The new KWTV Wilson Research poll is out, and it shows that the Presidential race in Oklahoma has narrowed from 30 points to 12 in just one week, while in the Senate race, Carson's lead over Coburn remains at 2 points.

18 points is a lot of movement for one week and that got me to dig a little further. There's a question asking what political label the respondent would give to himself -- liberal, moderate, or conservative, and if liberal or conserative, whether very or somewhat. The previous two weeks, the liberal number was at 12%; this week it's at 17%. The previous two weeks, the conservative percentage was 48% then 50%; this week it's 44%. What's more, the very conservative number had been at 23% and 24%, then dropped this week to 18%. (Moderate numbers stayed about the same all three weeks -- 34, 33, 31.)

Either we had a seismic shift in Oklahoma politics, with the whole political spectrum shifting six points to the left, or more likely, this poll is an outlier. When you take a random sample, the results most of the time will be representative of the larger population, within the margin of error, but there's always a chance (about 1 in 20) that you will get unlucky and pick an utterly unrepresentative sample. If that is the case with this week's Wilson poll, and in fact the real presidential numbers are closer to last week's, it makes you wonder what the real Coburn-Carson numbers are, if a liberal-skewed sample has them nearly tied.