Energy policy-making in the European Union between past and present

We use instruments from sociology of quantification to investigate a persistent paradox: the problematic technical quality of modelling studies when these are aimed at policy analysis. We review new and old instances of this paradox, focusing specifically on two energy assessment studies: one performed in the eighties and one recent. We apply sensitivity auditing, an existing checklist for the quality of quantification at the science-policy interface, which appraises the overall model framing process and application. The old energy assessment case is a study that was the subject of a pointed criticism by sociologist of science and technology Brian Wynne in a paper published in 1984. The recent impact analysis is from the European Commission, aimed to inform the selection of greenhouse-gases emission targets for 2040 in the European Union. The similarities in the shortcomings identified by Wynne and those we see in the European Commission one using sensitivity auditing seem to indicate that the lesson from that controversy was not learned, or that at a more fundamental level the purpose of these relevant institutional analyses is not to trace a path to a sustainable energy future but to reassure and confirm present policy agendas and visions, along the lines of a policy-based evidence strategy well known to sociologists of quantification. We investigate how modelling plays a key role in facilitating these analytic distortions, and conclude with some suggestions for progress, anchored to the double nature—technical and normative—of quantification and on the need for more analytic lenses to be systematically deployed when reading a policy assessment that builds on models. • Sociology of quantification helps assess model-based impact assessment quality. • Sensitivity auditing frames modelling at the science-policy interface. • We contrast a 1980s energy study and a 2024 EU impact assessment on emissions. • Auditing reveals persistent flaws when models are used to inform policy. • Key uncertainties and policy-relevant dimensions are often overlooked.