Instead of using long questionnaires administered in person, researchers are increasingly turning to phone surveys, which require shorter instruments but can be administered over multiple, shorter interviews. A limitation of high-frequency phone surveys is study attrition, where individuals enrolled in a baseline survey may not be reachable or willing to complete follow-up interviews.
This brief shares some evidence on phone survey attrition calculated from existing data collected in the early 2010s in Tanzania and Senegal. In these cases, the researchers distributed devices to respondents, ensuring the best-known conditions for minimizing attrition. In addition to presenting attrition rates calculated over multiple survey waves, the brief explores whether there is differential attrition by respondent type, examining changes to the sample composition. Differential attrition can lead to bias in the parameters that researchers are trying to estimate.
The results show that attrition was low after the initial drop-off from in-person baseline to first phone follow-up, but there were small but statistically significant differences between the characteristics of attriters and non-attriters in multiple follow-ups. These data provide a useful benchmark and some cautions for planning future studies.
Authors: Janina Roemer, Michael Rosenbaum