So I've gotten a fair amount of pushback that Donald Trump's job approval (which is probably about to hit an alltime high in our average) is due to differential response. Maybe? 1/
So differential response is when poll toplines change, but underlying attitudes haven't changed. Instead, people willing to respond to surveys change. So in a bad news cycle for Trump, Rs shy away from polls, in a good cycle they rush to the polls. 2/
This clearly happens. A lot of what we see in convention bounces, or short term reactions to news, probably reflects this. Clinton's Access Hollywood bounce here was probably to some degree ephemeral. 3/ 
But there's a couple of problems. At a certain point this becomes too nihilistic. We know attitudes *do* change, so chalking things up to differential response is tricky. We can try to control for partisanship, but soft partisans switch between Independent and R/D labels. 4/
What makes me suspicious of the explanation here is that this doesn't look like the sort of bounces we saw in the last chart. It's been a gradual improvement in his job approval over the past few months.
Likewise, the decline in his job approval from 2/17-8/17 is probably "real," as was his improvement from 12/17-3/18. The other thing is that in most cases, the DA explanation is pretty straightforward -- after a convention ppl are tuned in, and really fired up about their party.
Trying to sell a DA response during impeachment in January seems just sort of post hoc. Maybe we'd have predicted this change in the polls in response to the news, but I doubt it. Anyway, we'll find out more in a month. And of course, Trump is perfectly capable of tanking his JA.
