I’ve been so immersed in the Houston elections that I forgot to give a warm welcome to the Texas Tribune, which made its debut on Tuesday. I really like the look of the site, I like their lineup of writers, and I like what they’re aiming to do. Once their RSS feeds become available, I’ll be really happy. So welcome aboard, y’all. I look forward to seeing what you can do.
One thing in particular I’m interested in is their polling center, which assimilated the Texas Politics Project. Their first effort has drawn some criticism, much of which boils down to what Paul Burka says:
I will tell you up front that I do not know enough about statistics to know whether [their methodology] is reliable or not. I do know enough to know that this methodology is not truly random, because everybody who signed up has manifested enough interest in politics to want to be surveyed.
[…]
Internet polling is probably the future of polling, and the UT/Tribune poll is our best hope for a regular flow of campaign information, so I’m going to have to get used to it. But my confidence level is not very high.
The good news is that the more of this they do, the more of a track record they’ll build by which we can judge them. Remember that SurveyUSA, whose absence in Texas Burka rightly laments, was once viewed skeptically because it used an automated interactive script to get people to push buttons on their phone in response to questions instead of talking to a live person. If the TxP polls prove to be as accurate as SUSA has been, we’ll look back at this some day and wonder what we were afraid of.
But that’s a few years, or at least a few dozen polls, away. What we have to go on now is their October 2008 polls of the Presidential and Senate races. The results they got – McCain 51, Obama 40, Barr 1; and Cornyn 45, Noriega 36, Schick 5 – aren’t bad; they did come pretty close to the actual margin of victory in each race. Here’s what I wrote at the time.
I’ll note that whatever else one may think, the results are in line with most other recent polls, the last Rasmussen Senate poll being an exception. The (too) high number of undecideds skews things a bit – in particular, for the one bit of sample breakdown that we do get, the poll claims 16% of black respondents and 17% of Hispanics are undecided in the Senate race. I can just about guarantee you that a large majority of each will ultimately cast their ballots for Rick Noriega. On the flip side, I think the five percent showing for Libertarian Yvonne Schick is too high – I believe she’ll ultimately get two to three percent, with the rest mostly going back to Cornyn.
In case you’re curious, Yvonne Schick ultimately got 2.34% of the vote. Sometimes these predictions are easy to make.The point is that after they’ve polled the gubernatorial primaries in February, the general election races in October, and the (special?) Senate election whenever, we’ll have a much better idea if we’re dealing with reliable data or for-entertainment-purposes-only stuff. My advice is to poll as many races as you can, close to the election whenever possible, and let the chips fall where they may.