Partisanship, leader evaluations and the vote: Disentangling the new iron triangle in electoral research

This article provides an empirical assessment of the causal structure underlying the core dependent variable of electoral research (the vote) and two of its most notable predictors (partisanship and leader evaluations). A critical review of traditional models of voting highlights the need to account for the reciprocal relationship between the main predictors as well as for the potential feedback stemming from the dependent variable. In the light of these considerations, a new 'iron triangle' of electoral research would seem to take shape, with partisanship, leader evaluations and the vote tight to each other by a strong link of reciprocal causation. Making use of pre-/post-election surveys from Britain and Italy, the empirical analysis provides evidence for a strong effect of past behavior on political attitudes. However, past behavior seems to exert its effect mainly on partisan attitudes, whereas party leader evaluations appear only slightly affected. The results point to the considerably weakened role of partisanship as attitudinal anchor of vote choice. Leader evaluations, on the contrary, emerge as a crucial component in the voting decision.