Occasionally, there appear to be a few misconceptions about the methodology of Polimetrix/YouGov. In order to improve upon some of the well-known issues with traditional phone surveys, they have developed a methodology by which you can identify and interview a representative sample online (see appended information for details).
But one of the keys to this approach is to have a very large pool of people from which you can "match." Polimetrix/YouGov currently has well over a million people in their online pool. To get a large number of people with diverse age, income, educational, ethnic, and racial backgrounds, they have solicited people for their pool in a variety of ways. These efforts are all about improving the ability to achieve quality "matches" in a given sample rather than relying on a "shallow" pool with little diversity among harder to reach sub-populations.
We remain confident that the matching approach produces accurate estimates of the underlying distribution of preferences in the population.
Also, to clear up some apparent confusion: Hank Gilbert was included on the survey. It is the case that the initial survey did not include a response option for Gilbert on the Democratic Primary vote choice item. Upon learning of this error, we went back into the field with a "correct" set of response options and re-interviewed those respondents who said they intended to vote in the Democratic Primary, and conducted a number of additional interviews to bolster our overall Democratic Primary sample. The Democratic Primary trial ballot we report here is thus a statistically accurate and reliable accurate estimate. More broadly, we hope that we have been true to our policy of being upfront about our mistakes and to publicly acknowledge and correct them. As always, the datasets from the survey will be available at the Texas Politics website and (now) the Tribune site in the next couple of days.
[Editor's note: Gilbert's campaign blasted the results in an early morning press release, noting our openness about the error but failing to note that the pollsters went back and made certain to include him in a sound poll. From spokesman Vince Liebowitz's press release: "I cannot believe I waited up until 4:40 in the morning for this poll. It is a complete and total joke. In addition to Hank Gilbert having been left off part of the poll-a fact the Tribune fails to mention but which is clearly mentioned in a footnote of the poll-the company conducting the poll is one that conducts unreliable internet-based polls. The results of this poll across the board are universally unreliable. Just looking at the alleged religious breakdown of polling participants should tell any Texan that."]
Texas Tribune donors or members may be quoted or mentioned in our stories, or may be the subject of them. For a complete list of contributors, click here.
- Rivers: Internet polling (203.5 KB) DOWNLOAD
- October 2009 Poll Methodology (151.1 KB) DOWNLOAD