Semiparametric and Nonparametric Methods in Econometrics

Northwestern University, Evanston IL

Investigators

Abstract

This project has three parts. (1) Development of new semi- and nonparametric models that fit data better than existing ones; (2) Development of methods for semi- and nonparametric estimation and specification testing in the presence of instrumental variables, and (3) Development of methods for nonparametric estimation subject to shape restrictions obtained from economic theory. Part (1) is motivated by the observation that existing methods for achieving dimension reduction and, thereby, greater estimation precision in semi- or nonparametric estimation do not necessarily fit the data of interest in applications. The research will develop models and methods that are more flexible than existing ones are but still achieve the dimension reduction necessary to obtain useful results with data sets of practical size. The new models will be applied to data that existing models do not fit well. Part (2) is concerned with nonparametric instrumental variables (NPIV) estimation. One objective is to develop methods of dimension reduction for NPIV estimation. The other is to develop methods for testing a parametric model against a nonparametric alternative in the presence of instrumental variables. There is a large literature on testing a parametric model of a conditional mean or quantile function against a nonparametric alternative, but testing with instrumental variables is a new area of research. Part (3) will develop methods for nonparametric estimation of demand functions subject to monotonicity constraints and the Slutsky conditions of economic theory. There is a literature on estimation under monotonicity and/or convexity restrictions, but the Slutsky conditions are nonlinear and, therefore, present a special challenge. A further objective is to develop methods for testing the hypothesis that the Slutsky conditions are satisfied in a sampled population. The research will have several broader impacts. First, its results will be incorporated into graduate and undergraduate courses that are taught by the investigator and others. Second, the project will support the research of Ph.D. students who will work on problems generated by the project. Third, the results of the project are likely to benefit society at large by being incorporated into policy studies for public agencies. This will be done, among other ways, through the investigator's service on committees and other groups that advise the Federal government.