| Abstract: | Explaining the outcome of presidential elections is central to any model of American government. Previous researchers have found that economic conditions explain a substantial portion of the variation in vote outcomes. We make two contributions to this literature. First, we show that state partisan predisposition is the most important explanatory variable for the period 1972–1992. Several states are simply out of reach for one of the parties, no matter how favorable is the information about their candidate. Second, we find that national economic indicators have an effect on votes that is an order of magnitude larger than state-level aggregates. Presidents who try to curry favor with certain states through pork barrel projects are unlikely to be rewarded with large vote margins. Our model does a reasonable job forecasting the state-level vote for the 1996 election when the actual economic conditions are used as regressors. None the less we are skeptical that these type of models can accurately forecast the Electoral College winner because of the wide confidence intervals on each state's vote forecast and the potential error in predicted economic conditions. |
