What does “Polls Say” in an Election even mean? How are they updated? No one has ever asked me my thoughts. Who are they asking and who is actually answering?
In: Other
Polling companies have different methods of gathering survey information. In the past they would call phone numbers randomly from the phone book, but that is difficult now given the use of cellphones.
Sometimes they will do in the street polls, literally just walk up and ask you.
The method of survey does impact results though. Phone polling skews whatever way the elderly vote, because the elderly are the ones more likely to still have a landlines phone. In the street may exclude views of the disabled or those who don’t get out much.
It’s actually quite tricky to get a random and representative sampling sometimes.
I had a pollster quiz me over the phone once in a political election. After several questions, it was obvious that all the questions had been designed to get a certain response, and was obviously funded by one of the parties involved.
Sort of like the old joke, “Does your Mother know you’re stupid?” It’s established you’re stupid whether she knows or not.
When I mentioned this to the pollster (probably a minimum wage working-his-way-through-college type) about how obvious this was, he said, “Yeah, I know. But I gotta ask them like they’re written”.
If I remember right, the polls said Hillary would win…
Pollsters ask a cross section of voters who they will vote for. They typically have to ask 1000 people to get a statistically significant answer. So if they poll voters in a state they try to reach a proportional amount of e.g. voters who are middle class, a proportional amount of female voters and so on. This method has shown to be accurate most of the time, though it can depend on factors such as how many of those polled don’t answer.
They ask random people, but make sure the sample is “representative”, wich means they check wether all age/ethnicity/sex/education demographics are properly represented (because people of some demographics are less likely to answer such a questionaire, wich they have to compensate for)
The total samplesize is usually around 1000 people wich leads to an error margin around 1% (that’s the “law of large numbers”, the more people you ask, the more their answers will approach the average of the whole population)
There is a science (that I don’t totally know) behind determining how many people you need to poll and how you select them in order to produce a sample that reasonably reflects the wider population’s opinion, and the math boils down to only needing a few thousand people at most to figure out the country’s leaning. News organizations share the flashy numbers, but there’s usually a lot more data getting collected, and if you dig a little more, you’ll find out a lot more about who was polled, what they were asked, and how people responded. A good survey will layout this methodology, even if the news doesn’t report on that.
For example, [this article from Politico](https://www.politico.com/newsletters/west-wing-playbook/2024/07/18/so-you-wanted-some-harris-polling-00169524) reports on what pollees are saying about a Harris run. If you follow the links on the article, you’ll find [the actual data from Public Policy Polling](https://www.politico.com/f/?id=00000190-c7b2-d90b-a5fc-d7bf7a790000), and the first footnote explains their methodology:
> On July 17-18, Public Policy Polling surveyed 650 Michigan registered voters & 624 Pennsylvania registered voters on behalf of Clean and Prosperous America PAC. In Michigan, the survey’s margin of error is +/-3.9% and 60% of interviews were conducted by text message and 40% by telephone. In Pennsylvania, the survey’s margin of error is +/-3.8% and 61% of interviews were conducted by text message and 39% by telephone. Detailed methodology is available on the Public Policy Polling website.
1. **Choosing People to Ask**: Pollsters (the people who make the polls) pick a group of people to ask. They try to choose a group that represents the whole population, like making sure they ask people of different ages, genders, and backgrounds.
2. **Asking Questions**: They ask these people questions about who they plan to vote for or what they think about certain issues. This can be done over the phone, online, or in person.
3. **Collecting Answers**: The pollsters collect all the answers and look at the data. They use math to figure out what the answers might mean for the whole population.
4. **Predicting Results**: Based on the answers they collected, pollsters make predictions about how the election might turn out. They might say something like, “Candidate A is leading with 55% of the vote.”
5. **Margin of Error**: Polls also have something called a “margin of error,” which tells us how accurate the poll is. For example, if the margin of error is 3%, and Candidate A has 55%, it means the real number could be between 52% and 58%.
