Trump is trailing Hillary Clinton in Texas, in a two-way poll conducted by the Washington Post over the past month.
This result is indispensible to those following the 2016 election — as a reminder of how easy it is to misuse data.
Seven percent of Texans want a border wall to protect them from Oklahoma, one poll told us. “Nearly 20 percent of Trump’s supporters say freeing the slaves was a bad idea,” Yahoo News reported. A dead gorilla from Ohio would finish ahead of Green Party candidate Jill Stein in a five-way race. And Scott Walker, during his presidential race, was 2.1 times as untruthful as Florentine doll-turned-boy Pinocchio. Or something.
The political media have hurled this “data” at readers trying to follow the 2016 elections. All of this data is useless at best and dishonest and misleading at worst. There are boatloads of useful and enlightening data available this election, more than in any previous presidential cycle. But data is very easy to abuse and misuse. Our pollsters and our media have spent plenty of time and column inches this year engaged in this sort of inhumane torture.
Among those making readers dumber are the pollsters, first among them the Democratic firm Public Policy Polling. PPP this week retweeted a comment mocking Green Party nominee Jill Stein as “polling worse than Harambe.” PPP you see, has repeatedly asked voters if they would vote for the Texas-born Western lowland gorilla who was killed in May at the Cincinnati zoo. The pollster’s occasion for mocking Stein: A Stein tweet arguing that she should be in the national debates this fall, and in which Stein wrote, “I do not take polls seriously.”
While it’s foolhardy for a politician to ignore polls, PPP seemed to validate Stein’s position — and not only by polling a dead gorilla.
“30% of Republican primary voters nationally say they support bombing Agrabah,” PPP declared on Twitter in December. “Agrabah is the country from ‘Aladdin.'”
Media outlets gobbled this up as if it told us something. The problems here were many. First, the media accounts almost all omitted that 19 percent of Democrats polled held the same nonsense view.
PPP trumpeted in a press release that 42 percent of Trump supporters think Islam should be illegal, which newspapers picked up. Unremarked on: 59 percent of “very liberal” voters said Islam should be illegal, as did 55 percent of “somewhat liberal” voters — far greater portions than among conservatives.
These and other dumb polling questions, combined with partisan press releases, then fueled the left-leaning media to create a self-satisfying but totally uninformative narrative. It was a microcosm of the abuse that data has endured this election.
Trump supporters pining for black slavery was a big media theme in February. “Nearly 20 percent of Trump’s supporters say freeing the slaves was a bad idea,” Yahoo News wrote. “”Nearly 20 percent of Trump’s supporters disapprove of Lincoln freeing the slaves,” Vox wrote.
This was more bad polling made worse by bad reporting. It turns out that the question didn’t mention the Emancipation Proclamation, Abraham Lincoln, the Civil War or the 1860s. Pollsters embedded the question in a series of questions about executive orders and executive overreach. So the poll primed respondents to be skeptical of executive orders, then omitted any historical context or significance for the executive order pertaining to slavery.
Making matters worse was the media’s selective, partisan, and misleading reporting of the question. Every article laughing or scowling at Trump supporters omitted that 5 percent of black respondents polled in this way expressed the same view, and that Bernie Sanders supporters were twice as likely as Marco Rubio supporters to call the end of slavery a mistake.
A serious look at these results and other polling data suggested something fairly banal: it was a trick question, and less educated white respondents were more likely to be confused by it than were more educated white respondents.
But the silliest and most persistent abuse of data in this election may be the conceit by media fact-checkers that their fact checks can be compiled into something like a batting average for politicians.
Philip Bump of the Washington Post last September tried to use the Washington Post’s checks (“more than 100 ratings of statements from the 2016 candidates”) to present us with “The least- and most-factual 2016 candidates, charted.”
Bump’s dump of data revealed Rand Paul averaging 3.2 Pinnochios on Kessler’s scale to Hillary Clinton’s 2.4 Pinnochios to Scott Walker’s 2.1.
The problems with this analysis should be obvious. Set aside the tiny sample size and the utter subjectivity of the Pinnochio scale (the Post’s Glenn Kessler, by the way, is perhaps the best of the professional fact-checkers). The fatal problem here is the selection process. No fact checker checks 100 percent of a candidate’s statements. They pick the ones they doubt, or that they kind of think might be worth checking. This checking may or may not be valid or enlightening, but compiling them into percentages is idiotic.
Yet Politifact has spent all of the 2016 cycle tweeting out the percentages of the candidates. How many “Pants on Fires” did Hillary have? What percentage of Trump’s statements are “True” as opposed to “Mostly True”? The head of Politifact even wrote an op-ed in the New York Times about it.
It’s easy to find falsehoods by both candidates that the fact-checkers missed. It’s even easier — and the absurdity of this statement demonstrates the absurdity of the batting averages — to find true statements the checkers never checked.
Early in Trump’s recent immigration speech, for instance, he said “I’ve just landed having returned from a very important and special meeting with the President of Mexico.” Nobody wrote up a formal check of that fact. Nor did anyone check Clinton, when she said in Detroit, “We all face choices in life.”
You get the point. There’s no consensus on which claims are checkable or deserve checking. So these averages are 100 percent meaningless. But if you’re really looking for some numbers you can mash together and turn into a wonky blog post, then why not scrape some fact checks together?
The Washington Post’s recent poll of 74,000 Americans—the one that showed Hillary winning Texas — probably tells us something. It certainly doesn’t tell us, as McLatchy reported, that “Texas is a tossup.”
Data can be a useful tool for readers who want to inform themselves. It can also, this election reminds us, be useful for writers who are happy to misinform them.
Timothy P. Carney, the Washington Examiner’s senior political columnist, can be contacted at [email protected]. His column appears Tuesday and Thursday nights on washingtonexaminer.com.
