Polls aren’t proven to be worthless, they just have error bars that people don’t pay attention to. They’re nothing crazy, just statistics. Problem is: most people do not understand statistics.
No, modern polling is mostly useless. The respondents skew heavily in one direction (old people that answer the phone).
There is no sample size big enough to account for a complete cultural shift away from answering phone calls and text messages from people you don’t know.
Most polls report a response rate of about 90%, as reported by a PEW analysis of available studies. However, that is a made up number which does not account for attrition (quitters) or non-response to panel recruitment (no answers). If you include those numbers the real response rate is about 3%. Which means between the initial contact and giving their opinion 97% of people asked don’t participate in a given poll.
This comment comes off as if you think the people working at Pew, one of the most respected orgs for surveys out there, are just plain stupid and are dumb enough to leave a giant hole in their data when concerning young people. What you’re pointing out is a challenge in modern surveys but this is stuff that Pew is actively working to correct and with the midterm polling, they were far more accurate.
That’s all due to incorrect weighting of the data but Pew notes that polls specifically like the ones referenced here when looking at national sentiment tend to be much more accurate.
Thank you for posting that. It was well-written, and did a great job clarifying both why previous polls were inaccurate, and why they’re likely better now. I was particularly interested to read that the margin of error may be about double what is reported.
You should make this a top level post in this group. It is worth discussing on its own.
This is a great article to discuss, and I think it’d make a fine post all by itself.
However, “Restoring people’s confidence in polling is an important goal” is at the top of the article for a reason.
We can expect that the answer “polling is not useful” will not be explored. Limitations to polling concepts will be minimized. The article has stated one of its goals, if not its main goal.
No, he’s got a point. Saying Candidate X is in the “lead” over candidate Y right now is like saying some athlete is in the “lead” of an Olympics event… in May when the games haven’t even started yet.
That’s still important if the actual race began now and the finish-line was November 5th. What I mean is that if you’re 5 kilometers back when you were even at the same point last time in the race, then no matter what it’s generally going to be harder to make up that ground by the end of the race. Something has to change where voters are now versus November no differently than making up the difference between runners in a Marathon.
And this data can (a) be useful to change strategy (e.g., Biden stepping down; Harris stepping up or less drastic: altering campaign messaging), and (b) positive momentum tends to excite the base. People like to see positive results. The beauty is that the Harris campaign is still framing themselves as the underdogs — which they are, but it also helps offset any risk to complacency with overconfident voters.
Understanding Polls:
Individual polls from reputable pollsters can be a barometer for a snapshot in time, but they may also be outliers.
An aggregation of many reputable polls during the same period of time is a more accurate snapshot in time.
An aggregation of many polling snapshots over a period of time can show a Trend.
Long-term trends can be very useful and give more extrapolative trajectories (e.g., the long-term downward decline of Biden’s aggregate national approval ratings and his steady decline in swing-states leading to a change in strategy and his stepping down).
Still, such polls may not accurately represent fringe groups (though many pollsters compensate in a variety of ways).
We shouldn’t just blindly follow the polls (blind-leading-the-blind mentality)—e.g., if the case is never made for something, then it never gets popular. Bernie Sanders heavily advocated for Universal Healthcare and we of course have seen an adjustment in polling instead of simply reacting to its initial unpopularity—but we also shouldn’t ignore trends.
Polls don’t dictate what people do in the moment, or say or do later; instead, they’re a reflection of where they say they’ll do in the moment.
Every advocate should have the mindset of trying to change polls to their advantage; this by active campaigning (canvassing, phone-banking, fundraising, etc.), change of messaging, etc.
Context should always be considered when discussing polling. (e.g., in isolation, Biden’s debate could be considered, “just one bad night, and we can swing polls back,” without considering the long-term concern that was already present over his immutable vice — age/cognitive-decline.)
