With a little help from the state: interest intermediation in the domestic pre-negotiations of EU legislation

Abstract

Theories of interest intermediation maintain that the formation of bargaining positions in the European Union follows a distinctive bargaining style. This article evaluates such claims empirically. It compares the predictive accuracy of Nash bargaining models which take the salient features of five types of interest intermediation into account. We show that the interaction between government agencies, interest groups and parties in the formation of EU legislation is largely étatiste. When important private interests are at stake, the pattern is also quite often clientelistic or corporatist. The dominance of the state in the less politicized decision-making processes is apparent in all four countries under consideration: Finland, Germany, Great Britain, and the Netherlands. Consociationalist arrangements are rare in this arena of public policy-making.

Keywords:

• Applied bargaining theory
• domestic politics
• European Union
• interest intermediation
• Nash bargaining solution (NBS)
• model evaluation

ACKNOWLEDGEMENTS

Previous versions of this paper have been presented at the Pan-European Conference on EU Politics, Bologna, June 2004, the Pan-European Conference on International Relations, The Hague, September 2004 and at seminars at the Free University of Amsterdam, Paris Science Po, and the Max Planck Institute in Cologne. We have received research support from the German Research Foundation and would like to thank Anette Arslantas, Ursula Klöpper, Christian Lau, Michael Weltin, and Carl Zimanky for their assistance. The source code for the calculation of the models, an illustrative case study and some additional material can be found on the replication homepage of the first author at: http://www.uni-konstanz.de/FuF/Verwiss/GSchneider/downloads/daten.htm

Notes

1 See also the discussion on the power of economic interest groups in Grossman (2004).

2 The overall rank out of 15 countries examined, the mean scores and the standard deviation of these four countries are as follows: Netherlands: 4th, 4.0, 1.0; Germany: 6th, 3.5, 0.9; Finland: 7th; 3.3, 1.0; United Kingdom 12th, 1.5, 0.8.

3 We acknowledge that none of the systems of interest intermediation studied here include EU-level actors. The exclusion of supranational institutions and intergovernmental bodies can, however, be justified through the agenda-setting power that the Commission most notably possesses. As the Commission introduces legislative proposals, it implicitly formulates the status quo position for the subsequent national debates.

4 For a similar classification, see Atkinson and Coleman (1989).

5 Schneider et al. (2006) use different versions of the NBS to assess the predictive accuracy of competing models of EU decision-making, in particular the interactions within the Council of Ministers.

6 These power and salience scales also range between 0 and 100.

7 Similar to Schneider et al. (2006), we constrained the possible values of the optimization function to the bargaining space between 0 and 100 and we relied on the constrained optimization routine in GAUSS for the calculation of the optimized solution. We used the Newton algorithm as optimization method and employed the mean as the starting vector in all models. The mean was calculated from the positions of all national stakeholders on each issue of a proposal.

8 We use these power assessments for two purposes: first, we pre-select the actors included in the corporatist model based upon their attributed power and, second, we rely on this measure to calculate the asymmetric NBS models.

9 We owe this suggestion to James D. Fearon.

10 To control for possible problems of independence we calculated each issues probability to ‘meet’ another issue dominated by the same model in the entire dataset which amounts to values between 7.6 and 25.8 percent (depending on the predominant model of interest intermediation for that particular issue). As compared to these numbers, the average probability to ‘meet’ an issue of the same ‘kind’ within one proposal is 43.7 percent. Based on these results, we recalculated our results by averaging for proposal level first. The results do not change significantly. We therefore decided to keep the individual issue as our predominant level of analysis.

11 We have also conducted some sensitivity analyses to identify the factors that contribute to the forecasting error of the models. The main finding is that a polarization in the preferences of the actors exerts a strong influence on the accuracy of the predictions. Other measures of the preference distribution, like the variance or the skewness, are only significant for some models.

12 We also report in the webappendix to this article how frequently a model provided a ‘point’ prediction or was at least very close to the final outcome.

13 The results obtained by using the reference point differ only slightly: the best predictions still offer the étatiste model (MAE over all issues = 0.16), followed closely by a clientelistic model taking only those actors on the right side of the position of the leading ministry into account (MAE = 0.165) and the asymmetric pluralist model (MAE = 0.21). The winners in this contest are followed by the clientelistic model with groups on the left side of the leading ministry (MAE = 0.23) and the symmetric pluralist model (MAE = 0.27). The corporatist model still offers the worst predictions (MAE = 0.29). Another and more direct comparison between the models is the number of times that one model forecast the outcome better than another one. We do not present the results from this prediction error measure because its results corroborate in general those gained by the MAE measure.