Whereas there seems widespread agreement among politicians and journalists that the impact of political leaders on elections is substantial, electoral researchers again and again fail to observe such effects in their analyses. In this paper we revisit this issue and adopt a novel method to estimate the electoral effects of party leaders. Our approach rests on two presumptions: (1) Against the traditional view in which partisanship is considered a stable long-term predisposition, we contend that party leaders influence voting behaviour indirectly and the primary effect that leaders have is by influencing how voters think and feel about their parties. (2) If party leaders matter, then leadership changes should be accompanied by relatively strong changes in party attachments and subsequent changes in electoral support. In order to identify effects of political leaders on the vote an appropriate strategy is therefore to focus on shifts in partisanship immediately after a leadership change. We test our ideas on data from the Comparative Study of Electoral Systems. We compare the change in party attachments and electoral support of cases with and without leadership change. Furthermore, we analyze if such changes are moderated by the party system, and whether the size of leader effects is related to specific party characteristics. We discuss the results and also discuss the possibility of a reversed causal order: if parties loose support, they change the leadership.