No matter what the polls say, winning, tying, or losing… Always and I mean always Register and VOTE. Not just this, but drag 3-5 other people to register and vote with you.
Most of that after the first paragraph is valid, but it can only mean a candidate “is favored” or something like that (in the same sense, to continue the analogy, that an athlete who won a bunch of previous events in the lead-up to the Olympics “is favored” at the start of the Olympic event itself). It can’t be “in the lead,” because the actual race event doesn’t begin until the polls open.
The point is, being the favorite doesn’t actually mean you’ve made progress towards winning. It is not like being 5 km ahead in a marathon! It is still extremely possible for the favorite to choke at the event itself and lose badly, and all the prior favorability in the world is completely moot and confers no actual advantage at all.
Polls aren’t proven to be worthless, they just have error bars that people don’t pay attention to. They’re nothing crazy, just statistics. Problem is: most people do not understand statistics.
No, modern polling is mostly useless. The respondents skew heavily in one direction (old people that answer the phone).
There is no sample size big enough to account for a complete cultural shift away from answering phone calls and text messages from people you don’t know.
Most polls report a response rate of about 90%, as reported by a PEW analysis of available studies. However, that is a made up number which does not account for attrition (quitters) or non-response to panel recruitment (no answers). If you include those numbers the real response rate is about 3%. Which means between the initial contact and giving their opinion 97% of people asked don’t participate in a given poll.
This comment comes off as if you think the people working at Pew, one of the most respected orgs for surveys out there, are just plain stupid and are dumb enough to leave a giant hole in their data when concerning young people. What you’re pointing out is a challenge in modern surveys but this is stuff that Pew is actively working to correct and with the midterm polling, they were far more accurate.
That’s all due to incorrect weighting of the data but Pew notes that polls specifically like the ones referenced here when looking at national sentiment tend to be much more accurate.
If you’d like to read more about the problems with polling Pew has a whole write up on it
Thank you for posting that. It was well-written, and did a great job clarifying both why previous polls were inaccurate, and why they’re likely better now. I was particularly interested to read that the margin of error may be about double what is reported.
You should make this a top level post in this group. It is worth discussing on its own.
This is a great article to discuss, and I think it’d make a fine post all by itself.
However, “Restoring people’s confidence in polling is an important goal” is at the top of the article for a reason.
We can expect that the answer “polling is not useful” will not be explored. Limitations to polling concepts will be minimized. The article has stated one of its goals, if not its main goal.
No, he’s got a point. Saying Candidate X is in the “lead” over candidate Y right now is like saying some athlete is in the “lead” of an Olympics event… in May when the games haven’t even started yet.
That’s still important if the actual race began now and the finish-line was November 5th. What I mean is that if you’re 5 kilometers back when you were even at the same point last time in the race, then no matter what it’s generally going to be harder to make up that ground by the end of the race. Something has to change where voters are now versus November no differently than making up the difference between runners in a Marathon.
And this data can (a) be useful to change strategy (e.g., Biden stepping down; Harris stepping up or less drastic: altering campaign messaging), and (b) positive momentum tends to excite the base. People like to see positive results. The beauty is that the Harris campaign is still framing themselves as the underdogs — which they are, but it also helps offset any risk to complacency with overconfident voters. Understanding Polls:
Most of that after the first paragraph is valid, but it can only mean a candidate “is favored” or something like that (in the same sense, to continue the analogy, that an athlete who won a bunch of previous events in the lead-up to the Olympics “is favored” at the start of the Olympic event itself). It can’t be “in the lead,” because the actual race event doesn’t begin until the polls open.
The point is, being the favorite doesn’t actually mean you’ve made progress towards winning. It is not like being 5 km ahead in a marathon! It is still extremely possible for the favorite to choke at the event itself and lose badly, and all the prior favorability in the world is completely moot and confers no actual advantage at all.
The other important thing to consider: polls are a snapshot in time.
They are not a predictor. The only poll that matters is election day.