[One America News Network Telephone Poll] National Polling Results(Trump 50%, Clinton 50%D+7 Sample,

[One America News Network Telephone Poll via Gravis Marketing ^ | October 20th through the 23rd

[stockpirate]

Trump is leading that’s why they sample more dems, he is up by abt 5-7 points.

[Georgia Girl 2]

Clinton will likely get a D+2 turnout at best. So this poll is meaningless.

[xzins]

Otherwise, you could pull all of the sample from California.

[Rockingham]

The demographics of a poll matter. The general population and registered voters have different demographics, as do likely voters and those who will actually vote in the election. In addition, response rates vary by demographic.

Pollsters claim to have formulas and adjustments that compensate for such factors, but exceptional candidates, issues, and campaigns tend to escape such calculations when one side heavily declines to participate in surveys or generates a wave of new voters. When that happens, the polls get it wrong. Reagan in 1980 and the British Brexit vote are examples, and Trump in 2016 may prove to be another.

Notably, in Presidential contests, at the end of the campaign, polls usually close to within the margin of error. The pollsters can then claim to have been close enough in the end. In the meanwhile, many feel free to indulge their biases and preferences -- or even to put themselves up for sale. Thus many polls are more propaganda than an honest expression of the pollsters' art.

Internal polling -- meaning expensive, high quality polling paid for by the campaign and kept private -- tends to be far more reliable. Notably, even then, such polls are used mostly to allocate campaign effort and gauge relative trends. Every now and then, a campaign losing in the polls ends up winning unexpectedly. When pressed for an explanation, pollsters offer the excuse that they measure voter sentiment only when the poll is taken and do not claim to have a crystal ball as to election results.