Greg's Opinion Greg's big blog of whatnot


Turnout Obfuscation

Before I get knee deep in the hoopla of turnout in the Garcia v Morman series, I wanted to take a tangent and look at turnout in general from a very micro- perspective. While there is certainly a fair amount of benefit to reviewing turnout levels at the county level or subdivisions thereof, I wanted to point out how individual precincts can sometimes obfuscate turnout differences.

There are a number of Harris County examples I can point to for this, but for the sake of using something different, I thought I'd point out the area I grew up in Euless, TX. That's Tarrant County for those of you keeping score at home. The current precinct that encompasses my old 'hood is 3247. It combines my old Junior High, the Sotogrande apartment complex where my family lived, and the very nice, tucked-in neighborhood of Morrisdale Estates just to the west of the apartments.

By way of background, I'll attest to my own childhood memory of Morrisdale as being somewhat similar to what Houston's Briarmeadow neighborhood is today - by no means River Oaks, but still a very nice neighborhood that is something of an island to the area surrounding it. While Morrisdale may not necessarily be quite what it used to be, a quick check of housing prices on indicates that it's still at the upper-end of home prices in Hurst, Euless, and northern Arlington. In my day, the neighborhood was on the side of the HEB school district that sent kids to LD Bell High School, while the apartment kids like me went to Trinity High School. Back then, that may as well have been a demilitarized zone for football politics. I don't profess to know where the school boundaries are drawn today.

Tracking Houston and Harris County elections, it's not difficult to see the impact that turnout differentials between areas with single-family homes and those with multi-family homes can have a big impact on an election. Granted, I don't have to look outside my own present precinct, House District, or City Council District to see this. But here's the overview for an example outside of that world. First, the geography ...

The western portion, where you see a lot of trees and nice homes is Morrisdale. The eastern part - where you don't see any of that - is Sotogrande.
Precinct 3247 - Tarrant County

View Precinct 3247 in a larger map

And now, the numbers ...

2010 Registered Voters: 1,732
Ballots Cast: 546
Turnout: 31.52%
Bill White: 206 (38.4%)
Rick Perry: 309 (57.6%)

The precinct to the north (3327) is very similar to the Morrisdale side of 3247. It's comprised more completely of homes - if not 100% so, it's pretty close. That precinct went for Perry over White 66.3 - 31.0 with 47% turnout of 1,653 Registered Voters.

VAN identifies 1,751 people in the precinct with the following breakdown for each half:
- 596 people in apartments (34%)
- 1,155 people in homes (66%)

There's no perfect way of isolating the voting performance of each half, but some quick, back-of-the-napkin algebra for turnout models (assuming that the 2006 ratios are more or less instructive) has the western half voting with 5% more turnout than the eastern half. The math I get for this election is somewhere in the ballpark of 33% turnout in the west, 28% turnout in the east. If the western half of the precinct voted entirely like 3327, algebra dictates that all of the votes would come from that side of the precinct. I can't quite imagine that being the case. But I'd also argue that using 2006 overstates turnout in the apartments since it over-represents more stable apartment dwellers who have been in the same location for over 4 years. With that in mind, I feel fairly comfortable using some easy-math rounding to get to 35% and 25% as the assumption for turnout in each half.

The next assumption involves "how they vote". If we assume the western half votes similarly to Precinct 3327 (66-31) Perry, then that leaves the apartments voting at roughly a 60-65% clip for Bill White. Even if they voted 5 points below that mark (61-36), that would put the apartments as voting around 50-50 for White-Perry.

Either way, a dramatically different outcome for both sides. And the differences don't show up easily in the final results. The overall precinct shows turnout lower than many of the single-family-home based precincts in the area and the Democratic performance is a few ticks better. Overall, it seems a good bet to assume the turnout is ~10% lower in the apartments and that Dem performance is anywhere from 10-15 points better in the apartments.

So, the precinct-level summary doesn't offer up enough detail to paint a complete picture. Making matters even worse in this case, the apartment complex is cut in half, as the southern portion of apartments is included in Pct. 3166 (which again include an awkward mix of homes and apartments).

There are a host of precincts like this that work both ways. But it's useful to keep examples like this in mind when interpreting turnout. I believe it's more true in areas where demographics are changing quickly, as it's not usually the case that demographics change evenly throughout a precinct. Instead, a new subdivision or apartment complex often fuel those changes.

I don't get terribly vested in the political conspiracies that may or may not be behind why the boundaries get drawn like this. But I do think it's important to keep examples like this in mind for field teams as they look for areas to cover. Too often, it's just done on nothing more than a precinct-level analysis that leads to a lot of work in blue precincts and very little outside of that. I'd argue that the rationale for a field team is to know the turf better than that. A timely and active voter registration drive, as well as a willingness to work in the blue areas of some red precincts can dramatically alter some outcomes and lead to more lasting changes for the better.

Related Posts:

  • No Related Posts
Comments (0) Trackbacks (0)

No comments yet.

Leave a comment

Trackbacks are disabled.

Fatal error: Call to undefined function xwp_footerx() in /nfs/c01/h12/mnt/4748/domains/ on line 18