9321024 Honore This project contributes to our understanding of the properties of statistical models that are widely used in economic research. The research will provide whole new classes of econometric estimators that are not subject to some of the limitations of existing methods. The results from this project will be especially useful for a more thorough analysis of economic panel data such as the survey data from the Panel Study of Income Dynamics and for the econometric analysis of the duration of unemployment. More specifically, the project continues research on the estimation of semiparametric models in three areas. The first part of the project shows that it is possible to modify the large battery of estimators for the censored regression model in the econometrics literature in such a way that they can be applied when the censoring points are random and not known for all individuals. The advantage of this approach is that the resulting estimators will make weaker assumptions than the proposed in the literature. For example, the investigator will be able to construct estimators that require only conditional symmetry or quantile independence of the errors rather than independence between the errors and the regressors. The second part of the project concerns the estimation of discrete choice panel data models. It is known that a conditional logit approach can be used to estimate panel data logit models when all the explanatory variables are exogenous or when the only explanatory variables are lagged variables. Thus far, it is not known how to estimate such models when the explanatory variables consist of lagged dependent variables as well as exogenous variables. The approach taken in this project will be used in a logit as well as a semiparametric setting. The third part continues work on econometric duration models. The investigator demonstrated that it is possible to modify his estimators of the parameters of a mixed Weibull distribution i n such a way that rates of convergence arbitrarily close to root-n.
