The Trib’s strange samples

Here’s the UT/Texas Trib poll result the Trib should have reported:

Q19. If the 2012 general election for U.S. president were held today and Mitt Romney was the Republican nominee, would you vote for [randomize] Barack Obama, Mitt Romney, someone else, or haven’t you thought enough about it to have an opinion?

1. Mitt Romney              46%
2. Barack Obama             38%
3. Someone else              9%
4. Haven’t thought about it 
enough to have an opinion    8%

That’s in line with other polling. It translates to a 55-45 Romney win in November, which I daresay would not surprise anybody. But that’s not the result that was reported. This is the result that was reported:

Likely Voters (n=511; MOE=+/- 4.34%)

1. Mitt Romney              55%
2. Barack Obama             35%
3. Someone else              7%
4. Haven’t thought about it
enough to have an opinion    3%

The poll data is here and the Trib story is here. As was the case with their previous poll, I’m at a loss to understand how they arrived at that second sample. This time they at least give an explanation for their screening methodology:

The UT/TT internet survey of 800 Texas voters was conducted May 7-13 and has a margin of error of +/- 3.46 percentage points. Questions asked only of Republican or Democratic voters have larger margins of error, as indicated. And “likely voters” were defined as those who indicated they were “somewhat” or “extremely” interested in politics and who voted in “every” or “almost every” election in recent years.

Eighty-six percent of their sample claimed to be “somewhat” (38%) or “extremely” (48%) interested in politics, while 67% of the sample said they voted in “every” (32%) or “almost every” (35%) election “in the past two or three years”, which is how the question was phrased. Given that so many more people vote in Presidential elections than in any other kind of election – turnout for the just-completed Austin Mayoral election was expected to be around ten percent; I’m pretty sure it will be considerably higher than that this November – it makes no sense to me to exclude the respondents who had voted in “about half” (13%) or “one or two” (11%) of the elections in the past two or three years. Those were non-Presidential year elections. Does anyone really believe those people would have voted in 2011 or 2010 but won’t vote this year? It beggars the imagination. I understand the reasons why pollsters want to construct a “likely voter” screen. I’m just saying that this one looks awfully restrictive to me. On a side note, unlike last time this poll did not ask people how they voted in the 2008 Presidential election. The February sample went for John McCain by a 46-39 margin. Steve Singiser has more.

Finally, I see that the Burnt Orange Report has finally done what I’ve been begging some pollster to do, which is to ask Republican voters if they’d vote for Rick Perry or Greg Abbott in a hypothetical 2014 cage match. Click over to see their answer. Since this is apparently my week for getting poll-related requests fulfilled, let me now implore BOR to do a Presidential general election survey so we can have a third data source besides UT/TT and PPP. Thanks.

Related Posts:

• What the Trib’s pollsters miss about their own polls
• What the future may hold for Wendy Davis
• We will have exit polls this year