Replication data for: The Limits of Prediction: Incorporating Uncertainty in a Normal Vote Model

In the short term, an electoral gain for strongly Republican states like Texas and Arizona may benefit Republican presidential candidates. But what can these population forecasts tell us about the major parties' long term electoral prospects?

Burmila (2009) develops an updated model of Converse's (1966) "normal vote" to make inferences about the effect of electoral vote changes on the 2012-2028 presidential elections. Here, we test Burmila's model against known election results to derive a measure of the model's uncertainty. Our results suggest that a normal vote model (1) does not take into account the variance in state voting patterns, (2) does not make clear the limitations of its predictive power as it is applied farther in the future, and (3) is overly dependent on recent election results.