U.S. presidential election forecasts are of widespread interest to political commentators, campaign strategists, and the public.
We argue that most fundamentals-based political science forecasts overstate what historical political and economic factors can tell us about the likely outcome of a forthcoming presidential election. Existing approaches generally overlook uncertainty in coefficient estimates, decisions about model specification, and the translation from popular vote shares to Electoral College outcomes.
We introduce a novel Bayesian forecasting model for state-level presidential elections that accounts for each of these sources of error, and allows for the inclusion of structural predictors at both the national and state levels. CONT. – pdf
Benjamin E. Lauderdale (LSE) & Drew A. Linzer (Emory)