Building on theoretical models of retrospective voting (Alesina and Rosenthal, 1995), and focusing on the principal-agent relationship between voters and politicians (Ferejohn, 1986), the paper proposes a theoretical model that relaxes the commonly made implicit assumption that the effects of positive and negative changes in voters' (expected) utility are symmetric.
Specifically, in line with the literature on prospect theory and the "negativity bias" in psychology, the model assumes that voters respond more strongly to negative information, e.g. bad economic news, than to positive information.
The theoretical model is implemented as an agent-based model, and comparative statics results in terms of (i) selection of quality candidates, and (ii) duration of tenure in office and alternation in government are derived based on Monte Carlo simulations.
The results show that while retrospective voting induces higher average levels of government competence, this is conditional on the possibility of retaining high quality types, which in turn is undermined if voters exhibit strongly asymmetric behavior, i.e. fail to punish and reward evenhandedly.
Finally, some of the empirical implications of this theoretical model are tested using time series data on macroeconomic performance and popularity of incumbent governments in the United States and Europe.
