NPR’s odd claim about Latino voting in Texas

NPR takes its turn with the obligatory “can Texas turn blue” story, and in doing some provides a number I had not heard before.

A new group called Battleground Texas is planning to spend tens of millions of dollars trying to turn Texas blue by 2020. It is made up of Obama campaign veterans, like former field director Jeremy Bird, with lots of experience targeting and turning out minority voters for Democrats. Bird wants to do the same thing for Texas.

“If you look at the 2012 electorate,” he says, “only 38 percent of all eligible Hispanics turned out to vote. Compare that to Florida, where that number is 62 percent. If 62 percent of Hispanic voters who are eligible to vote turn out and vote in Texas, it’s a battleground state.”

The math makes the Democrats’ strategy sound simple. President Obama lost Texas last year by 1.2 million votes. And, says Hinojosa, there are currently 3 million to 4 million eligible Hispanics in Texas who didn’t vote in 2012.

“That’s where our problem is,” Hinojosa says. “The moment that they get engaged and start voting, it’s all over but the crying for the Republican Party in Texas.”

Not so fast, says Steve Munisteri, the chairman of the Texas Republican Party. Munisteri says the Democrats’ plans for Texas are based on a string of bad assumptions.

“The first false premise is that Hispanics will vote overwhelmingly for Democrats. We know that’s not true,” he says. “Republican elected officials are rational enough, a long time ago, to recognize how important Hispanics are to our state, and [they] actively sought the vote from Hispanic communities, with great success.”

Republicans have been averaging about 35 percent of the Hispanic vote in Texas — just enough to stay competitive. Even Mitt Romney did better in Texas — getting 36 percent of the Hispanic vote there, compared with his dismal 27 percent among Hispanics nationwide.

Wait, Republicans have been averaging only 35% of the Hispanic vote in Texas? Has anyone informed Karl Rove and PolitiFact about this?

I actually think that number is a reasonable guess, more so than Rove’s “averaging 40% of the Hispanic vote” claim. But where in the world did they come up with 36% for Mitt Romney? No source is cited for this number, despite the fact that there was a pre-election poll of Latino voters in Texas in 2012, which put Romney’s support at 29%. It’s just thrown out there, as if of course everyone knows this.

If you don’t want to believe that poll, and you don’t have sufficient trust in the more general polls and their small sub-sample sizes, then you can do what I have done before and take a look at actual election results to see if the numbers you observe fit with your hypothesis. Here’s another look at the most heavily Latino State Rep districts in Texas:

Dist SSRV Romney Obama R Pct O Pct ============================================= 31 76.2 16,605 27,543 37.3 61.9 35 76.7 10,081 20,470 32.7 66.3 36 85.4 7,230 22,133 24.4 74.6 37 80.0 8,786 20,458 29.7 69.2 38 79.9 10,696 21,465 32.9 66.1 39 73.9 8,144 24,015 25.1 73.9 40 75.2 6,121 19,468 23.7 75.2 42 88.4 8,084 26,063 23.4 75.5 75 82.8 6,888 18,719 26.6 72.2 76 76.8 7,313 25,729 21.8 76.8 79 71.3 12,231 23,200 34.1 64.6 80 80.4 11,408 25,440 30.7 68.4 Tot 113,587 274,703 29.3 70.7

As before, all data can be found on each House members’ page. This time I included raw vote numbers so I could get a grand total at the end. Looks like that Latino Decisions poll, which pegged Romney at 29% among Texas Latino voters, might have been on to something, eh? Note that for the individual districts, the percentages are of the total vote, including third party candidates, whereas for the grand total it’s a straight R-versus-D comparison. Had I included third party numbers, the overall percentages for Romney and Obama would have been a bit lower.

As I’ve also said before, not all of the votes in these districts were cast by Latinos. Most of the ones that weren’t would have been cast by Anglos, and unless you believe that Anglo voters were especially pro-Obama in these districts, the implication of that is that the Latino numbers for Obama were almost surely even better, and thus even worse for Romney. This is a pretty huge sample, probably about 20 or 25% of all Latino votes in Texas once you factor out the Anglos from the numbers above. To get from here to a 36% performance among Latinos, the remaining ones would have had to vote for Romney at about a 40% clip, which would then give rise to the question why they were different from Latino voters in these districts. The Occam’s Razor answer says that 36% is just too high a number to be real. You tell me which is more likely.

Look, I could be wrong about all this. Maybe Latinos really did vote 36% for Romney in Texas. Maybe they voted 40%, as Mike Baselice would have you believe. But if you want me to believe, you have to address the following:

1. The fact that three out of four polls in September and October of 2012 that publicly released their data showed Latinos at 32-33% for Romney.
2. The Latino Decisions poll that pegged support at 29% in Texas.
3. Romney’s actual performance in these heavily Latino districts.

