The authors contribute to the existing literature on the determinants of legislative voting by offering a social network-based theory about the ways that legislators’ social relationships affect floor voting behaviour. It is argued that legislators establish contacts with both political friends and enemies, and that they use the information they receive from these contacts to increase their confidence in their own policy positions. Social contacts between political allies have greater value the more the two allies agree on policy issues, while social contacts between political adversaries have greater value the more the two adversaries disagree on policy issues. To test these propositions, we use social network analysis tools and demonstrate how to account for network dependence using a multilevel modelling approach.
* Department of Political Science, University of Wisconsin, Madison (email: firstname.lastname@example.org); Department of Public and International Affairs, George Mason University and Department of Political Science, University of North Carolina, Chapel Hill, respectively. The authors wish to thank the European Union Center at the University of Wisconsin for its support. They are grateful for data collection efforts and research assistance from Peter Truby, Katie Renze, Ashleigh Baker and Christina Boyes. They also thank Stacy Bondanella and Jean-Dominic LeGarrec for translating the survey into French, and Jason Koepke for his technical advice and assistance. They are grateful to Simon Hix for providing the EP roll-call vote data used in the analysis, and to Giacomo Benedetto for sharing the burden of processing the data. Finally, they thank David Canon, Scott Gehlbach, Elisabeth Gerber, Jonathan Hurwitz, Scott Morgenstern, Laura Wills Otero, three anonymous reviewers and the Journal's editor Hugh Ward for valuable comments and suggestions. The authors regret that they cannot make replication data publicly available, given the sensitivity of their data about the personal connections between political actors. All respondents were assured complete anonymity, and the small sample of (actual and potential) respondents prevents the release of data including general attributes such as party affiliation and nationality instead of proper names. Two Supplementary Appendices containing additional information are available online at: http://dx.doi.org/10.1017/S0007123412000518.