Saturday, October 01, 2016

The Puzzling Polls (by William Hausdorff)

By William Hausdorff

One of the signature features of this bizarre political campaign was evident within the first few minutes of the presidential debate: Donald Trump simply lacks the knowledge and temperament to be President, in stark contrast to the eminently qualified Hillary Clinton. This raises what for many people is a most disturbing and puzzling question: how is it that Trump’s maddeningly obvious lack of fitness is not recognized by approximately half of the voting population?

The short answer appears to be that, actually, it IS apparent to a significant portion of Trump supporters, based on public polls. For example, in the mid-September Fox News poll 61% of respondents thought Clinton was “qualified” to be President, vs only 45% who thought Trump was. In the same poll, 59% said Clinton had the “right kind of temperament and personality” to be a good president, while only 31% said Trump did. Furthermore, 47% said electing Clinton as President was a “safe choice for the country” while only 30% said that Trump was. It should be noted that these results don’t significantly differ from those obtained from polling conducted weeks and months earlier on the same questions, nor from a NY Times/CBS News poll also conducted in mid-September.

It is striking that that these large differences were recorded at a time when the candidates were each polling in the range of 43-48%. Even when polled on which candidate would do a better job in specific areas, such as immigration, or national security, or managing the economy, respondents usually favored one candidate over another by relatively small margins.

So perhaps the disturbing and puzzling question is, in reality, how can a significant proportion of people who admit Trump is unfit for the job, if not a real risk, nonetheless say they will vote for him?
These seemingly contradictory polling results make me wonder about the predictive value of polls in this race. Last June, many Europeans mournfully pointed to the apparent polling failure regarding the outcome of the UK vote to leave the European Union (Brexit), and suggested it could be a harbinger of what will happen with the US Presidential race. So I looked to see whether the Brexit post-mortems by British analysts indicated that their polls had also offered contradictory signals that, in hindsight, had been missed.

However, the explanations I found centered around technical aspects, such as differences in accuracy between phone polling vs internet polling, or around unpredictable events. The latter included a lower than expected turnout of young voters, undecided voters not splitting as predicted, and the confusion engendered by the unfortunate death of MP Jo Cox.

While these explanations were interesting, and some (e.g, what the young and the undecided voters actually do) are clearly relevant to this race, none highlighted the existence of contradictory polling results.

I then consulted Nate Silver’s polling analysis website, , as he is the most critically thinking poll analyst I have read. He posted a recent panel discussion of what the polls might be missing that could affect the presidential election outcome.  One possibility the panel considered is that the true Trump vote may actually be underestimated by polls. This would occur if some Trump voters consider it socially unacceptable to admit to a pollster that they favor Trump, thereby constituting a version of the famous “Shy Tory” or “Bradley effect.” (Note: they dismissed this idea for various reasons.)

Nonetheless, the Bradley effect may be a useful concept here. The phrase derives from the California gubernatorial race in the early 1980s, in which the black Mayor of Los Angeles Tom Bradley was well ahead in the polls but nonetheless lost the race narrowly to his white opponent. Some attributed this “polling failure” to a hypothetical unwillingness of a portion of white voters to admit to pollsters that they wouldn’t vote for a black candidate.

As a parenthesis, I should confess that as a then-resident of North Carolina I was on the lookout for evidence of this phenomenon in the 1990 Senate race. That race pitted the liberal black Democrat Harvey Gantt against the incumbent Jesse Helms, then considered the most conservative US senator. The polls depicted a surprisingly close race, with some putting Gantt a few points ahead immediately prior to election day.

Gantt ended up losing by a few points, but locally it seemed evident at the time that the shift could be attributed, not to a re-run of the Bradley effect, but to a notoriously racist political advertisement the Helms campaign ran in the last few days of the campaign. The ad depicted a pair of white hands ostensibly holding a job application rejection notice, with the ominous voiceover saying, “You wanted that job—and they got it...due to racial quotas.” If the Trump-Clinton race is really polling that close in November, perhaps this will be a model for last minute Trump campaign ads.
Although the fivethirtyeight crowd didn’t consider it, I wonder if the polling contradictions in the present day presidential race might be explained as a variant of the Bradley/Shy Tory effect, but this time in favor of Clinton. In other words, some people are telling pollsters they will vote for Trump, even though they don’t necessarily intend to.
I have argued in an earlier post that Trump’s support is largely visceral in nature, and that it is doubtful that most rational supporters believe he will or can implement his few, wild ideas. But the same people are seriously concerned about the uneven economy, about immigration, about “the rigged system” and want to send a message that they are ANGRY and want a CHANGE. They appreciate that he has raised the issues to the forefront.

