Polls are all the rage in political commentary. They make for easy targets for news stories (even if they are bad or fake polls). One week, Joe Biden is leading the presidential polls. The next week sees Elizabeth Warren on top. Now suddenly Pete Buttigieg surges or Bernie Sanders makes a big push. And on and on. The “horse race” aspect of polling is what most people are interested in, but not so much the very nitty-gritty aspects. People tend to simply be more interested in a straightforward answer as to who is “winning.”
Polling is also a huge, expensive business. Good polls are critical for well-oiled campaign machines in heated races to determine campaign strategies. They’re also intricate, difficult, and expensive to make, which is why the highest echelons of the business rake in so much money from campaigns seeking their services. Well-crafted polls don’t always need a lot of money, but they are very limited in scope if there isn’t much financial backing. Each successive question asked gets a clearer understanding of the race, but it also requires more time and resources.
One of the details of polls you may have heard is the term “likely voter.” If you’re interested in who’s going to win, you usually don’t want to know what just anyone thinks if they aren’t going to come out to vote. The people who do show up to the polls are going to be picking the winner(s) and loser(s) of the contests. (Of course, there are other instances where we would want to know what nonvoters think and why, such as looking to increase civic engagement. It’s rare, however, to look at such groups when trying to win an election because they are so unlikely to turn out.)
But the problem is: What exactly is a “likely voter?”
Definitions matter a great deal in politics, and it’s no different here. If you went up to five different pollsters from different companies in different electoral contests and asked for exact definitions of this term, they may very well have five different ideas. The basic concept is straightforward: It’s somebody who is probably going to vote. But how do you determine that?
Those pollsters are almost guaranteed to not tell you how they specify the term because it is proprietary knowledge. Remember that polling is a very lucrative business. Firms want to give as little information that competitors could use against them as possible, so they don’t disclose all of the details of how they conduct polls. So the exact formulas for calculating likely voters are going to be kept secret.
As long as we are dealing with statistics and human behavior, figuring out who the “likely voters” are will never be a perfect, exact science. Despite this fact and pollsters’ reluctance to reveal their methodology, we do have some ideas of how likely voters are measured.
Ways of Measuring the “Likely Voter”
Good polls are carefully crafted with every single word of every single question being poured over to ensure that they contain no bias or influence certain responses. Some of these questions can be used to get a sense of who is likely to vote.
The most straightforward way to do this is to straight-up ask how likely the respondent is to vote in the upcoming election. The set of answers can be set up in many different ways. It could be a scale of 1–10 or 1–100, where higher numbers mean more likelihood of voting. Or it could be a set of specific worded answers such as “very likely/likely/unlikely/very unlikely.”
Perhaps this seems like the obvious way to tease out the likelihood of voting, but there are many reasons why it is not. For one, there could be a social desirability bias, where people are influenced to pick a response that would be looked upon approvingly by others. Many political pundits, academics, pollsters, etc. have concluded that this was a major reason why support for Donald Trump in the 2016 election was underestimated. In the case of the likelihood of voting, it is usually socially frowned upon to not go out and vote, as many people treat it as a civic duty. Thus, to be more favorably looked upon, some respondents will simply say they are likely to vote, even though their answers in the polls are confidential.
For two, people’s feelings can change over time. Maybe the respondent heard something that Donald Trump or one of the Democratic presidential candidates say that really incensed them, and at that moment they are charged up and ready to vote against them. As time goes on, however, that feeling could dissipate for various reasons.
There are other questions that can also be used to gauge how likely a voter will make it to the polls. Maybe a question asks how “interested” or “excited” they are about the election. The more intense such feelings are, the more likely they are to vote. This can still run into the same problem as asking them how likely they are to vote, but they are other methods available.
Or maybe pollsters will use seemingly unrelated questions, such as opinions on specific political figures or issues that they think will be particularly important to the election. If a pollster thinks that a strong opinion of Donald Trump is a key indicator of someone’s likelihood of voting, he or she may use that in their calculation.
Maybe there is some conjunction of these possibilities. In any case, there are a lot of options to choose from when wording questions to gauge the likelihood of voting.
Past History of Voting
The voter’s past history of voting is a strong indicator of whether they will vote in the next election. Polling firms will have voter rolls at their disposal, which will have which recent elections that the voters participated in (it will NOT, however, show who they voted for or in which particular races that they made a choice).
How often do voters need to vote before being considered “likely” to vote again? Pollsters will usually look at what types of elections that that voter participates in. If a voter has only participated in a bunch of presidential-year elections but never a midterm, they’d probably be a likely voter in 2020 but there is little chance they will be considered as such in 2022, the next midterm election.
There is also the issue of gauging brand-new registered voters because they have no previous electoral history to speak of. Those who just turned 18 and registered to vote aren’t necessarily going to participate — they may have been automatically registered when they went to get a license, for example. Yet they may be very civically engaged and raring to cast their first-ever vote.
Certain demographics are more likely to vote than others. For example, middle-class and upper-class voters tend to vote at higher rates than lower-class voters, so each of these groups’ likelihood to vote might be weighted differently by certain pollsters.
There could also be some other criteria we don’t think about or are unaware of. It’s difficult to tell since pollsters keep their voter likelihood formula close to their vests.
Likely Voters and the 2020 Election
It is still early on in the presidential election season, with the Democratic field still being whittled down on the way to the first primaries. It is useful to keep an eye on those “likely voter” opinions, but don’t read too much into it. Likely voters in primaries as a whole are a different group than likely voters in general elections. Many voters only participate in general elections but not in primary elections.
As the Democratic race for the first primary votes heats up more and more and polls show constant changes in which presidential candidate is leading, it’s hard to get a read of how exactly Iowa, New Hampshire, etc. will shake up. Yet even if they were more predictable, polling can only tell you so much. Giving an opinion to a pollster does not inherently mean that that voter is likely to actually cast a ballot. Of course, this does not mean that the polls should be dismissed or anything of that sort. It means that a bit of caution should be taken before making predictions.
And with so much focus on the Democratic primaries, Democratic voters, controversies centering on Donald Trump (particularly the impeachment process), and presidential approval ratings, it might be easy to start thinking that voters are going to easily vote for the Democratic nominee against Trump. There is still plenty of time until the general election season, however, when the “likely” voters for Trump start getting ready to cast ballots.