Inference in Conditional Moment Restriction Models When There is Selection Due to Stratification

University Of Wisconsin-Madison, Madison WI

Investigators

Abstract

When estimating or testing economic relationships, economists often discover that the data they plan to use is not drawn randomly from the target population for which they wish to draw an inference. Instead, the observations are found to be sampled from a related but different distribution. If this feature is not taken into account when doing statistical analysis, subsequent inference can be seriously off the mark. This phenomenon is commonly called selection bias. An important class of models in microeconometrics is characterized in terms of conditional moment restrictions. Thhis develops a new efficient semiparametric approach for conducting inference in models with conditional moment restrictions when the target population is stratified.