Off the Kuff Rotating Header Image

The UT/TT primary polls were completely useless

Wrong!!!

I expressed my contempt with the UT/Texas Trib’s Democratic primary poll result for the US Senate race last night, which they richly deserved. Sure, pollster Jim Henson admitted that “the first person to raise some money and run some ads could really move this”, and that’s largely what happened, but that got lost in all the national attention that was paid to Kesha Rogers being proclaimed the frontrunner in a poll where basically nobody had an initial preference. They had a “result” that was guaranteed to get them a ton of attention, and that’s what they got even though their track record in past Democratic primaries was shaky at best.

Well, now it’s time to pay them a bit of negative attention, because their Republican primary polls, which I originally noted had a decent track record based on previous results sucked eggs, too. Let’s take them one at a time and assess the damage. I’ll even be generous and start with the one poll they basically nailed, just to give them credit where it’s due. Here’s the poll story from which I’ll be quoting:

U.S. Sen. John Cornyn, facing a field of seven other Republican primary candidates in his bid for re-election, won the support of 62 percent of the likely Republican primary voters, followed by U.S. Rep. Steve Stockman, R-Friendswood, who got 16 percent. Support for the rest was in single digits: Linda Vega, 7 percent; Dwayne Stovall and Ken Cope, 4 percent each; Reid Reasor and Chris Mapp, 3 percent each; and Curt Cleaver, 1 percent.

Actual result: Cornyn won with 59.44%, Stockman came in second with 19.13%. Dwayne Stovall was actually in third with 10.71%, but I won’t crime them for that. From here, it’s all downhill.

In the heated Republican primary for lieutenant governor, incumbent David Dewhurst leads the pack with 37 percent of likely Republican primary voters at his side, followed by state Sen. Dan Patrick, R-Houston, at 31 percent; Agriculture Commissioner Todd Staples at 17 percent; and Land Commissioner Jerry Patterson at 15 percent.

Actual result: Dan Patrick led the pack with 41.45%, followed by incumbent David Dewhurst with 28.31%. Staples had 17.76% and Patterson 12.47%, not that it mattered. That’s a pretty big miss, but it’s not their biggest.

The Republican primary for attorney general is a statistical dead heat between state Rep. Dan Branch of Dallas, at 42 percent, and state Sen. Ken Paxton of McKinney, at 38 percent — a difference smaller than the poll’s margin of error. Railroad Commissioner Barry Smitherman got 20 percent. When they were initially asked about the race, 47 percent expressed no preference between the candidates.

Actual result: Paxton 44.44%, Branch 33.49%, Smitherman 22.06%. They did get Smitherman’s level of support correct, but they had the wrong frontrunner and the race wasn’t as close as they said. Oh, well.

In the race for comptroller, that group of initially undecided voters accounted for 54 percent, perhaps an indication of continuing flux in the race. Debra Medina, the only candidate who has been on a statewide ballot (she ran for governor in 2010), got 39 percent after voters were asked whom they would support in an election now, followed by state Rep. Harvey Hilderbran, R-Kerrville, at 26 percent; state Sen. Glenn Hegar, R-Katy, at 24 percent; and former state Rep. Raul Torres, R-Corpus Christi, at 11 percent.

Actual result: Hegar came thisclose to winning outright, with 49.99%. He was 151 votes short of a majority with four precincts still uncounted. Hilderbran was second with 26.01%, Medina third with 19.30%, and Torres last with 4.68%. I’m sorry, but that’s just embarrassingly inaccurate.

So in all three downballot Republican races as well as the Democratic Senate race, they incorrectly identified the frontrunner, with the extra indignity of having the almost clear winner of the Comptroller’s race not in the cut for a runoff. Well done, fellas. Well done.

Now you may say “c’mon, polling primaries is especially tricky”, and if you did I would agree. I’d also say that maybe their self-selected-sample-plus-secret-sauce methodology is especially poorly designed for polling in these specialized races, and I’d point to these very results as proof of that. You may also say that no one else was providing poll information on these races so at least they were telling us something, and I’d say we would have been better off with no information than we were with their badly wrong information. I’d also say they owe us an explanation for why they were so wrong, and a public examination and reconsideration of their methods given how badly wrong they were. If they can screw these races up so badly, why should anyone believe their general election polling? The ball’s in your court, guys.

I should note that I’m saying all this as someone who likes the Tribune and who thinks they generally do a good job. On this, however, they did a terrible job, and I’m not the only one who noticed. They should be embarrassed by this, and they should want to figure out where they went so far off track. I would advise them to be quick about it. Steve Singiser has more.

Related Posts:

One Comment

  1. Jim Fung says:

    Kesha Rogers probably was ahead before Alameel started running TV ads and mailing. That poll did a service because it woke people up. She might well have been the “Jim Hogan” if people had been complacent. The poll also correctly showed how weak Maxey Scherr was doing and how she was not making much of an impact.

    Also, even respected polling outfits like the Field Poll in California do poorly at polling downballot primary races. Those races are all about name recognition until the final weeks. The polling does give somewhat of an idea of the frontrunners with the most name recognition or the best sounding names, and you just take the results with a grain of salt but know what to fear like from a Kesha Rogers. Most of those polls did at least identify the runoff contenders, so they did better than random. I don’t think they need to self-flagellate and revisit their methodology. No amount of refining will make downballot primary polls accurate unless they are done a few days before the election and by then it’s too late for them to have the “watch out for ____” effect so I’m glad they were done earlier.

Bookmark and Share