From May 6-9, Public Policy Polling compared Donald Trump to hemorrhoids, root canals, lice, traffic jams, and cockroaches. Not surprising it was the one poll showing Hillary Clinton still ahead.
This is an old polling trick. When a supporter of one candidate is answering a survey, and suddenly starts having one insult after another hurled at that candidate, some of them start to get fed up and hang up – thus not completing the poll.
Another recent PPP poll gave the false impression that North Carolinians oppose Governor Pat McCrory’s attempt to prevent men identifying themselves as transgender from going into women’s restrooms. Their press release blasting McCrory showed opposition to HB 2 at 36% to 45%. Ironically they never asked the very straightforward question in a poll run by a North Carolina TV station showing strong support for the bathroom provision:
“How do you feel about HB2 requiring people to follow their birth certificate in using a restroom?”
Agree 55% (45% strongly agree)
Disagree 34% (25% strongly disagree)
Not Sure/Neither/Rounding 11%
I have run polls with live calls to hundreds of thousands of voters, and we had to be very careful to balance “negative” questions about competing candidates in order to make sure the result accurately reflected the results at the ballot box. Obviously the more Trump supporters who get fed up during questions 22 to 31 in the PPP poll – all of which insult Trump – the fewer complete the overall survey.
The insults start after the “Trump vs. Clinton” question is asked, but polling methodology traditionally involves only counting those who complete the whole survey so that you have a breakdown of age, gender, and race.
While PPP did not return an email from me Tuesday asking if they dropped results from voters who did not complete all 38 questions, the methodology is biased either way:
- If they included the partially completed polls by Trump supporters who quit during the insults, then they do not really know the demographic breakdown of their responses because they did not get to question 38.
- If they did not include the partially completed polls, then they dropped a lot of results AFTER the voter stated their support for Trump, thus skewing their poll toward Clinton.
Either way, the questions constitute terrible methodology being used to create a press release trashing Trump rather than learn where the race is.
Even among the 40 percent of PPP’s respondents who say they voted for Romney in 2012, their sample likely includes more Romney supporters who do not like Trump and therefore are more likely to stay on the call through the battery of insults.
Polls may move up and down through the campaign, but it is important to distinguish the bias polls that PPP runs that are designed to create sensational press releases and be self-fulfilling prophecies, vs. polls attempting to measure where the public is. In 2012, it was many Republican polling firms that biased their polls toward Romney by only calling landlines, which led many of us to give “garbage-in-garbage-out” projections.
Perhaps bias polling from sources such as PPP was the reason that this year it was Nate Silver eating crow because of his wildly erroneous projections on Trump — giving him a 5% chance of the nomination. Silver made his erroneous projection months after Trump redefined the race in an early debate by stating, “Give (politicians) money, and they’ll do whatever they hell you want.”