A few weeks ago, I wrote an article suggesting that Donald Trump's poll numbers had slipped. The next day, the candidate called me.

"I'm not complaining," Trump said. "But we have actually been going up. I know a lot of the online polls have been phenomenal, too, but I know that you don't cover that stuff as much."

Both parts of that are true. Some of the online polls have been phenomenal, and we don't cover them too much.

We don't cover them much because The Post relies on polls "based on the principles of inferential statistics," in the wonky words of our pollsters. The worry with online polls is that they "typically base their samples of respondents on groups of people who join a panel to take repeated surveys." If you take out the randomness of the sample, our pollsters say, the results can't be overlaid on the overall population "using traditional methods and without strong assumptions about participation in such surveys."

AD
AD

But there's something interesting about Trump's broader point. And it's that Trump actually does much better in online polls than in live-caller ones.

Polling guru Charles Franklin noticed this and tweeted about it.

Using data from Huffington Post Pollster, we can recreate the difference for Trump among the two types of polls. The lines here are polynomial regressions of the data -- averages of the polls, if you will.

The gap between live-caller and Internet polls is obvious here. It helps explain why we regularly get feedback contrasting the shape of the Real Clear Politics polling average -- which excludes most online polls -- and the curve displayed by Pollster. The latter shows an uptick for Trump where the former does not, which mirrors the uptick you can see in the online polls above.

Interestingly, the patterns for other candidates are different. There's less of a gap between the two survey methods for Ben Carson and Jeb Bush -- and in the case of Carson, he does better with live calls.

So what's happening here? One theory is that it's a function of there being an interviewer or not. (Live phone polls obviously include a person on the phone.) The implication being that there is a percentage of the population that doesn't want to talk about its support for Trump with another person. But, then, what does that say about Carson?

AD
AD

Donald Trump's goal is not to carefully analyze his position with the electorate and then share those findings with the world, of course. His goal is to reinforce that he's holding on to a big lead in the polls, and the gap between live-caller and Internet polling means that he naturally emphasizes the latter.

Whether or not those higher numbers accurately reflect the Republican electorate is something that will be determined only once people start going to the polls.

AD
AD