Much has been made of the apparent dip in Donald Trump’s support in recent few weeks. While the aggregate change in support levels is itself a fact that should not be discounted, the interesting things are taking place in the way different demographic subsets are changing (or not changing).
Let’s compare two polls by the same organization, CBS News, if only to pick a set of polls that are about 4 weeks apart. In mid May, the news was how Trump was seemingly closing the gap and the Republican voters were closing ranks behind him, as indicated by this poll. Almost a month later, Trump’s support seems to be cratering, as shown by this poll.
First, notice what is and what is not changing. Trump’s support is indeed falling, but so is Clinton’s. What demographic is accounting for Trump’s falling support? One fairly obvious source of his flagging support is the large dip in support for him among the Republicans (from 85% to 73%) Among the Democrats, the picture has not changed as dramatically, as the support for Clinton has fallen somewhat (88% to 81%), not so insignificant, but not so dramatic. Among the independents, in contrast, hardly any change–except for a marginal, probably meaningless (which I cannot assess due to lack of raw data).
Also notable is the drop in Trump’s support among women, especially white women. Peculiarly, one might have thought, Trump actually was effectively running even among white women in May. This should not have come as surprise, though: Republicans outnumber Democrats among white women, especially among the educated and upwardly mobile. This should worry Trump: this is core of the Democrats’ so-called Rendell strategy at work: Clinton campaign is (or was, at least) confident that it can peel off affluent women who find Trump’s tasteless antics offputting. However, while they may, for now, have been lost to Trump, they have not added to Clinton’s strength–her numbers remain essentially unchanged among this demographic.
This is a bit of chicken-and-the-egg problem for the Republicans: Trump does reasonably well with the Independents, who apparently find him no less distasteful than Hillary Clinton. His problem is that he offends Republicans: the educated and the affluent, who make up the core of the Republican coalition (in terms of the numbers combined with the turnout rates), especially the women (who do, after all, turn out more than men, especially among this demographic.) It is tempting to suggest that, since partisanship dominates the behavior of this group more than any other, the Republicans who have issues with Trump will still show up, hold their nose, and vote for Trump (and whoever is down the ballot) In practice, they might not…and the question then becomes, how many will stay home?
Years ago, I had proposed to NSF for a project to “stress-test” partisanship of the voters, to identify what it takes to get partisan voters to drop partisanship from their vote choice. I didn’t get that grant: the proposed experiment was way too clunky for its own good and it’s not clear what we would have gained, not to mention that similar studies had been done (if only to show that voters choose whom they would otherwise have rejected if they carry the right party label–this contrasts with another test that a coauthor and I had run, where, when faced with very obvious “wrong” positions (abortion, taxes, etc–framed in a very “obvious” fashion), voters flipped their choices promptly, regardless of the party label. We could only find very minimal–you’d miss, unless you expected find it, and yes, very statistically insignificant–effect from party affiliation.) Ironically, real life seems to have thrown us this stress test after all….