A Bayesian Approach for Validation of Election Results

Abstract

Election process and results in many countries have resulted in both political and economic instability of that country. Fair and credible election process and results must be evidence-based and statistical proven. This study employed a Bayesian procedure for the validation of election results. Based on Nigerian 2011 and 2015 presidential election results, Bayesian credible intervals were obtained to assess the credibility of Nigeria presidential election results. The study explores Bayesian methods using a Bayesian model called beta-binomial conjugate model to compute posterior probability of electoral votes cast and confirm if these votes are within Bayesian credible intervals. The results obtained showed that election outcomes for the two major political parties in Nigeria 2011 presidential election are not within Bayesian credible bounds while 2015 presidential election results are within computed Bayesian credible bounds. Also, in contrast to frequentist approach, applied Bayesian methodology exhibited smaller variance which is an indication that Bayesian approach is more efficient. Thus, for election to be fair, credible and acceptable by the electorates, Bayesian approach can be used to validate electoral process and results.

Keywords: Bayesian Methods, Bayesian Credible Intervals, Beta-Binomial Model, Empirical Bayes, Nigeria Presidential Elections.