I’ve noticed most U.S. polls query around 1,000 people and sometimes even less. Somehow that qualifies for headlines like “Americans say…” or “Most Americans…” How is it acceptable that 0.0002% of the population is accepted as representative?
In: Other
Because although people like to think they are very special and unique, we aren’t. *If sampled correctly,* 1000 subjects is more than adequate for *most surveys, depending on the outcome measure.*
We can design such surveys so that they are “representative” by recruiting subjects from different demographics and areas.
They’re only interacting with a certain type of person in the first place, someone that would actually take the poll. Then there’s also the method the poll was given in that will further diminish the people that will actually participate.
**ELI5:** It’s like guessing what’s in a giant jar of jellybeans by looking at just a small handful. If you pick your handful carefully to get all the different colors in there, you can make a pretty good guess about all the jellybeans in the jar. When polls ask questions to 1,000 people from all over the place and with different backgrounds, it’s like getting a handful that tells us what millions of people might think.
**Adult Answer:** Surveys with about 1,000 respondents can accurately reflect the views of the entire U.S. population due to strategic sampling and statistical principles. By choosing a sample that represents the population’s diversity (age, race, gender, geography), researchers can extrapolate the findings to the broader public. This method is supported by the central limit theorem, which indicates that the average of sample estimates will approximate the population average as the sample size increases, making the survey results a reliable microcosm of national opinion. This approach is scientifically validated and includes a margin of error to account for variability, ensuring the conclusions drawn from these samples are statistically sound.
A single person is dynamic and hard to predict. People in large groups on the other hand are pretty damn predictable. You don’t need to ask many to get a good idea of trends.
I’ve been citing this piece for years: [How can a poll of only 1,004 Americans represent 260 million people with only a 3 percent margin of error?](https://www.scientificamerican.com/article/howcan-a-poll-of-only-100/)
“The margin of error depends inversely on the square root of the sample size. That is, a sample of 250 will give you a 6 percent margin of error and a sample size of 100 will give you a 10 percent margin of error.” It doesn’t matter how many people the survey represents, because as long as the sample is truly random/representative (as other commenters have explained), the *percentages* will stays the same.
One bag of M&Ms will (within margin of error) have the same percentage of blues as a million bags of M&Ms all dumped into a bowl together.
“we surveyed 980 people who have a land line or answered a call from an unknown caller on their cell. Here’s how 350,000,000 people will vote in November” – Media.
Polling – or any attempt at broad general information gathering – is a function of math and statistics. If the audience polled is a fair estimate (through extrapolation) of the population at large then you should get a reliable result.
Something that you may not be accounting for is margin or error. Most statistical analyses state a margin of error, which accounts for any/most/some issues that a small field might throw off.
If you see a poll that states, for example, “margin of error +/- 3-5%” then that’s a lot more reliable than a poll that states “margin of error +/- 10-15%” .
Adding to the other answers, let’s see an example where having a lot of people answer your poll won’t give you an accurate prediction *because* it wasn’t representative of the population. I’ll talk about the 1936 US presidential election, and how a publication called The Literary Digest conducted one of the biggest polls ever (if not the biggest).
This publication managed to get more than 2 million answers, for an election where 80 million people were eligible to vote. They predicted that Alfred Landon would comfortably win against Roosevelt, but in the end the margin of victory was the other way around.
The main reason why their result was so off, was that they polled their own subscribers, plus people on two public lists: automobile owners and telephone users. All three lists are not representative of the population: it’s 1936, and the only people on those lists are the ones that have enough disposable income to keep buying the magazine, have a car, or own a telephone.
If you flip a coin 30 times and see that the coin is heads 49% of the time, do you expect doing another 970 flips will change your mind that the coin is balanced?
Because the findings don’t typically change too much if you poll more than that. So a survey of 1000 people that shows support for some issue at 55% has as much validity as a survey of 10,000 that shows support for the issue at about the same percentage.
Basically, you can get defendable data without having to waste time with more surveys.
Latest Answers