An earlier post took a look at the Census Bureau’s estimates of the ethnic makeup of 2012 voters in Texas.
But, as mentioned in a follow up post, the Bureau’s estimates tend to overstate actual turnout because people don’t like saying they didn’t vote, especially with something as important as a presidential election.
So, how accurate were the Census figures for Texas?
Well, using the Census’ breakdown of Texas voters by ethnicity (as opposed to citizen voting age population), we get a sense.
Bear in mind when looking at the chart that we know that only around 67.5% of all non-suspense voters and 58.5% of all voters in Texas took part in the 2012 election. In other words, turnout rates like those in the chart are mathematically impossible.
Luckily, there are some ways, albeit imperfect, to cross check - starting with looking at the list of actual voters.
The imperfection comes in the fact that Texas - unlike Georgia, North Carolina, and a number of other states - does not track the ethnicity of registered voters. That means that someone wanting to figure out turnout by ethnicity has to use markers (such as last name and census block data, supplemented by commercial info) to model the likely ethnicity of a voter.
But the data won’t be perfect in every instance.
Take, for example, an Anglo man named John Brown, who may well live in a neighborhood that is 99% African-American. While, statistically, it might be safe to assume he is African-American, it would be off in that individual case. Women who use married names also present issues - though there are things like maiden name searches that can help reduce error.
All that said, for the vast part of voters, we can be certain of ethnicity with a high degree of probability. Looking at just that universe, here’s what turnout rates look like:
What’s reassuring is that the figures on the two charts are broadly consistent at a high level.
Although the percentages are lower across the board in the voter-file analysis (accounting for a universal fudge factor with Census data), both sets of estimates have Anglos and African-Americans outvoting Hispanics and Asians - and Asians slightly outvoting Hispanics.
There two notable differences, though.
First, while the Census figures show African-Americans outperforming Anglos, the voter-file look has African-Americans a few points behind.
That’s perhaps not surprising. If any group would seem likely to be reluctant to own up to not voting in 2012, it probably would be African-Americans.
The Census data also has Hispanics trailing Anglos by only about 12 points, while the voter-file data has Hispanics trailing by nearly double that. (Ditto Asians, but the state’s Asian population is small enough that the impact is minimal.)
Both variances have a significant impact on the look of the universe of actual 2012 voters.
Using the Census figures, the mix of actual 2012 Texas voters looks like this:
On the other hand, applying the voter-file turnout rates to the Census’ universe of registered voters (for a rough back-of-the-envelope comparison), the turnout breakdown looks like this:
And some commentators have suggested that the Census’ estimate of the ethnic breakdown of registered voters may overstate Hispanic and African-American registration rates - because, well, people fudge being registered just like they fudge whether or not they voted.
Indeed, that view seems to be supported by the fact that the number of registered voters in Texas has been relatively flat for the better part of a decade despite burgeoning Hispanic and African-American growth during that same period - both in total and citizen voting age populations.
If that’s the case, the Hispanic and African-American shares of the turnout universe may be even lower yet. All of which suggests the state’s turnout challenge may be all the more so of one.