WILL WE HAVE TO SHOW A COLLEGE TRANSCRIPT TO VOTE?

There's something strange about the new Channel 3 poll on the Cleveland Mayor's race. It's not in the headline results; there's nothing self-evidently strange about Jackson 30%, Campbell 23%, Draper 15%. Debatable, yes, but not strange.

But if you look at the "crosstabs" that break these results out by age, ethnicity, ideology, etc., the strangeness appears. The 456 "likely voters" in the sample are a strikingly well-educated bunch of Clevelanders.

In the 2000 Census, only 35% of Cleveland residents age 18 and over had ever gone to college, and only 15% had degrees of any kind. Fully two-thirds of the city's voting-age adults said they'd never been inside a college classroom.

In the Channel 3 sample, however, college graduates are 32% of the "likely voters", and people with some college bring the postsecondary share of the sample to 59%. People who've gone to graduate school are 15% of the sample -- almost five time their share of the adult population!

Is it possible that this reflects the actual turnout a week from Tuesday? I suppose, but it's not very likely. The college-educated share of Cleveland's population has probably gone up slightly in the five years since the Census, and it's generally accepted that better-educated people are more likely to vote.

But this is a big skewing of the sample compared to the city's actual demographics -- a full 25% shift of the assumed likely voter base toward college-educated people, in a non-college-educated town.

At the very least it calls for some explanation by SurveyUSA, the pollsters.

Update: democracy guy sees some additional problems with the poll.