Lost but not forgetten: Attrition and follow-up in the Indonesia family life survey

Data from three waves of the Indonesia Family Life Survey (IFLS) are used to examine follow-up and attrition in the context of a large scale panel survey conducted in a low-income setting. Household-level attrition between the baseline and first follow-up four years later is less than 6 percent; the cumulative attrition between the baseline and second follow-up after a five-year hiatus is 5 percent. Attrition is low in the IFLS because movers are followed: around 12 percent of households that were interviewed in the first follow-up had moved from their location at baseline. About half of those households were "local movers." The other half, many of whom had moved to a new province, were interviewed during a second sweep through the study areas ("second tracking"). Regression analyses indicate that in terms of household-level characteristics at baseline, households interviewed during second tracking are very similar to those not interviewed in the follow-up surveys. Local movers are more similar to the households found in the baseline location in the follow-ups. The results suggest that the information content of households interviewed during second tracking is probably high. The cost of following those respondents is relatively modest in the IFLS. Although the analytical value of reinterviewing movers will vary depending on the specifics of the research, we conclude that, in general, tracking movers is a worthwhile investment in longitudinal household surveys conducted in settings where communication infrastructure is limited.

Duke Scholars

Author Duncan Thomas Economics