They don’t survey 10000 people in one town. They try to get a randomly chosen sample of 10000 people, or even 1000 people, across the entire United States.
If the entire United States is 50% men and 50% women, then a randomly chosen sample of 10000 will likely contain no more than 5100 men and no more than 5100 women. A sample of 1000 will likely contain no more than 530 men and no more than 530 women.
Now replace “men” and “women” with “Democrat” and “Republican”, or any other demographic. That’s how you end up with a group of people that reasonably represents the entire United States.
They don’t survey 10000 people in one town. They try to get a randomly chosen sample of 10000 people, or even 1000 people, across the entire United States.
If the entire United States is 50% men and 50% women, then a randomly chosen sample of 10000 will likely contain no more than 5100 men and no more than 5100 women. A sample of 1000 will likely contain no more than 530 men and no more than 530 women.
Now replace “men” and “women” with “Democrat” and “Republican”, or any other demographic. That’s how you end up with a group of people that reasonably represents the entire United States.