Optimizing Protocols for Random Digit Dial Surveys: Evidence from Low-and Middle-Income Countries

Abstract

In low-and middle-income countries (LMIC), Random Digit Dial (RDD) telephone surveys are increasingly relied upon to reach populations where face-to-face, web or postal surveys are too costly or infeasible. Implementing an RDD survey requires choices about an enumeration team's distribution of effort and objectives in maximizing data quality. We provide evidence on the tradeoff between a survey protocol that increases contact effort by increasing the maximum attempts per case versus decreasing maximum call attempts to increase number of cases attempted. We also compare survey performance by time of day and day of week to optimize timing of the first attempt per case. Our results are drawn from nine RDD surveys in Africa, Asia, and Latin America. We estimate the effect of these protocol tradeoffs on survey contact, completion, and sample composition. We find: (1) Repeat calling and rescheduling for a given case is costly, increasingly for each attempt, but does generate detectable changes in sample composition; (2) Compared to morning or evening calls, early afternoon calling reduces cost per observation by a small but statistically significant amount, and produces no detectable sample composition tradeoff; and (3) The day of week has no effect on completion rates nor sample composition, including weekday versus weekend.