Answers are grounded in this paper and cite their sources.
Response rates in labour force surveys have declined over time in many countries, posing risks to the quality of the collected data. Comprehensive analyses of trends in participation and respondent composition are therefore important for evaluating labour force data quality. In this article, we analyse cross-sectional response rate and nonresponse bias trends over a period of twenty-one years (2003–2023) in the Icelandic Labour Force Survey (IS-LFS), a quarterly telephone survey of individuals. Specifically, we decompose nonresponse into three parts: (1) unfielded noncontacts that lack a known telephone number, (2) fielded noncontacts that have a known telephone number, and (3) refusals, and evaluate their relative contributions to response rate and nonresponse bias trends. Our results show that response rates have declined from 81.5 percent in 2003 to 49.7 percent in 2023, averaging 1.5 percentage points per year, while the average absolute relative nonresponse bias (unweighted) more than doubled over the same period. These trends are primarily driven by noncontacts (fielded and unfielded) rather than refusals, particularly since 2013. In recent years, having an immigration background has been the strongest predictor of nonresponse, followed by age and education. Individuals not in the labour force are consistently less likely to participate relative to employed persons. The IS-LFS overestimates the share of the population that is employed by 5.6 percentage points on average relative to register benchmarks, which is reduced to 3.8 percentage points by weighting. We conclude with a discussion of the practical implications of these findings for labour force surveys and sample surveys more generally.
No takes yet. Share an insight, caveat, or question.
Einarsson et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: