Cutting Through the Noise: Tracking the Ups and Downs of the 2004 Presidential Race

The 2004 US Presidential campaign was a remarkably close race between President George W. Bush and Senator John Kerry. As a result, tracking the daily movements of the race was of exceptional interest, and much news coverage focused on day-to-day fluctuations in individual state and national polls. To reduce this variability I used publicly available state polling data and the rules of the Electoral College to calculate a probabilistic snapshot of the campaign (http://election.princeton.edu). This meta-analysis reliably captured swings in sentiment within a few days of major news events. The analysis could predict the electoral vote total and individual state voting margins under the assumption that aggregated polls were an accurate measure of voter sentiment. Meta-analysis was also useful in quantifying an individual voter's power to affect the overall outcome. This valuation corresponded well with efforts by voter turnout organizations and candidate itineraries. Overall, meta-analysis of polls provides a simple, precise tool for tracking future election campaigns.