Likely voters split between both parties as many Americans don’t know who they will vote for or won’t vote

Washington, DC, November 7, 2022 — How do survey results of the entire American adult population differ from that of the registered voter population? Or of the likely voter population? Recent survey data from the Ipsos Core Political weekly survey shows a close race in the generic ballot between Democratic and Republican candidates for Congress (37% and 36%, respectively) among the most recent survey of adults, with 23% of respondents reporting they don’t know who they will vote for or will not vote.

Registered voter samples and likely voter samples taken from the Ipsos Core Political Survey data still show a close race between Democrats and Republicans in the race for congress, but show that registered voters and likely voters are more likely to support either Democrats or Republicans than report that they do not plan to vote or are still unsure. To help those interested in better understanding how we arrive at results among registered and likely voters, this analysis will not just compare results between adults, registered voters, and likely voters but also explain the methods used to create these samples.

First, it is important to understand the difference between the three populations that are being surveyed. The adult sample refers to people who are over the age of 18. The registered voter sample refers to people who say they are registered to vote at their current address. The likely voter sample refers to those who we believe are likely to vote in the upcoming election.

This analysis uses data from the Ipsos Core Political Survey, a weekly tracker of how Americans view the President’s job performance, issues that Americans view as important, and the direction the country is headed. The survey also allows us to view how the population is likely to vote in an upcoming election.

When controlling for registered voters in the most recent Ipsos Core Political Survey, 43% of respondents say they will vote for the Democratic candidate for Congress, 40% say they will vote for the Republican candidate, 4% say they will vote for a candidate from another party, 2% say they will not vote, and 10% say they don’t know who they will vote for. Among registered voters, the generic ballot has remained competitive since Ipsos included the generic ballot question in early August.

While this still shows a close race nationwide, registered voters are less likely to say that they do not intend to vote than the adult sample. Among adults, 23% report they will not vote or don’t know who they will vote for. Among registered voters, only 12% of respondents report the same.

For those interested in how a specific election may turn out, the registered voter sample would be more relevant than the adult sample as respondents who are registered to vote are more likely to vote in an election than those who are not registered.

But those interested in discussing voter turnout, voter engagement, or voter enthusiasm in a certain election would more than likely find the adult sample more relevant as it is meant to reflect the opinion of those who are eligible to vote, not just those who have registered.

Another population often discussed when conducting a poll related to an election is the population of likely voters. Determining a likely voter population from a survey can be simple. Most tools used to determine this population identify key questions in a survey which would indicate whether a respondent is likely to vote.

For example, if a respondent says they voted in 2020 this may indicate whether they are likely to vote in 2022. In this analysis, we also used age to determine whether a respondent was likely to vote as older potential voters are more likely to vote in elections.

Finally, based on recent election turnouts we projected what voter turnout might look like in the 2022 midterms. We used that projection to determine what share of the total adult sample would be used for our likely voter sample. To give an example, if we projected that half of Americans would vote in the 2022 midterms then we would use the half considered most likely to vote out of our adult sample as our likely voter sample.

Compared to the survey of adults, significantly larger portions of the likely voter sample are supportive of either a Democrat or Republican for Congress and likely voters are significantly less likely to report not voting or being undecided (4%) than among all adults (23%).

Among likely voters, the race remains tight with likely voters growing more decisive as the election gets closer with 13% of likely voters being undecided in mid-September and only 4% saying they are undecided in late October. Over the same time period, Democratic candidates have gained six percentage points and Republican candidates gained three percentage points.

Likely voter models can be beneficial tools used to project what an election may look like and who may turn out to vote based on reported voting patterns. They can be used to show a difference between how society feels about a given election compared to how the potential electorate feels about the election.

But this doesn’t mean that these analyses are perfect. Controlling for likely voters often leaves more room for error compared to the original sample. If we were to be wrong about the impact age would have on likelihood to vote, incorrectly project how many Americans will vote in the next election, or use the wrong questions to determine who was likely to vote our model could be even more inaccurate than the margin of error would suggest.

Also, the likely voter sample has a larger margin of error due to there being a smaller sample of likely voters than adults. Among adults, the margin of error is at 3.8% with the margin of error among likely voters being 5.6%.

This means that, while we report that Democratic candidates have support among 48% of likely voters in the generic ballot, their actual support could range anywhere from 42.4% to 53.6%. With a wide confidence interval and a reported margin of two percent this would indicate a tight midterm election.

Controlling for the likely voters in surveys provides useful data when attempting to project an outcome of an election to get the best sense of who might show up to vote come Election Day. For this election, applying likely voter models to Ipsos Core Political Data confirms that the upcoming midterm election is likely to be close.