I agree with your point. “People don’t know what they don’t know.”
Only from reading, not from any personal experience, I have learned the following: (to me it has the ring of truth)
1. Back in pre-1990 a pollster could get a 50% response rate to a phone survey. So, to get a sample of 800 voters, they could call 1600 to 1800. Obviously quite doable.
2. The make-up of those 800 answering the phone (again in 1990) cut across the demographic lines ... Republican, democrat, independent, men, women, young, old, Black, Hispanic, wealthy, poor, etc.
3. So a pollster had a lot of data to work with. He could ascribe a certain number to this group, some others to this group and he had a pretty comprehensive swath of the voting public.
4. Today, everything is different. With cell phones, caller ID, and much higher suspicion about the polls, and pollsters, I have read the response rate is 2%. This means to get that same sample of 800 respondents, you must call 40-45,000 phones.
5. And perhaps even more important, the leading demographic that answers these calls is much narrower. Namely, young women aged 24 and under.
6. So now, even as an honest pollster, project that narrow demographic across the spectrum of all voters and try to prognosticate election day results.
I would say it is nearly impossible which is why the polls are so flawed, fake, and bogus.
I agree with your point —it is really hard to get an accurate poll.
There are companies that do it for private customers, but it costs a lot of money, a very high level of professionalism and objectivity, and requires a lot of validation questions to make sure the sample is a fair sample.
There are some folks here who have worked with these companies—and they have great posts from time to time.
When they post, I pay _close_ attention.