Understanding Elections: Measuring Electoral Determinants with Electronic Prediction Markets

While significant research has demonstrated that prediction markets are highly accurate indicators of the real-time status of an election, few studies within political economics or political science have used them as a tool to understand voting behavior. Assuming their accuracy, we utilize these markets to measure the primary determinants of the 2008 United States election. First, we use electronic prediction markets to show that the financial crisis had a decisive impact on the outcome of the presidential election, and conversely that Obama’s victory had a sizable impact on equity markets. Next, we use prediction markets to measure the scope of the impact of the presidential campaign. We find that campaign events had a significant – but small – impact on state-by-state outcomes, but campaign spending had no significant local-level effects. Finally, we use data from electronic prediction markets to measure the size of presidential “coattail effects” on congressional elections, finding strong effects in House elections but weak effects in the Senate. Throughout each of these chapters, our use of data from electronic prediction markets enables us to circumvent the typical measurement problems for each of these questions by increasing the sample size and eliminating endogeneity biases with a real-time metric of election status. Acknowledgements This paper would not have been possible without the invaluable ideas and support of Yiling Chen, whose expertise with prediction markets and feedback were instrumental throughout the writing process. I am thankful to Dan Wood, whose extensive advice and comments were formative in shaping my paper from the beginning. I am also appreciative of the ideas I gained from conversations with James Stock and Andrei Shleifer. Additionally, I am grateful to Intrade for its generous sponsorship of electronic prediction market data and DiskTrading.com for providing financial futures market data. Publically-available data from Iowa Electronic Markets were also instrumental in performing this analysis. Finally, I am extremely grateful for the stylistic support provided by Greg, Eva, and Holly, who were extremely generous with their last-minute editing and feedback.