Is This New Texas Senate Poll Accurate Or An Outlier ?

This chart reflects the results of the newest Quinnipiac University Poll on the U.S. Senate race in Texas. They questioned 807 likely voters in Texas between September 11th and 17th, and the poll has a margin of error of 4.1 points.

The poll has some Democrats upset because it shows Cruz with a lead of 9 points -- significantly more than any of the other polls have shown. In fact, Quinnipiac's previous polls on the race showed Cruz with leads of 3 and 6 points. What happened? Is Cruz extending his lead?

No. The difference is in the sample. Quinnipiac's previous polls questioned all registered voters. This latest poll questioned only those they consider to be "likely" voters. Likely voter samples are based on voting patterns in similar elections in the past (in this case it was probably the midterm elections of 2014 and 2010). I'm not at all sure basing likely voters on those elections is very accurate. Neither of those elections had an extremely unpopular Republican in the White House.

Note in the chart above that both Democrats (94% to 4%) and Independents (51% to 47%) preferred Beto O'Rourke. Only Republicans (6% to 94%) preferred Ted Cruz. That means to give Cruz a 9 point advantage, the sample had to contain significantly more Republicans than Democrats and Independents combined.

I think this poll is an outlier (even though I normally respect the Quinnipiac University Poll). I think this is an unusual year, and basing likely voters on past year voting patterns is a mistake -- at least for this election.

So, Democrats and Independent O'Rourke supporters, don't get depressed or give up. This is a much closer election than 9 points. I still think O'Rourke has a reasonable chance to win if the turnout is large. And even if Cruz does win, it won't be by as much as 9 points.