Essays on Econometric Models for Program Evaluation.
详细信息
- 作者：Park ; Byoung Gun.
- 学历：Ph.D.
- 年：2014
- 毕业院校：Yale University
- ISBN：9781321056310
- CBH：3580795
- Country：USA
- 语种：English
- FileSize：5917434
- Pages：151
文摘
The dissertation consists of three chapters. The first chapter develops a new identification and estimation method for the extended Roy model,in which the agents maximize their utility rather than just outcome. The identification results substantially relax conventional functional form restrictions. No functional form restriction is imposed on the distribution of the potential outcomes. Hence it allows for nonseparable functional forms and/or disturbances of an arbitrary dimension. The utility functions are allowed to be nonlinear so that it can accommodate important features of utility functions such as the concavity of the utility functions. The identification strategy does not use the identification-at-infinity approach,and some features of the parameters,e.g. the utility function and the median treatment effect,can be identified without the large support of the instrument variable. I show that i) when the instrument is continuous,possibly having a bounded support,the model is point-identified on a certain identification region,and that ii) when only discrete instruments are available,sharp bounds for the model are obtained. The key assumption of the method is the monotonicity of the selection with respect to the instrument. Based on the identification result,I propose a nonparametric estimation procedure that builds upon a simulation-based method proposed by Dette et al. 2006). The estimator is easy to implement in practice because it only uses a closed form formula and straightforward simulations. I show that the estimator possesses a standard nonparametric rate of convergence,and examine its efficacy in finite samples by Monte Carlo simulations. Finally,I estimate a model for farmers decisions to adopt a new technology using data from Malawi. While the first chapter considers the extended Roy model,the second chapter introduces a new identification method for the generalized Roy model,in which unobserved heterogeneity in utility is allowed. In contrast to the identification-at-infinity method which focuses on the identification of the marginal distribution,this chapter focuses on identification of the joint distribution of potential outcomes. It is assumed that there exists a special regressor that is independent of the unobserved heterogeneity both in utility and potential outcomes and enters the utility linearly. I show that the model is identified under a large support assumption of the special regressor. The identification method is constructive. An application of the model is concerned with panel data. Consider a dynamic model in which preferences may depend on the selection in the previous period and the heterogeneities may be correlated across periods and/or across sectors. If there are as many available special regressors as the number of periods and they jointly have a large support,the full joint distribution of potential outcomes is nonparametrically point-identified. The third chapter presents identification results for a nonseparable model of causal mediation analysis. I consider the causal effects of a treatment e.g.,smoking status) on an outcome of interest e.g.,baby birth weight). Direct causal effects are the effects of the treatment on the outcome when other covariates are held fixed. However,in many applications,there are some covariates e.g.,mothers health status) such that the treatment affects the covariates,and in turn,the covariates affect the outcome,which is the indirect effects of the treatment on the outcome. I provide a nonseparable model for causal inference in such applications and find sufficient conditions to identify the model. I also propose a quantile treatment effect parameter of total effects and a decomposition of the quantile treatment effect into direct and indirect effects.