A retrospective on retrospective voting

This paper critically reviews the extensive literature on retrospective voting in response to economic conditions. Each of the major types of analyses which have been performed — time-series analyses of national vote totals, presidential popularity, and cross-sectional analyses of individual survey responses — has raised several interesting and important questions. The answers that have been obtained, however, are only partial and limited, as each of these approaches entails serious problems of estimation and interpretation. Further progress in this area, we argue, requires explicit treatment of conceptual and statistical issues that have hindered previous research: the dynamic formulation of expectations and preferences, the incidence of policy (and nonpolicy) effects across the population, and notions of incumbency and political responsibility.