This proposal furthers the investigator?s previous work on identification and inference in econometric models and involves a detailed research agenda. The emphasis in this agenda is on the identification and estimation problems which arise in parametric models with plausible and easily interpretable assumptions. This proposal consists of four projects that study identification and inference in empirically relevant models. The proposal provides empirical researchers with a richer menu of approaches to tackle problems using more robust methods, and hence advances the econometric literature in relevant and important directions.
The proposed research projects in this proposal provide an array of econometric models that are practical, useful, and of immediate relevance to empirical work. The proposed econometric models advance research on identification and estimation of the identified features in important classes of parametric models. For example, in the first project generalizes the classical selection model of Heckman and Gronau to the multivariate case with multiple decision makers where the participation stage represents a game of complete information. This model is highly relevant to empirical economists estimating differentiated product demand for example in that it allows them to examine the effect of selection, or endogenous product qualities, on parameter estimates. The second project plans on providing methods to conduct inference in semiparametric likelihood models that are robust to failure of point identification. This project is relevant to a wider class of models that are used in both labor and industrial organization. This sensitivity analysis approach to inference is important and can be used to answer similar questions in a wide variety of partially identified models. The third project introduces the novel concept of median uncorrelation that parallels the familiar and heavily used concept of mean (linear) uncorrelation. We then easily show that this concept leads to transparent estimators for linear instrumental variable regressions with quantile restrictions. This is a widely used area in empirical economics. The fourth project furthers the work of the investigator on inference in nonlinear models with generalized forms of censoring by providing new ways to estimate Roy type selection models with panel data. This class of models is very important and empirically relevant and illustrates this approach in an empirical example using real world data.
Broader impacts of the proposal: The research described in this proposal provides a new framework for inference, and outlines practical empirical strategies that can be widely used. It provides applied economists and policy makers with a richer set of tools to conduct robust inference and influence policy. This proposal also integrates the research agenda on identification and inference in econometric models into an educational plan that will not only directly benefit graduate students, but will also provide research opportunities for outstanding undergraduates, and involve and interest computational economists and computer scientists.
