Heading down the home stretch, it's time to make a prediction.
I'm basing it solely off of the USC poll, which was the RAND poll in 2012. It's the only panel poll, following the same sample of individuals over time, so that any changes reflect true changes of heart and not a shift in the make-up of the respondents depending on who feels like participating. (See this earlier post for details.)
It's also an internet poll, which the Brexit results revealed were more accurate than phone polls. They even provide tablets and internet access to people without them, to get as representative of a slice of America as possible. And the sample is part of an ongoing, longer-term study independent of the election, so it won't be affected by who initially felt like joining an election-themed study during an election year.
Under its RAND branding in 2012, it was the most accurate -- more so than the IBD poll, which performed worse in 2012 than it did in '08 and '04. (IBD is not a panel but ad hoc sample, and is phone-based with about 65% being cell phones.)
The day-of estimates were Obama 49.4, Romney 46.8, for a spread of 2.6 points. Actual outcome was Obama 51.1, Romney 47.2, for a spread of 3.9. The RAND poll was off by just 1.3 points in the margin of victory, only 0.4 for the lesser candidate, and 1.7 points for the greater candidate.
I interpret the smaller error for the loser to reflect that fact that whoever admits to voting for the lesser candidate is all there is, while the greater candidate will enjoy some "everybody likes a winner" support among those who are not strongly decided.
When we look at the full series, an even more accurate picture emerges (click the earlier links for the interactive graphic):
After the summer stage was over, Obama clearly pulled away from Romney. Throughout the various ups and downs, Obama's ceiling was around 51-52, the peak being 51.2 on 10/27. That is only 0.1 point different from what he actually got. Romney's peaks throughout the series never broke 47, and that is what he ended up getting.
This indicates that the peaks are what the candidate will get -- a long-term maximum support level -- and the fluctuations below them are just people feeling wishy-washy from day to day.
Now we turn to the 2016 USC poll, which is the same methodology as before only in the present context:
After the summer conventions were over, Trump was slightly ahead or tied around Labor Day, after which he pulled clearly ahead. Trump is visibly the greater candidate, Clinton the lesser candidate.
Despite becoming tied again during the final phase of "uncertain about Trump," he's pulled away again. It's clear that his line has remained mostly above her line, and at worst tied. Unlike Obama in 2012, Trump is not the incumbent President and does not enjoy as decisive of an advantage. Other than that, its his line that obviously resembles Obama's line from last time.
So Trump will win, but by what margin?
Using the same insight of looking at peaks along the way, Trump's ceiling of support is 47-48, with the all-time peak of 47.8 on 9/18. Since Labor Day, Clinton's ceilings have been at 44-45, with a max at 45.1 on 10/23.
I'd say that makes the final outcome look like Trump 48, Clinton 45. I don't think Trump will get quite the last-minute bandwagon effect that Obama did because undecideds this time have consistently been uneasy about Trump the outsider, vs. Obama the incumbent. Even supposing he gets 49, I don't think that will come at the expense of Clinton's 45, but from the third party voters.
Speaking of whom, there are 7% left after Trump and Clinton. Stein has been polling around 2-3%, and her voters are going to be stickier because they're voting for her platform rather than merely against others or as an empty protest vote. Unlike her run in 2012, this time there was the Bernie phenomenon, many of whom are turning to her.
Johnson abandoned his campaign over a month ago, and his supporters cannot explain what he stands for, so he is just a vote-parking space for people who are uneasy about both Trump and Clinton. He polls at around 7%, but will be lucky to get half of that -- although probably close to it, given the wimpy cuckiness in the electorate about Trump the change agent. The other half just want to wash their hands of the election already and will simply abstain.
Final prediction, then: Trump 48, Clinton 45, Johnson 4, Stein 2, All Others 1. If the cucks for Johnson man up, it will shift 1 point from him to Trump (pretty iffy at this point, though).
Hope I'm wrong and it's a wider victory, but that's what the best poll suggests.
There's no evidence of a landslide, though, so please get that idea out of your head, lest you feel let down that we only won by 4 points. Remember: everyone said we were supposed to go down in flames during the primaries! Trump winning at all is a massive accomplishment for him and us the voters -- don't cry about it just because we're not going to win by 10 points!
But they say that conservatives have a lower tolerance for ambiguity, so it's either landslide or implosion in most of his supporters' minds, and we shouldn't take their claim about predicting a "landslide" literally, but simply meaning "we're going to win rather than lose".
Just one more week -- we got this!