ABSTRACT
This article points out a paradox: most political scientists apparently subscribe to a party identification model of voting behaviour and yet there is widespread concern about how that concept is currently measured. In particular there are doubts about whether the traditional measure captures the long-term nature of party identification. The article presents evidence from an experiment, carried out by the Gallup Organization, which compares the traditional measure with a new one. The results suggest that the traditional measure of party identification overstates the level of identification and leads to an exaggerated impression of the relationship between this variable and vote. The article ends with an appeal for more research into the measurement of party identification.
