2024 POTUS - polling only

[Anonymous]

Anonymous wrote:Watch that clip with the pollster until the end. The comparison to 2023 OH prior to abortion ballot. In Ohio, there was a 70% increase in women voter registration and the pro-choice side won in OH by 7-8pts. Very Interesting.

https://pbs.twimg.com/ext_tw_video_thumb/1828556487842582528/pu/img/YCPcSWI-u_z1N-CM.jpg

That was an interesting clip. The stats guy confirmed that the Dobbs decision has materially affected election outcomes since that date, partially from newly registered voters that are not easy to capture in polling.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Voter registration comparison between these past few weeks of 2024 vs. the same weeks in 2020.

Deeper dive into this:

This should be overlaid with where the registrations are happening. It's all well and good to have higher registration in New York, but it won't move the needle one bit.

They looked at 13 states including Michigan and North Carolina.

[Anonymous]

Anonymous wrote:After 2016 I'm skeptical of the accuracy of polls. I've never been polled before and I don't have a landline. I don't even answer my cell phone unless I know who is calling. What is the breakdown of Democrats vs. Republicans vs. Independents when it comes to these polls? Oversampling of one can obviously skew the results making a candidate appear more popular than what they really are.

Pollsters will oversample if they want to do a deep dive into a particular subgroup's views. But oversampling is expensive because you need to keep calling more and more people until you have your oversample numbers.

Instead, what they will do is adjust the weighting of certain groups. For example, lets say a Congressional district's voter turnout is traditionally 30% Latino/Hispanic. Well, you've talked with 1000 survey respondents but you only got 150 responses from Latinos/Hispanics (15% of sample). You could keep calling and get more Latinos, but that is expensive. Instead, what pollsters will do is take the 150 responses from Latinos and over-weight them to 30% of the overall "Total" results.

Basically you're doubling the power of the 150 Latinos you spoke with...but what if there's something different from those 150 Latinos relative to the rest of the Latino population? Maybe more likely to be unemployed? Older? Etc.

Pollsters have to do this re-weighting with multiple demographic groups in any given poll. A typical one is the 18-29 age bucket - pollsters have an extremely difficult time reaching this group. So the few interviews they get with Gen Z will be re-weighted and scaled up when calculating the overall results sometimes using just a few dozen interviews in a 1000 person poll. This age group also has large swings in voter turnout every four years, so its difficult to model what percentage will compose come Election Day.

In short, polling is broken for lots of reasons - the weird self-selection of people who will talk to pollsters, difficulties in modeling turnout among a diversifying electorate, etc.

[Anonymous]

Anonymous wrote:Where can one see reliable swing state polls? The overall polls don’t really mean much.

Right, just as folks relied on NYT in '16 and they were so shocked when results came out. What does Prof Lichtman predict for '24 POTUS?

[Anonymous]

Anonymous wrote:

Anonymous wrote:After 2016 I'm skeptical of the accuracy of polls. I've never been polled before and I don't have a landline. I don't even answer my cell phone unless I know who is calling. What is the breakdown of Democrats vs. Republicans vs. Independents when it comes to these polls? Oversampling of one can obviously skew the results making a candidate appear more popular than what they really are.

Pollsters will oversample if they want to do a deep dive into a particular subgroup's views. But oversampling is expensive because you need to keep calling more and more people until you have your oversample numbers.

Instead, what they will do is adjust the weighting of certain groups. For example, lets say a Congressional district's voter turnout is traditionally 30% Latino/Hispanic. Well, you've talked with 1000 survey respondents but you only got 150 responses from Latinos/Hispanics (15% of sample). You could keep calling and get more Latinos, but that is expensive. Instead, what pollsters will do is take the 150 responses from Latinos and over-weight them to 30% of the overall "Total" results.

Basically you're doubling the power of the 150 Latinos you spoke with...but what if there's something different from those 150 Latinos relative to the rest of the Latino population? Maybe more likely to be unemployed? Older? Etc.

Pollsters have to do this re-weighting with multiple demographic groups in any given poll. A typical one is the 18-29 age bucket - pollsters have an extremely difficult time reaching this group. So the few interviews they get with Gen Z will be re-weighted and scaled up when calculating the overall results sometimes using just a few dozen interviews in a 1000 person poll. This age group also has large swings in voter turnout every four years, so its difficult to model what percentage will compose come Election Day.

In short, polling is broken for lots of reasons - the weird self-selection of people who will talk to pollsters, difficulties in modeling turnout among a diversifying electorate, etc.

Didn't the OP say if you have more than a few lines, discuss on another thread??

