Greg's Opinion Greg's big blog of whatnot


The Mother of All Redistricting Hearings (Part Two)

Day two of the San Antonio court proceedings are underway. I'm merely watching twitter and checking email to see if anything interesting happens. One particularly interesting point made in the testimony (as tweeted by Nolan Hicks) ...

MALDEF: 80 percent of Democratic primary voters in 2010 were Anglo.

You definitely get some interesting findings when you search in VAN for any kind of electorate and then see what the racial makeup of it is. And if more of that information was out in the public eye, it would definitely alter how many people understand their own districts. Just take it from this resident of Houston City Council District J ... ya know, the "Hispanic Opportunity District" as described by everyone except the guy who drew it and me.

TANGENTIAL, DEMOGRAPHIC-RELATED ADD-ON: Reading Jason Stanford's post over at HuffPo ...

Add to that the anomaly of Texas Hispanics' miserable rates of voting participation, and you understand why Texas is a Republican state and not a swing state. Hispanics are 37 percent of the Texas population -- and rising -- but they're only 15 percent of those who show up at the polls.

"There is a mystery to Texas," Andre Pineda, the late Democratic pollster, told me last summer. "Why is it that Latinos turn out less there? I think the Democratic performance of Latinos is a big difference but not a defining difference because it's when you throw in the turnout part that explains why Texas is red." Translation: Not only are Texas Hispanics less likely to vote than California Hispanics, but they are more likely to vote Republican when they do.

Right about now, the sound of my head slamming against my desktop should make for a handy metronome to guitar practice tonight. How is that we're having the umpteenth bazillion conversation about Hispanic population share anywhere in Texas and omitting the impact that citizenship has on that share. In fact, the state is roughly 25% Hispanic when you break it down to Citizen Voting Age Population.

The point about 15% of turnout being Hispanic, I've seen differing numbers with 15 being on the low end. But it's certainly close enough to whine about. In general, there isn't a "Hispanic problem" with regard to turnout, so much as there is an "economic problem" with regard to turnout. Ignoring the latter and assuming you can fix the former independent of that is a bit of a mistake. And believe me ... it's one we've tried many times in the past.

The bigger dispute I have with the rest of Jason's take is over the fact that Hispanics are voting 1/3 Republican. I've seen the exit polls and I've seen the equally questionable studies that suggest its much lower. But if you take a hard look at precincts that are overwhelmingly (90% or more) Hispanic, you don't see much evidence that 1/3 of the votes are going the other way. And when you factor in the Hispanic population surge in the suburbs, you're left with a very tough rationale to explain why the Democratic vote in those areas rises as they get more diverse if those Hispanics are voting more Republican than they are back in the 'hood.

I don't quarrel with Jason's conclusion. But the points used to illustrate the problem are flawed. I'm not convinced that getting Hispanic turnout up to 20% of the electorate is realistic. But doing the work to increase it would be a nice thing to see. Fortunately, you can see it in every competitive election. It was certainly done in 2010. So what else is needed?

TIME-KILLING SIDENOTE: Some quick and extremely dirty regression analysis from a data geek who rarely does regression analysis except to calculate Quarterback Ratings. Here's what I get from looking at House Districts and using CVAP shares for Anglo, Afr-Am, Asian, and subbing out Hispanic with the SSRV share to calculate what share of the vote each demographic supported Susan Strawn for in 2008:

108% African-American
82% Hispanic
58% Asian
22% Anglo

Obviously, you can't vote 108% for anyone. That bit of data does suggest that turnout was above and beyond for African-Americans, though. If I do some even dirtier extrapolation and back that number down to 97% among Afr-Am voters, the question is what to do with the remainder.

Among the problems with taking these numbers as gospel is that I'm still overstating Hispanic vote share among the actual electorate. Ideally, I'd have a fairly reliable count of the share for each demographic among actual voters. Even using the lower metric of SSRV instead of H-CVAP, I'm still overstating the share of Hispanic vote. If I just pick some percentage at random and say that the electorate share of Hispanics is 90% of SSRV, I feel obliged to define an "Other" category that covers both the unattributed share of voters lost from this calculation and the true "Other" for non-covered demographic groups in this simple breakdown. When I do that, I get the following:

108.5% Afr-Am
77.4% Hispanic
18.7% Anglo
49.6% Asian
78.1% Other

The instant you compress Afr-Am numbers down to 97% or whatever comparable share of support for Dem candidates you opt to select, that starts to raise the performance of other groups. Where anyone goes with that, be my guest to invent your own mathematical concoction. But what it doesn't do is anything that would lower the vote performance among Hispanics. Which means they likely voted somewhere in the ballpark of 80% Dem in 2008.

Based on my understanding of urban areas, that 80% mark sounds about right. I'm not going to pretend to be an expert on voting patterns in other areas. But look at HD31 - it's the most Hispanic district in the state: 95%. SSRV is over 91%. I think a good ballpark assumption for any turnout differential might peg the electorate share at about 85%. Let's call the 2008 average Dem performance at about 80%. If those Hispanics voted 1/3 Republican, that would mean that Dems would get to 55% on the back of Hispanic votes alone. If they got 100% of everyone else, they would get 70% of the vote. And yet, in the much more difficult year of 2010, Dems got an average of 75% of the vote in that district.

So, needless to say, I'm not buying the 1/3 GOP vote for Hispanics. The work above isn't the best math in the world for getting to the core of this, but I think its good enough for outlining the ballpark. And to me, that ballpark looks more along the lines of 75-80% of Hispanic voters voting Dem.

