Where Do We Go from Here? Nonresponse and Social Measurement

Surveys undergird government statistical systems and social scientific research throughout the world. Rates of nonresponse are rising in cross-sectional surveys (those conducted during a fixed period of time and not repeated). Although this trend worries those concerned with the validity of survey data, there is no necessary relationship between the rate of nonresponse and the degree of bias. A high rate of nonresponse merely creates the potential for bias, but the degree of bias depends on how factors promoting nonresponse are related to variables of interest. Nonresponse can be reduced by offering financial incentives to respondents and by careful design before entering the field, creating a trade-off between cost and potential bias. When bias is suspected, it can be countered by weighting individual cases by the inverse of their response propensity. Response propensities are typically estimated using a logistic regression equation to predict the dichotomous outcome of survey participation as a function of auxiliary variables. The Multi-level Integrated Database Approach employs multiple databases to collect as much information as possible about the target sample during the initial sampling stage and at all possible levels of aggregation to maximize the accuracy of estimated response propensities.