I’m not super trusting of polls anymore
Like, do you not believe the people responded the way they say they did?
No, I’m just not certain it’s an accurate sample. Polls were way off in 2020 and 2016.
They actually weren’t though. Trump lost the popular vote by a huge margin in both cases, which is what was predicted.
They weren’t. Aggregated polling in 2016 gave Trump a 1/3 chance of winning. That’s not low. It’s actually quite likely. Him becoming president was invariably within the margin of error in many polls.
Eg. 2016
https://projects.fivethirtyeight.com/2016-election-forecast/
2020:
https://projects.fivethirtyeight.com/2020-election-forecast/
https://projects.fivethirtyeight.com/2020-election-forecast/senate/
https://projects.fivethirtyeight.com/2020-election-forecast/house/
2022:
https://projects.fivethirtyeight.com/2022-election-forecast/
The media reporting on polls (and anything scientific for that matter) is universally abysmal, that’s why you mistakenly think otherwise.
Define “accurate”. Define “way off”. What do you think a poll tells you?
What you’re upset about is how you’re interpreting polls - and I guarantee you you’re doing it wrong.
90% chance is not a guarantee success. 30% chance is not a guarantee fail. They’re probabilities.
This poll, taken alone, tells you what these people think. It’s not a prediction and by itself doesn’t really say much. Taken in aggregate with other polls you can start to form an idea. But NO POLL will ever tell you the future.
Not the responses themselves but the methodologies of collecting responses don’t result in accurate representation of the population.
Using collection methods that skew demographics in one direction or another, like older people being more likely to pick up a phone call.
Failing to account for other potentially major variables. Like the 2016 and 2020 elections, pollsters failed to account for negative voter turnout, people who were motivated to vote against a specific candidate, which had major impacts on the elections.
In this poll responses were collected online and the sample was weighted to reflect census demographics
I like that, it’s a pretty good breakdown of controlled variables. And it looks like they’re factoring in socioeconomic factors too, which is always a good thing.
After the last however long of bad polling, especially in the last 8 years, it’s refreshing to see some better methodology but it’s still going to take a while to get that general trust back.
Most polling is done via landline phone. Thus polling does reflect well on the actual voting population.
I’m not so sure that’s true anymore. I don’t go looking at every single study but I usually see a mix of landline, cell/sms, and online samples
That is why I said most, not all.
About the Study
This Ipsos poll was conducted August 18-21, 2023 on behalf of Politico Magazine, using the probability based KnowledgePanel®. This poll is based on a representative sample of 1,032 U.S. residents, age 18 or older, including 272 Republican respondents, 321 Democratic respondents, and 319 independent respondents.
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. The weighting categories were as follows:
Gender (Male, Female) by Age (18–29, 30–44, 45-59 and 60+) Race/Hispanic Ethnicity (White Non-Hispanic, Black Non-Hispanic, Other, Non-Hispanic, Hispanic, 2+ Races, Non-Hispanic) Education (Less than High School, High School, Some College, Bachelor or higher) Census Region (Northeast, Midwest, South, West) Metropolitan status (Metro, non-Metro) Household Income (Under $25,000, $25,000-$49,999, $50,000-$74,999, $75,000-$99,999, $100,000-$149,999, $150,000+) Party ID (Democrat, Republican, Independent, Something else)
The margin of sampling error is plus or minus 3.2 percentage points at the 95% confidence level, for results based on the entire sample of adults. The margin of error takes into account the design effect, which was 1.08 for all adults. The margin of sampling error is plus or minus 6.2 percentage points at the 95% confidence level, for results based on the sample of Republicans. The margin of error takes into account the design effect, which was 1.08 for Republicans. The margin of sampling error is plus or minus 5.8 percentage points at the 95% confidence level, for results based on the sample of Democrats. The margin of error takes into account the design effect, which was 1.11 for Democrats. The margin of sampling error is plus or minus 5.7 percentage points at the 95% confidence level, for results based on the sample of independents. The margin of error takes into account the design effect, which was 1.09 for independents. In our reporting of the findings, percentage points are rounded off to the nearest whole number. As a result, percentages in a given table column may total slightly higher or lower than 100%. In questions that permit multiple responses, columns may total substantially more than 100%, depending on the number of different responses offered by each respondent.
