It’s easy enough to get more than 1,500 people to answer your survey. But every pollster is plagued by the fact that people who voluntarily answer surveys are usually not representative of the broader population on the relevant issue.

Basically any method for sampling the population introduces bias. The sample of people who pick the phone, who respond to texts or emails, who come to the door for a stranger, are all subpopulations with their own biases. A pollster’s ability to correct for these biases is what separates the good from the bad.