The poll numbers for Clinton and Trump are not quite as variable as the poll numbers **not **for Clinton and Trump.
Right now, Johnson and Green are averaging about 9% in the polls. They will not get 9% of the actual votes. People want to support third parties in theory, which is what polls are, but when they actually get to the voting booth they vote for a major party. Additionally, about 8% are undecided. (RealClearPolitics four-way average as as now, noon on 10/12.) In October 2012, Obama and Romney were polling at about 92%. They actually gathered about 98%. The missing voters broke around 3 to 1 for Obama.
Where will those 17% go this year? Johnson and Green will get half of their current numbers, say 4% and 1% to be generous. Of the remaining 12%, another 3 to 1 split makes it 9% for Clinton and 3% for Trump. That adds up to 53% Clinton, 42% Trump. Is a 3 to 1 split too high? Possible, since that was for a sitting President. A more even split makes it 50% Clinton, 45% Trump. I think that’s a bit low since that merely leaves the spread where it is today, and it will clearly be higher.
I think the polling average will be accurate. Or at least, that it’s not significantly more likely to be inaccurate in one direction than the other.
It’s based on the idea that we can guess how people will vote by… asking them how they’re going to vote. And then aggregating the results and correcting for known inaccuracies in the sample. That makes sense to me.
I went with the +/-1 answer, but I’m not really happy with that. I don’t think the polls will be quite that close, but I also don’t think they’ll systematically miss for either candidate in a predictable way.
The differences we’ll see will be because:
polls are always at least a few days behind reality. Plenty of unpredictable things could happen in a couple of days that might move the results a few points, from scandals to weather conditions… but there’s no predicting which candidate will be more affected.
polls already have margins of error higher than +/-1. I know that sites like 538 try to aggregate polls to further reduce the accuracy, but still…
I am assuming that the margins listed refer to the difference between the two candidates, in which case, I am going with the “Clinton home run.” I have seen elections in both Canada and the UK seriously off the mark against the polls (in the UK, it was in double figures).
I see two things contributing to this; one, that Clinton will reach 270 relatively early, and Republicans will be more likely than Democrats to stay home in the west; two, a considerable number of college (and 18-year-old high school) students will decide at the last minute, “If I can get the Democrats elected in Congress, the promised ‘free tuition’ will be much easier to achieve - and who knows; maybe Bernie will have enough say to push through a reduction on student loan interest,” resulting in a last-second swing to Clinton, even if a lot of them either vote Johnson or write in Sanders.
I think that a lot of Republicans are going to go in, vote for their Senataor and Congressman, and fail to fill out the field for President. In particular, the women will fail to do it.
Republicans are only about 30% of the population and women something like 40% of that. If half of them decline to vote for their party’s nominee, that’s already a 5 point loss for Trump. Independents are going to be a lot less merciful.
I suspect that after the election, a lot of pollsters are going to be taken to task for oversampling older, Caucasian men and improperly weighting to correct for that. If you look at ethnic, gender, and party polls and aggregate them together along with demographics, you simply can’t reconcile them with the national polls that are being reported.
For example, I think Trump is polling something like 5% with African Americans and they’re 12% of the population. So how is he less than 10% away from Hillary? The math don’t work. I can see if you’re talking electoral votes or voter participation rates, but not national polls.
Oh some will answer merely out of how they want things to go, or how they fear things will go … but personally I have long been of the belief that the polls themselves have been a pretty stable story, staying pretty much Clinton +4 +/- 3 (and still are), and that the guess of who actually will turnout, i.e. the art of the LV screen, will be the big polling story of the cycle. FWIW I will be surprised if the aggregates on election eve fall outside that range, even as the Trumptanic keeps headbutting the iceberg.
The biggest way in which this cycle is actually quite different than recent past cycles is the demographic shifts that make up each side, with increased concentration of White non-college-educated, especially male, voters, on the Trump side, and more support for especially White college educated voters, especially women, and Hispanics, on the other. The difference in ground game infrastructure and investments is also significant.
Recent cycles do not give a good dataset to reliably predict how those will play out in terms of who actually comes out on election day, or minimally give conflicting signals. Those non-college educated males are often expressing a strong intent to vote while also often not having voted in the last election. Which should get more weight? Strongly stated intent or past behavior? Some who currently prefer Clinton are less enthusiastic about her and less 100% confident of intent to vote … but they have voted in the past and will be reached out to.
Without good past similar election datasets the pollsters are making best guesses and might even adjust their methods to not become the outlier. Thus there actually is a real potential for a systematic error that skews the result. IMHO.
If they get it very close to right I will be very impressed.
The ground game is the thing that I think will really be distorting. In previous elections, one party or another had the better ground game, but both parties had a ground game with serious people working on it. In this election, my impression is that Trump has almost entirely ignored the ground game. And I don’t know how polls can account for that.