The Institutional Power of Western European Parliaments: A Multidimensional Analysis

Abstract

Parliaments are more than legislative bodies. However, we lack an adequate understanding of the theoretical relationship between different facets of parliamentary activity or ‘parliamentary functions’. Relying on the principal–agent framework, this article argues theoretically that parliamentary power is a multidimensional concept comprising three distinct mechanisms to ensure policy outputs in line with the collective preferences of parliaments: direct influence on policymaking, the ex ante selection of external officeholders, and the ex post control of the cabinet. These mechanisms mirror the classic legislative, electoral, and control functions of parliaments. Empirically, the paper uses factor analysis of newly developed indicators for electoral powers and established measures of legislative and control resources to show that the institutional powers of 15 Western European parliaments comprise four distinct dimensions. These dimensions match the three theoretically derived mechanisms with committee power as an additional factor. Locating the 15 parliaments in this multidimensional space of parliamentary powers demonstrates that classifications based solely on lawmaking lead to biased assessments of parliamentary strength and weakness. Instead, the paper provides a more nuanced picture of the ways in which Western European parliaments can influence policymaking under the conditions of delegation.

Acknowledgements

I gratefully acknowledge very helpful comments and suggestions by John Carey, Steffen Ganghof, Marcelo Jenny, André Kaiser, Richard S. Katz, Bernhard Miller, Wolfgang C. Müller, and two anonymous reviewers.

Notes

1. The name is somewhat misleading because the British cabinet assumes office without a formal investiture vote in parliament.

2. Only the general importance of different functions and the question of whether control and recruitment became more important compared to lawmaking in general was discussed in the context of the decline of parliaments thesis (Bryce 1921; Loewenberg 1971).

3. As stated above, political parties play an essential role in aligning actor preferences across institutional arenas. A high level of party unity decreases the impact of institutional powers of parliamentary actors vis-à-vis the cabinet. Thus, the institutional powers covered in this paper present the maximal potential for parliamentary power but are not fully applied under all circumstances.

4. My argument does not touch on functions like representation, legitimisation and providing a forum for public debate that cannot be conceptualised as institutional powers. In addition, some functions discussed in the literature, e.g. Packenham's (1990 [1970]) ‘recruitment, socialisation, and training functions’, entail elements beyond the institutional powers covered here.

5. These countries are: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, and the United Kingdom. I treat France as a parliamentary democracy because it satisfies the most important defining characteristic that the cabinet can be forced to resign by a majority vote in parliament (Müller et al. 2003). Extending the analysis to the settler democracies or Eastern Europe would require massive data collection and was not feasible in the context of this project even though it offers a valuable perspective for future research.

6. Italy is an extraordinary case in this respect because the cabinet is accountable to both the Chamber of Deputies and the Senate. Due to data problems, my analysis covers only the Chamber of Deputies. Furthermore, the focus on lower chambers might be problematic in countries such as Australia where the second chamber is specifically designed to control the cabinet and the majority in the first chamber. However, none of the cases covered in this paper follows such a pattern.

7. Budgetary powers are sometimes considered a separate function from lawmaking. I treat them as a subcategory of lawmaking powers in line with Bagehot's original formulation.

8. As these measures are taken from existing studies, they refer to different points in time ranging from the mid 1990s to today. Unfortunately, time-specific data is not available. Thus we can neither measure them all at the same point in time nor capture institutional change in parliaments. However, the basic institutional parameters of parliamentary competition covered here tend to be rather stable over time in most countries. Thus the lack of time-sensitive measures should not lead to massive distortions in the empirical analysis.

9. No-confidence procedures do differ in details, especially the question of whether a cabinet has to be replaced by an alternative in a constructive vote of no confidence and of whether no-confidence votes can be directed against individual ministers. For more information on these differences and their relationship to electoral powers with regard to the cabinet see Sieberer (2010: Ch. 5).

