I wish Nate Silver would do seminars teaching journalists how to report polls beyond just cutting and pasting the results.

Today, the Columbus Dispatch reported on a poll reported in the D.C. political paper, The Hill, which reportedly shows bank lobbyist Steve Stivers sitting on a nine-point lead over freshman Democratic Congresswoman Mary Jo Kilroy.  Incidentially, the poll was funded by the American Natural Gas Alliance.  No, I don’t what to make of that, either.

But I know that the poll has a fundamental flaw.  Like most of the polling we’ve seen in Ohio this year, it failed to include third-party candidates.  In this particular instance, that could be enough to cut Stivers’ lead entirely.

One of the most underreported facts about the 2008 race was the entire results:


Everyone was so preoccupied with reporting the vote margin, that most have simply overlooked that the two third-party candidates, both of whom ran to the right of Stivers, took…

roughly 9% of the vote from Stivers.

In other words, when you look at 2008 results, if you assume that the votes for the Libertarian and pro-life/social conservative independent candidates went mostly to Stivers (which is not that unreasonable of an assumption since you’d be hard pressed to believed they’d vote for Kilroy in any significant numbers given the obvious ideological differences) what would have happened?

Stivers would have won in 2008 by the same margin shown in today’s Hill poll but for two other right wing candidates in the race.

Oh, and did I mention that this year the Libertarian and the Constitution Party both have candidates in this race running to the right of Stivers?

Are you starting to see why this Hill poll’s failure to include them might be a BFD?

Let’s not forget that back in July the Dispatch noted that the Libertarian candidate in the nearby 12th District race was polling between 5% to 10% in that race, too.

Just something to keep your eye on Election night.  A poll that leaves out half the candidates in a race which recent history has shown has affected the outcome is fundamentally flawed and cannot be believed.

Tagged with: