Bicameralism and the balance of power in EU legislative politics

ABSTRACT

While there are many studies that focus on the changing power of the European Parliament (EP), the shifting relationships between European Union (EU) institutions has received far less attention. Few consider the impact of the transformation of the EP on the other core EU institutions, particularly the Commission and Council. Together these three institutions determine the policy outcomes of the EU, but how they work together and the changing balance of power between them is largely a mystery. This research seeks to fill this lacuna by examining the relative policy influence of these institutions through the lens of policy preference congruence and inter-institutional policy coalitions. Utilizing the DEU II dataset to measure institutional policy preferences and policy outcomes it is possible to discern patterns of policy preference congruence between these institutions and assess their relative influence over policy outcomes. This research finds that previous expectations of a stable policy coalition between the two “supranational” actors are no longer accurate. Increasingly the EP and the Council share policy congruence and form effective policy coalitions. Moreover, the historical dependence of the EP on the support of the Commission has diminished, while the Council continues to exert largely independent policy influence

KEYWORDS:

• European Union
• European Parliament
• inter-institutional relations
• policy-making

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Amie Kreppel is a Jean Monnet Chair (ad personam) and Director of the Centre for European Studies at the University of Florida. She has been a Fulbright-Schuman Chair at the College of Europe and a Fernand Braudel Senior Fellow at the European University Institute (EUI) in addition to serving as international visiting faculty at universities in Austria, Belgium, France and Italy. Dr Kreppel has written extensively on the political institutions of Europe in general and the European Union, Italy and the UK. Her publications include a book on the Development of the European Parliament and Supranational Party System (Cambridge University Press, 2002) as well as articles in a wide variety of journals, including Comparative Political Studies, the British Journal of Political Research, European Union Politics, the European Journal of Political Research, Political Research Quarterly, the Journal of European Public Policy and the Journal of Common Market Studies. She is currently co-editor of the Italian Political Science Review. Email: [email protected]

ORCID

Amie Kreppel http://orcid.org/0000-0001-8657-7326

Notes

1. There are other analyses that examine the development of potential partisan linkages between the two institutions (via national parties) and the potential implications for EP–Council relations and EU policy-making (Kreppel, 2013; Mühlböck, 2013).

2. There is some analysis that suggests changes to the rules of the EP to provide greater transparency regarding EP positions during the informal trilogues that proceed early agreements have actually led to information asymmetries that effectively reduce the bargaining power of the EP (Héritier & Reh, 2012).

3. There are specific policy areas not fully governed by the ‘ordinary’ procedure as outlined in Article 294 TFEU. These are decided under ‘special procedures’ (Article 289 TFEU).

4. This data was collected through a laborious process that included nearly 350 semi-structured interviews with key actors within the EU legislative process from across the three relevant institutions over a period of several years.

5. The DEU II data set includes information for the EP and the Commission as unitary actors. In addition, it provides information on the preferences of each of the member states within the Council rather than a unified position for the Council as a whole (Thomson et al., 2012). In this analysis, following Kreppel and Oztas (2016), we use a single measure for the Council derived from the mean of their individual preferences.

6. For a more in-depth analysis of these and other concerns regarding the DEU II data set see Slapin (2014).

7. See below for a description of how the DEU II data set is utilised in this analysis through the creation of categories rather than nominal distances.

8. This information was collected from the EP's website on legislative activity (available at http://www.europarl.europa.eu/plenary/) by the author. Each piece of legislation adopted by codecision in the DEU II data set was checked to determine whether a decision was reached at first reading. If so it was coded as an early decision.

9. Recitals provide justifications and explanations of proposed legislation. Thus, the number of recitals included in a proposal is often used as a measure of relative saliency/controversy (Häge, 2007; Häge & Naurin, 2013). In the sample included here the number of recitals ranges from 5 to 49.

10. The respective authors graciously provided their data sets for use in this research.

11. All of the analyses discussed hereafter were also run with a 10-point margin with similar results. The use of a margin for victory is particularly important because while the EP and Commission have unitary scores, the Council score is an aggregate derived from the positions of the various individual member states, resulting in a greater diversity of institutional preference scores.

12. A ±5-point scale was also created and tested with similar results. The decision to utilise the 10-point scale was made because it resulted in fewer omitted cases.

13. A median preference holder was only identified when the three institutions held distinct preferences. The author’s thanks go to an audience member at the 2017 EUSA conference who pointed out the need to make this an explicit test within the model.

14. Note that the data provided in both cases are for the legislative proposal as a whole rather than the specific policy issues provided in the DEU II data set.

15. With dependent variables that are discrete (in this case dichotomous) and measure only membership in a group or category standard ordinary least squares methods are inappropriate as they can lead, among other things, to predicted values of greater than one and/or less than zero and such values are theoretically inadmissible with a dichotomous (0,1) variable.

16. The number of cases in each model across the three institutions varies based on the inclusion (or not) of non-codecision data (Models 1 and 2) and the level of missing data points for actor preferences in the DEU II data set. Note that because of the comparatively high level of missing data for the ‘status quo’ variable for codecision policies, this is only included in Model 2. The dummy variable for the codecision procedure is of course omitted from Models 3–5, which only include codecision legislation.

17. Note that the Median = EP dichotomous variable is significant in just two of the five models (Model 1 and 5). Model 1 is all cases, including consultation, while Model 5 (discussed below) incorporates a measure for ideological distance between the EP and the Council. The limited impact of the Median = EP variable on these two models suggests that ‘luck’ matters more for the EP under consultation and when the EP and Council are ideologically distant, however the limitations of the data restrict further analysis to confirm these results.

18. Likelihood of Commission success decreases by 3.7 per cent for every one point increase in the distance between the average institutional policy preference and the outcome.

19. Although the decline in third reading agreements is usually attributed to the time costs associated with the drawn-out nature of the legislative process, it may well be that the EP and the Commission try to avoid third readings because they tend to do less well when the process reaches this stage.

20. Recall that Models 3–5 only include legislation under the codecision procedure due to the limitations of the available data.

21. This variable falls just short of statistical significance in Models 4 and 5 – probably as a result of the reduced number of cases.