Polls just give us an idea of what might happen, they aren’t always 100% accurate.
Idk why we (the US) don’t have a nationally funded and 24/7 (well, i suppose business hours at least) polling stations. Tie it in with the Post Office, have it have new poll questions every week/day/whatever. People walk up (or have it done thru the mail), show their ID/voter registration and boom, they answer the poll questions and a couple days later the polling people report on what was answered.
Sometime, like after trump got shot by the republican kid or after Biden fumbled the fuck out of the debate, the polling stations will be busy all day long. Other times it would be empty. Either way, the ‘polls’ would be way more accurate and/or precise. And localized…and all that.
Plus it would not have the trolls and bots issue that modern polling has. Otoh it would be just another thing to argue over “Who gets to decide what questions are asked/how they are formatted?” or ” Is it fair to the poor who don’t have the time to get to the polls?” or “Since it is not voting, should non-citizens be allowed to answer poll questions?”…
Polling is an art and a science. The idea is to ask a limited number if people and be able to *reliably* use that to predict how the population will act.
so that means kniw a great deal about the people who are being polled: age, sex, if they are married, level of education, what they do for a job, income level, where they live, etc etc etc.
and then have an accurate break down of the population – how many people their are, how many are male/female, income levels, race, ethnicity, job, religion, car they drive, etc etc etc (the more the merrier). So the pollsters kniw that 48% of Americans are male, 52% are female, what percentage is what age, how much they make, their education levels etc etc etc,
so when someone is polled the pollster also asks for some of this demographic information (or has already gotten it from other sources) – the the answer from the polled person becomes – white, make, age 37, income 60,000, advertising rep, BS in computer science, lives in NYS. – PERSON POLLED WILL VOTE STRAIGHT LINE DEMOCRAT IN NATIONAL ELECTIONS.
wash rinse repeat for – white male, age 67, retired, etc etc etc VOTE TRUMP NATIONAL, VOTE DEMOCRAT LOCAL
each of these answers are plugged into the demographic information and used to extrapolate how the population is going to vote. The demographic information is used to limit the predictions from the peopled polled only to people have some kind of overlap with the person polled.
This is actually a big problem in modern polling. It used to be regular people would answer their phone, answer the door, and answer questions from pollsters on the street.
Now, the only people who answer unknown numbers are old people and idiots. The same goes for people who answer the door.
Here is an example of phone survey responders breakdown: 85% +65yo, 7% 55-64yo, 4% 45-54yo, etc. There are ways to extrapolate poll results from underrepresented demographics, but these numbers have a high error rate.
Social media surveys try to account for these underrepresented demos, but the people who answer social media surveys are usually lower educated with extreme views. Their responses only skew the results, adding to the error.
So yeah, this is a huge issue and it hasn’t been solved yet. It won’t be solved before this election. Consider all polling results with a grain of salt.
I get called and texted by pollsters all the time but I hang up on them. A lot of them do “push polls” and I’m done with it.
A push poll is when someone pays to disguise a political ad as a poll.
Example: for my third question, when Joe Biden said babies are delicious, how did that make you feel?
The “answer” isn’t the point. The fake information in the question is meant to influence you. This happens a lot.
Your question as to who is answering the questions is valid. For awhile, polling was done mostly with landline phones which meant a certain age demographic. They do need to change and validate their methodologies if their goal is to match what the public actually thinks.
Think of it like this: if you run a polling company and your numbers don’t even come close to what actually ends up happening when the election comes around, that’s bad for business. You want to be accurate and you want your methodology to be good, unless you’re deliberately trying to fool people. An exception is if you are consistently off by a certain amount.
Imagine you never match the real outcome but you are always 5% off in the same direction every single time, then there is some value to your polls. They just need to be adjusted which is what people like Nate Silver do. They grade polls on accuracy but polls like I described are still used, just weighted accordingly.
Not Push Polls though, those are always unethical.
Latest Answers