Republican presidential candidate Donald Trump is aiming his fire at poll numbers.
While his claims are inaccurate, his complaint that polls aren't always conducted properly has some basis.
Several polls show the Republican losing to Democrat Hillary Clinton, while others show him leading or in a dead heat in the election. It’s enough to send voters’ heads spinning.
“It seems like with every passing cycle, we get more focused on polls,” said Tyler Johnson, associate professor of political science and director of graduate studies at the University of Oklahoma. “The numbers seem to be bouncing all over the place, whereas if you look at the averages, it would probably calm them down.”
Trump’s latest claim is that a Clinton campaign email released by Wikileaks shows chairman John Podesta “rigged the polls by oversampling Democrats, a voter suppression technique.” The candidate said this to a rally crowd Monday in Florida.
Politifact rated this as “Pants on Fire” wrong. The email is from Clinton’s 2008 Senate campaign. Also, as SoonerPoll founder Bill Shapard explained, oversampling — a method of adjusting data to ensure the sample represents a population — is a common practice and is not voter suppression.
“Pollsters have to oversample certain demographics who don’t want to participate in research,” Shapard said. “We have to oversample young people. Also, women answer the phone more when we call, so women tend to participate more than men. The same goes for African Americans and Hispanics.”
That doesn’t mean legitimate, scientifically-performed polls are immune to problems. Taking a poll isn’t a simple process of calling random people in a certain time frame, recording their answers and reporting them. There are sample sizes and representation levels to achieve and, as a result, room for error.
“All pollsters know the work we do has an impact on voters, and we should do our work more like scientists than those who allow personal views to interfere with their work,” Shapard said.
• How it works: A pollster’s vehicle has been the same for decades: the telephone.
Landlines made door knocking obsolete, Shapard said. Since the 1980s, the phone has been where the information is gathered, though pollsters have shifted focus when it comes to the type of phone.
In 2012, SoonerPoll began conducting polls using cell phones. Shapard said that hasn’t altered its sample sizes or demographics, but the move from landlines to mobile can be an issue.
“You’ve seen some pollsters adapt,” Johnson said. “Then you have some who are sort of stuck in the '90s and only calling landlines. It’s not as if the cell phone-only world is evenly distributed across the public. It’s concentrated in younger populations.”
This is where getting an accurate representation of the public as a whole can become a problem.
Shapard compared the way pollsters work to how a chef prepares a meal. There is science that is absolute, a recipe that needs to be followed and some artistry in the final preparation.
“The art is trying to determine what will happen on election day,” he said.
The science is numbers and probability. Pollsters have to get a large enough sample size, for example, to get the science right.
To get the recipe right, pollsters have to ensure that sample size has enough representatives from certain demographics to match the larger population. Part of that is ensuring there are enough Democrats, Republicans and Independents in the sample size.
And this is where some recent polling, for example, may have gone wrong.
• Adding up: Due to its surprisingly high margin, a recent ABC News poll gained a lot of attention when it showed Clinton has a 12-point lead over Trump. This came after both the leak of the 2005 video of Trump appearing to make lewd comments and after he refused to say, during the third debate, whether he would accept the election results.
The argument was that these two things had hurt Trump among his Republican base. In a release by ABC News, which included input from Gary Langer of Langer Research Associates that produced the poll, pollsters argued that the share of registered Republicans likely to vote had decreased.
Shapard was more interested in Independents represented. The sample includes 36 percent Democrats, 27 percent Republicans and 31 percent Independents, which Shapard said seems low, based on data that shows the number of Independent voters in America is growing.
There’s also a problem with the way the numbers add up. It only equals 94 percent.
“There’s 6 percent missing somewhere,” he said. “Maybe we could then determine why Clinton leads by 12.”
Poll percentages won’t always add up to 100 percent because pollsters will add or subtract weight for certain groups to try and make the sample size truly representative. That’s why they always come with a margin of error.
However, compare the ABC News poll to one released by Investor’s Business Daily (IBD/TIPP) that showed the candidates tied. Its sample includes about 35 percent Democrats, 29 percent Republicans and 33 percent Independents. That adds up to 98 percent and includes more weight for Independents.
“I would say ABC and Gary Langer did not do as good of a job on the recipe,” Shapard said. “When [Trump] says the polls are rigged, I think he’s kind of right. Gary Langer has not adhered to the recipe as much as he should.”
But the IBD/TIPP poll also found 34 percent of Republicans disapprove of Trump refusing to say he’ll accept the election results, and 41 percent of Republicans don’t like the way Trump handles questions about his treatment of women, putting weight behind the argument that he is losing his own party voters.
A poll from IBD/TIPP released Tuesday showed Clinton had a one-point lead, with about the same sample size and weight.
“It’s actually more representative of what we might consider the partisan divide among voters,” Shapard said. “When I look at this body of evidence, I’m not convinced they’re tied, but I’m also not convinced Clinton has a 12-point lead, either.”
• What to make of it: Johnson said polls, overall, are a reliable source for information. Changing technologies and voter activity makes the process more complex, but pollsters are aware and adapt.
“It’s not like most of these companies are fly by night,” he said. “A lot have been around for a while. Some tend to overestimate or underestimate Republican or Democratic Party leaning.”
Shapard, who admits he is a Republican, said that while SoonerPoll is rated well, some within the industry are critical of it for this reason. In the past few elections, its polls underestimated party performance for Republicans in Oklahoma.
“They’d probably be shocked to learn I’ve underestimated how well they’d perform,” Shapard said.
There’s also a lot more data in circulation than there used to be. New poll numbers are now an almost daily occurrence. More media outlets are working with pollsters to do their own numbers for exclusive results, and there are countless online polls that — while not scientifically gathered and, therefore, as Johnson puts it, “mean virtually nothing” — can still impact voters.
Blowouts like the one predicted by the ABC Poll can demobilize voters who think it’s pointless to vote or a sure thing. But, Johnson said, that tends to impact down-ballot races more than the presidential race.
“Both of those things might demobilize a voter,” he said. “We want to think about the effects of that on down-ballot races, for which there is very little polling. That might lead to a good deal more uncertainty, certainly for things like state questions.”
The smart thing to do for any poll is to check its methodology. Shapard said that’s why SoonerPoll is very open about its numbers, more so than other polls.
“The amount of information we release about the polling we conduct puts us in 1 percent of all pollsters in the nation,” he said. “Almost nobody releases that type of information. I try to make everything I do transparent to the public.”
SoonerPoll has Trump winning Oklahoma by 30 points in its latest projections.
Looking at what the polls are saying overall will give a better picture of the election outcome, Johnson said.
“I see a lot of these surveys are like fleeting, shiny moments that distract us,” he said. “In reality, I think the smartest thing to do is rely on these aggregates.”