Shielding free movement? Reciprocity in welfare institutions and opposition to EU labour immigration

ABSTRACT

Public attitudes towards the free movement of workers in the European Union vary substantially between countries and individuals. This paper adds to the small but growing research literature on this issue by analysing the role of national welfare institutions. We investigate the relationship between the degree of ‘institutional reciprocity’ in national systems of social protection and attitudes to EU labour immigration across 12 European countries. We do not find evidence of an effect of institutional reciprocity on opposition to EU labour immigration among the public at large. However, institutional reciprocity appears to matter for economically vulnerable groups. We identify an interaction effect indicating that higher degrees of institutional reciprocity in national social protection systems, and in unemployment insurance systems specifically, are associated with lower levels of opposition to EU labour immigration among unemployed people. Hence, reciprocity in welfare state institutions appears to shield free movement from opposition, at least among vulnerable groups.

KEYWORDS:

• European Union
• free movement
• public attitudes
• reciprocity
• social protection systems

Acknowledgements

For their helpful comments, we would like to thank three anonymous reviewers and the editors of JEPP. We are also grateful to our colleagues in the Horizon-2020 funded REMINDER project and participants in workshops at Oxford, Uppsala, and the EUI.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Statistical replication materials and data

Supporting data and materials for this article can be accessed at https://doi.org/10.7910/DVN/HTYQVK.

Notes

1 Our expectations apply to the entire population of citizens and long-term residents in host countries, including persons with a migrant background.

2 Our data indicate that these two aspects of reciprocity are empirically distinct and hardly even correlated (Pearson’s r –0.10, see Appendix Figure A5).

3 Only 12 clusters can be an issue when using clustered robust standard errors. We follow the recommendation by Cameron and Miller (2015) and use a correction for the degrees of freedom. Additionally, we run a robustness test using the wild cluster bootstrap procedure.

Additional information

Funding

The research for this paper was supported by the European Uni, Horizon2020 ‘REMINDER’ project [grant number: 727072].

Notes on contributors

Moa Mårtensson

Moa Mårtensson, researcher, Department of Government, Uppsala University, Uppsala Center for Labor Studies (UCLS).

Marcus Österman

Marcus Österman, researcher, Department of Government, Uppsala University, Uppsala Center for Labor Studies (UCLS).

Joakim Palme

Joakim Palme is Professor at the Department of Government, Uppsala University, the Uppsala Center for Labor Studies (UCLS).

Martin Ruhs

Martin Ruhs is Professor of Migration Studiesat the Migration Policy Centre, European University Institute, and Associate Professor of Political Economy (on leave) at the Department for Continuing Education, University of Oxford.