This article discusses the use of Bayesian analysis in the evaluation of temporal volatility and information flows in political campaigns. Using the 2004 US presidential election campaign as a case study, it demonstrates the utility of a model with two volatility regimes that simplifies the task of associating events with periods of high information. The article first explains why prediction markets are able to aggregate information such that the prices of future contracts are reflective of the event’s actual probability of occurring before analysing data from futures on ‘Bush wins the popular vote in 2004’, or the traded probability, of Bush winning the election. These data are used to build a measure of information flow. The results show that information flows increased as a result of the televised debates, and that these debates, along with the selection of the vice presidential candidate, increased prediction market volatility.