Methodological evaluation of respondent driven sample in Istat LGB experimental survey

Abstract

The Respondent Driven Sampling (RDS) technique was first applied by Istat (the Italian National Institute of Statistics) in the 2022 “Survey on Labour Discrimination against LGB (lesbians, gay and bisexuals) people not in Civil Union”. The RDS is a valuable approach for studying populations that are difficult to reach, such as LGB people, thanks to its robust theoretical basis. However, the validity of the samples it produces depends on strict assumptions about network structure, the recruitment process, and the sample/population. Furthermore, its implementation is particularly sensitive to operational constraints, including privacy concerns. This work provides a methodological evaluation of the RDS sample obtained in the aforementioned survey. The aim is to identify critical issues related to design choices, implementation limitations and other factors, such as network characteristics and recruitment dynamics. The analyses focused on sample convergence and dependence on seeds, along with potential sources of recruitment bias, including network bottlenecks and homophilic behaviour among participants. The results highlighted several factors that undermined the inferential validity and representativeness of the sample. However, the Italian experience demonstrates the RDS’s ability to engage with populations that are usually under-represented in probability-based surveys. It also contributes to the wider debate within official statistics on the use and enhancement of non-probability sampling methods, and on combining these with probability-based techniques.