‘With or without you’? Revisiting territorial state-bypassing in EU interest representation

Abstract

Both the number and the powers of sub-state entities in the European Union (EU) have grown. These sub-state entities represent their European interests using both intra- and extra-state channels. The increasing use of the latter has encouraged scholarly literature to focus on the emerging ‘paradiplomacy’ of these entities. Sub-state paradiplomacy, however, can be both conducted in tandem with its member state or bypassing it. This article seeks to better understand such patterns of interaction between state and sub-state interest representation. Using original survey data, it tests five different hypotheses about the determinants of state bypassing and non-bypassing. It argues that devolution of powers and party politics are relevant factors explaining the frequency of bypassing and co-operative interest representation. Other factors, including size, financial resources and length of exposure to the integration process do not seem to play a role.

Keywords:

• Devolution
• interest representation
• paradiplomacy
• party politics
• state-bypassing

ACKNOWLEDGEMENTS

The author would like to thank Chris Hanretty, Adrienne Héritier, Michael Keating, Charlie Jeffery, Jaap Dronkers, Gary Marks, Patrick Le Galès and the two anonymous reviewers for feedback and advice on a draft version of this paper. A debt of gratitude is owed to the organisers and participants of the CONNEX workshop on ‘Representation’ (EUI, 23–24 April 2008) and more particularly Peter Mair, Jacques Thomassen, Alexander Trechsel and Beate Kohler-Koch for their suggestions on this paper's research design. The author is also thankful to all those – too numerous to list here – who helped test and improve the survey instrument. The author would also like to acknowledge financial support from the Lavoisier programme (French Ministry of Foreign Affairs). Usual disclaimers apply.

Notes

This listing is accessible through the BELO's search engine: http://www.blbe.be/directory/find.asp (last accessed October 2008).

Hence only relative – and not absolute – frequency was sought to be measured. The absolute and relative frequency of these two types of paradiplomacy may not correlate.

In both cases the Kaiser-Meyer-Olkin measure of sampling adequacy is above the minimum threshold of .5 recommended by Kaiser (1974), while Bartlett's test of sphericity is significant at the .001 level (indicating that the R matrix is not an identity matrix) and the determinants are superior or equal to .001 (indicating that multicollinearity is not a concern for these data). The relative size component explains 97 per cent of variance while the absolute size component accounts for 74 per cent of variance. Cronbach's Alpha (based on standardized items) is in both cases above .8, indicating that each scale is reliable.

The attentive reader will notice that the expected signs for the ‘year office opened’ and ‘date joined the EU’ variables are the reverse from that of H5a and H5b. This is because the longer the length of exposure the smaller the value on these variables.

For reasons of presentation, the first two hypotheses are tested simultaneously in the table, as results hardly differ in terms of coefficient and significance when these two explanatory factors are tested separately.

The ‘regionalist/nationalist party’ variable is a survey-based dummy variable.

The United Kingdom (UK) dummy was the only country dummy significant in both models, hence suggesting that the full models fail to account for a UK-specific factor.

Spearman rank correlation tests (Gujarati 2003: 406–8) for the full models (residuals against fitted values) and for each independent variable (against the model residuals) were all insignificant while non-constant error variance tests (Breusch and Pagan 1979; Cook and Weisberg 1983) were insignificant for both full models. Similarly, the Breusch-Pagan-Godfrey (Gujarati 2003: 411–12) tests were insignificant as well as White's general test (Gujarati 2003: 413–14). The tests were performed using Rgui 2.7.1.