Estimating Prevalence of Forced Labor in the Brick Kiln Industry in Pakistan: A Comparison of Two Sampling Methodologies

Author: Hameed, Waqas; Wirtz, Andrea; Lakhani, Sohail; Mahnoor Rizvi, Syeda; Nafees, Asaad Ahmed; Shah, Sajid; Fatmi, Zafar & Robinson, W. Courtland

Abstract: Forced labor is a global human rights and development issue. Population estimates are critical to monitoring forced labor, but estimates are elusive due to its hidden, complex nature and there is little research into optimal sampling methodologies. We compared sampling methods from two parallel surveys to estimate the prevalence of forced labor among workers in the brick kiln industry in Sindh Province, Pakistan. We compared a multi-stage cluster random sample survey with probability proportional to size (PPS) to respondent-driven sampling (RDS), with N = 800 per method. Overlapping confidence intervals of weighted population estimates generated by each methodology were interpreted as no evidence of difference by methodology. We observed differences in estimates for three of seven domains of forced labor, specifically violence, employment practices and penalties, and recruitment, with higher estimates produced by RDS. Thirty-three percent (RDS) and 16.5% (PPS) of brick kiln workers were estimated to experience at least one of two thresholds used to operationally define forced labor (RDS: 33.6%, 95% CI: 29.0–38.7; PPS: 16.5%, 95% CI: 12.1–22.2). Differences in estimates are potentially attributed to differences in sampling methodology or factors associated with its implementation. Further research is needed to determine which methodology is optimal to measure population prevalence of forced labor.

Keywords: human trafficking, forced labor, prevalence estimation, respondent-driven sampling, probability proportional to size, Pakistan