A Bayesian Model for Election Forecasting: A Case Study of Voting Intentions of Students of University of Lagos

• Activity Number: 352
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Ray Okonkwo Okafor*+ and Rotimi Kayode Ogundeji
• Companies: University of Lagos and University of Lagos
• Keywords: Empirical Bayes ; Beta-binomial model ; Estimation ; Election-forecast ; University of Lagos ; Lagos state-Nigeria

Abstract

Bayesian methods provide more intuitive and meaningful inferences than likelihood-only based inferences. This is simply because of the fact that the Bayesian approach includes prior information as well as likelihood. In empirical Bayes (EB) methodology, we use data to help determine the prior through estimation of the so-called hyperparameters. In an attempt to predict the outcome of the April 2011 gubernatorial election in Lagos state, Nigeria, we employed a Bayesian model namely; Beta-binomial conjugate model to forecast the voting patterns of students of University of Lagos, Nigeria. On a weekly basis for four weeks prior to the election, data were collected by administering questionnaires to members of the target population to gather information about the voting intentions of the registered student voters. The results of analysis compared estimated sample and EB proportions sequenced over the weeks. Also the standard deviation of these two outcomes were computed and compared. Ultimately, our forecast substantially agreed with the results of the elections as announced by the national electoral body.