Polling got way harder after everyone ditched land lines. Nate Silver went from near perfect handicapping in 2008/2012 to abysmal predictions from 2016 onward.
I genuinely wonder how polling is going to recover. Pollsters are having to rely more and more on the few remaining landlines, online polls of questionable quality, and the rare millennial who is willing to answer calls and texts from unknown numbers. To make up for it, they have to pile on more and more assumptions from things like demographics data.
I don’t see any of these factors changing any time soon, and unless pollsters figure out ways to work around them without compromising the quality of their work, they’re kind of fucked.
Yep and now you’re also getting into the arena of, “okay well you’re GETTING people in those age groups and so forth… So what kind of subset of that demographic is actually responding to pollster’s calls and texts…?” Is that really a random sample anymore?
E.g., NYT Siena does get youth to respond, but I’m curious how many more calls they have to make to get one versus, say, someone in their 60s.
The West Wing Weekly touched on this in one of their episodes. Basically they are turning more to focus groups where participants are paid a small amount.
At this point, polling should assume a conservative bias.
Polling got way harder after everyone ditched land lines. Nate Silver went from near perfect handicapping in 2008/2012 to abysmal predictions from 2016 onward.
I genuinely wonder how polling is going to recover. Pollsters are having to rely more and more on the few remaining landlines, online polls of questionable quality, and the rare millennial who is willing to answer calls and texts from unknown numbers. To make up for it, they have to pile on more and more assumptions from things like demographics data.
I don’t see any of these factors changing any time soon, and unless pollsters figure out ways to work around them without compromising the quality of their work, they’re kind of fucked.
Yep and now you’re also getting into the arena of, “okay well you’re GETTING people in those age groups and so forth… So what kind of subset of that demographic is actually responding to pollster’s calls and texts…?” Is that really a random sample anymore?
E.g., NYT Siena does get youth to respond, but I’m curious how many more calls they have to make to get one versus, say, someone in their 60s.
They’re going to have to start figuring out sentiment without polling, which is a lot harder.
The West Wing Weekly touched on this in one of their episodes. Basically they are turning more to focus groups where participants are paid a small amount.
At this point, polling should assume a conservative bias.