Here’s an example. Believe it or not I’m a registered Democrat, Wife’s a Republican ( she said she could marry a democrat but could never be friends with one...lol. Line taken from Garrison Keillor). We’re a house divided. Every night I get at least one, if not several, opinion polls, via text message, asked to participate ( which I never reply) Wife nada. Why’s that????
So sample size determines margin of error. A larger sample of randomized voters mean a lower margin of error. Weighting is a different story and that comes down to the pollsters best guess of what the electorate will look like. Basic categories like democrat republicans and independents but the sub categories like race and gender and then third level like education and income. Now for the sake of simplicity, let’s say a pollster expects an electorate of 50% dem and 50% republican. He grabs a list of 200 likely voters 50/50, and randomly calls the list and gets in contact with 60 democrats and 40 republicans. The pollster would then have to “weight” the dem respondents lower and Republican respondents higher. So a dem response might be assigned a weight of 0.9 whereas a Republican respondent would get something like a weight of 1.1. Then you multiply those weights to the response total of each and now you have your weight. But high samples are a good thing just understand that if more democrats are reached than the pollster expects to be in the electorate then their responses have less value individually than the republicans.
There also is a thing called Shy Trump Voter Syndrome. Trump will win this election in an electoral college landslide.
So instead of calling it over sampling. lets call it overweighting. The polls which gave biden larger leads Over weight Dems.
As the saying goes: "no one will ever win the war between the sexes because there is too much fraternizing with the enemy."
If that’s the case, so should my wife. I’m being continually chosen because pollers assume my responses will be liberal in nature. My wife is avoided in the opposite context.