Ambiguous Attribution: Theory and Evidence

Clarity of responsibility is an essential element of political accountability. We develop a rational model of Bayesian updating in the presence of ambiguous attribution and we test its predictions using an original survey. We show that respondents' partisanship, assessment of public healthcare quality, and beliefs over which layer of government is responsible for healthcare are correlated as predicted: good-assessment voters attribute responsibility to the layer governed by their preferred party, while bad-assessment voters blame the layer governed by the party they dislike. These partisan patterns of credit and blame, often interpreted as evidence of motivated reasoning or partisan bias, can thus arise from rational Bayesian updating under attribution ambiguity. No such partisan patterns exist where the same party is in charge of regional and central government. A survey experiment in which we inform subjects of the official quality of healthcare has them update in the predicted, partisan, direction. Model and empirical results show that partisan priors are extremely hard to dislodge when attribution is ambiguous.