Telling the pollsters that you are going to vote for Trump is a clear message that has indeed shaken up the establishment. But when push comes to shove, will all those angry people who believe that Trump is unqualified and risky even make it to the polls? If they do, will they actually vote for him? I’m wondering.

I freely admit that entertaining these doubts could be self-deluding, fueled by wishful thinking. Indeed, the most convincing explanation I read as to why the bettor market for the Brexit vote made Remain an overwhelming 85% odds favorite on election day was simply that: wishful thinking. More than one post-mortem highlighted that, although a majority of polls had Remain winning, there was actually a sizeable number pointing in the other direction, especially in the days immediately prior to the June vote.

In other words, despite the closeness of the polls, it was just too difficult (for the bettors) to imagine that UK voters would actually pull the plug on EU membership. Sounds frighteningly familiar.

The post-debate polls may show a sizeable, possibly transient, boost to Hillary Clinton. However, if for many of his supporters Trump is deemed more of a highlighter of issues, rather than a serious candidate, it is all the more reason why Clinton should now openly acknowledge that Trump has raised some important, if not legitimate, concerns.

For example, on jobs moving overseas she could agree that there are significant problems with the Free Trade agreements, describe how she became aware of them and changed her views, perhaps even acknowledge the influence of the Sanders campaign in the latter. She then could offer her own solutions, pointing out that scrapping free trade altogether, erecting significant tariffs, or demonizing China, won’t, however, revive the coal industry.

Similarly, she could acknowledge that there remain serious problems with the current immigration system, and that Trump’s apocalyptic rhetoric about crumbling US infrastructure is not completely delusional. And that as highlighted by the recent Wells Fargo scandal, she also recognizes that many executives in the financial industry/Wall Street appear unrepentant and continue to act with complete impunity. Therefore, despite earning a considerable amount from her Goldman Sachs speeches, she understands that significant, painful reforms are still needed. In other words, it may be the perfect time for her to take the issues raised by Trump (and Sanders) seriously.


William Hausdorff received his PhD in Biology from the Johns Hopkins University/National Institutes of Health and conducted post-doctoral research in biochemistry at Duke University.  For the past 25 years he has worked and published widely in the field of international public health, initially with the US Centers for Disease Control/US Agency for International Development in Washington DC and Cairo, Egypt, and more recently within the vaccine development divisions of two pharmaceutical companies.  At present he is a freelance consultant based in Brussels, Belgium. He has closely followed presidential politics since the days of McGovern/Nixon.  His special interest is in the intersection of science and society, dating from his undergraduate thesis on the health effects of Agent Orange. His prior DoL posts can be found here.


Asher Steinberg said...

"I then consulted Nate Silver’s polling analysis website, , as he is the most critically thinking poll analyst I have read."

Sam Wang, a Princeton neuroscience professor who does gerrymandering research and polling analysis on the side at, is quite a bit better. Anyway, I think the answer to your question is that people can, and many people do, prioritize their ideological and partisan preferences over the personal qualities of the candidates. (They may also think that Clinton suffers from crippling personal defects that they view as more disqualifying than Trump's; a voter might view Trump as incompetent, Clinton as wildly dishonest and corrupt - indeed, I think a significant share of the population holds both views - and prefer to vote for an incompetent candidate over a corrupt candidate.) Just imagine, if you would, that someone of Trump's unpreparedness, ignorance, and lack of impulse control were the Democratic candidate, while Ted Cruz was the Republican candidate. Would it surprise you if half of the country told pollsters they intended to vote for this liberal Trump? I would suspect that the majority of commentators who claim they can't understand how anyone could support Trump would support a liberal Trump over Cruz.

Shag from Brookline said...

And a voter might view Trump as not only incompetent but also corrupt.

Fred Raymond said...

"....the true Trump vote may actually be underestimated by polls. This would occur if some Trump voters consider it socially unacceptable to admit to a pollster that they favor Trump...."

At least in my workplace, I feel it is quite the opposite. I'm surrounded by loud Trump supporters and so I claim to them that I just like to keep my politics private when actually I'm afraid to voice my HRC support.

One anecdote, I realize.