ESRA 2027 sessions by theme

Back to Overview of Sessions

Session Details

Probability-based surveys remain a cornerstone of population inference, but declining response rates, increasing data-collection costs, and difficulties in reaching population groups put traditional survey designs under increasing pressure. Consequently, non-probability samples (NPS), including opt-in web panels, convenience samples, and digital data sources, are increasingly considered as complementary sources for statistical production. However, unknown inclusion probabilities and potentially strong selection mechanisms pose fundamental challenges for population inference.

Addressing these challenges requires not only appropriate estimation methods but also suitable auxiliary and reference information. Reliable inference from NPS often depends on integration with probability surveys, registers, administrative data, or other data sources. The availability, quality, and predictive power of this information may be as important as the choice of adjustment method itself.

This session focuses on methods for integrating and analysing NPS, with particular attention to the interplay between estimation methods, auxiliary information, and reference data. We invite methodological contributions and empirical applications addressing topics such as:

- integration of probability and non-probability samples, registers, and administrative data;
- identification, selection, and evaluation of auxiliary and reference information;
- propensity score adjustment, inverse probability weighting, calibration, post-stratification, and generalised regression estimation;
- machine-learning approaches to weighting and auxiliary-variable selection;
- doubly robust, model-assisted, and model-based estimation;
- data integration and statistical matching;
- sensitivity analyses and evaluation of assumptions underlying NPS adjustment;
- simulation and empirical comparisons of estimators, auxiliary information sets, and integration strategies;
- applications in official statistics and other fields.

Particular attention is given to studies examining when NPS adjustment and integration succeed or fail, including the role of auxiliary information, selection mechanisms, model specification, and differences between reference and NPS populations. The session aims to improve understanding of the information and methodological conditions required for reliable population inference using NPS.