Ideal auxiliary variables for use in post-survey nonresponse adjustments are associated with both survey variables of interest and response propensity. Auxiliary variables having these properties will generally reduce the bias and variance in survey estimates. Unfortunately, auxiliary variables available for both respondents and nonrespondents to a survey request seldom have strong associations with key survey variables in practice. As a result, large face-to-face household surveys have started to request that field interviewers record estimates and judgments about selected characteristics of all sampled housing units. Although these auxiliary variables may be associated with survey variables of interest in theory, they will be prone to measurement error. Large amounts of measurement error in these observations may have negative implications for survey estimators in terms of the bias and variance introduced by the nonresponse adjustments. Practical techniques for reducing the error in these observations are therefore needed in the field. This article presents results from an analysis of an intervention that was implemented prior to the 15th quarter of the recently completed Continuous National Survey of Family Growth (NSFG). The intervention was designed to provide field interviewers with observable predictors of a key auxiliary variable for which they were recording observations. Analysis of the intervention shows evidence of a significant improvement in the quality of the observations. The article concludes with a discussion of directions for future work in this area.
Country of focus: United States.