ABSTRACT
If European Union (EU) member states realise differentiations in EU Treaties, what effect do we see on public and political support for future integration? We argue on the basis of a two-tier integration theory and postfunctionalism that differentiations of member states do lead to a preference for slower future integration by its citizens and parties. Once citizens and parties are used to opting out, they demand more of the same in the future. We test our arguments with time-series cross-sectional data for 1994–2018 on all voluntary primary law opt-outs in the EU. Our panel matching estimates demonstrate that opt-outs decrease integration support. After a differentiation, parties become more Eurosceptic on average and publics express a lower preference for future integration. This suggests that differentiated integration is not a cure against Euroscepticism that leads to a unified EU in the future but rather reinforces two-tier integration.
Acknowledgements
Previous iterations of the paper were discussed and improved by Melike Akkaraca Kose and Martin Moland at the 79th MPSA Annual Meeting in Chicago (2022) and the 11th Biennial Conference of the ECPR Standing Group on the European Union (2022). We thank the two referees and Frank Schimmelfennig for very helpful comments as well as Han-Nuri Kim for research support.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Notes
1 Since at least Jensen & Slapin, Citation2012 used the term multispeed integration together with DI, ambiguities exist how the dimensions time, space, and matter should be translated into realisations of differentiation. Whereas Telle et al. (Citation2022) and De Blok & De Vries (Citation2023) equate the two-speed concept with enhanced cooperation (the coalition of the willing) as opposed to a two-tier concept that is manifested in opt-pus (for the laggards), other authors are less deterministic in the demarcation (Holzinger & Schimmelfennig, Citation2012). We stick to the broader definition of Leruth et al. (Citation2022, p. 10) that uses differentiation as an umbrella term ‘covering a wide range of (dis)integrationist techniques such as multi-speed Europe, variable geometry or à la carte Europe’. We however call the mode of DI that we are focusing on in accordance with Telle et al. (Citation2022) ‘two-tier’. Nonetheless, the two different tiers produce also different speeds of integration, which we use as a measure of the dependent variable.
2 We discuss the implications of our results as research agenda that could be applied to enhanced cooperation in the conclusion.
3 We thank referee 1 for that thought.
4 One caveat emerges if we apply the responsiveness logic to DI. The thermostatic model assumes the possibility of governmental overshooting, i.e. too few or too many realised opt-outs. Until the recent cancelling of the Danish opt-out from Common Security and Defense Policy through a referendum in June 2022, there were hardly any examples for a governmental overshooting in terms of too many opt-outs leading to a preference for less DI.
5 We think that the desired speed indicator has similarities to survey items about EU integration support in the European Social Survey (e.g., regarding whether EU integration ‘has gone too far’ or ‘should go further’). Compared to this data, however, the desired speed indicator from the EB covers more countries and years, and taps more directly into the notion of multiple speeds of integration.
6 To echo footnote 1: Whereas there exists a distinction between temporary (two-speed) and permanent (two-tier) DIs in the conceptual literature, we do not think that this distinction is meaningful for our analysis, since we assume that citizens neither know nor perceive a difference between the two.
7 The lagged window for matching is also considering whether a control condition was treated in the past. We restrict the selection of control units as to avoid including control cases that were treated in the three-year lag window. This, at least partially, accounts for the potential confounding role of past treatments. The underlying bias-variance trade-off of expanding the lagged window only allows us to partially account for this source of bias.
8 A logit regression of missingness in the speed variables using these 11 covariates, country-fixed effects, and a second-degree polynomial – as in our MI imputation approach outlined above – reports a value of area under the ROC curve (AUC) of 0.86, demonstrating a good predictive power.
9 This limitation is also related to the inability of our panel matching method to model cumulative treatment effects, meaning the accumulation of opt-outs over time. This limitation, however, most likely underestimates the effect of DIs on popular perceptions, rather than inflating our results. Indeed, cross-sectional analysis demonstrate a robust link between a MS's DI history and popular preferences for DI (Moland, Citation2022).
Additional information
Funding
Notes on contributors
Thomas Malang
Thomas Malang is head of the research group ‘Legislatures in International Politics’ at the Department of Politics and Public Administration, University of Konstanz.
Dominik Schraff
Dominik Schraff is an Associate Professor in the Political Sociology Group at the Department of Politics and Society, Aalborg University.