Well they could try to not be so daft about these challenges and act in acccordance with living in the 21st century. (I know, very difficult for Biden and his team, but it is possible.)
Phone-based MFA, spend a tiny fraction of campaign dollars to advertise on social media platforms so respondents can opt-in for the poll, applied stats to randomly select for the target demographic you’re after. This should satisfactorily solve for the above and improve poll quality.
So first off, you seem to be confusing direct campaign activities with polls. Campaigns do sometimes run internal polls, but you should never trust the numbers they put out. Generally, they do want them to be accurate because they want to make strategic decisions around them. However, they aren’t obliged to release those numbers to the public, and if they do, it’s often for a specific reason; they may want to spin a narrative that they’re in a powerful position, or perhaps paint themselves as the underdog. Either way, you don’t want to rely on those numbers.
FiveThirtyEight has not historically included campaign polls in their Presidential model. They sometimes do for Congressional campaigns, because those don’t get polled as heavily and there would be a lack of data if they weren’t included (which also means they have to use other factors to correct the results to get a good model).
Most polls aren’t like that, though. They’re run by private companies.
Second, any little road bump you do to polling means fewer participants. Need to verify MFA through a text message? Whole lot of people are going to see that and promptly stop and go back to what they were doing.
Third, advertising for opt-in? No way you’re getting a randomized sample out of that.
Turns out, polling companies are not run by idiots. This is not an easy problem.
You’re totally right about the distinction between campaign and official polling, fair point.
Still, I do think there are means available for pollsters to get more accurate results from younger voters if they so choose. MFA can solve the identity problem well enough, but there are better solutions if we want them. For example, I’d love to have an anonymous, secure, unique voter id which could be used by individuals to verify voting results independently after an election concludes. If we had it, we could use that instead.
We absolutely should not have such an ID. People buying votes would ask you to show them your verified vote on your phone before you get paid. There’s a long history behind anything that could let you show your vote to another person after the fact, even voluntarily, and we’ve banned them for a reason. It’s one of those problems that we’ve solved so well that people forget why those rules are there.
Morning Consult does online polling, and they seem to do OK. If you look through the FiveThirtyEight polls, you’ll notice they sometimes have an unusually large sample size, like 10k registered voters when most others have between 1k-3k. That’s because they gather a whole lot of people in their polls, but the result isn’t particularly random. They then have to apply weights to get something like a random sample.
Where do they get those weights and how do we know they’re valid? That’s a very good question. They match up with other polls, but those other polls have problems that we’re trying to get away from.
Well they could try to not be so daft about these challenges and act in acccordance with living in the 21st century. (I know, very difficult for Biden and his team, but it is possible.)
Phone-based MFA, spend a tiny fraction of campaign dollars to advertise on social media platforms so respondents can opt-in for the poll, applied stats to randomly select for the target demographic you’re after. This should satisfactorily solve for the above and improve poll quality.
So first off, you seem to be confusing direct campaign activities with polls. Campaigns do sometimes run internal polls, but you should never trust the numbers they put out. Generally, they do want them to be accurate because they want to make strategic decisions around them. However, they aren’t obliged to release those numbers to the public, and if they do, it’s often for a specific reason; they may want to spin a narrative that they’re in a powerful position, or perhaps paint themselves as the underdog. Either way, you don’t want to rely on those numbers.
FiveThirtyEight has not historically included campaign polls in their Presidential model. They sometimes do for Congressional campaigns, because those don’t get polled as heavily and there would be a lack of data if they weren’t included (which also means they have to use other factors to correct the results to get a good model).
Most polls aren’t like that, though. They’re run by private companies.
Second, any little road bump you do to polling means fewer participants. Need to verify MFA through a text message? Whole lot of people are going to see that and promptly stop and go back to what they were doing.
Third, advertising for opt-in? No way you’re getting a randomized sample out of that.
Turns out, polling companies are not run by idiots. This is not an easy problem.
You’re totally right about the distinction between campaign and official polling, fair point.
Still, I do think there are means available for pollsters to get more accurate results from younger voters if they so choose. MFA can solve the identity problem well enough, but there are better solutions if we want them. For example, I’d love to have an anonymous, secure, unique voter id which could be used by individuals to verify voting results independently after an election concludes. If we had it, we could use that instead.
We absolutely should not have such an ID. People buying votes would ask you to show them your verified vote on your phone before you get paid. There’s a long history behind anything that could let you show your vote to another person after the fact, even voluntarily, and we’ve banned them for a reason. It’s one of those problems that we’ve solved so well that people forget why those rules are there.
Morning Consult does online polling, and they seem to do OK. If you look through the FiveThirtyEight polls, you’ll notice they sometimes have an unusually large sample size, like 10k registered voters when most others have between 1k-3k. That’s because they gather a whole lot of people in their polls, but the result isn’t particularly random. They then have to apply weights to get something like a random sample.
Where do they get those weights and how do we know they’re valid? That’s a very good question. They match up with other polls, but those other polls have problems that we’re trying to get away from.