Predicting U.S. Elections: an Alternate Approach Using Monte Carlo Simulations

Previous literature has shown the importance of economic growth and the incumbency status of candidates in predicting the results of U.S. Elections. We created a joint model that leverages these data– all of which are available months before the actual election and are largely independent of a candidate’s charisma or experience– to predict the results of Presidential, Senate, and House of Representatives elections. By running a series of Monte Carlo simulations at the state level, our model accurately predicts the results of U.S. elections dating back to 1982. We show that the possibility of a unified government (one party controlling the Presidency and majorities in both chambers of Congress) is almost entirely predicted by these exogenous factors: major power swings are largely predictable even before campaigns formally begin.