Respondent-driven sampling: A new method to sample businesses in Africa

Abstract

Much of our understanding about contemporary African economies relies on survey data from small and medium enterprises. In this study, we apply a new method for sampling enterprises: respondent-driven sampling, or RDS. RDS is a modified method of chain-referral or network sampling, in which survey participants recruit other enterprises in their social network to the study. It incorporates a mathematical model to minimise biases inherent in network sampling. RDS has the potential to complement existing sampling methods, such as household listing, random walks and using existing frames. This study has three objectives: it evaluates the feasibility of using RDS to study enterprises, tests the statistical assumptions underlying the RDS approach and compares the sample characteristics with external data. We applied RDS in a survey of small and medium enterprises in Addis Ababa, Ethiopia. To our knowledge, this is the first time RDS has been applied to enterprises. We find that RDS is a feasible, efficient method for obtaining a high-quality sample of enterprises: 608 enterprises were interviewed within 6 weeks and the statistical assumptions underlying RDS generally held. We also show that RDS captures less established businesses that are less likely to be in surveys based on government and commercial sampling frames. These findings lead to the conclusion that RDS is a viable complement to existing sampling methods.