10. If a member of a collective body is elected, the relevant status quo is the ideal point of the median of the remaining members as he or she dominates decision-making if the vacancy was to prevail (Krehbiel 2007). If parliament elects a single officeholder, we can treat the incumbent's ideal point as status quo because he or she often remains in office until a successor is elected or procedures established by the incumbent are carried out by administrative staff.

11. I subsume political parties as organisations under parliamentary actors because parliamentary party groups often play an important role within party organisations (Heidar and Koole 2000). Distinctions within the group of cabinet actors between the prime minister, a specific minister, or cabinet as a collective body are not pursued here, even though they could be relevant in coalitions or highly factionalised one-party cabinets.

12. An index based on the alternative ordering of the two dimensions (AM3 <AM2 <AM1 <CEM3 <CEM2 <CEM1) is highly correlated with the one I use in the cases analysed (r = 0.94; p = 0.000; N = 65).

13. These values give the three variables (number of candidates, nominating actor, and majority requirement) approximately equal impact on the rank ordering of the index scores. Coding the majority requirement as 1; 2; 3 leads to very similar results (r = 0.98; p = 0.000; N = 65).

14. Using an additive link instead leads to very similar results (r = 0.93; p = 0.000; N = 65).

15. Using values of 0.25 or 0.50 instead leads to virtually identical indices (r = 0.99; p = 0.000; N = 65).

16. Some of these external officeholders are not agents in the strict sense as they are not accountable to parliament after assuming office. Nonetheless, parliamentary actors can affect their preferences and thus expected behaviour via the election. Therefore, they can use electoral powers as means to indirectly affect policy outputs and reduce agency loss.

17. In some countries the number of offices is lower because there is no constitutional court (e.g. in the UK and the Netherlands) or ombudsman (in Germany and Italy).

18. I use the number of existing offices because not electing a non-existent officeholder does not indicate a weak parliament.

19. The unweighted index is highly correlated with an index assigning double weight to the election of the prime minister and cabinet (r = 0.97 for PFCI and r = 0.98 for ICI; p = 0.000; N = 15) and an index weighing the prime minister and cabinet by a factor of three, heads of states and constitutional judges by two and heads of audit institutions and ombudsmen by one (r = 0.90 for PFCI and r = 0.94 for ICI; p = 0.000; N = 15).

20. The principal factors method does not assume that the factors completely capture the variance in the variables which seems appropriate given the sometimes rough measurement of the theoretical concepts. Using principal component factors instead leads to very similar results. The varimax rotation ensures statistical independence of the factors and facilitates their interpretation. The estimation was performed in STATA 10. The use of factor analysis is methodologically problematic for some variables with few values because the assumption of a multivariate normal distribution is violated. Despite this problem, scholars frequently apply factor analytic procedures to this kind of data (Mattson and Strøm 1995; Tsebelis 2002). I follow this established procedure.

21. The confidence intervals are estimated using the bias-corrected and accelerated method because the simulation indicates bias in the eigenvalues.

22. The percentage of variance captured by a factor does not constitute a measure of its substantial importance because it depends both on the choice of input variables (the more variables relate to a factor the higher its explained variance) and on the rotation procedure.

23. Two variables have loadings higher than 0.5 on a second factor in addition to the one they belong to theoretically. The control structures index correlates with factor I. A plausible explanation is that the index includes elements that are relevant for both control and lawmaking (e.g. the size and jurisdictions of committees). Committee powers to rewrite government bills also load on factor II. I have no theoretical explanation for this result. Note, however, that the variable is also correlated with factor I and retains a relevant uniqueness value which indicates a weak overall fit with the factor solution.

24. A value of 1.10 was added to the scores for the third factor because the size of the circles must be positive. This linear transformation does not affect the ordering of the countries.

25. However, the strong institutional prerogatives of the cabinet in the legislative process do not keep parliament from amending bills quite frequently (Brouard 2011) which reinforces the point that we cannot infer behaviour directly from institutional rules.

26. By necessity, the factors are uncorrelated across all cases because of the varimax rotation.