Predicting the 21st General Election Results Through Polling-Based Methods

This paper attempts to derive broad inferences about the 21st General Election through polling-based methods. Despite the challenges (e.g., small populations, in-person campaigns constrained by the gravity of COVID-19, the period of prohibition on publication of survey results), this paper predicts the election results by using various analytical techniques and statistical models including the Bayesian predictive model. Overall, the predictive model shows that: (1) regional polarization (Yeongnam vs. Honam) via the two major parties will emerge again in the nominal tier; (2) third parties will decline; and (3) a solid group of voters will support the current government and the president. They will play a decisive role in the remaining regions, including the metropolitan area. The predictive model was applied to two battlefield districts in Seoul. It predicted the victory of candidate Lee over Hwang in Jongno District. In Gwangjin Eul District, Koh and Oh competed within the margin of error. In terms of rates of party support, supporters for the two major parties were found to be gathering, while the Justice Party was on a downward trend.