I have a related degree. The reason people distrust polls, is because the media frequently misreports or misrepresents them.
Eg. aggregated polling from the 2016 suggested Trump had a 1/3 chance of winning. If you believed some media coverage every poll said Clinton was certain to win. That was how the media reported on the polling, not the polling itself. Invariably Trump winning in 2016 was within the margin of error.
that whopping MoE
Not a large margin of error. You’re extrapolating from 1000 people to 300 million. It’s astonishing it’s that low if you think about it.
because they’re usually done by telephone
Not that common anymore. Often they’ll do a a telephone poll then supplement it with online or other methods. Here’s IPSOS’s article about this poll:
The study was conducted online in English. The data for the total sample were weighted to adjust for gender by age, race/ethnicity, education, Census region, metropolitan status, household income, and political party affiliation. The demographic benchmarks came from the 2022 March Supplement of the Current Population Survey (CPS). Party ID benchmarks are from recent ABC News/Washington Post telephone polls.
I remember reading 538 leading up to the 2016 election, and hearing them say repeatedly that if Trump has a 1 in 4 chance (or whatever amount) of winning the election, not only is it possible for Trump to win, but in fact it means you actually expect it to happen in 1 out of 4 times.
Yeah I remember this too. I think the problem is that people simply don’t understand statistics and don’t realize a 70% chance of winning is totally different from getting 70% of the vote. I like what 538 has been doing in recent years by presenting odds rather than percentages, but people like echo chambers that confirm their biases so idk if this “polls don’t work” narrative is going to go away any time soon.
That’s the article that has caused me to trust 538 above any other election prediction source. When HRC was doing a preemptive victory lap in Texas and HuffPo was publishing articles that said she had a 99% chance of winning, Nate Silver and Co were the only ones willing to admit the possibility of what would later become reality.
Indeed, the problem isn’t polls themselves, assuming they’re well constructed they’re generally sound data, it’s the interpretation and packaging of it as reported to the larger populous that gets in the way. Sometimes it gets to the point of funny when someone does an infographic where 30% and 60% somehow appear to have the same weight.
My biggest issue with polls is that the media tout them as predictions, ignoring the fact that even if the data is 100% valid, circumstances can change dramatically in just a couple of days.
I maintain that polls are not actionable data for voters. They can help campaigns see trends and gauge the effectiveness of messaging, but they are useless to voters.
They can and do change in just a couple of days, but the real issue is that the media invariably fails to mention the margin of error or confidence interval.
It’s always Candidate A 51%, candidate B 49%. When in reality it’s inevitably something like “There’s 19/20 chance that candidate gets between 48.5-53.5% of the vote, and that candidate gets between 46.5-51.5% of the vote.”
And then when candidate B wins, the media will go “Why did the polls get it wrong?” when the election was always to close to call definitively.
Oh, and this is obviously ignoring the far more sinister use of misrepresented polling data, micro-demographically targetted thanks to big data harvested from social media. Think Cambridge Analytica algorithms which have determined that women in village X with one child and dog, being more likely to vote party Y, and then targetting them on social media with stories about the polls showing the result is a foreglone conclusion and that there’s no point voting.
I have a related degree. The reason people distrust polls, is because the media frequently misreports or misrepresents them.
Eg. aggregated polling from the 2016 suggested Trump had a 1/3 chance of winning. If you believed some media coverage every poll said Clinton was certain to win. That was how the media reported on the polling, not the polling itself. Invariably Trump winning in 2016 was within the margin of error.
Not a large margin of error. You’re extrapolating from 1000 people to 300 million. It’s astonishing it’s that low if you think about it.
Not that common anymore. Often they’ll do a a telephone poll then supplement it with online or other methods. Here’s IPSOS’s article about this poll:
https://www.ipsos.com/en-us/politico-indictment-august-2023
I remember reading 538 leading up to the 2016 election, and hearing them say repeatedly that if Trump has a 1 in 4 chance (or whatever amount) of winning the election, not only is it possible for Trump to win, but in fact it means you actually expect it to happen in 1 out of 4 times.
Yeah I remember this too. I think the problem is that people simply don’t understand statistics and don’t realize a 70% chance of winning is totally different from getting 70% of the vote. I like what 538 has been doing in recent years by presenting odds rather than percentages, but people like echo chambers that confirm their biases so idk if this “polls don’t work” narrative is going to go away any time soon.
That’s the article that has caused me to trust 538 above any other election prediction source. When HRC was doing a preemptive victory lap in Texas and HuffPo was publishing articles that said she had a 99% chance of winning, Nate Silver and Co were the only ones willing to admit the possibility of what would later become reality.
And then the big companies came in and wrecked 538.
Ok, fair enough. I defer to your expertise.
Indeed, the problem isn’t polls themselves, assuming they’re well constructed they’re generally sound data, it’s the interpretation and packaging of it as reported to the larger populous that gets in the way. Sometimes it gets to the point of funny when someone does an infographic where 30% and 60% somehow appear to have the same weight.
Lies, damn lies, and statistics…
Not that it matters here, but in case you want to use them somewhere serious:
Populous is an adjective
Populace is a noun
Thus why I’m not a writer, but at least the intent was there 🙂
I’m doing good if it doesn’t look like a drunken baboon wrote it sometimes due to fat finger typing.
Considering people tend to view probability as 100% A, 100% B or 50-50, I’m not sure showing a 30-60 split as the same weight is really a bad choice…
My biggest issue with polls is that the media tout them as predictions, ignoring the fact that even if the data is 100% valid, circumstances can change dramatically in just a couple of days.
I maintain that polls are not actionable data for voters. They can help campaigns see trends and gauge the effectiveness of messaging, but they are useless to voters.
They can and do change in just a couple of days, but the real issue is that the media invariably fails to mention the margin of error or confidence interval.
It’s always Candidate A 51%, candidate B 49%. When in reality it’s inevitably something like “There’s 19/20 chance that candidate gets between 48.5-53.5% of the vote, and that candidate gets between 46.5-51.5% of the vote.”
And then when candidate B wins, the media will go “Why did the polls get it wrong?” when the election was always to close to call definitively.
Oh, and this is obviously ignoring the far more sinister use of misrepresented polling data, micro-demographically targetted thanks to big data harvested from social media. Think Cambridge Analytica algorithms which have determined that women in village X with one child and dog, being more likely to vote party Y, and then targetting them on social media with stories about the polls showing the result is a foreglone conclusion and that there’s no point voting.
Thanks! I really like how Lemmy users with expertise in their area can add nuance to a lot of reporting, it really matters.