The striking number on the Democratic side is not that 90% of the erstwhile Sanders supporters say that they prefer Clinton over Trump, but the fact that almost 10% of them say that they actually support Trump over Clinton. One should not expect numbers like that, in general, unless there are serious problems between the general election candidate and particular segment of the electorate. This, of course, complements a very peculiar description of the Sanders supporters: his support apparently came from “Democratic leaners, those self-identified as “very liberal” and the religiously unaffiliated,” i.e. a very peculiar set of voters–especially the shared presence of “very liberal” and “leaners.” It begs the question where the 10% of the Sanders voters favoring Trump over Clinton are, and how many of these leaners there are in the panel.
The proportion of the leaners is critical: internet-based panels systematically overrepresent potential voters who like and pay attention to politics. These voters tend to be partisan and ideological. The share of independent and/or otherwise apolitical voters in the electorate tend to be larger than in the samples. This is significant because the independents played a crucial role in making the Sanders movement possible in the past primary season. If the 10% of the Sanders voters voters are favoring Trump over Clinton, and if they are overwhelmingly among the leaners, especially in the age 30-49 cohort, and if the leaners among the Sanders supporters in the Pew sample are far fewer than what we know to have been the composition of the Sanders coalition from exit polls and such, we’d be talking about something like 15% defection rate. My guesstimate of the share of independents among the Sanders supporters was approximately 1/3 to 40%, based on exit polls. 15% would represent almost half of this total, which I thought would be enough to bring Trump a narrow victory, IF the bleeding among the regular Republicans can be stopped (which is a giant unknown, for the reasons explored in the other half of the Pew survey). This should scare the Democrats shitless. Repeat after me: that the average Sanders voter is certain to vote for Clinton is not an assurance that ALL Sanders voters will stay with Clinton. Lose a significant chunk of Sanders voters to Trump (and 15% is a big chunk) Trump runs a good chance at a narrow win.
Now, I would have liked to see the equivalent numbers on the Republican side: how many of the Trump skeptics would actually consider voting for Clinton? Democrats have been campaigning quite hard to capture at least some of a significant component of the Republican electoral coalition: educated, affluent white women. There have been many polls suggesting that they are the least favorable towards Trump, by big margins. While Pew’s tables suggest that this demographic segment is indeed where many of the Trump skeptics come from, it would have been nice to have at least some cross-tabs. How many of them would actually prefer Clinton over Trump? Have they even asked the question? This is critical: even if Trump can peel off a significant chunk of apartisan/apolitical Sanders voters–which the Democratic side of the article suggests is a possibility, losing a significant chunk of the usual Republican voters will bite into his totals badly. The large number of skeptics suggest that many voters will be lost to the “none of the above” or to Gary Johnson. But how many will actually take the ultimate step and go Clinton?
Neither is especially revelatory: they confirm what has been known about these candidates already–both are very, very weak and have significant problems with very important potential pieces of their electorates, and the Sanders coalition is much more strategically important because of its variable nature than people seem to think.