Pew has a set of new survey results out on the 2016 campaign. It’s not really all that interesting if you are interested in aggregate numbers. The overall numbers, once you look under the hood, are pretty much the same across all surveys: Trump is not very popular overall, and he is especially unpopular among what normally would be fairly reliable Republican voters–upper income, educated white men and women. Among women especially, Trump is lethally unpopular. The difference among the surveys is largely due to how “popular” Hillary Clinton is among these disaffected Republicans: if Clinton is expected to win them over in significant numbers, especially the women (and Pew expects Clinton to actually carry the upper income bracket by a significant margin), Clinton wins by double digits (or close). If they remain disaffected and vote for third party or stay home (or just not vote for a presidential candidate), Clinton is nearly as unpopular as Trump and the election becomes close again. In other words, we are talking the same thing.
The really interesting question that sums up the state of the campaign at the moment is the very last question, Q. 49. Upon asked whether Clinton represents a better future, a worse future, or more of the same, a majority (53%) of the respondents said “more of the same.” When asked the same about Trump, only 21% did. All the affect aside, this is the real difference that Clinton and Trump will be fighting over: is the status quo good for you? Clinton is the low variance candidate: she will give you more of the same with at most incremental changes. Trump is the high variance candidate: he will bring big changes, or so people believe, but they could be very bad or very good. This is where things can and may get interesting: Trump, figuratively speaking, is definitely the shorter population. But depending on the luck of the draw, the random draw from the “Trump” population has a significant chance of being taller than a random draw from the “Clinton” population–even though, on average, it won’t be.
The question about people’s attitudes towards the status quo does not have an obvious answer: for all the campaign rhetoric aside, the status quo is not terrible. For many, it is not good, but there are plenty for whom this is acceptable and for a fairly sizable minority, it is pretty good. In a sense, things are pretty good, ironically, for traditionally Republican voters (read youngish, educated, and affluent white men and women) than for traditionally Democratic ones (especially the minorities), but, under the present circumstances, minority voters voting for a Republican is nonexistent and Trump has not exactly helped his chances–not that any Republican could have realistically gotten any more than a marginal share of additional minority votes anyways. (Yes, Rubio could probably have carried a few more Hispanic voters, but most Hispanic voters reside in non-competitive states anyways and the kind of Hispanics Rubio could have captured tend to be residents of Republican states like Texas anyways.) So Trump is down to only one voting block that he could count on–the working class whites, at least the subset who are inclined to support Republicans but are not reliable in their turnout. (Which, again, has been showing up consistently in poll after poll–Trump does somewhat better with them–but not much–than other Republicans of late and this support is much more consistent over time, unlike that of educated and affluent Republicans.) I would imagine that much of the 33% who believe Trump represents the better future comes from this bloc of voters. The significant probability that Trump could outdo Clinton comes from the possibility that, motivated by their belief that Trump is the path to the better future, they will be turning out in huge numbers, while those who believe that Trump is the worse future, for various reasons will not. This is not too unreasonable. Turnout among African-Americans will fizzle: Clinton is no Obama. That among the young will be likely relatively small: the status quo is not very good for the youth, after all. The Hispanic turnout is a great unknown, but, as much as many Hispanics dislike Trump, Hispanics are too diverse a population and “immigrants” are not as numerous as one might think–not to mention they have no special love for the Clintons either. On the other hand, a lot of people who don’t like the kind of change Trump might bring are among the core Republicans, and a sizable drop in their turnout will kill prospects of any Republican, Trump or no, no matter how high the turnout among the believers might be. This, in other words, represents the other side of the high variance.
As a party strategy, actually, Trump is still probably a better bet for the Republicans than almost any other candidate they could have chosen. The basic demographics run against the Republicans. A low variance candidate for the Republicans in a conventional election year, I think, can reliably get 47% or thereabouts but nothing more or less. For all the baggage that the Clintons carry, Clintons start and stay at the Democratic mean, more or less (Heck, the Democrats probably could have nominated the skull of Stephen Douglass and still get the same number). The Republican strategy, effectively, rests on winning the gamble that their candidate could raise turnout heavily among the missing white voters AND winning the other gamble that their candidate would not lose many (or, indeed, any) of the usual Republicans. Both are risky (thus the high variance), but, even with a far lower mean, Trump has a better chance of winning that Romney ever did. When I was still teaching stats, I used to throw a question about comparing heights (or the probability thereof) between two possible pairs of populations. There’s a population with some known mean and variance in heights (that are normally distributed), and two other populations that are shorter, on average. One is only marginally shorter, but has small variance (but still normally distributed). The other is much shorter on average, but with far larger variance (again, still normally distributed). In a comparison of randomly picked individuals, the taller population will always be taller, on average. BUT the question was always set up so that the high variance even if shorter on average population was going to wind up with a higher probability that a random individual drawn from this group would be taller than the first (the taller population) than a random person from the first of the shorter group. Fairly straightforward explication of how the average is not necessarily the representative and can mislead, in fact…but this bothered a lot of the students. The intraparty disruption that Trump brings is, for the Republcans, actually the smarter strategy: they can stay the same forever and lose forever, or they can risk a change, and try to win. Of course, if they were cleverer and forward looking in the past decade or two, they wouldn’t be in this mess to begin with.