Politics Divided from Society? Three Explanations for Trends in Societal and Political Polarisation in the Netherlands

Abstract

A growing political polarisation on ethnic integration policy is characteristic of current discussions in Dutch politics. The preferences of Dutch citizens, by contrast, remain fairly stable over time. Thus, polarised politics in the Netherlands is assumed to grow apart from the preferences of ordinary citizens, leading to a gap between politics and society. The present article describes and compares trends in societal and political polarisation on ethnic integration policy in the Netherlands between 1994 and 2006. Three mechanisms are explored that explain a discrepancy between trends in political and societal polarisation: (a) parties' responsiveness to political elites, (b) mean partisan representation, and (c) issue salience. Analyses of data from Dutch election studies and party manifestos reveal the existence of a discrepancy in trends. Political polarisation appears to be associated with trends in mean partisan polarisation and in issue salience, and not with trends in political elite polarisation.

Acknowledgements

Previous versions of this article were presented at the Annual Conference of the Dutch and Flemish Political Science Associations, Antwerp, 31 May–1 June 2007; at the winter meeting of the European ‘Polarization and Conflict’ (PAC) project, La Palma de Mallorca, 10–11 December 2007; and at the Groningen-Utrecht Workshop on Polarization, Groningen, 16 January 2008. We thank participants of these workshops for critical remarks and comments. We would also like to thank the editor and two anonymous reviewers for their helpful suggestions. Torenvlied acknowledges support from the Polarization and Conflict Project CIT-2-CT-2004-50604, funded by the European Commission DG Research Sixth Framework Programme and the Netherlands Organization for Scientific Research (NWO), VIDI grant 452-06-001.

Notes

1. Esteban and Ray derive their measure from three axioms that should be satisfied, as well as a requirement of invariance to population size. The first axiom is that polarisation increases when two small groups are pooled while keeping the average distance with a third group constant (Esteban and Ray 1994: 832–3). The second axiom is that polarisation increases when the distance between a small intermediate group and a larger (but not largest) group becomes smaller (Esteban and Ray 1994: 832–3). The third axiom is that polarisation increases when a coalition supporting a moderate position is split into two separate groups (Esteban and Ray 1994: 832–3). These axioms and requirements result in a measure of polarisation which: (a) takes into account differences between groups in preferences, and (b) includes the relative group sizes. See Appendix I for a more detailed description of the ER measure of polarisation.

2. Van der Eijk (2001) also points to the lack of the standard deviation as a measure of dispersion or polarisation, by showing that the standard deviation is highly sensitive to skewedness of distributions. The standard deviation is therefore inappropriate as a measure of dispersion (Van der Eijk 2001: 328).

3. An alternative is to use the data of the Manifesto Research Group (Budge et al. 2006). However, these data cannot be used because they do not include all parties, which is necessary for a measure of polarisation. Furthermore, these data are based on one single and highly unstable item. For example, for the year 2002 the Manifesto Research Group places the Fortuyn party at a middle position, while the Green party is placed at a conservative position, promoting a mandatory cultural assimilation of minorities.

4. The WRR reported about the political discourse on ethnic integration in the Netherlands. The report includes a content analysis of the Party Manifestos of Dutch parties between 1998 and 2006.

5. Parties with similar positions are merged into one policy coalition, with party proportions aggregated. As a criterion for a similarity of mean partisan positions we applied a maximum of one percentage point difference on the scale. Also note that we apply the same proportions – based upon the number of seats in parliament – to calculate political polarisation and mean partisan polarisation. Consequently, differences in values of these two measures are due to differences in positions and not to differences in proportions. We checked our results to find out whether similarity is driven by the weighting – computing trends with and without weights for party proportions – and found no evidence for this. Results without the weighting replicate those with weighting.