Flawed methodology misleads poll watchers.


Last week, the Texas Tribune —in conjunction with the University— released its second comprehensive poll of the 2014 election cycle. The results paint an unflattering portrait for the Democrats, with state Sen. Wendy Davis, D-Fort Worth, trailing Attorney General Greg Abbott in the race for governor by 12 points and state Sen. Leticia Van de Putte, D-San Antonio, behind 15 points against state Sen. Dan Patrick, R-Harris County, in the contest for lieutenant governor. Races up and down the ballot from attorney general to railroad commissioner also met similar fates.

However, upon closer examination, the poll does not hold water. Like its counterpart from February, the polling methodology consists of an opt-in survey, supposedly "weighted" for population groups not properly sampled. Furthermore, this month's poll did not even pretend to correct for the gaping geographic inconsistencies. Almost 1/3 of those polled reported being from the Austin area, despite the region holding barely 7 percent of the state's population. And rest assured, my concerns over this poll are not just theoretical. In the March primary, following the February poll, most all of the Tribune's primary predictions proved to be completely worthless.

Like I wrote in this paper last March, serious politicians should not take these polls seriously. Instead, we should wait for a reputable polling firm, one actually worth its weight in paper such as Public Policy Polling or Rasmussen, to gauge the temperature of the race. Davis and Van de Putte being the odds-on underdogs is still likely, but at least this way we can actually shine a light on the state of the race rather than shooting in the dark.

Horwitz is an associate editor.