Why do some polls show Texas Democrats and Republicans in close races, but others show the opposite?
Is Democrat Beto O’Rourke in a “statistical tie” with incumbent Greg Abbott? Or is he facing a double-digit deficit with less than two weeks to go until Election Day? Is incumbent Attorney General Ken Paxton headed toward a third term, or is challenger Rochelle Garza within striking distance?
Texas voters who look to more than one poll for answers won’t have an easy time — a late-season flurry of surveys tell different stories about what Texans can expect on Election Day, Nov. 8.
In the governor’s race alone, three polls conducted within a week of one another showed considerable differences.
A poll conducted by The Texas Politics Project at the University of Texas at Austin, released October 21, showed Abbott up with likely voters by 11 percentage points, 54% to O’Rourke’s 43%. That’s an increase of 6 percentage points from a survey the same group conducted in early September. But a Beacon Research poll released Sunday showed O’Rourke within two percentage points of Abbott, as the incumbent garnered 48% of support compared to O’Rourke’s 46%. Just two days later, a Spectrum News/Siena College poll showed Abbott with a 52% to 43% advantage over O’Rourke.
So why are we seeing such different numbers? Joshua D. Clinton, a political science professor at Vanderbilt University who has studied polling and political trends for more than two decades, said polls are an “art and a science” but lately lean toward being more of an artform.
“[Pollsters] make assumptions about who is going to vote and what the electorate is going to look like,” he told The Texas Newsroom. “If you think that there's so-and-so percentage of Republicans and Democrats who are voting, or those with a college degree, or a postgraduate degree, or less than high school [education] or by race, by age. So, as a pollster, you need to make your data fit an electorate that you're making assumption about, which may or may not be accurate.”
He said that means results vary depending on who is contacted and who responded, on top of the assumptions pollsters make.
“When you see polls of wildly different estimates, it's not necessarily the data,” explained Clinton. “But that could show how there are different pollsters or assuming different things about who's going to vote or not.”
Take the UT poll that showed Abbott up by 11 percentage points: The sample size was 1,200 self-declared registered voters but was narrowed down to about 880 likely voters, which the pollsters defined as “respondents who indicated that they have voted in every election in the past 2-3 years; or those respondents who rated their likelihood to vote in the November elections on a 10-point scale as a 9 or a 10.” In the Beacon poll that showed a narrower race, the sample size was about 1,125 U.S. citizens who "definitely plan" to vote. But that poll also noted the margins change when pollsters focus on about 950 people who are “extremely motivated” to vote. Among that group, Abbott’s lead is just one percentage point, at 48% to 47%.
Clinton said, historically, there hasn’t been a big gap between how accurate polls of likely voters have been compared to polls of registered voters. But the difference in how someone identifies themselves presents a challenge for pollsters.
“The general point is that likely voter screening requires you, as a pollster, to make decisions about who you're going to count, who you're not going to count in ways that you don't have to make for registered voter polls,” he said, adding that some so-called likely voters might not end up casting a ballot even though they say otherwise.
“If I'm taking a survey and you ask me that question, [no one] wants to say, ‘I'm a bad American. I'm not going to vote.’ So, everyone says they're going to vote,” he said.
An additional challenge is that it’s getting harder to convince voters to take a poll, according to a Pew Research Center analysis of polling conducted in 2021.
“This is mostly a consequence of lifestyle changes – people seem to have busier lives – coupled with a growing wariness about cooperating with requests for information from a person or organization they are not familiar with,” said Scott Keeter, a senior survey advisor at the Pew Research Center.
And when pollsters find respondents eager to take a survey, they don’t necessarily represent a larger sample of voters.
“People with higher levels of education tend to be more willing to participate in surveys, resulting in samples that overrepresent those with college degrees,” Keeter said. “The same is true for people with higher incomes. Perhaps not surprisingly, people who are engaged with politics and public affairs are also more willing to take polls that deal with these subjects.”
So, at the end of the day, are polls important? Clinton said he thinks current polls are, for the most part, accurate predictors of what will happen next month. But he said what could swing any race is turnout, which is the most important factor.
“Even in the largest presidential election turnout … in 2020, turnout was only like 65%. Like 35% of people did not vote. And if you think about the margin of victory that was happening, more people [voting] could swing massively. So, if people chose to vote and participate, then they could completely upend what you know and change a lot of what we see in kind of politics,” he said.
To that end, Clinton said polls shouldn’t determine whether a person decides to cast a ballot.
“The only thing that really matters is who actually turns out to vote on Election Day. So if you let a poll, basically determine whether or not you're going to vote or participate, then I think you're doing a disservice to one another,” he said. “Polls are interesting to talk about and they give you some sense of what's going on in the world. But you should never let them drive your own decision to vote or not.”
Got a tip? Email Julián Aguilar at email@example.com.You can follow Julián on Twitter @nachoaguilar
Copyright 2022 KERA. To see more, visit KERA.