Quinnipiac University has developed a reputation for outlier polls that consistently deliver bad news for Hillary Clinton and good news for Donald Trump. When we deconstruct their polls, Quinnipiac’s errors are glaringly obvious.


On June 29, we took apart a Q poll that showed Donald ahead of Hillary nationally by two points.

rpoll


My colleagues Anthony Reed (predictive modeling expert and founder of the highly respected Benchmark Politics) and Eric Kleefeld pointed out that Quinnipiac was alone in its depiction of the race as particularly close. Most other polls at the time indicated a Hillary lead of between 4 and 6 points, and some polls placed her advantage at 8 points or above. Unsurprisingly, the Q poll got a wave of breathless media coverage. But a “unique poll result” is typically an indication of a bad poll result.

Qp2

Here’s Quinnipiac’s fundamental problem: They tend to under-represent minorities. In their June 29 analysis, Reed and Kleefeld argued that polls like PPP, IBD/TIPP predict white turnout will be around 70 percent in 2016, down from 72 percent in the 2012 exit poll. Quinnipiac, on the other hand, has white turnout pegged at 73%. Another dubious finding in the June 29 Q poll is Latino support for Donald at 33% when most polls place it around 20%.

Quinnipiac’s reputation was already very shaky, in light of their dishonest, discredited, and widely disseminated 2015 poll announcing that in a word-association exercise, voters immediately thought of Hillary as a “liar.”

While Quinnipiac presented the poll as evidence that voters associated “liar” with Hillary, we demonstrated that it was Republican and Republican-leaning respondents to the Q-poll who linked Hillary to liar and other derogatory terms (including “bitch”). It is a vastly different thing for Republicans, parroting Fox news and talk radio, to hurl misogynistic insults at Hillary than for all voters to believe Hillary is a liar.

As Mediaite explained at the time: “As far as we know, only 35% of the people asked had something negative to say about Hillary Clinton.” But the way Quinnipiac and the media played it, Hillary was seen as a “liar” by the majority of American voters.

With Quinnipiac’s questionable history of faulty 2016 polls, their latest outlier, showing Donald ahead of Hillary in battleground states, should be taken with a grain of salt.

Reed explains that Quinnipiac seems to consistently produce different results from all other polls in Florida, Pennsylvania, and Ohio. Of all major polling groups, Quinnipiac consistently gives Donald his best results in Florida and Ohio. In fact, Quinnipiac is the only major poll that shows Donald winning in Pennsylvania. You would be right to suspect something is off.

Indeed, something is off. Until recently, we have not been able to get a good handle on how Latino voters plan to vote in 2016 because sample sizes have been very small. Recently though, an important poll with a sample size of over 3,000 respondents gives us essential data. This poll shows Hillary with a resounding +67% lead — which obviously doesn’t square with Quinnipiac giving her a +24% lead among non-white voters in Florida.

Here’s how The Atlantic frames the issue:

Is it possible that Trump could break with recent precedent, and win a historically high share of Hispanic voters? Yes, under certain quantum-theory conditions, all sorts of things are possible: Time can move backwards, tornadoes can build neighborhoods from pieces of homes floating in the sky, and a cracked egg can leap into its original unbroken shell. But unless you think that “the Hispanics” love Trump as much as he swears he loves them, you might want to question the poll.

We are left with this conclusion: Quinnipiac is inaccurate, undersamples minorities, and underestimates Hillary’s support among them.

The corporate media, desperate for a horse race, won’t tell you that. So we will.