I listened intently to a couple of pollsters on local NPR affiliate WCPN’s Science Cafe yestereday, as they discussed election polling. As insightful as it was, it also did not discuss adequately several issues that I find to be much more insightfully highlighted on Five-ThirtyEight.Com. Run by a founder of baseball prospectus, number geek/statistician, Nate Silver does an exemplary job of describing in plain language a variety of issues regarding polling. Most notably, he persistently challenges the so-calle “Bradley Effect”–in which white voters say one things about their voting preferences to pollsters and head into the polls and vote against the black candidate.  See, the most recent post on this: FiveThirtyEight.com: Electoral Projections Done Right: The Persistent Myth of the Bradley Effect.

FiveThirtyEight also has some really interesting discussions of likely votes and registered voters, and how they youth vote figures into this equation. Short version: Anyone creating a poll with “likely voters” does not count the youth vote because the under-30 crowd turns out to vote in such low numbers. However, evidence suggests that this election might be different–who knows. Even a modest turnout among the “kids”–who will tend strongly toward Obama–could really tip the election. We can only hope, but in either case we should be skeptical of those “likely voter” models.