[Anonymous]

Anonymous wrote:

Anonymous wrote:Where can one see reliable swing state polls? The overall polls don’t really mean much.

Right, just as folks relied on NYT in '16 and they were so shocked when results came out. What does Prof Lichtman predict for '24 POTUS?

He said that while he will put out an official prediction after Labor Day, for now he thinks "a lot would have to go wrong for the Democratic Party to lose." Take that for what it is.

https://www.miamiherald.com/news/local/news-columns-blogs/andres-oppenheimer/article291223915.html

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:After 2016 I'm skeptical of the accuracy of polls. I've never been polled before and I don't have a landline. I don't even answer my cell phone unless I know who is calling. What is the breakdown of Democrats vs. Republicans vs. Independents when it comes to these polls? Oversampling of one can obviously skew the results making a candidate appear more popular than what they really are.

Pollsters will oversample if they want to do a deep dive into a particular subgroup's views. But oversampling is expensive because you need to keep calling more and more people until you have your oversample numbers.

Instead, what they will do is adjust the weighting of certain groups. For example, lets say a Congressional district's voter turnout is traditionally 30% Latino/Hispanic. Well, you've talked with 1000 survey respondents but you only got 150 responses from Latinos/Hispanics (15% of sample). You could keep calling and get more Latinos, but that is expensive. Instead, what pollsters will do is take the 150 responses from Latinos and over-weight them to 30% of the overall "Total" results.

Basically you're doubling the power of the 150 Latinos you spoke with...but what if there's something different from those 150 Latinos relative to the rest of the Latino population? Maybe more likely to be unemployed? Older? Etc.

Pollsters have to do this re-weighting with multiple demographic groups in any given poll. A typical one is the 18-29 age bucket - pollsters have an extremely difficult time reaching this group. So the few interviews they get with Gen Z will be re-weighted and scaled up when calculating the overall results sometimes using just a few dozen interviews in a 1000 person poll. This age group also has large swings in voter turnout every four years, so its difficult to model what percentage will compose come Election Day.

In short, polling is broken for lots of reasons - the weird self-selection of people who will talk to pollsters, difficulties in modeling turnout among a diversifying electorate, etc.

Didn't the OP say if you have more than a few lines, discuss on another thread??

I think this post is about how polling is conducted, and therefore relevant. Once someone starts going off on tangents unrelated to polling, that should be reported to keep the thread from deteriorating. I don't think OP literally meant don't write more than a few sentences, but rather that larger paragraphs are indicative of non-polling related discussions.

[Anonymous]

BREAKING MICHIGAN poll... Trump wins indies by 56 POINTS?
🔵Harris/Walz (D): 50.1% (+0.2)
🔴Trump/Vance (R): 49.9%
🟡Independent: 78% Trump-22% Harris

Weighted sample: D55/R30/I15

[Anonymous]

Anonymous wrote:BREAKING MICHIGAN poll... Trump wins indies by 56 POINTS?
🔵Harris/Walz (D): 50.1% (+0.2)
🔴Trump/Vance (R): 49.9%
🟡Independent: 78% Trump-22% Harris

Weighted sample: D55/R30/I15

Do you have a link for this poll?

[Anonymous]

Anonymous wrote:

Anonymous wrote:BREAKING MICHIGAN poll... Trump wins indies by 56 POINTS?
🔵Harris/Walz (D): 50.1% (+0.2)
🔴Trump/Vance (R): 49.9%
🟡Independent: 78% Trump-22% Harris

Weighted sample: D55/R30/I15

Do you have a link for this poll?

ActiVote / July 28-Aug. 28 / n=400LV

[Anonymous]

Anonymous wrote:BREAKING MICHIGAN poll... Trump wins indies by 56 POINTS?
🔵Harris/Walz (D): 50.1% (+0.2)
🔴Trump/Vance (R): 49.9%
🟡Independent: 78% Trump-22% Harris

Weighted sample: D55/R30/I15

this cite is missing a critical point: pollster?

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:BREAKING MICHIGAN poll... Trump wins indies by 56 POINTS?
🔵Harris/Walz (D): 50.1% (+0.2)
🔴Trump/Vance (R): 49.9%
🟡Independent: 78% Trump-22% Harris

Weighted sample: D55/R30/I15

Do you have a link for this poll?

ActiVote / July 28-Aug. 28 / n=400LV

Link doesn’t work

[Anonymous]

Activote.net

[Anonymous]

https://www.activote.net/harris-and-trump-essentially-tied-in-michigan/

[Anonymous]

keep going like we're 20 points behind, friends