Abstract

Representativity is one of the most frequently invoked, and least precisely defined, concepts in communication of statistics. Despite its central role in survey methodology and official statistics, the term is often used without a clear formal framework, leading to questionable interpretations of statistical output and, in some cases, misguided policy decisions. At the same time, declining response rates in surveys are widely cited as evidence of a general crisis of data quality. Yet the relationship between response rates, bias, and inferential validity is far more complex than commonly assumed.
This session critically examines how notions of representativity, and quality are mobilized in the transition from data acquisition (or gathering) over analysis to policy-making. Rather than focusing solely on technical indicators, we argue that a more fundamental question must be addressed: For what purposes do we require data? Data collected for exploratory research, public debate, or media attention may be evaluated differently from data intended to inform legislation or regulatory interventions. The level of evidentiary robustness required depends on the stakes involved.
Recent experiences have highlighted the risks associated with an oversimplified understanding of representativity. Public and political debates have relied on statistical findings without sufficient reflection on technical aspects, such as sampling design, coverage errors, or model-based adjustments. At the same time, the increasing availability of new data sources, including administrative records, digital trace data, and other non-probability sources, challenges classical survey-based paradigms. These data may offer unprecedented scale and timeliness, but they also complicate traditional concepts of inference and population representation.
The session invites a nuanced discussion on how we define and assess evidence in a rapidly evolving data environment. It aims to bridge methodological, practical, and policy perspectives by addressing three core questions: (1) What do we mean by representativity in different empirical contexts? (2) How should data quality be evaluated relative to intended use? (3) Under what conditions can statistical results legitimately serve as a basis for policy decisions? By reframing the debate from technical issues to “fitness for purpose,” the session seeks to contribute to a more responsible and transparent use of data in policy-making.

Speaker 1: Ralf Muennich (University of Trier)
Speaker 2: Denise Britz do Nascimento Silva (Science)
Discussant: Stefano Campostrini (Ca’ Foscari University of Venice)