An Evaluation of Nonresponse Bias Using Paradata from a Health Survey

Abstract

Auxiliary variables that are available for all sample units and related to both the probability of response and the survey variables of interest are potential candidates for nonresponse adjustment variables. The National Health Interview Survey (NHIS) paradata file includes two potentially important sets of auxiliary variables measuring the cooperation and contactability of households in the NHIS sample. In an initial study, these paradata variables were found to have moderate correlations with the probability of response, but weaker correlations with a subset of variables on the NHIS family file. This paper expands this previous study by testing social scientific theories that provide a link between the cooperation and contactability paradata variables and a few key health variables on the NHIS family and sample adult files.