Use of Paradata in a Responsive Design Framework to Manage a Field Data Collection

In many surveys there is a great deal of uncertainty about assumptions regarding key design parameters. This leads to uncertainty about the cost and error structures of the surveys. Responsive survey designs use indicators of potential survey error to determine when design changes should be made on an ongoing basis during data collection. These changes are meant to minimize total survey error. They are made during the field period as updated estimates of proxy indicators for the various sources of error become available. In this article we illustrate responsive design in a large continuous data collection: the 2006-2010 U.S. National Survey of Family Growth. We describe three paradata-guided interventions designed to improve survey quality: case prioritization, "screener week," and sample balance. Our analyses demonstrate that these interventions systematically alter interviewer behavior, creating beneficial effects on both efficiency and proxy measures of the risk of nonresponse bias, such as variation in subgroup response rates.