OPINION: Can We Trust the Polls?

“The electoral vote count from 2016 and 2020” Photo by Shayanne Gal, Business Insider.

Zach Richards is a senior studying history and an opinion writer for The New Political.

Please note that these views and opinions do not reflect those of The New Political.

With Democratic candidates doing relatively well in recent polls, it is reasonable to look at polling inaccuracies in the 2016 and 2020 elections and ask if they will be just as inaccurate in the midterms this year. However, recent special election results and the accuracy of polling from 2018 suggests that the polls may be more precise than many assume.

The Democratic Party has undergone a massive change in fortune in the last two months. The generic ballot, which measures the electorate's opinion on which party should be in power, is three points better for the Democrats than it was in July, giving them the lead for the first time since November 2021. Polling in individual races looks even better. Democratic candidates currently have polling leads in the Senate races in North Carolina, Georgia, Nevada, Wisconsin, New Hampshire, Arizona, Pennsylvania and even Ohio. If these polls are accurate, Democrats could have as many as 54 seats in the Senate next year and possibly hold the majority in the House of Representatives.

It would be highly unusual for the Democrats to gain seats in either house of Congress this year, as they are the party in power. The party in power almost always loses seats in midterm elections. When presidents are popular, there are exceptions to this rule. However, President Joe Biden’s approval rating is low. Thus, it is reasonable to ask if the polls will be accurate this year.

In both 2016 and 2020, the polls underestimated support for former President Donald Trump and other Republicans. It could very well happen again in 2022. Even with their polling deficits in various races, if the polls are as inaccurate as they were in 2020, Republicans will be favored to take the Senate.

Still, if the Democrats defy the odds and over perform in November, the reasoning will be obvious in hindsight. A common explanation for why parties in power usually lose seats in midterms is thermostatic public opinion. The idea is that when a party in power implements its agenda, a change-resistant public will react negatively and vote them out. However, even though Democrats control Congress and the presidency, the conservative Supreme Court has handed down the Dobbs decision overturning Roe v. Wade. The decision hands the responsibility to the states to dictate abortion access.The Dobbs decision is arguably more unpopular and disruptive than anything the Democrats have done, so it would make sense for Republicans to face more backlash than Democrats.

Special election results have supported the thesis that Dobbs created a backlash against Republicans. In special elections before the Dobbs decision, Republicans generally did better than Trump did in 2020. This changed almost as soon as the Supreme Court handed down Dobbs, notably Democrats have outperformed Biden in every special election since. Democrats' overperformance in special elections suggests that polls predicting such might be accurate.

To know if polling errors from 2016 and 2020 will repeat, it is helpful to examine why the polls were inaccurate to start. One hypothesis points out that Trump does particularly well with voters who have low trust in societal institutions. Because of this, these voters are less likely to answer polls, making the overall polling less representative.

This hypothesis further states that low-trust voters tend to view the Republican Party as just another institution and are far more loyal to Trump personally than to the party. They are less likely to vote in midterm elections where Trump is not on the ballot, making the polls more representative. If these people are not voting, their lack of response will not be an issue. This could explain why the polls were much more accurate during the last midterm in 2018 than in the previous two presidential elections.

2016 and 2020 are reminders that the polls will not always be accurate. However, the 2018 polls and special election results suggest that polling errors do not always repeat themselves. Either way, polling can be a useful tool for forecasting which electoral outcomes are within reason, as long as no one accepts them as gospel.