Work all that into your hypothesis, I still may not believe you, but at least I’ll listen. And maybe I won’t write yet another blog post that tries to prove you wrong. Show me the math, in other words. Is that so much to ask?

Anyway. NPR did a whole series of stories on Texas last week, and my complaints about this one notwithstanding, they’re all good stuff. Check ’em out.

Related Posts:

This entry was posted in Show Business for Ugly People and tagged , , , , , , , . Bookmark the permalink.

3 Responses to NPR’s odd claim about Latino voting in Texas

  1. That said, if NPR is wrong, it means BT’s even wronger to rely on Hispanic demographics to cause a major vote swing. GOTV, for both Hispanics and white suburbanites (especially relatively recent moves to Texas from most places not called, say, Orange County, Calif.) is the key.

  2. Greg Wythe says:

    A little bit more number-crunching on the 36% hypothesis shows some impossible results if it were to be applied evenly. I put in the numbers from your chart and expanded it out a little. Knowing that SSRV doesn’t equate with electorate share, I assumed a drop of 5% for Hispanic vote share. That number is essentially pulled out of thin air, but it errs toward the conservative side of the arithmetic. Border districts may not even lose that much when it comes to the electorate, but I wanted to throw in something to account for the fact that voter rolls aren’t administered evenly. And my own hunch is that several counties may have over-inflated rolls. But there’s nothing wildly out of sync with CVAP data and citizenship in all of these districts is above 90% of VAP, with most clocking in north of 95%.

    In HD31, if Romney were to have obtained 36% of a likely Hispanic electorate, it would mean that he would have gotten about 40% of the Anglo vote in that district. Not necessarily improbable, I suppose.

    But going to HD37, for Romney to have earned 36% of the Hispanic vote would mean that he would have gotten a negative amount of votes from non-Hispanics. This, of course, is a mathematical impossibility … even in South Texas. Negative results turn up in 6 of the districts you list. Results below Hispanic vote performance turns up in another 5. The only district where non-Hispanic vote performance by Romney would have been better than Hispanic vote performance would be in HD31. Clearly, this doesn’t match with any anecdotal evidence. So it can’t be taken seriously despite whatever imperfections may be in my hypothetical electorate modelling.

    If you take Hispanic performance from 36% to 20%, the results are a bit more realistic.

        Romney Vote %
    ----------------------
    Dist   Hisp   Non-Hisp
    ----------------------
    HD31   20.0%   82.7%
    HD35   20.0%   66.8%
    HD36   20.0%   43.3%
    HD37   20.0%   60.4%
    HD38   20.0%   73.5%
    
    HD39   20.0%   37.1%
    HD40   20.0%   33.0%
    HD42   20.0%   41.2%
    HD75   20.0%   50.9%
    HD76   20.0%   26.7%
    
    HD79   20.0%   63.7%
    HD80   20.0%   65.3%
    

    I’d be curious to know what results turn up in VAN for % of turnout that’s Hispanic in some of the districts to see what might be a more accurate dropoff to account for in vote share.

    And while there should be some expected differences in how Hispanics vote in border districts, urban minority districts, wealthier suburban districts, West Texas farming districts, and so on, it’s hard to see the share approaching 35-40% on average. At some point, the rising vote share in the suburbs is hard to explain with that: if suburban Hispanics are making up some of the difference for the Rove argument and voting more GOP than their border-district cohorts, then that would mean suburban whites are trending far more Democratic than anyone else has ever acknowledged. I’d expect to find a likelier scenario of suburban Hispanics voting slightly less Dem than in districts such as these, but nowhere near enough to raise the average to 36%.

  3. Greg Wythe says:

    On a sidenote to the above, I ran the numbers under the hypothesis that Hispanic vote share was actually higher than SSRV shares. It’s not likely that this averages out to be the case, but it’s not entirely improbably in some instances where you have SSRVs north of 85% and HCVAPs north of 90% as in HD42. At those points, there just aren’t enough numbers of other voters to go around and it’s worth pointing out that this approach lumps together whites and African-Americans, who generally provide two very different vote performance results.

    Anyways, assuming an across-the-board 5% increase in vote share over SSRV, you end up with only a 25% average for each of these districts to show a non-Hisp result over 0% and most of the districts showing non-Hisp vote performance above that of Hispanic vote performance. So it’s not like a change in assumptions greatly alters the outcome.

    All told, I think any strict mathematical calculation based on SSRV and Presidential performance will show a result between 25-30% statewide. The results may be higher based on polling due to MOE and non-surname Hispanics being picked up in polling. I think that puts the Latino Decision findings pretty close to “on the money.”

Comments are closed.