644 – Managing Nonresponse Bias in Sample Surveys
Using Paradata to Understand Effort and Attrition in a Panel Survey
Andrew Mercer
Westat
Current practices in survey research prioritize high response rates as the key indicator of survey quality. Survey organizations often expend large amounts of time and resources to attain responses from more difficult to interview respondents, particularly those that require refusal conversion, tracking or are otherwise hard to reach. Longitudinal surveys face the additional challenge of maintaining a high response rate over multiple rounds in order to minimize the effects of attrition. In this paper, we use contact record paradata from the 2009 cohort of the Medical Expenditure Panel Survey in order to construct three scenarios that simulate the effects of restricting recruitment effort spent on locating, contacting and obtaining cooperation from respondents respectively. For each of these scenarios, we examine differences in the pattern of attrition over the course of the survey and the changes that occur in estimates of key indicators as a result. For this survey, we find that restricting contacting effort results in the largest increase in overall attrition. Reducing effort put toward locating and contacting are found to produce sizeable changes in the resulting estimates. Reducing refusal conversion effort produces a comparable level of additional nonresponse to locating effort, however the impact on estimates is minimal.