Issues in Estimation of Structural Economic Models

Princeton University, Princeton NJ

Investigators

Abstract

This project is mostly methodological, and the main contribution will be to provide tools and insights that will be of use for researchers in many areas of economics (and related fields). The focus will be on the estimation of statistical models, which are designed to incorporate and test restrictions implied by economic theory. Applied researchers face many issues when estimating such models. The project will focus on two of them. The first issue is that it can be very time-consuming to produce convincing statistical evidence from these models. The project will introduce and explore a new statistical method, which will dramatically reduce this computational burden. The second issue of this project regards the extent to which it is, even in principle, possible to estimate some standard economic models without arbitrary and untestable assumptions. The findings of this project will be useful for researchers in many areas of applied economics, including labor economics, public finance, industrial organization and macroeconomics, and therefore advances science. Because these economic models and statistical methods are now widely used to measure the effects of business methods and government policies, the research also contributes to the national prosperity. The most important new econometric tool is a bootstrap-like method. The main idea behind the tool is to apply the usual bootstrap method to a number of one-dimensional estimators. The project will exploit the fact that the resulting joint distribution contains all the information that one needs to estimate the standard errors associated with the parameter-vector of interest. This bootstrap-like method will dramatically lower the computational burden associated with estimating complicated models such as structural econometric models. The second intellectual contribution of the project will be to provide a set of new results about Dynamic Roy Models, which will be useful for understanding what one can and cannot hope to estimate from observational data in these models. The project will also deliver a new statistical test of an independence assumption, which is commonly made by researchers estimating duration models. Finally, the project will estimate a dynamic discrete choice model designed to shed light on household retirement decisions. The empirical focus will be on the extent to which one spouse?s retirement time affects the retirement timing of the other.