Discussion: Clinton Holds On To 3 Point Lead Over Trump In Final Bloomberg Poll

I just find this whole polling thing – particularly national polling – a little bizarre. For example, the polls – including the NY Times Upshot – continues to have Clinton ahead in Nevada by 2 points. (Nate Silver/538 predicts that Clinton will carry Nevada by 0.3%) But the news out of Nevada has huge Hispanic turnout, particularly in Clark County (Vegas), which has 60% of the Nevada voters. According to some reports, Dem early voters outnumbered Repub early voters by 57,000 voters, which alone probably is more than Trump can recover in the rest of the state, even if all the Repubs were going to vote for him.

Given this fact, of what consequence is a poll, particularly a live poll which notoriously undercounts Hispanics?

Polling has been in something of a state of progressively expanding crisis since about 2004. Everyone used to think the problem was people dropping landlines, but the problems are much greater than that. Increasingly, it seems like traditional polling methodology depended on a set of work and lifestyle expectations to be functional that no longer apply. Besides the sporadic nature of modern working hours, the fact that past voting behavior and expressed voting intent are no longer predictive (though pollsters seem stubbornly unwilling to acknowledge that), I also think telemarketing, and the defense mechanisms people developed in terms of being willing to just hang up, plus caller ID, have really screwed them up and they don’t know what to do.

Meanwhile, political campaigns are now relying on a “post polling” model of data collection that are far more accurate because they’re basically working with “samples” so large that they might as well be the whole population.

1 Like

Sure. There was a fascinating back and forth between HuffPost’s poll leader and Nate Silver. HuffPost was challenging Silver’s “adjustments” to poll results based on his analyses of the quality of the polls (judging by past results). Nate was more than a little annoyed. The bottom line, though, is that once early voting starts in states where people register by party, and I’m thinking of Nevada, Florida, and North Carolina specifically, of what value is a poll when there’s already a ton of data on who has voted before election day?

The problem in North Carolina, at least, is that there was a huge upsurge in early voting by non-aligned voters and no one seems to know what that means.