Randomized Respondent Driven Sampling: A Cellphone Based Approach

Abstract

Respondent-Driven Sampling (RDS) is a widely used method for accessing hidden populations when more traditional survey techniques may not be feasible. However, the reliance on non-random peer recruitment introduces substantial bias, particularly in the presence of homophily. This paper introduces Randomized Respondent-Driven Sampling (RRDS), a novel, cellphone-based adaptation that incorporates researcher-controlled randomization into the recruitment process. RRDS preserves the network-based advantages of RDS while mitigating selection bias by decoupling recruitment from respondent preferences. Through simulation on synthetic networks with high homophily and an empirical application among Bangladeshi garment workers during the COVID-19 pandemic, RRDS demonstrates superior performance in sample representativeness, recruitment efficiency, and convergence to population parameters. The empirical study also reveals gendered constraints in referral behavior, underscoring the importance of context-sensitive implementation. RRDS offers a scalable, remote-compatible alternative for sociological research in hard-to-reach populations, or in populations that are not traditionally hard to reach, but become temporarily inaccessible, such as the case of garment workers during the pandemic.