Donald Trump speaks in Youngstown, Ohio, Monday, Aug. 15, 2016. (AP Photo/Gerald Herbert)

I come here not to praise polling rumors, but to bury them.

In an interview on British television this week, Donald Trump's campaign manager Kellyanne Conway said that Trump "performs consistently better in online polling where a human being is not talking to another human being about what he or she may do in the elections." That's why Trump is trailing in the polls, she argues: His voters are wary of telling pollsters that they support the guy. I already debunked this, but sometimes you have to put a very fine point on such things. So, let's.

1. "Online polls" here does not refer to stupid Twitter surveys or anyone-can-click-a-button polls at media websites.

Polling is a science that uses statistical analysis to gauge sentiment across a broad population by taking a small, representative sample from that population and asking specific questions. The analogy I like to use is having tests done at the doctor: She doesn't need to remove all of the blood from your body to figure out why you're sick. She just needs a small vial of it.

Conway is, by trade, a pollster. So when she says "online polls," she doesn't mean something like the poll below, which I put on Twitter.

My poll is obviously ridiculous, but no more ridiculous than any other poll that allows anyone from anywhere to vote as much as they want or to get all their friends to vote, too. As I write, Elvis Presley's ghost is outperforming Trump by a wide margin, but I don't think that's what an actual election would show. People could be voting in my poll from literally anywhere, for any reason. I'm supposed to think the results mean something? These online "polls" are like if you went to the doctor to figure out why you were sick and he used two pints of your blood, a gallon of his own, 10 drops from Vladimir Putin and also two Coronas.

Conway means scientifically weighted polls that are conducted using an online mechanism under rigorous circumstances, as opposed to traditional phone polling, in which a pollster calls various voters and asks them a battery of questions. It's those online polls — like those from YouGov, SurveyMonkey or Ipsos — that Conway suggests show Trump doing better, the theory being that having a real person ask if you like Trump is more unnerving than pressing a button indicating your support on an iPad. This is called the "Bradley effect," after Tom Bradley, who ran for governor in California in 1982. Polls showed him winning, but he lost, a change blamed on his being black. The idea was that people told pollsters they supported him so they wouldn't appear to be bigoted, but then didn't actually vote for him.

There's just one problem ...

2. Donald Trump is not performing better in online polls.

I pulled data on online polls (the real kind) and live-caller polls since the convention and going back several months. In the former case, the numbers showed that Trump was doing better on average in live-caller polls than online ones both during and after the conventions.


Over the longer time period, the trend was similar: Trump did better in live-caller polling.


This uses data from Huffington Post Pollster, so feel free to run the numbers yourself. In fact, here's what they show as I write: Trump's average in live-caller polling is 40.8 percent. In online polling? 40.6 percent.


Are there some online polls where Trump does better? Sure. But on average, he doesn't.

3. Since when are Trump fans bashful?

I'm not going to point out that Trump's supporters have at no point appeared to be timid wallflowers, because that doesn't need to be pointed out. But he led consistently in primary polling for months and ran close with Hillary Clinton for a long time. If the idea is that people are worried about expressing public support for Trump because he said contentious things, he said those things on literally the first day of his campaign. If you're worried about people who don't like Trump viewing you negatively for backing him, why is that something that just emerged?

Trump's electoral problem has always been that he brought a base of support into the general election and hasn't been able to expand it outward. The recent shift in polls after the conventions does reflect softer support from Republican voters, and it is the case that even big chunks of his own party see Trump as biased against women and minorities.

But is it more likely that hundreds of individuals independently felt nervous about telling a pollster in an anonymous survey that they were supporting Trump? Or is it more likely that Trump actually saw his support dip after a Democratic convention that hit Trump hard and a Republican convention that Gallup found turned off more voters than attracted them? Especially given the lack of difference between online and live polling, it's safe to assume the latter.

Conway's job, at its heart, isn't to be truthful. It's to win the campaign. And to win the campaign, she needs people to believe it's winnable. And that means explaining away Trump's bad poll numbers.

But we don't have to believe it.