[piva00]

Do you know how to calculate sample sizes? For a population around 300m people you can survey some 7500 (if those are representative of the overall population, not just a homogeneous cohort of the population) and get a confidence level of 99% with 1.5% margin of error.

I recommend studying a little of basic statistics.

[blue_light_man]

Does this also balances out the fact that these are people who choose to respond to their survey and not random sample of all Americans? Are you saying the list of people who choose to respond to their survey is an accurate sampling of the entire American population?

Also why not say we surveyed 6000 people in our list instead and 30% of them have this view? That is more true than saying 30% of Americans have this view which is quite a big jump. There is ofcourse some lying going on here. What are they trying to achieve by lying here is my question.

[defrost]

It's good to have questions. Here is some useful background reading:

https://www.pewresearch.org/our-methods/

https://www.pewresearch.org/topic/methodological-research/su...

https://www.pewresearch.org/topic/methodological-research/

There are literal books devoted to survey methodology.

[froh]

No. this is not how statistics work.

survey statistics work by computing the margin of error for a specific sample size.

for example, if you toss a coin, say 10times and it shows heads 9 out of ten times, would you say "oh this was completely coincidental, you have to throew it more times before you can know!" or would you already say the coin is biased after ten tosses?

And the key thing is: there is math for that. "How probable is it that this result that I see is representative for all the possible coin tosses that I could ever make with this coin?" or, in our case here "How probable is it that the result of _these_ 6000 people here is close enough to what I'd get if I ask all Americans with a College degree?"

and let me assure you, a sample size of 6000, properly controlled for biases, gives a very representative result.

If you'd cry foul for 30 respondents, I'm with you.

but for 6000? and a professional research institute which klnows how to do their homework? nah.

If you doubt what I say here, feel free to brush up your statistics (udemy, udacity, ...) and then review the Pew methodology. I'll be curious about your findings.

[piva00]

> Also why not say we surveyed 6000 people in our list instead and 30% of them have this view? That is more true than saying 30% of Americans have this view which is quite a big jump. There is ofcourse some lying going on here. What are they trying to achieve by lying here is my question.

Because their methodology is designed to work through sampling, and sampling can determine results with very high confidence (95-99%) from a quite small sample size of the overall population, they don't need to qualify further given Pew has a trustworthy methodology for sampling.

It's not a big jump, it's how sampling works. I really recommend you read about it, it might be very helpful to understand and stop thinking "they're lying!", they are not.

It's basic statistics, you are trying to argue against very established mathematics.

[ohmyiv]

It seems you might have a misunderstanding on how surveys and statistics work.

You can learn more about surveys and statistics through the following links:

Understanding and Evaluating Survey Research:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4601897/

Fundamentals of Survey Research Methodology:

https://www.mitre.org/sites/default/files/pdf/05_0638.pdf

What is survey sampling: Understanding methodology and sampling techniques:

https://blog.surveyplanet.com/what-is-survey-sampling-unders...

OXFAM: Understanding survey sampling:

https://oxfamilibrary.openrepository.com/bitstream/handle/10...

Some general knowledge on surveys:

https://en.wikipedia.org/wiki/Survey_methodology

Statistics learning course on Kahn Academy:

https://www.khanacademy.org/math/statistics-probability

These links will hopefully help you understand surveys and statistics more. If you need more links, let me know. I can find a lot more courses on these topics if you need.

[Brybry]

If you think they're lying and don't trust their methodology: there are other polls that show a similar trend in sentiment [1][2]

I couldn't find any polls that showed the opposite.

[1] https://www.wsj.com/articles/americans-are-losing-faith-in-c... https://archive.md/1SrWr

[2] https://news.gallup.com/poll/508352/americans-confidence-hig...