Scholars have long assigned a key role to party identification as an explanation of voting behaviour. In doing so, they have assumed that individuals' partisan affiliations remain unchanged for long periods of time. But is partisanship sufficiently stable to justify this assumption? At the very least, to be considered a long-term force party identification cannot change during an election. Yet the intra-election stability of party affiliations has been accepted on faith, rather than examined empirically.
Our analysis tests this assumption by looking at the evolution of partisanship over the course of the 1980 election. We find that many citizens do alter their partisanship over a single electoral period. These changes in party identification follow a systematic - and not a random - pattern. Both cognitive and affective factors account for this intra-election partisan lability. These findings suggest that much of the previous research on voting behaviour has been seriously misspecified.