There is little difference between a poll of 2,858 people and 100,000 people
The poll of 2,858 has a margin of error of 1.83% while a poll of 100,000 has a margin of error of .31%.
The 97,000 extra respondents don’t make all that much of a difference.
“The poll of 2,858 has a margin of error of 1.83% while a poll of 100,000 has a margin of error of .31%.”
If someone lies to a pollster, it is not reflected in the “margin of error” numbers, it makes them much worse.
Many people lie to pollsters. Additionally, the act of getting a representative sample also is not reflected in the margin of error numbers.
Political bias is also not reflected in statistical error.
So, 100,000 is much better, assuming you know who people are, statistically speaking.
The margin of error should be considered a “minimum ideal error” for best case statistical samples, with the maximum error being some multiple (5x or 10x?) Of the statistical rate to account for many factors that amplify the ideal statistical error rate.
There is no cheap way to poll acurately anymore. A great deal of money and effort must be spent finding demographically relevant statistical samples. That’s hard, and most pollsters are lazy, so you cannot draw too many conclusions from most polls these days. Pollsters simply pull stuff out of their wazoo and call it science.
Easier to cherry-pick with a smaller sampling - depends on the integrity/agenda of the pollsters.
The margin of error has to do with the quality of your sampling — when all else is equal, and not all polls share the same margin of error.
Today it’s about profiling voters, and they’re just going to get better and better at it.
The difference is in the selection and the reporting. The online poll doesn’t throw out a portion of the replies on some idea of “fairness.”
“The 97,000 extra respondents don’t make all that much of a difference”
Silly wabbit, of course it does. It’s a difference of about 95,000.
You site the margins of error from supposed “crooked” polls and believe it is accurate.
The only reason for so called “scientific” polls is so that the polling company can contact as few people as possible, read expenses/money. That’s it!
A random poll of so few people could never be accurate. but a random (which this poll is not completely random) of a YUGE cross section of voters is more accurate because of the sheer numbers. The good, the bad, and the ugly all mixed together averages itself out.
A poll of 100,000 various voters over 1,000 cherry picked voters gets my vote any day, by a HUNDRED to ONE!