It seems that the argument that I’d been making on this site is increasingly going mainstream. I suppose I am not the only one who’d been thinking along these lines and looking at the data.
Do note that the argument that I’d been making is subtly different, and anticipates the likely counterargument: the average Sanders supporter is indeed a highly liberal young person who is not very likely to jump to Trump. Clinton is quite right to expect that, come the election day, the average Sanders supporter will indeed hold his or her nose and vote Democratic. If Trump does win, it will not be because of the average “Bernie Bro.”
The real danger for the Democrats is that Sanders’ support base is more heterogeneous that people think it is: there is a sizable minority of oldish, liberalish (in the sense that they are willing to vote Democratic), “independent,” working class folks among his voters. These folks have been showing up as the real source of majority for Sanders in the primaries and caucuses that he won and they are least likely to stay in the Democratic camp come November. I would not expect 20% of the Sanders supporters to defect as per Zeits: I figure that’s the % that might be open to voting for Trump and Trump might win over half of them–but, as I noted before, that will be enough to swing the election if all the other voters stay put, more or less.
Thinking in terms of averages is natural to humans: that’s dangerous–that is exactly how we are lied to by statistics. The core statistics is that it quantifies how “wrong” our suppositions are–so even if the average Sanders voter might be so liberal that it is unthinkable that Trump will win over him or her, the variance is so large (i.e. the average has very high probability of mischaracterizing a random Sanders voter) that a surprisingly large proportion might wind up behaving unexpectedly. This, not making stupid predictions, is why you study statistics.