No one asked me. Who conducted the poll?
Basic statistics. To get a representative sample of a group you don't ask the entire group. You don't even need to ask that large a number of people of a group. As long as they were randomly polled from multiple locations you wouldn't likely need to ask more than a few thousand people out of the entire population to get a representative sample. The margin of that method drops mighty quickly as the sample size inches upwards.
During the last presidential election, due to the incredibly long lines, I even hard some people suggest that we use this method for voting. A random sample of people would be selected to vote and the outcome would depend on the sampling. While I don't like that idea at all (it's just too open to abuse, and even if it were perfect, something just wouldn't feel right about everybody not having a vote), but the fact remains: you take a fair small, random fraction of a population and results within that fraction will typically mirror that of the whole.