Randomized Recruitment Driven Sampling

TL;DR

Through simulation and an experiment surveying Bangladeshi garment workers during the COVID-19 pandemic, it is demonstrated that RRDS produces less biased estimates and improved confidence interval coverage compared to traditional RDS.

Abstract

Surveys are critical inputs for research and policy, yet, enumerating a sampling frame is logistically infeasible or financially nonviable in many circumstances, such as during pandemics, natural disasters, or armed conflict. Respondent Driven Sampling (RDS) does not require a sampling frame, yet non-random peer recruitment often introduces substantial bias, particularly under high homophily. We introduce and evaluate Randomized Recruitment Driven Sampling (RRDS), a cellphone-based adaptation of RDS that incorporates researcher-controlled randomization into each recruitment wave. While standard RDS is necessary for stigmatized groups where network transparency is infeasible, RRDS is designed for low-stigma populations that become difficult to access due to logistical barriers. In these contexts, RRDS enforces the random recruitment assumption that traditional RDS relies upon but rarely achieves. Through simulation and an experiment surveying Bangladeshi garment workers during the COVID-19 pandemic, we demonstrate that RRDS produces less biased estimates and improved confidence interval coverage compared to traditional RDS. RRDS offers a scalable, remote-compatible alternative for studying low-stigma groups in challenging contexts where large-scale probability sampling is unsafe or infeasible.

Figures (8)

• Figure 1: Simulation results comparing RDS (as practiced: homophilic referrals, no replacement) to RRDS across 12 recruitment waves. Dashed lines indicate true population parameters. Starting from biased seeds (young men), both methods show similar convergence toward population means for age (A) and gender (B), but RRDS achieves substantially larger sample sizes (C) by avoiding repeated sampling of high-degree nodes.
• Figure 2: Comparison of RDS and RRDS estimates across six demographic characteristics. The Tree Bootstrap estimator generally reflects greater uncertainty due to network structure and sampling variability. Randomized recruitment tends to yield less biased estimates.
• Figure 3: Each of the RDS estimates more closely approximate the population mean age from the 2017 baseline survey. The Tree Bootstrap estimator captures additional uncertainty from sampling and network structure, leading to wider intervals than the traditional RDS estimator.
• Figure 4: Proportion of respondents who are native born to the Dhaka region across recruitment waves. The horizontal dashed line represents the population parameter from the representative 2017 survey. This is a very small proportion of the population, and due to the limited sample size of from the network-based sampling it is not possible to recover a reliable estimate of this quantity with either method.
• Figure 5: Proportion of married respondents across recruitment waves. RRDS point estimates are far less biased than traditional RDS. With traditional RDS weights, the RRDS estimate is not statistically different from the true population parameter, while the RDS estimate is.