According to Nate Silver’s polls plus forecast, Clinton’s chances of winning the election are now just a little under 70%. In early August she was at around an 80% chance of victory. Who are the poll respondents that have caused this shift? I doubt that Trump has managed to manufacture some new white males without a college education out of thin air. Have some of the Sanders people lost their desire to support Clinton now that we’re a month out from the Democratic convention? Is it moderate Republicans that can’t make up their mind about who they dislike less? Is there another group of voters that I’m overlooking that is responsible for Clinton’s slow slide since early August? I’m interested in hearing opinions about this demographic. Is there anything Clinton should be doing to win them back to her side?
There’s almost always a convention bounce in the polls. And since the Democrats went second this year, it probably took longer for Clinton’s bounce to wear off, compared to Trump’s.
The fact that convention bounces happen doesn’t erase the question though; who are the people who get fired up after a convention and then gradually change their minds?
Polls are also using different samples and different voters. There’s no reason to expect unanimity.
They are the people you find perusing the menu at McDonalds.
Ha! This is great.
They are the people who hate both Trump and Clinton. Their answers depend on which one has pissed them off most recently.
The difference between 72% chance of winning and a nearly 90% chance of winning is the difference between a 8% lead and a 4.6% lead, so a net of maybe 1.7% changing from Trump to Clinton. I’m imagining its the people who have forgotten about Trump’s treatment of the Goldstar mother, and thinking that there was something scandalous about Clinton meeting with some important people who gave money to help aids in Africa. No need to panic yet. It isn’t that Trump is beginning to win, its that he’s not losing by quite as much.
It just so happens that an article just came out about this issue. This is a profile of voters in Muncie, Indiana.
Polls also move as a result of response rates rather than changes in voter preference.
So if you energize your core supporters, you can get a temporary bounce resulting from their eagerness to tell a pollster they support you instead of hanging up, even if no one has changed his mind. There is some political science research suggesting that this phenomenon accounts for the majority of short-term poll movement.
If you look at FiveThirtyEight’s Florida predictions (which they predict is most likely to tip the election), it appears Trump’s predicted vote share is going up faster than Hillary’s is going down. Which means that much of Trump’s gains are from previously undecided people deciding to support Trump.
In America, you move polls, in post-WW2 Prussia, Poles move you!
It doesn’t mean that, necessarily.
Imagine you have three buckets of marbles: 100 red, 100 blue, and 50 yellow. You pour them into a pot and stir. Then you pick 20. You draw 10 blue, 5 red, and 5 yellow. So you report that based on your sample, 50% of the sample is blue, 25% red, and 25% yellow. Then you do a second draw. This time, you draw 10 blue and 10 red. Your sample now shows that the pot is 50/50 red and blue. Does this mean that the yellow ones turned red? No.
Of course, ordinarily, you take a large enough sample and do it often enough that random chance is unlikely to explain the changes. Moreover, pollsters adjust their numbers to try to make the sample more representative.
But, one of the differences between drawing marbles and calling people is that people have that pesky free will. The marbles don’t become easier and more difficult to pick out of the pot as political winds shift. But humans do become easier or harder to get to finish a telephone poll. You have to spend a lot of time and money to get a representative sample unaffected by the differential response rates resulting from political events–something many public polls don’t really do–and even with those resources it is hard to control for.
Some clever people have set up ways to assess what percentage of short-term trends are a result of differential response rates, and the evidence is mixed. But it is certainly possible, even likely, that short-term movements in polling averages do not reflect people changing their minds.
That’s why at least one pollster keeps on calling the same people. I believe Nate Silver says that’s the LA Times poll, the only one showing a Trump lead. While the poll is almost certainly a big fat outlier, it is useful for tracking which demographics are swinging.
Ah yes, it is the LA TImes poll:
Can anyone find the internals, see who is changing their mind?
Reuters/Ipsos also does the tracking bit and weird, just as I am trying to figure out the internal of its apparent narrowing over several weeks they must’ve released this weeks. Now in the 4-way from about even last week to 11 up Clinton almost all from the don’t knows. About tied in the two-way.
No time to play with internals now but maybe later. Still very odd shift on the four-way. Too odd to believe.
If this election has shown anything, it’s that the media really, really loves a horse race and that knocking down the front-runner to make for a close finish is taken really seriously.
Hopefully both parties will learn from that in the future and not nominate candidates where there’s such a target rich environment. If this was Tim Kaine vs. Mike Pence the media would have to work harder.
As I mentioned a while back, Trump had already once before erased a double digit polls deficit – and he’s now done it yet again. I was wondering if the Reuters/Ipsos polls, one of the more volatile ones I’ve seen, was just an outlier, but the CNN poll pretty much confirms the voter shifts. As I also mentioned, Clinton’s (and Obama’s) decision to go on a fundraising tour and skip out on helping white flood victims in Baton Rouge was a painful tactical error. As I predicted, white voters in other parts of the country were watching. She can recover from it, I think, but she can’t make those kinds of mistakes again.
Trump could be two good debate performances away from the presidency.
The irony is that if Trump wins, he will probably kill off the republican party once and for all.
Obama’s Gallup numbers are the same 52% they were before.
But certainly Reuters is good for identifying trends and the trend is right now encouraging for Trump.
Clinton has just not done a great job working the media. Depending on Trump to keep shooting himself in the foot is not enough to rely on.