From uninformed to informed choices: Voters, pre-election polls and updating

Pre-election opinion poll results for U.S. presidential contests have large variation in the early parts of the primary campaigns, yet pre-election opinion polls later in the campaign are typically within several percentage points of the actual outcome of the contest in November. This paper argues this trend demonstrates that voters are beginning to poll “correctly” – that is, to ascertain their most-preferred candidate. This convergence process is consistent with boundedly rational voters making decisions with low information. We examine the process by which voters can use opinion polls to guide their candidate choice. We undertake a series of laboratory experiments where uninformed voters choose between two candidates after participating in a series of pre-election polls. We demonstrate that voters update their beliefs about candidate locations using information contained in the opinion polls. We compare two behavioral models for the updating process and find significant evidence to support a boundedly rational Bayesian updating assumption. This assumption about the updating process is key to many theoretical results which argue that voters have the potential to aggregate information via a coordination signal and for their beliefs to converge to the true state of the world. This finding also indicates that uninformed voters are able to use pre-election polls to help them make correct decisions.

► We isolate the effects of exposure to pre-election straw polls on uninformed voters through a series of laboratory experiments.
► Individuals rely upon these polls as a source of information as part of a Bayesian updating process.
► Although voters do appear to be relatively ignorant about politics, they can rely upon these informal institutions as shortcuts to cast “correct” votes.