There’s only one close to accurate poll, the election. Given that, how can polling be presented to candidates and voters in a meaningful manner? Personally, I watch the trends in polling numbers, because what you’re saying is right, but I’m trying to make decisions with polling as one of the strongest arrows we have…
Polls are one of several inputs we can use, if we're inclined, to INFORM a prediction about the election. The quality and quantity of polls, and size of possible confounding variables all influence the level of strength and confidence we allow those inputs to have.
But in the absence of proper understanding, experts have convinced themselves, and the public, that those polls ARE predictions.
Polls are blurry snapshots. They're imperfect pictures of the race at some point before the finish line. I'm not sure polls have practical use to the average person - they're pumped by the media as content because (due to the belief that they're predictions) gives them something to talk about.
To qualified practitioners, we can use polls in our models to build a good, probabilistic range of outcomes to optimize spending and messaging in close or otherwise important races.
Given the fact that most people only want to know about election chances, 538 deserves a lot of credit for presenting things in probabilistic terms. But as you probably know, those probabilities are wrongly interpreted by the media and individuals as "calls"
(You said this person had a 70% chance to win, you must have been calling that outcome)
When that 70% hits, their model is praised, they accept it.
But when their 70% chance to win doesn't hit, "polls were wrong." No, not our forecast, it's just a probability, you're just too stupid to understand it.
As unscientific as it is, I find that sitting back in my chair, drinking a cup of coffee and looking at the RCP averages, informs me to the point where my gut feeling is usually right. All of the prediction game is an art form in my mind, more than it is a science…
Currently, there's a ton more art (and guesswork) than science, for both polls and forecasts
Pollsters are getting increasingly creative with their methods but introducing lots more error
I believe it is scientifically possible to make polls so precise that it would be ethically questionable as to whether they should be released to the public for fear of suppressing turnout.
There’s only one close to accurate poll, the election. Given that, how can polling be presented to candidates and voters in a meaningful manner? Personally, I watch the trends in polling numbers, because what you’re saying is right, but I’m trying to make decisions with polling as one of the strongest arrows we have…
Polls are one of several inputs we can use, if we're inclined, to INFORM a prediction about the election. The quality and quantity of polls, and size of possible confounding variables all influence the level of strength and confidence we allow those inputs to have.
But in the absence of proper understanding, experts have convinced themselves, and the public, that those polls ARE predictions.
Polls are blurry snapshots. They're imperfect pictures of the race at some point before the finish line. I'm not sure polls have practical use to the average person - they're pumped by the media as content because (due to the belief that they're predictions) gives them something to talk about.
To qualified practitioners, we can use polls in our models to build a good, probabilistic range of outcomes to optimize spending and messaging in close or otherwise important races.
Given the fact that most people only want to know about election chances, 538 deserves a lot of credit for presenting things in probabilistic terms. But as you probably know, those probabilities are wrongly interpreted by the media and individuals as "calls"
(You said this person had a 70% chance to win, you must have been calling that outcome)
When that 70% hits, their model is praised, they accept it.
But when their 70% chance to win doesn't hit, "polls were wrong." No, not our forecast, it's just a probability, you're just too stupid to understand it.
As unscientific as it is, I find that sitting back in my chair, drinking a cup of coffee and looking at the RCP averages, informs me to the point where my gut feeling is usually right. All of the prediction game is an art form in my mind, more than it is a science…
Currently, there's a ton more art (and guesswork) than science, for both polls and forecasts
Pollsters are getting increasingly creative with their methods but introducing lots more error
I believe it is scientifically possible to make polls so precise that it would be ethically questionable as to whether they should be released to the public for fear of suppressing turnout.
But that's for another day
I wanna read that!