Politics

There are a ton of midterm election polls. Here’s a guide for how to read them

Polls often measure whether voters' have a favorable or unfavorable view of candidates, as well as asking specific questions testing candidates' strengths and weaknesses.

alt = white voting booths with American flags on them in a wood-paneled gymnasium
Anna Watts/The New York Times


Pre-election polls came in for a drubbing following the 2020 election after underestimating former president Donald Trump’s vote margin by about five percentage points, the highest error in at least 20 years according to the American Association for Public Opinion Research.

2022 Massachusetts elections

One possible cause of 2020 errors was something the experts called nonresponse bias, with Democrats participating in polls at higher rates than Republicans. Many polls have attempted to measure and correct for partisan nonresponse, but these adjustments have been far from perfect, and it will take time to know how effective they are in reducing error over the long run.

The accuracy of 2022 polls won’t be clear until votes are counted on Nov. 8. But even imprecise polls can provide useful information about how and why voters are making their decisions this fall.

Advertisement:

Here is a guide on how to get the most out of pre-midterm 2022 polls:

1. Don’t sweat small “leads” or changes.

While you might see reports of candidates “leading by three points” in individual polls, that has little meaning beyond saying the contest is very competitive. The same goes for polls finding a candidate’s support has increased or decreased by a few points from a previous survey. The reason for such small discrepancies is that very few pre-election polls are precise enough to measure such differences.

While the Census Bureau interviews 60,000 households monthly to measure changes of less than 1 percent in the country’s unemployment rate, most pre-election polls interview 1,000 or fewer voters.

Even if we assume (unrealistically) that a survey reached a perfect random sample of Americans or registered voters, typical election polls have margins of sampling error between 3 and 5 percentage points. Readers also need to remember to apply the error margin to both candidates in a question about voter preference, so the lead actually needs to be nearly twice as large as that error margin to be statistically significant. And sampling error is only one type of error; nonresponse error and error associated with how voters’ preferences are measured are also factors, but are more difficult to estimate.

Advertisement:

2. Browse a poll’s full results page.

News stories and poll reports focus on the buzziest results from a poll – including who’s leading overall and among key groups. But polls ask many other questions that can give you a richer understanding of voter attitudes.

You can often find more questions about opinions by clicking on a poll’s full results. There, you might find opinions on campaign issues like inflation, abortion or climate change. The Post’s polling archive links to overall results and cross tabs by group for all of our surveys.

3. Look at what voters think about the candidates.

Polls often measure whether voters’ have a favorable or unfavorable view of candidates, as well as asking specific questions testing candidates’ strengths and weaknesses.

Those findings illustrate how voters are reacting to the campaign and help explain why candidates are leading or trailing in overall support, as well as the challenges they face in the final weeks of the campaign.

Favorable ratings may be an indicator of which candidate undecided voters will swing to, But they’re also important in understanding whether they are supporting their favored candidate because they like them – or because they dislike their opponent.

For instance, a Pennsylvania CNN poll released Monday found double-digit unfavorable ratings for both Republican Senate candidate Mehmet Oz and Republican gubernatorial candidate Doug Mastriano. Meanwhile, Democratic Senate candidate John Fetterman received middling favorability ratings and Democratic gubernatorial candidate Josh Shapiro was in double-digit positive territory. Shapiro led Mastriano among likely voters by 15 points in the governor’s race, and Fetterman held a narrower six-point edge over Oz in the race for U.S. Senate.

Advertisement:

In late September, The Post’s Aaron Blake compared favorable ratings of Democratic and Republican candidates across 20 polls, finding Republican candidates trailing on this metric in most key states.

4. See whose supporters are more motivated.

Turnout reached its highest levels in a century in the past two general elections. But even so, half of eligible voters did not vote in the 2018 midterms. Pre-election polls typically ask respondents how likely they are to vote, providing a useful indicator of what types of voters are most committed to turning out.

In a Washington Post-ABC News national poll last month, 76 percent of registered voters who supported Republicans for Congress said they were certain to vote compared with 74 percent of those supporting Democrats, a statistically insignificant difference. Other recent polls have shown a slight Republican edge in motivation to vote.

In 2018, when Democrats won control of the House, the final Post-ABC poll before the election also found relative parity in this measure, with 83 percent of voters who supported Republicans and 79 percent of those who supported Democrats saying they were certain to vote or had already done so.

But in 2014 when Republicans expanded their House majority, Republicans had an 11-point advantage on this question – with 72 percent of their party’s supporters saying they were certain to vote (or had voted early) compared with 61 percent of Democrats.

5. Compare results to polls from past elections.

Comparing new polls with those from previous elections can point to significant shifts among different groups of voters.

The Post’s Sabrina Rodriguez recently highlighted recent polls showing Democrat Stacey Abrams is averaging 83 percent support among Black voters in Georgia’s election for governor. That’s an overwhelming majority, but it is down from 2018 when Election Day surveys showed between 93 and 94 percent of Black voters supported her, according to AP VoteCast and network exit polls. CNN’s Harry Enten found a similar pattern in national polls, with Democrats receiving 10 points less support among Black voters than in 2020.

Advertisement:

The 2018 midterms saw women support Democratic House candidates over Republican House candidates by 19 percentage points, the largest margin in the history of exit polls. Democrats held a smaller eight-point advantage among women voters in a September Post-ABC poll, despite signs that women have reacted strongly to the Supreme Court’s decision eliminating the right to abortion.

6. Be wary of outliers and campaign-sponsored polls.

Each new poll usually adds more information to our understanding of public opinion. But there’s a substantial amount of random variation in polls – in addition to differences in methodology – and it is worth being skeptical of surveys showing a sharply different result than other recent data.

It is possible that such a poll is capturing a significant shift in voter attitudes, though more often than not they will remain outliers. Poll averages are a useful solution to this, incorporating information from multiple polls while smoothing out bumps. But if most polls are overestimating one party’s support, an average of polls may not correct for this.

Lastly, be wary about polls sponsored by candidates or advocacy groups. Even if such polls are sometimes conducted using quality methods, they are often selectively released when results are beneficial to their favored candidate, unlike independent polls commissioned by news organizations or other nonpartisan sponsors.

Conversation

This discussion has ended. Please join elsewhere on Boston.com