Off the Kuff Rotating Header Image


Does getting to 40% make you likely to win the runoff?

Anna Eastman

I was talking with some fellow political nerds last week, and one of the topics was the forthcoming runoffs. As is usually the case, this year we have some runoffs between candidates who finished fairly close together in round one, and some in which one candidate has a clear lead based on the initial election. The consensus we had was that candidates in the latter category, especially those who topped 40% on Super Tuesday, are basically locks to win in May. The only counter-example we could think of off the tops of our heads was Borris Miles beating Al Edwards, who had been at 48%, in the 2006 runoff for HD146.

So, later on I spent a few minutes on the Secretary of State election archive pages, looking through past Democratic primary results and tracking those where the leader had more than forty percent to see who went on to win in the runoff. Here’s what I found:


Winners – CD03, CD10, CD23, CD31, Governor, SD17,
Losers – CD27, HD37, HD45, HD64, HD109*, HD133*


Winners – CD15, HD27
Losers – SBOE6


Winners – Senate, SBOE13
Losers – HD105

Winners – CD34, HD95, HD137
Losers – CD23*, SBOE2

Winners – CD10, HD76*

Winners – CD32, RRC

Winners – Senate, Lt Gov, HD42, HD47*
Losers – HD146

In each of the cited races, the leading candidate had at least 40% of the primary vote. Races that have asterisks indicate that the runnerup also had at least 40%. As you can see, up until 2018, having forty percent or more in the primary was indeed a pretty good indicator of success in overtime. The last cycle provided quite a few counterexamples, however, including one incumbent (Rene Oliveira, who had been busted for a DWI earlier) who went down. So maybe 40% isn’t such a magical number, or maybe it’s harder now than it was before 2012. Or maybe this is just a really small sample and we should be careful about drawing broad conclusions from it.

Fortunately, we have quite a few races this year to add to this sample:

CD03 – Lulu Seikaly 44.5%, Sean McCaffity 43.8%
CD10 – Mike Siegel 44.0%, Pritesh Gandhi 33.1%
CD13 – Gus Trujillo 42.2%, Greg Sagan 34.7%
CD17 – Rick Kennedy 47.9%, David Jaramillo 35.0%
CD24 – Kim Olson 40.9%, Candace Valenzuela 30.4%
SBOE6 – Michelle Palmer 46.8%, Kimberly McLeod 34.6%
SD19 – Xochil Pena Rodriguez 43.7%, Roland Gutierrez 37.3%
SD27 – Eddie Lucio 49.8%, Sara Stapleton-Barrera 35.6%
HD119 – Liz Campos 46.1%, Jennifer Ramos 43.7%
HD138 – Akilah Bacy 46.7, Jenifer Pool 29.3%
HD142 – Harold Dutton 45.2%, Jerry Davis 25.3%
HD148 – Anna Eastman 41.6%, Penny Shaw 22.1%
138th District Court – Gabby Garcia 48.0%, Helen Delgadillo 31.0%
164th District Court – Cheryl Elliott Thornton 41.3%, Alexandra Smoots-Thomas 33.1%

I’ll be sure to do an update in May, when we can see if the leading candidates mostly held serve or not. Place your bets.

Precinct analysis: Abbott versus Perry in Latino districts

District level election data for 2014 has been available for a few weeks now. Seems like as good a time as any to return to a favorite topic, namely how Greg Abbott did in heavily Latino areas. An exit poll from November claimed Abbott drew 44% of the Latino vote, which would be a very impressive accomplishment. My complaint whenever I read a story like that is that no one ever bothers to go back and check the actual election results later to see if that kind of number makes sense. No one but me, of course, because I’m a crank about that sort of thing. Now that we have this data, how does it look? Here’s a comparison to Rick Perry in 2010 in the most heavily Latino districts:

Dist SSVR% Perry Abbott ============================= 031 76.46% 42.01% 44.80% 035 76.58% 37.19% 39.11% 036 87.34% 29.55% 31.21% 037 81.21% 36.96% 38.13% 038 80.92% 39.11% 40.39% 039 85.14% 27.03% 32.12% 040 88.14% 25.37% 28.59% 041 71.98% 46.69% 47.84% 042 88.70% 22.58% 29.69% 075 83.70% 29.04% 30.84% 076 84.73% 23.57% 24.32% 079 72.70% 38.89% 39.26% 080 80.84% 34.79% 37.78%

SSVR data is from here. I’d like to think that this would put those 44% assertions to rest, but I know better by now. Abbott clearly did better than Perry, though by only a point or two in most districts. Some of that may simply be due to Perry doing worse overall than Abbott. Still, his actual number among Latino voters is nothing to sneeze at. But as I’ve said before, while the actual results provide a reality check on exit polls and from-the-ether assertions, they’re more suggestive than conclusive. We don’t know what percentage of actual voters in these districts was Latino. To see what I mean, consider a district with 10,000 voters and an SSVR of 80%. Imagine also that Abbott gets 70% of the Anglo vote, which is likely to be at least what Abbott would need to get to almost 60% overall. How does the vote break down if Abbott scored 40% (i.e., 4,000 votes) in that district?

If the actual mix of voters is 80% Latino and 20% Anglo, then Abbott got 1,400 Anglo votes, which means he needs 2,600 Latino votes to get to 40% overall. 2,600 votes out of 8,000 is 32.5%.

If the actual mix of voters is 70% Latino and 30% Anglo, then Abbott got 2,100 Anglo votes, which means he needs 1,900 Latino votes to get to 40% overall. 1,900 votes out of 8,000 is 23.75%.

