Analysis: Even in a close election, random chance means polls should be showing a broader range of results. That raises the question of whether we’re in for another polling surprise.
The #1 pain point for pollsters is the prediction of the election demographics.
Polls and statistics are such that a general simple random sample has too little power / weak error bounds (I’m talking like +/-10%, nearly useless).
The easiest way to improve your error bounds is to make assumptions about the electorate makeup. IE: if you know the election will be 50% male and 50% female, you can poll 50 men and 50 women (rather than 100 random people, which might end up as 60 men and 40 women due to randomness).
Lather rinse repeat for other groupings (Latino, Asians, black, 18Y olds, 55Y olds, Rural, Urban, etc. etc.) and you get the gist of how this all works.
Alas: the male / female vote this year is completely worked because abortion is on the ballot. All pollsters know this. Their numbers are crap because the methodology is crap this year. It’s impossible to predict women turnout.
The #1 pain point for pollsters is the prediction of the election demographics.
Polls and statistics are such that a general simple random sample has too little power / weak error bounds (I’m talking like +/-10%, nearly useless).
The easiest way to improve your error bounds is to make assumptions about the electorate makeup. IE: if you know the election will be 50% male and 50% female, you can poll 50 men and 50 women (rather than 100 random people, which might end up as 60 men and 40 women due to randomness).
Lather rinse repeat for other groupings (Latino, Asians, black, 18Y olds, 55Y olds, Rural, Urban, etc. etc.) and you get the gist of how this all works.
Alas: the male / female vote this year is completely worked because abortion is on the ballot. All pollsters know this. Their numbers are crap because the methodology is crap this year. It’s impossible to predict women turnout.