Did late deciders for Donald Trump impact the election polls this year?
The election polls missed again in 2020. While a few votes still have to be counted, it is clear surveys underestimated support for Donald Trump by about five points overall, with some greater misses in critical swing states like Wisconsin. The latest results have unleashed renewed criticisms of an industry that assumed it had addressed the errors of 2016, and there is no shortage of theories about why the polls performed so badly.
Before and after the election this month, many analysts speculated about the chance that late deciders broke heavily for Trump. Pointing to the exit polls, some in the industry have suggested this was indeed the case. After four years of learning about Trump and over $6 billion dollars in campaign spending, was there a noticeable portion of late deciders for him? We are not so sure. There were fewer undecided voters in 2020 than 2016 and no evidence these folks leaned toward Trump before the election.
We looked at data from the Reality Check Insights survey that interviewed respondents after the election to figure out if their intention had changed. The respondents first noted whether they planned to vote for Trump, Joe Biden, or neither candidate for three surveys conducted between August and the start of November. In the three days after the election, more than 300 of such respondents were interviewed once more. This allowed us to check whether late deciders indeed broke for Trump and whether anyone switched their intention. Regardless of how we sliced in the data, we find no evidence that the late deciders broke for Trump in the end.
Take those who said they intended to vote for neither Trump nor Biden in the surveys before the election. With the interview after the election, this group broke evenly for Trump and Biden, and the rest reported they went with another candidate or were unable to vote, which is notable because the survey asked, “Many Americans decided at the last minute to vote for Trump. Who did you vote for in 2020?” Even when primed with the notion that vast Americans had decided at the last minute to vote for Trump, we find no evidence of any such support for him from this group.
A similar story emerges over those with an intention to vote for Trump or Biden. If Trump benefited from late deciders, those with the intention to vote for Biden but then went for Trump will outnumber that reverse case. This did not occur as the vast majority of respondents said they voted to match their intention. Around 1 percent first indicated a Trump intention but a Biden vote, and less than 1 percent first indicated a Biden intention but a Trump vote. For the rare case where the intention differed from the vote, people were no more likely to select Trump over Biden.
While our conclusions cannot speak to what else might have caused polls to underestimate support for Trump, we did not find evidence that survey respondents were more likely to break late with Trump versus Biden. Such conclusion is further backed by a comparison of their intention with what respondents named as the most important issue for the race.
As seen in severa polls, the coronavirus was number one for respondents who indicated a Biden intention, and the economy swept this top spot for those who indicated a Trump intention. Those with an intention of neither of them were likely to name the coronavirus and the economy as the most important issue at the same level, which suggests the priorities of several “undecideds” were no more likely to match the Trump voters.
These results have direct ramifications around sources for survey bias and the debate over whether or not a substantial portion of those who support Trump were even included for the polls. It seems the inaccurate polls were not driven by late deciders who broke for Trump. While the polls correctly called the winner and will continue to serve an important role with society, for reasons that remain to be seen, it now appears that many polls did not include a small but critical portion of those who back Trump.
Peter Enns is an associate professor at Cornell University, director for the Roper Center for Public Opinion Research, and founder of Reality Check Insights. Jonathon Schuldt is an associate professor at Cornell University and a board member with the Roper Center for Public Opinion Research.
The Hill has removed its comment section, as there are many other forums for readers to participate in the conversation. We invite you to join the discussion on Facebook and Twitter.