2020 Election Polls: Should We Expect More Than 2016?

With November 3rd quickly approaching and election day anxiety intensifying, uncertain voters turn to the polls for some insight into what to expect. But, should we trust the polls? In 2016, national election polls had Hillary Clinton to win right up until Donald Trump’s surprise presidential victory. While polling ultimately results in a straightforward election prediction, there are many factors to consider when examining a poll. Often polling is more complex than simply comparing an election result to a poll’s prediction. 

What is polling?

Polling is used during an election to gauge public opinion about specific candidates or issues. Various organizations, including companies dedicated to national polling and research institutions such as universities, conduct polls. A few well-known examples include Pew Research Center, Survey USA, SSRS Research, Marist College and Emerson College. Pollsters ask respondents a series of questions and record their responses. Different organizations have different methods that affect the results and the accuracy of their polling.

The ultimate goal of a poll is to show the general opinion of a population. In reality, it is impossible to poll every member of the population so polls rely on a random sample to represent the opinion of the population. To accurately represent the population, the demographics of the population must be reflected in the sample. Different organizations use different polling methods leading to different results. Traditional methods include phone call polling, mail-in polling and in-person polling. Technological developments have allowed lower cost polling such as interactive voice response or robocalling and online polling (Hillygus). The data collected is analyzed and organized into the polling numbers we see every day. 

The information learned from polls demonstrates public opinion on issues and candidates during an election. A poll can predict election outcomes, describe voter behavior and influence campaign strategy (Hillygus). The results of the poll may influence voters to research or support a candidate. A political campaign may redirect resources to a specific area or demographic of voters based on the findings of a poll to more effectively campaign. 

So what happened in 2016?

While the 2016 presidential election may have led to the public questioning the legitimacy of polling, it was not an outright failure for pollsters as it may seem on the surface. Many factors contributed to the inaccuracies in the polls that made election night such a shock. These elements can be divided into two categories: sociological and methodological. 

A major sociological contributor to pollsters incorrect prediction of the 2016 election was the inaccurate assumption that Donald Trump would never win. This expectation is reflected in both voter behavior and political reporting. Ultimately, it was those who did not vote in 2016 who determined the election. In comparison to the 2008 and 2012 elections that won Obama the presidency, 2016 saw a significantly lower Democratic voter turnout (Ben-Shahar). The number of people who decided not to vote may have done so assuming Clinton did not need their vote (Westwood, Messing and Lelkes). Low turnout may also be contributed to the polls themselves. Pre-election polls may have given voters a false sense of confidence in Clinton, leading to low voter turnout. After the election, it was clear that undecided voters leaned towards Trump. While it is possible that these voters decided their candidate right before the election, there is also evidence suggesting that some undecided voters were actually committed to Trump, but identified themselves as undecided to pollsters (Cohn). The media also fell victim to the assumption that Trump would never win. The press dedicated more time to covering Trump than they did Clinton. Even though this coverage was mostly negative it was ultimately still attention for Trump’s campaign, contributing to his surprise victory (Azari). The effects of assuming Trump would never win were not considered when conducting polling, leading to imperfect election predictions. 

Other elements that contributed to an inaccurate 2016 election prediction were within the polling methods. Traditional election polling depends on phone polls. The switch from landlines to cell phones has made phone polls ever more unreliable in recent years. While landlines are associated with a specific street address, cellphones are more difficult to tie to a specific place meaning a pollster may inaccurately report the location of a respondent. The popularity of cell phones has increased the margin of error in phone polling. Since it is no longer reliable for pollsters to conduct polling by phone, there has been a surge in online surveying (Kennedy and Deanne). After the 2016 election, it was clear that education level played a large role in which candidate a voter chose. Pollsters did not take this into account, leading to a poll that overrepresented college-educated voters (Cohn). While most polls were ultimately unsuccessful in their predictions, many polls were accurate in predicting the outcome of the popular vote, but as Trump’s win made clear, the Electoral College decides the president. Polls did not effectively consider the role of the Electoral College. The factors that were overlooked in polling methods contributed to an inaccurate analysis and prediction of the 2016 election. 

Will 2020 be better?

Learning from the mistakes of 2016, pollsters have made changes in how they poll to more accurately represent the population leading up to the 2020 election. Polling has focused on representing groups who were overlooked in 2016. There has been a focus on representing mid-western states that were critical to Trump’s win (Skelley and Rakich). In order to ensure these rural voters are appropriately represented, respondents’ locations must be accurately reported. Poll organizations have also put effort into polling all educational levels. These changes are expected to more accurately show the influence of Trump’s supporters. Many of the sociological factors that influenced inaccurate 2016 polls are no longer at play in 2020 since it has been made clear that Trump can win.

There are some worries that the pandemic could lead to differences in the poll predictions and election results. Pollsters have expressed concern that COVID-19 may lead to unexpected turnout (Skelley and Rakich). For instance, a respondent may tell a pollster that they will be voting, but come election day there could be a spike in Coronavirus cases in their area that makes them feel too unsafe to go vote. There has also been an unprecedented surge in mail-in ballot requests leading to concerns that ballots may not be processed in time to count (Broadwater). 

While a poll can be a useful tool to understand what is going on during an election, it is important to not oversimplify election predictions. When looking at the results of a poll, one must consider the factors that go into the poll itself such as polling methods and analysis techniques. If 2016 taught us anything, it is not that we should not trust pollsters, but rather that there are innumerable complexities of reality that must be considered in a poll.

Works Cited:

Azari, J., (2016, September). Political Communication. Research Gate. https://www.researchgate.net/publication/308010264_How_the_News_Media_Helped_to_Nominate_Trump

Ben-Sharar, O., (2016, November 17). Forbes. Forbes. https://www.forbes.com/sites/omribenshahar/2016/11/17/the-non-voters-who-decided-the-election-trump-won-because-of-lower-democratic-turnout/#413db29d53ab

Broadwater, L., (2020, October 12). New York Times. New York Times. https://www.nytimes.com/2020/09/30/us/mail-voting-democrats-republicans-turnout.html

Cohn, N., (2017, May 31). New York Times. New York Times. https://www.nytimes.com/2017/05/31/upshot/a-2016-review-why-key-state-polls-were-wrong-about-trump.html

Kennedy, C., Deane, C., (2019, February 27). Fact Tank News in Numbers. Pew Research Center. https://www.pewresearch.org/fact-tank/2019/02/27/what-our-transition-to-online-polling-means-for-decades-of-phone-survey-trends/

Skelley, G., Rakich, N., (2020, October 13). FiveThirtyEight. FiveThirtyEight. https://fivethirtyeight.com/features/what-pollsters-have-changed-since-2016-and-what-still-worries-them-about-2020/

Sunshine Hillygus, D., (2011, December 1). Public Opinion Quarterly. Oxford Academic. https://academic.oup.com/poq/article/75/5/962/1830219
Westwood, S., Messing, S., Lelkes, Y., (2018, July 12). Social Science Research Network. SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3117054

Edited by Sean Francis