Elections under biased candidate endorsements — an experimental study

We construct an election game to study the electoral impacts of biased candidate endorsements. We derive a set of testable predictions. We test these in a laboratory experiment and find that observed election outcomes and vote shares are well predicted. We find no support, however, for our prediction that the relationship between election outcome and the endorser's bias is non-monotonic; i.e., ex ante, a candidate's winning probability will first increase and then decrease as the endorser becomes more biased towards her. Voter turnout is much less responsive to the bias than predicted. We argue that observed voting behavior can be explained, to a substantial extent, by three behavioral mechanisms: (a) distinct levels of rationality for candidate choice and turnout decisions, (b) conservative belief updating, and (c) ‘partial competition neglect’, where voters underestimate the correlation between the information released by an endorsement and the closeness of elections.