Assessing what roll call votes can tell us about legislator preferences has proved to be an important conundrum in legislative studies. This is due to two reasons: first, roll call votes are likely to occur for very specific votes in most parliaments, and second, having a roll call vote (i.e., a public record of how a legislator has voted) will influence a legislator's voting behavior. Drawing on a unique dataset comprising all votes from the Swiss parliament, some of which were automatically roll called while for others legislators requested them, we are able to identify the effects of both roll call votes and the effect of selecting particular topics for roll call votes. As we also have information on who requested the roll call, we can distinguish among motivations for calling a roll call. We find, based on an extension of the two cut-point IRT model (Clinton, Jackman and Rivers, 2004), that members of the Swiss lower house support more leftist positions when their voting behavior is exposed to scrutiny, compared to situations when their behavior is not visible. Thus, we can demonstrate that inferences drawn from roll call votes, when these do not cover the full universe of voting decisions, lead to biased inférences.
1 The partial financial support of the Swiss National Science Foundation (Grant No 100012-111909) is gratefully acknoledged.
2 Département de science politique, Faculté des
sciences économiques et sociales; Université de Genève; 40, Bd du
Pont d'Arve; 1211 Genève 4; Switzerland; phone ++41 22 379 83 78;
email: simon.hug@unige.ch