Before the first presidential debate was over, my colleague Aaron Blake noticed an unusual ad on The Washington Post's website. "Trump won," it declared in large type. "Contribute."
The timing was awkward, but this wasn't necessarily an abnormal response from a campaign: Our guy did great, ignore the haters. Trump didn't win; he was out-debated on nearly every conceivable metric. But campaigns aren't usually in the business of saying "We did badly."
Far more awkward is Trump's subsequent insistence that he actually did win, because of polls. It's awkward not only because it's nonsense, but because it's not clear whether Trump actually believes it.
There was one poll conducted in the wake of the debate that holds some statistical validity. Accurate polling depends on getting a sample of respondents that is representative of the population whose opinion you'd like to gauge. CNN and its polling partner ORC conducted a poll after the debate that found that Hillary Clinton won easily, by a more than 2-to-1 margin. The sample leaned slightly Democratic, which CNN noted, but it was generally a good snapshot of the views of the American public. FiveThirtyEight notes that CNN's survey has historically correlated to shifts in the polls.
There were also a lot of garbage polls conducted after the debate. There was the poll at the Drudge Report, a survey that you can take right now, if you wish. According to that one, Trump was viewed as the victor by 82 percent of those who replied, with about 570,000-plus having weighed in. A who-do-you-think-won poll at Time gives it to Trump with 54 percent of the vote; the same sort of thing at CNBC mirrors the CNN poll in reverse, 2-to-1 for Trump.
Some media outlets conducted polls on Twitter. Here's USA Today's. Go ahead! Vote! As of writing, Clinton's winning that one, but who knows!
— USA TODAY (@USATODAY) September 27, 2016
These online polls are, again, garbage, no more representative of the population as a whole than is the crowd at a Trump rally. That comparison is very apt, in fact. The crowd at a Trump rally 1) is open to all comers, 2) is geographically isolated, meaning that while anyone can attend, it doesn't include a huge swath of people who vote, and 3) it rewards enthusiasm in a way that tends to obscure actual interest. In other words, if 20,000 people in a state press into a Trump rally to cheer lustily for his stump speech, that's still only a tiny fraction of the population of even our smallest states.
The online polls are the same way. Open to anyone, meaning that anybody with an Internet connection can go and cast a vote. Anyone in Russia, for example, or in Canada. Anyone who is 12 years old or who is not a citizen. Literally anyone can weigh in at any time. And can do so more than once: Vote once from your phone and once at your desktop. No reason not to.
What's more, the audience at the Drudge Report is geographically centralized on the Internet. Lots of people read Drudge, but you might be skeptical of the results of a poll there for the same reason that you might be skeptical of the results of a poll conducted entirely within Trump Tower. Contrast with CNN's poll: geographically distributed and including a cross-section of the electorate, all of whom are registered voters.
Amusingly, a Twitter poll hosted by conspiracy theorist extraordinaire Alex Jones (whose links appear frequently on Drudge) shows Clinton with a lead. Why? Because Twitter isn't the same geography as his site, InfoWars, or as Drudge. And also because Twitter is built to share, and at least one high-profile Twitter user is intentionally gaming the system.
It’s very important to me that you vote…for Hillary on this Alex Jones twitter poll. It’ll be very funny if she wins https://t.co/SGpxIaAD9e
— joe mande (@JoeMande) September 27, 2016
A poll you can organize people to participate in is not a poll that is worth citing as fact. Enthusiastic Trump backers will, and no doubt are, pressing people to go vote in these polls to show he won, in the same way they might fill a car with friends to fill a convention hall at a Trump rally. But casting your vote more loudly doesn't make it worth more, and while enthusiasm may flood an online poll or fill a camera's viewscreen, it doesn't tell us much about who will actually win. (A fact to which Bernie Sanders and Ron Paul can attest.)
Just to put a fine point on it: Polls like the one at Drudge or Alex Jones's or USA Today's are entertainment, no more a poll that gauges reality than grabbing a lariat for a sepia-toned photo at the fair makes you an 1890s cowboy.
But when it's all you've got, it's all you've got. Those sympathetic to Trump, including supporters, conservative media outlets and Trump himself, are already citing these nonsense polls as evidence of his victory. Trump is even making up polls that didn't exist.
— Major Garrett (@MajorCBS) September 27, 2016
"I won every single poll other than CNN," Trump said on "Fox and Friends" on Tuesday morning, "and not many people are watching CNN." (Trump does; he referred to its coverage during the debate Monday night.) He continued, excoriating CNN's poll for being "very Democratic," skipping over the fact that every other poll he cited includes a wildly non-scientific and non-representative pool of respondents that also certainly far more heavily overestimates Trump's support. It's like saying that national poll numbers don't matter because he's got big crowds — which, of course, is exactly what he was saying a month ago.
The #1 trend on Twitter right now is #TrumpWon - thank you!
— Donald J. Trump (@realDonaldTrump) September 27, 2016
Does he really believe this? Could he? I ran an online poll once in which Elvis's ghost performed well against Trump in his (its?) bid for the presidency. I ran one once as part of a story about how these polls are garbage in which only three-quarters of respondents knew what 1-plus-1 equaled. Is it possible that Trump actually thinks that polls as scientific as that, polls thrown on websites to lure people to look at ads ... actually represent reality?
I guess I should note that the ads that run on websites are also not always accurate, either, including on this site. One I saw Monday night said that Trump won the debate.
Update: I'm getting a few responses on Twitter that run along the lines of this.
— JJ (@American_J2) September 27, 2016
The answer is: Yes, any poll that used a non-scientific methodology and allowed anyone to weigh in to generate its results is nonsense that should be ignored.
As I put it to someone else, throwing more garbage onto a garbage pile doesn't make it a mansion, it makes it a garbage dump.