Are Polls Overstating Donald Trump’s Real Level Of Support, Or Understating It?
There have been many arguments that polling has over-stated Donald Trump's actual level of support among likely Republican voters, but there's also a good argument that they are understating it and that Trump may do better when people start voting than many think.
Kentucky Senator Rand Paul, who is currently averaging around three percent or lower in the polls at the national level, as well as in early states such as Iowa and New Hampshire, contends that the polls showing real estate mogul Donald Trump leading the Republican field are overstating his support:
Republican presidential candidate Rand Paul (Ky.) on Sunday said it’s “deeply disturbing” that Donald Trump is still the race’s front-runner and predicted his rival would be “wiped out in a general election.”
Paul, who was in danger of being cut from the main stage of the last Republican debate because of low poll numbers, questioned the accuracy of public surveys.
“We’ve all let the polls consume us too much. I don’t think the polls are very accurate,” he said, noting that the polls were off by 13 points a week before the recent Kentucky gubernatorial race.
He said Trump is performing well in the polls because he’s getting a disproportionate amount of media coverage and many respondents who say they’ll vote for him aren’t likely to turn out on primary day.
It’s not uncommon for candidates who are behind in the polls to say that the polls aren’t accurate, of course, and as Jazz Shaw reminds us that the “skewed polls” argument is one that was made quite frequently back during the 2012 campaign by conservatives who didn’t want to admit to themselves that all of the available evidence leading up to the General Election was pointing toward the re-election of President Obama. On some level, of course, such arguments are understandable, especially from candidates themselves, but that doesn’t make them any more accurate. Indeed, in Paul’s case he isn’t really making the argument that the polls are skewed so much as he’s repeating the axiomatic point that Trump is doing well in the polls because he’s getting so much media coverage. There is, perhaps, some truth in that statement but it’s also the case that Trump is getting so much media attention because he’s not only leading in the polls, but because he’s leading by wide margins nationally and, with the exception of Iowa, at the state level in all of the early states, and that he’s been holding those leads virtually since he entered the race this past summer. Moreover, Trump has held these leads despite the fact that he’s often gotten exceedingly negative coverage because of the controversial things he’s said about Mexicans, John McCain, Megyn Kelly, Carly Fiorina, Ben Carson, a disabled New York Times reporter, and Muslims. To say, then, as Paul seems to be saying here that Trump’s success is purely a creature of media coverage is completely invalid and largely a sign of desperation on his part.
Viewing Paul’s comments in their most generous light, it’s certainly possibly to make the argument that Trump’s support in the polls may not translate into support in the polls, an issue I discussed myself earlier today. As I noted, there are signs that Trump may not have the kind of ground operation in early states like Iowa that is generally needed to get people to the polls and the caucus locations, and if that’s true then the polls could well be significantly over stating Trump’s support. That’s something we can’t know until people actually start voting, though, but yes it could be the case that Trump’s support is being overstated in no small part because of his celebrity and that it won’t translate into success at the polls but that’s a hypothesis that can’t be proven just yet.
On the other side of the coin, Andrew McGill at The Atlantic takes note of a study that suggests that Trump’s support in the polls, especially in the live voter polls that most major polling companies still rely upon, may in fact be understated. Namely, McGill notes, the study suggests that people may be embarrassed to say that they support Trump when talking to a live pollster and are more honest about their support when polled via an automated poll or in one of the new online polling methods that are becoming more popular in this election cycle:
Last week, research firm Morning Consult put this hypothesis to the test. Recruiting 2,397 registered Republicans and Republican-leaning voters online, the company split the sample into thirds—sending one group to answer election questions on a web site, another to an automated interactive voice response phone line, and the rest to a call center staffed by live interviewers.
Critically, this survey drew its respondents from the same general pool, which had all answered demographic questions beforehand. The only variation was the mode of interview.
The findings, released today: The Trump mode effect is definitely real. Just over 38 percent of people who answered via a web form said they supported Trump, compared to 32 percent of their peers who spoke to a call-center employee, a 6 percentage point gap. But that gap, among college-educated respondents, widened to 9 percentage points.
