Evaluation of Gaining Cooperation Methods for IVR Surveys in Low- and Middle-income Countries

Abstract

Interactive voice response (IVR) is gaining popularity as a data collection method for survey research. In low- and middle-income countries, IVR is used as a primary data collection mode. The system places an out-bound dial; when the individual answers, he/she hears a recorded greeting and invitation to begin the survey. This approach has the benefit of reducing labor costs, but without an interviewer, there is no one to help gain cooperation, answer questions, or identify the appropriate language in which to continue, resulting in low production outcome rates (e.g., cooperation rate, response rate). In this paper, we use experiments embedded in four studies in three countries (Ghana, Malawi, and Nigeria) to test how three design choices affect production and representativeness in IVR surveys in low- and middle-income countries. Specifically, (1) should we send an SMS (i.e., text message) prior to the first IVR contact? (2) Where should we place the language selector within the introduction? (3) Should we notify the individual that they are listening to a recording during the introduction? While some of these design choices resulted in different production outcome rates at different points in the survey, there was no clear effect on overall yields nor on the representativeness of the sample.