We've noted before that Donald Trump's poll numbers have often relied on the sort of poll that's being conducted. There are terrible "who won the debate??" polls that anyone can click on; Trump cleans up in those (as he likes to point out). There are the more scientific Internet-based polls, using set panels of respondents. Trump has consistently done better in those, too -- perhaps because people are more willing to click on his name than to say it. And then there are the gold-standard, live voice call polls, where Trump still leads, but generally by less dominant margins.

There's a tricky aspect to that latter set of polls, though: cellphones. With Americans increasingly using cellphones as their only phone line, pollsters can't rely solely on land lines for outreach. The Washington Post and ABC News, for example, include about 65 percent cell numbers in our joint polls, despite the increased cost. There are wide demographic differences between people who live in cell-phone-only households and those in homes with landlines or both. Pew Research highlighted that split last year.

Younger, poorer households are less likely to have landlines -- in part because they move more often.

A new survey conducted on behalf of the liberal group Democracy Corps finds that the method by which the survey was conducted makes a big difference. There was a 25-point spread in opinions on the presidential race (framed as Hillary Clinton vs. Donald Trump), and a 22-point spread in a generic congressional match-up.

The Washington Post/ABC News poll released this week found a similar split. Calls to landlines put President Obama's net approval at minus-16. In cellphone-only households, it was plus-1.

Remember: There are demographics at play here. Older, wealthier voters are more likely to have landlines and more likely to be Republican. They are also more likely to vote.

So Trump does better with landline households. But  again, Trump does better in online polling, too. If we take only the Trump-versus-Clinton match-up and compare Internet to phone-based polling, there's something of a gap -- though the results are pretty much all over the place.

(Reminder: You should usually ignore these polls. We're making an exception here to make a point.)

At the beginning of 2014, Pew found that younger people are more likely to be online than older ones; 97 percent of those under 30 were online, compared to 57 percent of those over 65. So Trump should do worse online?

But  then again, wealthier households were more likely to be online than poorer ones. So maybe he should do better?

But then again, Trump does better in general with less-wealthy voters. So ... who knows.

There's no indication at this point that any of the differences in methodologies have resulted in polling that's unreliable. Since, you know, no one is actually voting yet. But if you see a poll that's conducted online only or only with land lines and it shows Donald Trump rolling to victory, you should at the very least not be surprised.