogeneity produces smaller estimates of feedback effects than previously reported and leads to different policy recommendations.; In evaluating the effects of new social programs heterogeneity in the composition of treatment and comparison groups can be accounted for by using matching methods. Chapter 2 introduces a nonparametric estimator for one such method, the propensity score. My proposed estimator performs well when the underlying model is nonlinear, and reproduces the experimental values of treatment effect with high accuracy when used on National Supported Work Demonstration data.; Chapter 3, based on joint work with Steven Durlauf and Andros Kourtellos, focuses on the failure of conventional linear growth models to allow for heterogeneous production functions for aggregate economies. I address the heterogeneity by developing a local approach to the Solow growth model. The parameters of the local Solow model vary smoothly with some initial conditions such as per-capita income. I find that evidence of parameter heterogeneity is strongest for the poorer economies in the sample. The estimated correlation curve shows a monotonic tendency for the Solow growth model to better capture growth variation for richer than poorer economies.