Forecasting the 21st Korean National Assembly Election from Post-Election Surveys: Lessons from the Failure and the Change of Electoral Landscape in Korea

This paper attempts to predict ex ante the 21st National Assembly election in 2020. As a short history of elections in Korean makes it improbable to rely on aggregate-level data, this paper proposes an alternative approach that utilized individual-level data from post-election surveys in the past four National Assembly elections. First, we estimate how probabilities of each demographic group to vote for major parties had changed over time. Then, estimated probabilities are combined with the actual demographic composition of the Korean electorate to produce the prediction of how each group is expected to vote in an upcoming election. Then, we predict how expected vote shares of each party may translate into their seat shares in each region. In conclusion, we compare our forecasts with the actual outcome of the 21st National Assembly election and discuss the implication regarding the change in electoral landscape in Korea.