Forecasting South Korea's presidential election via multiparty dynamic Bayesian modeling

Forecasting a presidential election's outcome is a long-standing topic in statistics and political science. However, a lack of historical data and a complex multiparty political system make it challenging to apply models developed so far to South Korea's presiden-tial election. In addition, no suitable model has been proposed to address these issues, and there are no practical means by which to forecast presidential elections in South Korea. Here, we propose a flexible Bayesian framework for forecasting election outcomes at the provincial level by incorporating abundant pre-election polls into historical data. Hilbert spaces are employed to induce a multiparty forecast. Our framework provides numerous findings worth examining, such as long-and short-term opinion trends, the effect of fundamental conditions on vote share, and systematic bias in pre-election polls. The framework is applied to the 2022 South Korean presidential election, demonstrating that our framework is promising.(c) 2023 International Institute of Forecasters. Published by Elsevier B.V. All rights reserved.

Automated summary

This article proposes a flexible Bayesian framework for forecasting provincial election outcomes by incorporating pre-election polls into historical data. The framework is applied to the 2022 South Korean presidential election and demonstrates promising results.