Besides the fact that talking to a few hundred (or even thousand) knuckleheads can never be a precisely accurate representation of the beliefs of a nation of hundreds of millions, something called "biased sampling" on the part of the pollsters can further skew the results away from reality. Let's say you want to know who "the people" consider the greatest singer of all time. And because you're lazy and live next door to a nursing home, the only people you poll are on the closer-to-death side of 75 years old. That's biased sampling. Also, good luck quelling the riots between the Johnny Mathis Mafia and the Tony Bennett Gang.
The "fierce debate" we mentioned earlier, over whether we should still be incorporating the margin of error into poll discussions at all, was sparked by the recent advent of online polling. Because what could be more random than asking people on the internet? Surely, surveys taken online are going to result in the most accurate outcomes possible, one argument states, so therefore the whole margin of error concept has been rendered practically moot. To which we can only make a simple counterpoint: Mountain Dew once asked people online to name one of the their new flavors, and the winner -- by a landslide -- was "Hitler Did Nothing Wrong."
Hillary Clinton could have displayed her trust in the the average American voter
if she had used this system to name her recently-born granddaughter.