Cutting the costs of attrition: Results from the Indonesia Family Life Survey.

Attrition is the Achilles heel of longitudinal surveys. Drawing on our experience in the Indonesia Family Life Survey, we describe survey design and field strategies that contributed to minimizing attrition over four waves of the survey. The data are used to illustrate the selectivity of respondents who attrit from the survey and, also the selectivity of respondents who move from the place they were interviewed at baseline and are subsequently interviewed in a new location. These results provide insights into the nature of selection that will arise in studies that fail to track and interview movers. Attrition, and types of attrition, are related in complex ways to a broad array of characteristics measured at baseline. Our evidence also suggests attrition may be related to characteristics that are not observed in our baseline. We draw on data from a Survey of Surveyors and describe characteristics of both the interviewers and the interview that predict attrition in later waves. These characteristics point to possible strategies that may reduce levels of attrition and may also reduce the impact of attrition on the interpretation of models estimated with longitudinal data.