Quote (dxlightning @ Jun 18 2016 05:09pm)
Statistically significant polls take 30 000 people (or fewer) and apply results to 3,000,000 people. I've never once found that convincing in politics, my apartment complex is literally a 180 difference from the one across the street because we are largely populated by students abroad and they're rich white Caucasians. 46% is a joke statistic that means nothing if one building was sampled more heavily than another.
Well sometimes they use much smaller polls, 1000-1500 people to apply to national results. Some polling is very flawed, but it's not flawed in the way they select people. For the most part if the polling company is anything remotely capable (and most are, they make all their money on private contracts not political polling) their sample is very close to representative. Where political polls falter is they ask the wrong questions, or simply it's impossible to get a truly random sample because people don't really want to answer those questions so they can't be as picky as who they ask.