Basically, the share of the Anglo vote, even though it is relatively small in a district like this, has a large effect on the share of the Latino vote. Changing the assumption that Abbott got 60% of the Anglo vote in this district instead of 70% doesn’t make that much difference. In scenario 1, Abbott needs 2,800 Latino votes instead of 2,600, or 35%. In scenario 2, it’s 2,200 instead of 1,900, or 31.4%. Even in a scenario where you assume the Latino vote exceeds the SSVR%, you get the same kind of result. In a 90/10 situation with a 70% Anglo vote, the corresponding Latino percentage is 36.7%; with a 60% Anglo vote, it’s 37.8%. The only way for the Latino vote percentage to be higher than the overall percentage is if the Anglo vote is less than the overall. I suppose it’s possible Abbott could fail to break 40% of the vote in these districts, but I’ve yet to see anyone offer objective evidence of it. Therefore, the numbers I present above represent the upper bound for Abbott among Latinos in these districts. Anyone who wants to claim otherwise needs to show me the numbers.

(To be completely fair, one scenario under which the Latino vote could be higher than the overall would be if some other segment of the electorate was voting disproportionately against Abbott. A significant portion of African-American voters in these districts could do that. Take the first scenario above and change the voter demography to 80% Latino, 10% African-American, and 10% Anglo. Now assume a 70% Anglo vote for Abbott and 10% A-A vote for him. With those assumptions, 3,200 Latino votes are needed to get to 40% overall, and as it happens that’s a 40% share of the Latino vote. However, in the districts above, the largest African-American population is four percent; it’s less than one percent in most of them. As such, this variation pretty much can’t exist.)

Another way we can look at this is to see if other Republicans did better in these districts as well, or if the effect was limited to Abbott. For that, we turn to a comparison of David Dewhurst in 2010 to Dan Patrick.

Dist SSVR% Dew Patrick ============================= 031 76.46% 45.47% 40.46% 035 76.58% 37.99% 34.86% 036 87.34% 29.04% 26.67% 037 81.21% 35.77% 33.85% 038 80.92% 38.91% 35.40% 039 85.14% 26.44% 27.50% 040 88.14% 25.11% 23.00% 041 71.98% 48.27% 42.16% 042 88.70% 24.68% 23.67% 075 83.70% 30.16% 29.72% 076 84.73% 24.67% 23.37% 079 72.70% 41.50% 37.98% 080 80.84% 35.40% 34.59%

With the exception of HD39, Dewhurst did better than Patrick. Obviously, Dewhurst did better overall than Perry, while Patrick was roughly equivalent to Abbott. That suggests that while Abbott may have improved on Perry’s performance, he wasn’t necessarily a rising tide. To be sure of that, we should compare him directly to his comrades on the ballot. I’ve thrown in Perry as well for some perspective.

Dist Abbott Perry Patrick Paxton Hegar Bush ========================================================== 031 44.08% 42.01% 40.46% 41.36% 40.97% 45.24% 035 39.11% 37.19% 34.86% 35.93% 35.70% 39.45% 036 31.21% 29.55% 26.67% 27.89% 28.06% 32.42% 037 38.13% 36.96% 33.85% 34.16% 34.13% 39.77% 038 40.39% 39.11% 35.40% 36.30% 36.15% 41.98% 039 32.12% 27.03% 27.50% 28.58% 28.68% 33.18% 040 28.59% 25.37% 23.00% 23.92% 24.24% 29.45% 041 47.84% 46.69% 42.16% 44.51% 44.77% 49.92% 042 29.69% 22.58% 23.67% 22.48% 23.40% 33.23% 075 30.84% 29.04% 29.72% 29.33% 29.21% 28.75% 076 24.32% 23.57% 23.37% 23.52% 22.91% 24.76% 079 39.26% 38.89% 37.98% 37.94% 37.41% 37.76% 080 37.78% 34.79% 34.59% 34.14% 33.71% 39.13%

A few observations:

– Clearly, Abbott did better in these districts than anyone except Baby Bush. Playing up their own Latino connections – wife in Abbott’s case, mother in Bush’s – helped them, at least to some extent. We have seen this before, with several other candidates – Ted Cruz, Eva Guzman, Hector Uribe, and as you can see above, Leticia Van de Putte. The effect isn’t much – a couple of points – but it exists. It should be noted that since these candidates’ overall totals don’t differ much from their ballotmates’, there’s an equivalent but opposite effect elsewhere. Just something to keep in mind.

– Note that the effect for Abbott was greater in South Texas and the Valley, and lesser in El Paso (HDs 75, 76, and 79). Bush also did worse in El Paso, no doubt due at least in part to having former El Paso Mayor John Cook as his opponent. Consider this a reminder that the Latino electorate is not monolithic, even within the same nationality. What works well here may not be as effective there. This should be obvious, but I feel like we all sometimes act as if that’s not the case, and yes I include myself in that.

– Along those lines, I wish that the SSVRs were high enough in the urban Latino districts to include them here, but they’re not really comparable. Having written that, I’m now curious enough to do that comparison in another post, just to see what I get.

– At the end of the day, Greg Abbott in 2014 was a lesser known quantity than Rick Perry in 2010. He had a chance to introduce himself as a more or less clean slate. That won’t be the case in 2018, if Abbott is on the ballot for re-election. He’ll have a record to defend, for good or bad. We’ll see how much his wife and madrina can help him then.