Seriously, I don't judge any of you for what you choose to do on a lunch break. So don't judge me.

Related Posts:

Comments (6) Trackbacks (0)
  1. 2008 turnout in the Democratic Primary in District 31 was higher than in the general election. The voters favored Clinton over Obama more than they did Strawn over Price. Obviously voters are favoring female candidates, more significantly than race or partisanship.

    Zapata had 45.0% turnout in the primary, vs 36.0% in the general; Duval 54.2% vs 44.4%; Starr 32.8% vs 32.6%. Webb increased from 41.6% but 69% of the county is in District 42. Strawn received 84.4% of the vote in Webb-31, vs 68.9% in Webb-42, so I suspect that turnout for Webb-31 was more like the rest of the district, than the rest of the county (Webb-31 is 97.6% Hispanic vs 94.9% for Webb-45).

    In 2010, Zapata went from 44.7% in the Democratic primary, to 16.0% in the general election. On rank, this is from 2nd highest to 251st.

    I suspect that local political machines are pretty good at get out the vote, when there is reason to get them out to vote.

    You are also assuming that perhaps the poorest, most Hispanic district state is representative of the state, and that other economic status has no effect on voting behavior.

  2. Actually, I’ve said on a number of occasions that Valley Hispanics aren’t the same as urban Hispanics … and for all other geographic variations. Likewise, the regression analysis was done to capture a measurement of what that share of the vote might look like on a statewide level.

    My point only has to do with the % of Hispanic vote that goes to each party in General Elections. And on that point, I’m skeptical of the claims of 1/3 of the vote going to Republican candidates. While I’m sure that there are areas where the Hispanic vote does perform better than others for Republicans, I just don’t see any evidence for the 1/3 claim.

    In doing the same thing for other candidates, the difference isn’t terribly great. There may certainly be individual candidates that perform better (Abbott & Guzman in the case of 2010, Bush in the case of 1998 for instance). But on any kind of broad scale, it looks more like the ratio of statewide support for Republicans by Hispanics is somewhere between 20-25% and likely offsets the share of Anglo support for Dems at this point in time.

  3. The turnout in District 31 is so variable you are guessing about the Anglo share of the electorate, and presuming that the few Anglos in those areas turn out in numbers comparable to Collin County and vote similarly.

    Let’s take a look at the Webb County portion of District 31 and District 42 (The Webb County portion is 47% of D-31, and in 2010 48% of the vote for Guillen was cast in Webb-31).

    In Webb-42 Abbot 31.4%, Guzman 36.1%
    In Webb-31 Abbot 21.9%, Guzman 27.0%

    Webb-42 is 94.9% HP, 94.2% HVAP, 92.4% HCVAP, 84.5% SSVR,

    Webb-31 is 97.6% HP. If we assume that Webb-31 maintains the same percentage drops as 31 as a whole then Webb-31 is 96.9% HVAP, 96.1% HCVAP, and 92.3% SSVR.

    But is it possible that the SSVR/HVAP ratio in Webb-42 89.0% and Webb-31 is 94.5%? This implies a huge difference in either naming patterns or registration patterns within the same county. And if that is true – can we assume that voting patterns (ie the GOP/SSVR is a constant). That is, is the fundamental assumption for linear regression true?

  4. Well, at some point, a statewide average is a statewide average. And that’s the original purpose behind the whole exercise. There’s certainly some fuzziness in using CVAP – and even tweaking HCVAP downward (to approximate SSVR). My point in highlighting HD31 was merely to suggest that, if the 1/3 GOP vote were visible, that it’s certainly not visible in the highly concentrated areas – neither by House District nor by precinct in the big urbans. While I’d agree that that leaves some open room for discussion of how the less-concentrated Hispanic voters vote, I think it’s still difficult (assuming that the 1/3 GOP hypothesis is to be believed) to square the improving Dem vote in those areas with minority voter growth – either more suburban Anglos are going Dem than even I’m prepared to believe, or the minority growth is a large factor behind the Dem vote there and the patterns aren’t wildly different from the more concentrated areas.

    The difference in naming patterns may be something to look deeper at – you can certainly see some differences in family backgrounds due to long-term intermarriages. I can’t say that my sense is that it would do a lot of the explaining of different vote patterns, but it would certainly be an interesting variable to explore further.

  5. In the areas with the highest Hispanic concentration, there is not even a Republican primary if one wanted to be a Republican. If you want to choose any part of your local government, you have to vote in the Democratic primary. Under such circumstances you are about as likely to encounter a Republican as Rick Perry did growing up in Paint Creek.

    Take a look at the SSVR/HCVAP numbers (not percentage) for the 150 House districts. There is quite a large variation. Some may be accounted for by naming variation, but I suspect most is related to participation rates. If 70% of HCVAP are registered in some areas, and 90% in others, is it at all likely that those who do register will vote for Republicans in the same percentage?

    You’re more like to see the GOP vote ranging between 10% and 40%, than it to be tightly clustered between 20% and 25%.

  6. I don’t disagree that there would be a wider range that averages out to somewhere in the 25% ballpark. But it’s still a challenge for me to see areas where it will reliably hit 40% GOP among Hispanic voters. There may be enough swing vote for different candidates so that a Bush-98, Strayhorn-02, Bush-04, or even Guzman-10 may hit that peak in certain areas. But I don’t see it on any kind of baseline level.

Leave a comment

Trackbacks are disabled.