Effects on attrition and fieldwork effort of participation in a supplemental study: evidence from the Panel Study of Income Dynamics
Jul 25, 15:45
A key challenge for longitudinal surveys is to understand the trade-off between increasing the amount of information being collected—through changes in interview length or frequency—and respondents’ willingness to continue participating in the survey—assessed by measures of effort, reluctance, or non-response. However, relatively little is known about this trade-off between respondent burden and participation. We will provide an overview of research on the effects of respondent burden on subsequent survey outcomes; we will also present results from new research examining the effects on survey participation of conducting a supplemental study, a widely used approach to collecting additional information on survey subjects.
Supplemental studies are among the potentially most burdensome activities for participants in longitudinal surveys because they may require a separate interview, interaction with other family members (such as children or a spouse), or providing other new types of data (such as biological samples). We will examine the effects of a supplement on subsequent survey outcomes including attrition, initial refusals, and calls per completed case. Our results will reveal whether this supplement had a positive, negative, or neutral effect on core panel participation and will place these findings within the broader context of the existing literature.
A major challenge in analyzing the effects of respondent burden on survey outcomes is that burden is rarely randomized. Although this situation presents a challenge to our study, we apply a variety of statistical approaches to address the endogeneity of the supplement. In particular, we use a discontinuity approach to examine differences between those who met the supplement’s selection criteria compared to those who just missed the criteria (e.g., based on age-eligibility); we control for observed characteristics using regression analysis and a propensity score reweighting procedure.
We use survey data and paradata from the Panel Study of Income Dynamics (PSID) to examine the effects among householders of being invited to have one or more of their children participate in the original PSID Child Development Supplement (launched in 1997) on their own subsequent participation in Core PSID from 1999 to 2015. Outcomes include non-response on Core PSID as well as paradata-based measures of resistance to participating (such as fieldwork effort and an initial refusal). Our findings indicate that CDS has a modest negative effect on subsequent participation in Core PSID, but that this increase in attrition was offset by the large amount of valuable new data collected on children in the survey.