Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 476 - Innovations in Analytic Approaches for Survey Data
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Survey Research Methods Section
Abstract #310960
Title: Bayesian Hierarchical Models for the Prediction of American Elections
Author(s): Brittany Alexander* and Jeffrey Hart
Companies: Texas A&M University and Texas A&M University
Keywords: political polling; American elections; hierarchical models
Abstract:

We propose a series of hierarchical Bayesian models to predict the outcome of American elections using only toplines from pre-election polling. The models are applied to recent federal elections with a focus on the past three Presidential elections. The models are structured to allow the pooling of information between states deemed similar. The models are similarly structured but they have different distributional assumptions (i.e. Gaussian or beta) and likelihood structures. They are based on previous models that were found to be approximately 93% as accurate as the FiveThirtyEight Polls Plus model in terms of RMSE in predicting the 2008, 2012, and 2016 presidential elections and made nearly identical predictions of the winners of states as FiveThirtyEight. The methodology is implemented in the forthcoming bayesurvey R package and will be implemented to predict the 2020 Presidential and senate elections.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program