This paper examines the impact of polls on perceptions of the race, preferences, and vote choice under three different electoral systems: pure plurality (Canada), pure PR (the Netherlands), and a mixed member proportional system with a threshold (New Zealand). We test the hypothesis that poll effects are greatest in Canada, where the incentives for strategic voting are strongest, and smallest in the Netherlands, where the incentives are weakest. The data come from the 1997 Canadian Election Study (CES), the 1998 Dutch Election Study (DES), and the 1999 New Zealand Election Study (NZES). [ABSTRACT FROM AUTHOR]