***POLL**** Marquette National Poll (A+): Harris 53 Trump 47

[Bucky]

8. not exactly

After a certain population size, increasing the sampling size won't increase your accuracy or confidence rate (repeatability of the results). Weirdly enough, if you get too big a sampling, it can actually reduce the accuracy of the sampling because retrieving data (survey responses) may introduce new accessibility variables that distort the random selection process.

You can play with this website to see how big a sample you'd need for different population sizes. The last election had just under 160 million voters. This next one may see 180mil. That difference doesn't matter. At any national population over 5 million, you need 1068 respondents to get a margin of error of 3%. If you want a 2% MOE for the next election, you have to get 2400 surveys done. That would cost more than 3 times the amount of a standard 3% MOE survey.

No one will pay for that. Marquette Law School's poll only interviewed 879 registered voters and had a MOE of over 4%