The project has the main purpose of developing flexible statistical methods to analyze the effects of economic factors on the distribution of outcomes of interest. More specifically, our objective is to develop nonparametric distributional and quantile methods to estimate these effects in nonseparable models using cross sectional and panel data. Nonseparable models are important in Economics because they do not restrict the relationship between observable and unobservable variables. For cross sectional data, we analyze the properties of quantile regression series estimators. For panel data, we consider identification and estimation of average effects, quantile effects, and derivatives of structural functions in models with unrestricted individual heterogeneity. These methods can be applied to policy analysis. In particular, we develop inference methods to answer policy questions in rich economic models that allow for multiple sources of individual heterogeneity. For example, we can use panel data to test the hypothesis that the declining union premium across the wage distribution found by Chamberlain (1994) is explained by skill differences )unobserved heterogeneity) among unionized workers.
The project's duration is three years, and it is strictly focused on the following five parts: (1) Conditional quantile processes in large models (series, many regressors); (2) Average and quantile effects in nonseparable panel models; (3) Derivatives of structural functions in nonseparable panel models; (4) Local average and quantile treatment effects in nonseparable panel models; (5) Nonparametric policy analysis.
The nonparametric methods proposed are similar to methods commonly used to analyze mean effects, and expected to be quickly adopted and routinely used for practitioners. They can be implemented using standard software. The inference methods for policy analysis are also expected to have a broad impact since this type of analysis is commonly used in labor economics and other fields. A final purpose of the project is to produce public software in R that implements all the methods developed.
The project has focused on developing flexible statistical methods to analyze the effects of economic factors on the distribution of outcomes of interest. More specifically, the main objective is to develop nonparametric distributional and quantile methods to estimate these effects in nonseparable models using cross sectional and panel data. Nonseparable models are important in Economics because they do not restrict the relationship between observable and unobservable variables. For cross sectional data, we analyze the properties of quantile regression series estimators. For panel data, we consider identification and estimation of average effects, quantile effects, and derivatives of structural functions in models with unrestricted individual heterogeneity. These methods can be applied to policy analysis. In particular, we develop inference methods to answer policy questions in rich economic models that allow for multiple sources of individual heterogeneity. For example, we can use panel data to test the hypothesis that the declining union premium across the wage distribution found by Chamberlain (1994) is explained by skill differences (unobserved heterogeneity) among unionized workers. The main outcomes of the project are following papers: (1) Inference on counterfactual distributions: published in Econometrica. We have developed also a companion command in Stata that implements the methods of this paper. (2) Conditional quantile processes under increasing dimension. We have developed a companion package in the statistical software R that implements the methods of this paper. (3) Quantile regression with censoring and endogeneity: forthcoming in the Journal of Econometrics. We have developed a companion command in Stata that implements the methods of this paper. (4) Average and quantile effects in nonseparable panel models: published in Econometrica. (5) Panel data models with nonadditive unobserved heterogeneity: estimation and inference: published in Quantitative Economics. (6) Nonparametric identification in panels using quantiles: forthcoming in the Journal of Econometrics. (7) Individual and time effects in nonlinear panel models with large N,T. We have developed a companion command in Stata that implements the methods of this paper. (8) Program evaluation with high dimensional data.
