Constraining or enabling? The effects of government composition on international commitments

ABSTRACT

The effect of government composition on the ability to make international commitments is an ongoing debate in the literature. Using the veto players and clarity of responsibility theories, I introduce a nuanced understanding of government composition that considers the different types of multiparty governments and the degree of policy incongruence inside them. I test the effects of these variables on the intensity of international commitments with a post-Cold War events dataset of parliamentary European governments. The results of multilevel analyses show that oversized coalitions make more intense commitments than single-party majority governments through responsibility diffusion, though neither approach explains the commitment behavior of minimum winning coalitions. Finally, commitment intensity increases for minority coalitions if their ideological setup leaves the parliamentary opposition fragmented. Ultimately, a government's arithmetic and ideological composition conditions the intensity of its international commitments. I also find that European Union membership has no significant effect on commitment intensity.

KEY WORDS:

• Coalition
• commitment
• diffusion
• Europe
• foreign policy
• veto player.

ACKNOWLEDGEMENTS

I would like to thank Ryan Beasley, Jonathan Caverley, Terrence Chapman, Chad Clay, Miriam Elman, Jonathan Hanson, Margaret Hermann, Seth Jolly, Juliet Kaarbo, Ranan Kuperman, Hendrik Spruyt and Dominik Tolksdorf for their thoughtful feedback and discussions at numerous stages; panel participants at the 2013 European Union Studies Association meeting, as well as the seminar participants at Binghamton University and Northwestern University for their comments on the previous versions of this contribution. James Caporaso, Clifford Carrubba and Joseph Jupille provided valuable advice at the earlier stages of the project. Finally, I would like to thank two anonymous reviewers for their very helpful suggestions.

Notes

1 Using parliamentary seat share to measure coalition strength is misleading in proportional representation systems, where parties are more disciplined and ideologically coherent (Leblang and Chan 2003).

2 Pivotal party in a minimum winning coalition can ‘dissolve the government by defecting from the coalition and taking away the number of parliamentary seats it controls that are necessary for the maintenance of the cabinet’ (Kaarbo 1996b: 507).

3 For the present dataset, the correlation coefficient for minimum winning coalitions and number of parties is 0.18; for oversized coalitions it is 0.41. On average, minimum winning coalitions have 2.7, oversized coalitions have 3.9 parties.

4 Events datasets are generated by disaggregating newswires into actor, action and target. Frequently used news repositories include The New York Times (McClelland 1978) and Reuters (King and Lowe 2003).

5 Governments were coded using Browne et al. (1984) and Warwick (1992).

6 Switzerland's executive branch is structured differently than the other European parliamentary systems, and is therefore not included in the analyses. Data for Croatia begins at 2000, when it switched to parliamentarism.

7 An alternative measure extends from –4 to +4 (Callahan 1982a). Both the Callahan and Goldstein measures use the World Event/Interaction Survey (WEIS) event categories (McClelland 1978). The Goldstein (1992) scale is a better measure for commitment intensity, as it relies on the same event categories that are coded in the original King and Lowe (2003) dataset. The events used here were also coded using the Callahan (1982a) scale to check the extent of overlap between the two measures. The overlap is large enough to confidently use Goldstein's (Cronbach's alpha = 0.9).

8 A third alternative variable, ideological range, measures ‘the absolute value of the distance between the most extreme parties in a coalition’ (Tsebelis 1999: 599), and it overlaps with the standard deviation variable (Cronbach's alpha = 0.9). Tests using this measure yield results similar to those using the standard deviation, but are excluded owing to space limitations. They are available from the author upon request.

9 Using the weighted standard deviation assumes that the mean coalition position is proportional to each coalition party's parliamentary seat share. Since this contribution argues that a junior coalition partner's policy influence might not always be proportional to its size, an unweighted standard deviation measure is used.

10 Standard deviation variable is coded missing if there is no left–right position data for at least one party in the government. If the coalition has at least two parties from the opposite sides of the left–right political spectrum, or includes the center party, center is crossed variable assumes 1. This produces fewer missing cases.

11 Benoit and Laver's (2006) 20-point scale is divided by two to match the 10-point scale used in the other datasets. Left-wing parties assume less than 5 on this scale. Right-wing parties assume 6 or more; 5 denotes the center of the political spectrum.

12 A binary ‘EU membership’ variable would capture the East/West Europe differences instead of EU membership, since the present dataset ends at 2004 and the EU's 2004 enlargement was largely towards Eastern Europe.

13 I thank one anonymous reviewer for this suggestion.

Additional information

Biographical note

Sibel Oktay is a PhD candidate in political science at Syracuse University, USA, and currently a visiting pre-doctoral fellow at Northwestern University, USA.