Survey Experience and Nonresponse in an Online Probability Panel: A Survival Analysis

Abstract

We fit discrete time survival models to data from an online probability panel, where the outcome is the respondent first nonresponse to a survey invitation, following at least one previous survey completion. This approach has the advantage of utilising information about survey experiences over multiple survey waves, while accommodating the unbalanced data structure typical of OPPs, where the number, timing and content of survey invitations varies widely between panel members. We show that the nature and quality of previous survey experience has a strong influence on the propensity to respond to the next survey invitation. Longer surveys, reporting a survey as less enjoyable, a phone interview, and more days since the last survey invitation are found to be important predictors of nonresponse. We also find strong effects of personality on response propensity across survey invitations. Our findings have important implications for survey designers wishing to minimise nonresponse and attrition from OPPs.