Behind Poll Error: Review of the 2020 Presidential Election and the Cause of Poll Error

Abstract

After experiencing the poll error and the forecast failure in the 2016 US presidential election, pollsters' effort to improve the pre-election poll didn't work. The poll error of 2020 is even more significant than the poll error of 2016. By studying the poll error that appeared in history, I found three possible causes: late decider, shy Trump voter, and non-response bias. The thesis examined all three causes to see whether they are the cause of the 2020 election. Based on exit poll data, the late decider is hardly the main cause of the 2020 election poll since there are fewer people who decide to vote during the final week before the election compared to 2016, yet the poll error increased. The demographic data showed the opposite of the shy Trump voter hypothesis that the poll error is more significant in areas less likely to have social pressure for supporting Trump. The non-response bias caused by people within demographic groups who are less likely to take a survey seems to be one of the causes of the 2020 election poll error, yet furth study is needed to provide proof.