Just read to the end of the article to see the punch line: “This poll was conducted among a random sample of 671 adults nationwide.” Very funny! Not likely voters, not even registered voters, but adults. This is the same flaw that was present in the recent ABC poll that showed Hillary 9 points ahead. I do not understand why such polls are even conducted — even if it is cheaper, it is still a waste of money. The Gallup poll, which shows McCain narrowly ahead of both Democrats, sampled “likely voters.” So don’t panic.
There is no point in criticizing a poll for its results, but methodology is fair game. Even the use of “registered voters” instead of “likely voters” is highly suspect and tends to produce bad results. If you ask a small sample, as many as half of whom are non-voters, about what they think of a presidential race, you’ll get an answer the question of what non-voters think about an election in which they won’t be voting.
Moreover, “less likely” voters tend to talk about the candidates they are seeing on television or the ones they know the most about. Even some of the stricter polls that use bad screens to identify “likely voters” can suffer from this flaw.
Let me share just one polling story from 2006. Five months ahead of the primary election, polls suggested that former University of Nebraska head coach and Rep. Tom Osborne (R) was well on his way to becoming the state GOP’s gubernatorial nominee. After all, everyone knew who he was and liked him. Several months before the election, with public polls showing Osborne 12 points ahead, we wrote in the Evans-Novak Political Report:
Osborne should not be considered a strong favorite against popular incumbent Dave Heineman (R)…[W]e believe that Heineman actually begins as the slight favorite.
Matt Braynard, who was working for the third man in the race, had conducted a private poll of the state using an exacting and more accurate method of pre-screening likely voters, and Heineman was in the lead. A week before Election Day, he did it again and we repeated our prediction of Osborne’s doom despite surveys to the contrary, one of which was a useless poll of “registered voters.”
Four out of four Nebraska governors, including Gov. Heineman, agree that accurate polls are more useful than inaccurate ones.