Forecasting Elections from Partial Information Using a Bayesian Model for a Multinomial Sequence of Data

TL;DR

This paper develops a Bayesian model for early election outcome prediction from partial multinomial data, providing theoretical guarantees and demonstrating effectiveness with real-world election data from India and the US.

Contribution

It introduces a hierarchical Bayesian approach with convergence analysis for early election prediction from batch-wise multinomial data.

Findings

The model accurately predicts election outcomes early in the data collection process.

Theoretical conditions for asymptotic normality of estimates are established.

Simulation studies confirm the model's effectiveness in real-world scenarios.

Abstract

Predicting the winner of an election is of importance to multiple stakeholders. To formulate the problem, we consider an independent sequence of categorical data with a finite number of possible outcomes in each. The data is assumed to be observed in batches, each of which is based on a large number of such trials and can be modeled via multinomial distributions. We postulate that the multinomial probabilities of the categories vary randomly depending on batches. The challenge is to predict accurately on cumulative data based on data up to a few batches as early as possible. On the theoretical front, we first derive sufficient conditions of asymptotic normality of the estimates of the multinomial cell probabilities and present corresponding suitable transformations. Then, in a Bayesian framework, we consider hierarchical priors using multivariate normal and inverse Wishart distributions and establish the posterior convergence. The desired inference is arrived at using these results and ensuing Gibbs sampling. The methodology is demonstrated with election data from two different settings -- one from India and the other from the United States of America. Additional insights of the effectiveness of the proposed methodology are attained through a simulation study.