|
Originally Posted by wufwugy
I'd say you would be right if "approval" was the same thing as "will vote for". Lots of people disapprove yet still vote for.
Apparently that included Banana in 2016 so there you go.
Originally Posted by wufwugy
Also, it's important to note that reasoning from a statistic can only tell us so much. We have to also ask questions like how it makes sense given what we know about the topic. There is a very strong case to be made, which BStand alluded to, that the approval polls are not capturing any meaningful less-good-feel about Trump. Like we discussed yesterday, most people who voted for him are happy and a chunk who normally vote Republican yet didn't vote for him have warmed up. Personally, I have several friends who did not vote for him because they thought he was an ass, but my read on them is that they are likely to vote for him in 2020.
The statistic doesn't 'make sense' or 'not make sense' except to the subjective observer. Not agreeing with the statistic doesn't change the fact of the statistic itself. If one person thinks his 'real' approval rating must be higher than 35% this has no more meaning than that another person thinks his 'real' approval rating must be lower than 35%.
The most common problem people have with understanding statistics is appreciating that it's an estimate with various degrees of fuzzy around it. The value of any statistic is a ballpark one and the size of the ballpark is inversely exponentially related to the size of the sample (in simple terms, very small samples result in a very large ballpark, but the ballpark gets smaller as the size of the sample increases). So the world would be better off if all statistics were printed in grey.
That said, for a statistic of approval rating with > 1k samples to be off by more than a few % would be very very rare occurrence, less than 1/1000. Note also that this poll is updated daily and so larger samples with less uncertainty can be gained by averaging over a number of days (as Gallup does using a 3 day rolling average).
Another source of error would be a systematic bias in who is being polled, in how they are asked the question, or w/e. This might be the case, e.g. if Trump supporters were more inclined to hang up the phone or lie when they get questioned than were his detractors. There's no way of knowing if this is true or not, but again in the history of polling it would be an unlikely event for such a bias to exist to such a great extent as to make the poll off by more than a few %.
|