9810356 Newey This project develops improved methods for drawing inferences from economic data. The methods should help make more accurate margin of error predictions for economic data in several important applications. The proposal includes two specific projects and extensions: (1) Efficient bootstrapping in semiparametric models: Bootstrapping is a resampling method that can be used to improve the accuracy of margin of error predictions. The degree of improvement can depend on how the bootstrapping is done. This project develops a theory for determining the most efficient bootstrap available. Preliminary results indicate that there is large potential for improvement from an efficient bootstrap. New applications of the efficient bootstrap will be used to further illustrate the potential gain. (2) Selecting the number of moment conditions: Generalized method of moment estimators have many applications in economics, including estimation of program evaluation models and dynamic models of growth using a time series of cross-sections. Often there are many more moment conditions than parameters available and it is important to know how to choose among them. This research will develop a method based on the tradeoff between bias and variability. The theoretical properties of this method will be studied and applications provided. Also, the practical value of this approach will be evaluated in some small simulation studies.
