A recent new NBC News "Survey Monkey" online poll shows Hillary Clinton with a six percentage point lead over Donald Trump.
But polling experts says when it comes to polls, the devil is often in the unreported details. Not mentioned in this article is that substantially more people who identify as Democrats took part in the NBC poll. According to the poll, 26,925 likely voters were sampled. Digging into the methodology and details, 29% (7,808) consider themselves Republicans and 36% (9,693) consider themselves Democrats. The difference of 1,882 means about 24% more Democrats were interviewed than Republicans. Suddenly a 6 percentage point lead doesn't look as good for Clinton.
An LA Times poll strives to take a more politically-even approach: 25% of its sample is made up of voters who say they voted for Republican Mitt Romney in 2012, and 27% of the sample is voters who cast a ballot for Democrat President Obama. That poll has Trump leading Clinton by several points nationally. But here, too, there are many question marks. What is the political breakdown of the other 48%? Is the idea that Democrats voted for Obama and Republicans for Romney accurate, or did some Republicans cross over for Obama--but will not do so for Clinton?
Which poll is correct (if either one is)? NBC is most likely to be accurate if about 24% more Democrats vote in the upcoming election. The LA Times could be closer to accurate if the turnout breakdown is similar to that of 2012.
Polling is a difficult business, one in which accuracy can only be determined in retrospect. It's amazing that, considering all the variables, they often prove to be pretty on target. If this topic interests, you, please watch Full Measure, this Sunday when we examine the Weird Science of polling.