Letting silence speak: analysis of nonresponse in survey data from sub-Saharan Africa

Guy Stecklov, Hebrew University of Jerusalem
Alexander Weinreb, University of Texas at Austin

We study data normally seen as expendable or a statistical nuisance: survey nonrespondents. In each survey cluster, some proportion of intended survey respondents are not successfully interviewed. The underlying cause of nonresponse may vary, but nonrespondents are generally coded separately from households who express an overt refusal to participate in a survey. This practice of not incorporating in reported nonresponse rates nonrespondents except in cases when they clearly refuse leads to relatively small nonresponse rates. Nonrespondents, it is argued in the literature, may be very adept in avoiding interviewers using a variety of tactics that reduce the need for refusal. In this case, nonresponse rates themselves offer a unique indicator of willingness to contribute and engage with a program that is targeted at the national level. In this sense, nonresponse rates can be seen as indictors of civic engagement as well as other markers of political engagement. We explore these issues using DHS and Afrobarometer data.

Presented in Session 88: Data quality