The dynamics of aggregate partisanship

Despite extensive research on party identification, links between partisanship at the individual and aggregate level have largely been ignored. This leaves a gap in our understanding of the dynamics of aggregate partisanship. To remedy this, we identify a set of ideal types that capture the essential arguments made about individual-level party identification We then combine the behavioral assumptions for each type with existing results on statistical aggregation to deduce the specific temporal pattern that each type implies for aggregate levels of partisanship. Using new diagnostic tests and a highly general time series model, we find that aggregate measures of partisanship from 1953 through 1992 are fractionally integrated. Our evidence that the effects of a shock to aggregate partisanship last for years-not months or decades-challenges previous work by party systems theorists and students of ''macropartisanship.'' Our arguments and empirical evidence provide a conceptually richer and more precise basis for theories of issue evolution or endogenous preferences-in which partisanship plays a central role.