I want to first mention that I program stats, but I don't create or craft samples - people that are a lot smarter than me do that.
It does look ok in that they did more than just ask people if they watched it or planned to watch it, but these kinds of quick polls
don't reflect much in people's opinions.
I think any sample SHOULD look like the universe it draws from. Most of those numbers look good, but a few are not.
With ANY opinion poll - and I have said, they are the hardest thing to measure - you have all sorts of errors that can emerge.
With phone polls, a big one is non-response bias. People hang up. I do. Lots of people do. So you introduce a bias
right there - your poll consists of people who said they'd participate. (Yes I know they asked ahead of time - that's still a bias).
I also noticed a lower percentage of minorities and a higher selection of Republicans and those that lean right than is generally occurring
in the population. That by itself could account for a lot of the positive feedback.
I just don't see how it's possible that one speech has changed a year's worth of generally poor approval ratings.
It's either reflecting a knee-jerk reaction that will snap back (had I been interviewed after one of Clinton's SOTUs - I might have said
positive things as well) - or it's reflecting that the kind of person who watched it is just more likely to be positive about Trump.
LASTLY - polls don't mean a lot overall, unless it's around an election and it influences the low information voter.
We've had Presidents who over the years have been recognized as good Presidents - but not during their lifetime.
The classic case is Truman, who had some of the lowest approval ratings ever recorded.
How much you are liked doesn't say a lot about how good a job you are doing.