For good, reliable data, several orders of magnitude more than 1,000 and it would need to have the methodology and data published along with it.
Opinion polls in general are not reliable sources of information and the wrong approach anyway. Telling people that X% of their neighbors hold Y opinion is a well known and effective propaganda and marketing tool for influencing opinion and decision making.
The people conducting the polls use a technique called random sampling to select candidates from a pool that gives more accuracy. But it’s not perfect and the academics think it’s sus too (I dropped a few studies in another comment).
You have to dig for it a bit, but the actual survey can be downloaded (as a pdf) from CNBC. Their data show bias. The data over-representing people over 60. Their education numbers are biased towards the less educated. Their racial numbers are biased (slightly) towards white people. Their income numbers are biased towards wealthier people.
Their voting data shows a major bias towards people who voted, but I’m actually okay with that one, at least in the context of the political reporting. The people who didn’t vote’s opinion on the political situation in the US is not as important as the people who voted. As part of an economic survey about holiday spending that also asked questions about the recent election, it’s not so great though.
No one of those biases would be a big deal, but in totality they add up to a significant and misleading bias that favors the opinions of older, white, middle-class respondents who vote and graduated high school (but attained no further education). That demographic is also the biggest consumer of CNBC content, so the reason for the bias seems fairly obvious. And again, as an “All-America Economic Survey” that’s not really a big deal, especially considering the massive gaps in the data they polled. But as a barometer for political opinion it skews the data in very important and meaningful ways.
Most of them would be from an academic source most likely. That kind of polling would be very expensive and time consuming. There probably aren’t commercial, short term polls with that level of rigor.
A 2020 study published by Berkeley found that the accuracy of election surveys (which are conducted similarly to opinion polls) was grossly exaggerated.
A 2018 Cambridge study says “the level of error has always been substantially beyond that implied by stated margins of error.”
Okay, since that sort of polling would be very expensive and time consuming and people would like to know the opinions of their fellow citizens in aggregate, what would you suggest?
Nothing. That information is not actually useful for most people. But I fully acknowledge that’s just my opinion.
A better solution would be different metrics for different topics. Consumer faith in the economy can be measured by spending, especially if that data could be broken down by demographic. That data absolutely exists, whether businesses would make it public is abother thing entirely.
The results of the election, especially given it was less than six weeks ago, is a much more compelling data point for how Americans feel about the president elect and his policies. Just under half of all Americans voted, so that’s a pretty decent sample.
The “best solution” would be for news organizations to pool resources and do it more reliably. That would mean no more flash polls or opinion polls, and favor longer term tracking of public sentiment.
Social media companies also have much more robust sets of data that better encapsulate public opinion, they could share that quarterly or even just sell reports to news outlets.
But polls are so unreliable and so many people blindly trust and believe them, eliminating that entire class of reporting would be preferable to continuing to publish and circulate that information.
That’s pretty standard size for a national opinion survey. How large do you think they’re supposed to be?
For good, reliable data, several orders of magnitude more than 1,000 and it would need to have the methodology and data published along with it.
Opinion polls in general are not reliable sources of information and the wrong approach anyway. Telling people that X% of their neighbors hold Y opinion is a well known and effective propaganda and marketing tool for influencing opinion and decision making.
It’s essentially institutional peer pressure.
Yeah, I’m not sure how anyone can see “1,000 people accurately represent 330+ million people” and say say, “yep, sounds about right, that does.”
The people conducting the polls use a technique called random sampling to select candidates from a pool that gives more accuracy. But it’s not perfect and the academics think it’s sus too (I dropped a few studies in another comment).
You have to dig for it a bit, but the actual survey can be downloaded (as a pdf) from CNBC. Their data show bias. The data over-representing people over 60. Their education numbers are biased towards the less educated. Their racial numbers are biased (slightly) towards white people. Their income numbers are biased towards wealthier people.
Their voting data shows a major bias towards people who voted, but I’m actually okay with that one, at least in the context of the political reporting. The people who didn’t vote’s opinion on the political situation in the US is not as important as the people who voted. As part of an economic survey about holiday spending that also asked questions about the recent election, it’s not so great though.
No one of those biases would be a big deal, but in totality they add up to a significant and misleading bias that favors the opinions of older, white, middle-class respondents who vote and graduated high school (but attained no further education). That demographic is also the biggest consumer of CNBC content, so the reason for the bias seems fairly obvious. And again, as an “All-America Economic Survey” that’s not really a big deal, especially considering the massive gaps in the data they polled. But as a barometer for political opinion it skews the data in very important and meaningful ways.
I have never seen any sort of poll of Americans several orders of magnitude more than 1000. Can you give an example please?
Pew Research is pretty much a gold standard. In a recent survey on Ukraine they polled almost 10,000 https://www.pewresearch.org/short-reads/2024/11/25/wide-partisan-divisions-remain-in-americans-views-of-the-war-in-ukraine/
Also they post their dataset and methodology. Any poll/survey that doesn’t do that is reasonably suspect.
This Pew Research?
The one with this page entitled, “How can a survey of 1,000 people tell you what the whole U.S. thinks?”
https://www.pewresearch.org/short-reads/2017/05/12/methods-101-random-sampling/
Most of them would be from an academic source most likely. That kind of polling would be very expensive and time consuming. There probably aren’t commercial, short term polls with that level of rigor.
A 2020 study published by Berkeley found that the accuracy of election surveys (which are conducted similarly to opinion polls) was grossly exaggerated.
A 2018 Cambridge study says “the level of error has always been substantially beyond that implied by stated margins of error.”
Okay, since that sort of polling would be very expensive and time consuming and people would like to know the opinions of their fellow citizens in aggregate, what would you suggest?
Nothing. That information is not actually useful for most people. But I fully acknowledge that’s just my opinion.
A better solution would be different metrics for different topics. Consumer faith in the economy can be measured by spending, especially if that data could be broken down by demographic. That data absolutely exists, whether businesses would make it public is abother thing entirely.
The results of the election, especially given it was less than six weeks ago, is a much more compelling data point for how Americans feel about the president elect and his policies. Just under half of all Americans voted, so that’s a pretty decent sample.
The “best solution” would be for news organizations to pool resources and do it more reliably. That would mean no more flash polls or opinion polls, and favor longer term tracking of public sentiment.
Social media companies also have much more robust sets of data that better encapsulate public opinion, they could share that quarterly or even just sell reports to news outlets.
But polls are so unreliable and so many people blindly trust and believe them, eliminating that entire class of reporting would be preferable to continuing to publish and circulate that information.
Polls seemed pretty reliable when it came to the election.