Testing theories of electoral cycles

Abstract. A common form of electoral cycle theory asserts that support for government parties is dependent on the proportion of an inter-election interval which has passed since the last national election. Weekly opinion poll data for the Netherlands are used to test whether or not such 'cyclical' patterns of change can be detected in the inter-election periods of 1977–1981 and 1982–1986. The relative merits of two methods of analysis, polynomial regression and ARIMA, are discussed and demonstrated. The latter is shown to be the more suitable of the two; the former may yield misleading outcomes. The analyses reported do not display even a glimpse of empirical support for the alleged cyclical phenomenon. The ARIMA analyses can be used, however, to describe how processes and events in a society impinge on (noncyclical) developments in the electoral support for political parties.