Are Item Non-Response Problems Ignorable in Survey Data? An Analysis Using Current Population Survey Data

Abstract

This paper analyzes the wage non-response problem using the 2011 Current Population Survey. I find that the increasing wage non-response cannot be ignored, as including the imputed wages in a OLS regression leads to an imputation bias. In addition, the imperfect match criteria, such as not using union status to find a hot donor, is a main source of the imputation bias. I conclude that dropping the imputed values provides better estimates than simply including the imputed wage values in a wage equation.