Strand Leader: Professor Paul A. Smith , University of Southampton
This strand aims to establish the feasibility and advantages of individual-based sampling and respondent-driven sampling (RDS) to recruit of hard-to-reach groups, and to identify good practice in the use of non-probability sampling.
The research strand will establish the feasibility of using information from administrative sources to enhance the Postcode Address File (the current best-available sample frame) or as an alternative sample frame. This may facilitate greater use of online approaches for surveys, more cost-effective data collection (little or no need for screening) and improved inclusivity by facilitating boosts (e.g. based on ethnicity). Administrative data frames have been employed successfully on special population surveys (e.g. DfE cohorts of young people, COSMO) and health surveys (e.g. REACT, digitrials, clinical trials). RS1 will provide an overview of sources that have recently been used for sampling and a description of the criteria and process for gaining access (with examples) and will investigate the wider feasibility of these approaches, and how these frames can be made accessible to survey practitioners. RS1 will evaluate gains in efficiency and coverage.
Many surveys suffer from under-representation and biases within samples of minority groups, but these groups are often of particular interest. RDS using a probability-based ‘seed’ may provide a relatively robust, but cost-effective way to reach these groups. RS1 will review existing literature on use of RDS to recruit hard-to-reach groups and will produce a report summarising current knowledge and practice, including guidelines for practitioners.
Evidence continues to suggest that data from non-probability (NP) samples (in particular commercial online panels) are less reliable than those from P samples. However, NP samples also provide opportunities to collect data in a more timely and cost-effective manner and reach scarce populations. Combining NP samples with P samples can reduce bias while maximising the achieved sample for a given cost. RS1 will review current practice on collecting data from NP samples and combining them with P samples. RS1 will analyse data from NP samples used in parallel with Natsal and the NatCen panel to compare P and NP samples and experiment with different approaches for integrating them. Based on this RS1 will produce good practice guidelines and tool-kit for assessing and improving the quality and collection of NP survey data, and for blending P and NP samples.
Research Strand 1 (Phase 2 project): Under-represented Population Sub-Groups in Social Surveys. Methods for Respondent Driven Sampling with Probability-Based Seeds
Led by: Dr Olga Maslovskaya, University of Southampton
Co-Investigators: Dr Carina Cornesse, GESIS, and Curtis Jessop, National Centre for Social Research