I think the burden of proof goes the other way? Like, the default wisdom for polling is that each polling error[1] is another sample from a distribution centered around 0. It’s not very surprising that it output a R bias twice in a row (even if we ignore the midterms and assume it was properly twice in a row). It’s only two samples! That happens all the time.
If you want a positive argument: pollsters will have attempted to correct mistakes, and if they knew that there would be an R/D bias this time, they’d adjust in the opposite way, hence the error must be unpredictable.
pollsters will have attempted to correct mistakes, and if they knew that there would be an R/D bias this time, they’d adjust in the opposite way, hence the error must be unpredictable.
Exactly. Silver has discussed this dynamic in some of his old FiveThirtyEight articles. The key is to appreciate that polling error is not an effect one can naively predict by looking at past data, because it is mediated by polling agencies’ attempts to correct it.
I think the burden of proof goes the other way? Like, the default wisdom for polling is that each polling error[1] is another sample from a distribution centered around 0. It’s not very surprising that it output a R bias twice in a row (even if we ignore the midterms and assume it was properly twice in a row). It’s only two samples! That happens all the time.
If you want a positive argument: pollsters will have attempted to correct mistakes, and if they knew that there would be an R/D bias this time, they’d adjust in the opposite way, hence the error must be unpredictable.
That is, for a smart polling average; individual polls have predictable bias.
Exactly. Silver has discussed this dynamic in some of his old FiveThirtyEight articles. The key is to appreciate that polling error is not an effect one can naively predict by looking at past data, because it is mediated by polling agencies’ attempts to correct it.