May, 2010:

In response to my comments about the UT/Texas Trib poll, pollster James Henson was kind enough to send me the following:

The number of independents is a function of how we present the results, not the methodology per se. We ask our respondents to provide party id, then follow up with independents to ask how they lean. As I’m sure you know, most self-identified leaners exhibit behavior that in many cases is actually more similar to strong party identifiers — a tendency that is fairly evident in the cross tabs of this and other polls. As a result, we isolate what people tend to think of as “true independents” and group the leaners with their more likely tendencies in these breakdowns. There is of course bound to be some error in this, but we think that it’s a lot less than if one just keeps the leaners in the category. If you want to test this yourself, we will of course be posting the original and complete dataset later in the week, so you can conduct your own analysis. (For now, you can extrapolate the breakdown of the 7-point scale in the partial crosstabs posted at the Trib site.)

I appreciate the clarification, and I look forward to seeing the complete dataset. Henson also sent me this detailed explanation of the methodology random matched sampling method, which was written by Doug Rivers. I need to spend a little time with it, but there it is for you. You can also listen to Henson’s colleague Daron Shaw explain the methodology.

Henson also noted that he has only done this style of poll since July of 2008 – you can see all of those results here. I was thinking of Lyceum polls from earlier than that, which Henson did with Shaw, but which were phone polls. The results of interest to me from 2008 are the July (McCain 43, Obama 33) and October (McCain 49, Obama 38) polls, which were a pretty accurate reflection of the final spread.

I think that addresses all of my questions. My thanks to James Henson for the information.

UPDATE: On a tangential note, Phillip Martin has put together a handy dandy chart showing all public polling data on the Perry/White race since February. If you take the average, Perry has a 47-40 lead.

UPDATE: Here’s Dr. Murray’s take on the poll.

We finally have a non-Rasmussen result to discuss.

Republican Rick Perry leads Democrat Bill White by nine percentage points — 44 percent to 35 percent — in the 2010 race for governor, according to the latest University of Texas/Texas Tribune poll.

Those numbers are identical to the results of a “fantasy race” between White and Perry in the February UT/TT poll, which was taken before Democratic and Republican primary voters selected the two men as their nominees. This time, 15 percent said they don’t know yet who they’ll vote for, and 7 percent said they’d prefer “someone else.” That last finding could be consequential: Libertarians will choose their candidate for governor at a state convention next month, and that candidate or a write-in or some combination of the two could give those “someone else” voters a place to go.

Perry’s stronger with Anglo voters, leading White 55 percent to 25 percent. White is stronger with African American voters (69 to 5) and with Hispanics (43 to 32). Perry is ahead among both men (48 to 34) and women (40 to 35).

You can see the toplines here, the crosstabs here, and the methodology here. There are two main points to make here. One is that this result, and to a somewhat lesser extent all of the downballot results, were a reflection of party ID as the poll showed it. Republicans went strongly for Perry, Democrats went strongly for White, though there were a few more undecideds on the Democratic side, giving him a little bit more room to grow. Independents went for White by a small margin, with “Someone else” finishing ahead of Perry and “Undecided” finishing ahead of both of them. That was a fairly small subsample, however, so don’t read too much into it. The bad news for White here is that Perry’s basic strategy is based on there being more Republicans, and doing well enough with them to not have to worry about anyone else. This poll will not give Perry any reason to doubt that strategy. White doesn’t need to chip into that support too much to make the race competitive, but he’ll need more than what he’s getting here.

The other point is that this was not a phone poll. From the methodology description:

The PollingPoint panel, a proprietary opt-in survey panel, is comprised of 1.6 million U.S. residents who have agreed to participate in YouGov Polimetrix’s Web surveys. At any given time, YouGov Polimetrix maintains a minimum of five recruitment campaigns based on salient current events.

Panel members are recruited by a number of methods and on a variety of topics to help ensure diversity in the panel population. Recruiting methods include Web advertising campaigns (public surveys), permission-based email campaigns, partner sponsored solicitations, telephone-to-Web recruitment (RDD based sampling), and mail-to-Web recruitment (Voter Registration Based Sampling).

The UT Politics Project did some of these polls in 2008 and I believe in 2006. I’ll need to go back and try to find some of their prior results, which I recall being more or less reasonable, though I don’t recall if they did any polls after about June of 2008. I suspect their methodology is the reason for the very low number of self-identified Independents in the sample, about 10% of the whole. Because it’s not a phone poll, it sort of stands apart, and isn’t quite a direct comparison to Rasmussen. Burka has been skeptical of this kind of methodology, though he didn’t seem to notice it this time around. It’s a data point, and I’ve been waiting for more data points, so there you have it.

UPDATE: I should clarify that the main objection to polls like this one is that their sample is not really random. It’s necessarily a sample of people who are more engaged, and possibly more partisan. The methodology here is worlds better than the crazy and ultimately useless Zogby Interactive polls from 2006, but it’s still not widely accepted. Of course, a non-Presidential year will have a much higher proportion of such voters than a Presidential year does, and robocall polls were not widely accepted once upon a time, too. Ultimately, their track record will determine whether or not polls like this are taken seriously. Evan Smith emails to remind me that the UT/TT polls from the Republican gubernatorial primary were pretty accurate – they had Perry over KBH by a 45-21 margin in February, for example; their predicted spread was closer to the final 21-point difference than what Rasmussen had at that time. I’m still going to try to find older UT poll results, but they did do a good job with the primary.