Estimating the Prostitution Population in the Netherlands and Belgium: A Capture-Recapture Application to Online Data

Anahita Azam , KU Leuven
Jef Hendrickx, KU Leuven
Stef Adriaenssens, KU Leuven

People involved in prostitution and their activities often go unregistered due to the difficulty of acquiring representative data about them. Concurrently, evidence-based risk-prevention policies require that reliable population estimates are built. This study illustrates the feasibility and potential of combining internet data with an advanced application of the capture-recapture methodology to derive policy proposals based on empirically sound population estimates. We make a twofold contribution: deriving population estimates of female sex workers in the Netherlands and Belgium, and proposing a novel data source and an appropriate methodological framework for measuring such hidden populations in regions where the internet plays a predominant role in the industry. The count data are derived from reviews posted about sex workers on a commercial sex review website, from which detailed information about the population of interest is crawled. Since this data source results in a single list of counts based on one or more reviews per sex worker, it requires the relatively newer single-registration design in capture-recapture methodology. The Zelterman approach controls for unobserved heterogeneity bias present in traditional Poisson-based capture-recapture approaches by using logistic regression to model the binary probability of only the first and second counts. Inclusion of covariates further models heterogeneity in the population. The resulting estimates are lower than existing rough estimates. We find that relative to the overall population of the two countries, the proportion of sex workers are roughly identical despite differing legal environments.

See extended abstract

 Presented in Session 73. Issues in health data analysis