A similar split held true for registered voters who participated in previous elections, indicating that politically engaged people may also be more reluctant to tell a pollster their true opinion of Trump. One alternate explanation for the gap in levels of support for Trump registered by different polls has been their varying definitions of likely voters; live-interview polls tend also to use more restrictive definitions, making it hard to tease the effects apart. These results, though, imply that mode effects play a larger role than likely voter screens in the discrepancies.
To some degree, what this study is suggesting is a reverse version of the so-called “Bradley Effect,” which was named after former Los Angeles Mayor Tom Bradley, an unsuccessful candidate for Governor of California in both 1982 and 1986. According to polling in both elections, but especially in 1982, Bradley performed far better in the polling than he ultimately did at the polls. This was in an era when all polling was still conducted solely by live pollsters rather than any automated means. Given the discrepancy between the polls and the result of the election results, many pollsters and political scientists postulated that voters were telling pollsters that they would support Bradley, and African-American, or that they were undecided due to social pressure but that these voters ultimately ended up voting for Republican candidate George Dukemejian. Because it relied heavily on assuming what was on the mind of voters both when they talked to pollsters and when they voted, the entire theory was heavily dependent on assumptions of what was on the mind of voters when they were polled and when they voted, the “Bradley effect” remains simply a theory that voters felt social pressure to tell pollsters they supported an African-American candidate but no such pressure when they went to vote. However, the fact that there was evidence of a similar effect in other races where African-Americans were candidates kept the theory alive long enough that there were many who wondered if it would play a role n the 2008 Presidential election when polling was showing Barack Obama performing so well in the polls. If there ever was a “Bradley Effect,” the fact that President Obama won both election and re-election to the Presidency would suggest that changed attitudes about race since the early 1980s had greatly diminished this effect.
What this study is suggesting, then, is basically a reverse “Bradley Effect” when it comes to Trump which leads voters to be reluctant to admit that they support Trump when talking to a live person, but more willing to state their opinion when dealing with an automated phone poll, or one of the new, supposedly more scientific, online polls that are becoming more prominent as this election cycle goes on. To some extent, of course, accepting the results of this survey requires accepting that the automated and online polls are, generally speaking, as good at their demographic and other screening as live polling is. If that’s not true, then the discrepancies can’t necessarily be said to be due solely to the fact that people are embarrassed to say they support Trump. It’s worth noting, though, that this “mode effect,” as the study refers to it, is seemingly replicated in a separate study done by the Pew Research Center, which also found that this impact was more prevalent on questions where “social desirability,” such as the desire to not appear t support a controversial candidate or issue position, plays a significant role. In this sense, it is similar to the argument that the advocates of the “Bradley Effect” made that poll respondents felt some psychic pressure to appear to be open-minded by appearing to be open to the idea of supporting an African-American candidates when talking to a pollster, but to act differently in the anonymity provided by the polling place. In addition to the Pew study, Henry Olsen at The Atlantic, recently the possibility that Trump’s actual level of support may actually be greater than what we’re seeing in the polls, and used the example of the anti-immigrant right-wing parties in Europe as support for his argument.
If this “Trump Effect” exists, then, it could be the case that Trump’s actual level of support among Republicans, and perhaps outside the Republican Party, could be higher than it appears in the polls, and if that’s true then we could be in for an interesting 2016 campaign season. Much like the “Bradley Effect,” though, we won’t have any evidence to argue over regarding this potential phenomenon until people actually start voting. As McGill notes in his piece, though, if there is a “Trump Effect” then it could work against Trump in Iowa, where voters must get up in front of their friends and neighbors and state their support for their candidate publicly. As McGill says, if voters are too embarrassed to tell an anonymous pollster about how they really feel about Trump then it’s unclear how they’ll feel about doing it in such an open forum. Given that, the more likely location for demonstrating the existence of a “Trump Effect” would be in the early primary states such as New Hampshire and South Carolina, where voters can vote in privacy and anonymity without worry that people will know that they voted for an undesirable candidate. If the “Trump Effect” does exist, though, then Donald Trump could end up doing far better going forward than even the polls are saying right now.
Here’s the Morning Consult study for those interested: