Busy day today, but I couldn’t let go without comment this new poll by Missouri State University for KY3 of the Missouri Senate race.

Even if its findings weren’t entirely out of step with the rest of the polling world, the details of the poll carry a number of warnings that it’s not very accurate.

First off: the key top line: Republican Roy Blunt 48.8, Democrat Robin Carnahan 48.4, MoE 3.5. Yeah, a virtual tie when everyone else shows Blunt way ahead. But I’ll tell you why this poll is like a caricature of the typical university poll and all its problems:

• The poll was conducted over 16 days, from the 7th to the 22nd. It’s hard to say that this is even a sampling of the same voter pool when it was taken over that long of a period. The only good thing is they at least waited until after the primaries to end officially, but 16 days for 785 registered voters? What, did they have two interns each calling 25 people a day on work study in the evenings?
• The poll is of registered voters, and then weighted in order match certain census statistics. So not only is it the wrong kind of poll, but it’s not even a truly random sample, but rather one that’s tweaked in order to look right according to these researchers.
• The polling is heavily weighted toward women. 63.2% they report are women. Sure, it’s true that there are more women than men who vote, but 63-37 is a pretty bad result when according to CNN reporting of 2008 exit polls the voter split in Missouri is 54% women/46% men. I can only imagine how much the numbers were distorted to bring that back in line with the Census.
• There is no record that the orders of the candidates were rotated randomly as any serious poll does. Instead the questions just list the Democrat first every time. Ballot order matters, and this order skewed toward the Democrats in every race.

I can confidently say now that I will give this poll no weight in my next Senate projection update. It has a number of fatal flaws and I cannot take it seriously.

From Unlikely Voter