I think that he’s saying basic common sense here, including that nothing so far is a reliable predictor of where Trump’s approval rating will be in the future.
More interesting to me was the FIRST-TERM APPROVAL RATING table at the bottom of the article. I read that as an indicator of how much more polarized the country has become since the 1950s and how much less automatic approval the office of the president gets. Only one approval figure is over 60% and that’s Bush 43’s war-boosted midterm ratings. Presidents start out limping now. I’m sure an equivalent chart for Congress would be even more dismal. That makes this a systemic problem. Predicting that it won’t go away soon is far more certain than whether Trump will hit the 20s.
It was interesting if not wholly informative to see that a linear trend is the best fit, and good to highlight the “floor” around 23-25%, though that was easily findable from earlier sources.
In terms of predictions, a look at all the past presidential trends brings to mind the adage: “Events, dear boy, Events.”
As Silver has pointed out previously, Trump won despite historically high unfavorability ratings. He didn’t win because voters wanted to have a beer with him; many of his own voters don’t like him. This makes approval polls not only more uncertain but also a less useful metric overall.
I noticed the same thing when CNN presented a table of historical presidential approval ratings at the six-month point of administrations from Kennedy onward. (Maybe Ike was on there too, I forget.) It wasn’t exactly a linear progression, but the first thing that jumped out at you was that every single approval rating was less than or equal to the immediately previous one.
That would’ve been slightly pre-9/11 Bush, so it’s unaffected by wartime inflation.
One thing Silver left out is that election campaigns themselves tend to affect people’s approval ratings, in a cognitive dissonance type of way.
Meaning, suppose you absolute despise Trump, but would never vote for a liberal Democrat. If the topic of the day - and the choice you personally are facing - is “Trump vs liberal Democrat”, then you might be forced to reluctantly support Trump, and once you’re doing that anyway you would automatically find yourself coming up with ways to think of him positively. The reverse would apply to the guy you’re intending to vote against, and you would find yourself justifying that decision by considering that guy in the worst light possible.
The implication in the case of Trump is that while nothing is at stake anyway you get some people disapproving of him who might find themselves approving of him if there was an election at stake and they felt they had to support him.
All that said, Trump himself has no margin for error, and if his support dips by 1% or so as compared to 2016, he loses in 2020. But the more relevant question is to what extent he drags down the Republicans in 2018.