University of Miami Law Review


How do campaigns, media, and voting laws influence the outcome of U.S. Presidential elections? Political scientists often argue that these factors influence outcomes much less than commonly thought. To illustrate this argument, we show that we can predict the presidential vote in each state with a high degree of accuracy. Specifically, between 2004 and 2016, we correctly predict 94% of all state presidential vote outcomes. Our predictions are based on a forecasting model of the Electoral College, based primarily on each state’s approval rating of the incumbent president (using almost 90,000 survey responses from June and July of election years), current economic conditions in each state, and state votes in the previous election. We use these forecasts to help establish the upper bounds of campaign and media effects. We argue that identifying the limits of these effects is a critical step when trying to estimate their impact. We also show how our forecasts can be used to test the aggregate effects of election-related laws, such as Florida’s Amendment 4—which enfranchised hundreds of thousands of Floridians who previously could not vote due to felony convictions—and voter ID laws, whose effects are notoriously difficult to study. We have made our data publicly available to facilitate further research on these topics.