In the previous looks at how the new Voter ID law was administered, we saw that 8.3% of Harris County voters ended up initialing the affidavit acknowledging that the names on their ID were not a 100% match to their name on the voter file. There remains, however, an open question of what the ratio should have been. According to media reports, County Clerk Stan Stanart noted that “[a]bout 40 percent of registered Harris County voters have mismatched addresses that could require additional verification.” Emphasis mine.
The law states that the purpose of the ID is to compare the name on it to that on the voter roll, not the address. So while it’s quite believable that Stanart’s 40% number could be believable, it’s not quite the same as saying that 40% of the names on the voter roll fail to match the names on DPS records. So what might that number look like?
Included in the information request I got from the County Clerk was the match file for the voter roll to that of the DPS records. Each record has four name fields for the voter roll (first, middle, last, suffix) and four from DPS. In comparing the file, I ended up with 20.9% of records not matching. My own mismatch name is there, as is Kuff’s, and a few others who mentioned that their names didn’t match. I’m under the impression that this was the very file that the County Clerk’s office used to prepare for the 2013 election (I have a request for verification in with the office, but no response yet). One bit of information I’m hoping to clear up is why the file I have contains 1,871,369 records while the total Registered Voter count on the canvass lists 1,967,881. There could be a number of perfectly logical explanations, but I’m assuming for the moment that the file I have goes a long way toward helping identify what the rate of Substantially Similar affidavits should have looked like.
Countywide, the roster of voters from In-Person Early Voting contained 18,276 voters who should have initialed an affidavit – 20.7% of those who cast their vote early in this manner. That matches surprisingly close to the overall comparison to the DPS records – a finding I hadn’t expected. And it’s worth contrasting that to the 8.3% who actually did initial an affidavit. On the whole, that would suggest a 40% efficiency rate of capturing the initials of those voters who should have initialed.
But the overall result obfuscates a great deal of detail that includes locations that simply did not administer the Voter ID law as they should have … as well as several who over-administered the law to capture names that should not have been required to initial an affidavit. In the previous posts, I identified 40% of Early Voting locations that captured less than 5% of their voters on Substantially Similar Affidavits. In looking at “what should have happened”, I’m finding that out of the 480 location/day combinations (40 locations x 12 days), 20% of these location-days ended up with over-administered affidavit signatures.
In Tuesday’s preview, I noted the location of Pasadena’s Harris County Court Annex. That location “shoulda” seen 174 voters needing to initial an affidavit (10.7% of their Early Voters). In reality, the location got 398 (24.5% of their Early Voters). This gives the location an “Efficiency Rate” of 229%. That qualifies – easily – as the most out-of-whack finding for an “over-administered” location. Obviously, those locations that seemingly failed to get the memo on what the Substantially Similar Name process was ended up with Efficiency Rates at or near zero. In all, seven location ended up as “over-administered” for the entire run of 2013 Early Voting:
126C – 112% (Champion Forest Baptist Church)
128P – 229% (Pasadena – Harris County Court Annex)
130T – 188% (Tomball Public Works Bldg)
135M – 107% (Metropolitan Mult-Service Center)
138B – 186% (Bear Creek Park Community Center)
141H – 102% (Octavia Fields Branch Library)
149G – 131% (Glen Cheek Education Bldg)
Even among locations that did not clock in at over 100% efficiency, there are sporadic daily patterns. For instance, my location – Bayland Park (137B):
10/21 – 24%
10/22 – 94%
10/23 – 109%
10/24 – 164%
10/25 – 105%
10/26 – 88%
10/27 – 100%
10/28 – 94%
10/29 – 122%
10/30 – 85%
10/31 – 70%
11/1 – 50%
Some possible reasons for why we might see efficiency rates over 100% are:
– Not everyone showed a driver’s license or state ID. The DPS matches wouldn’t reflect the fact that a passport or military ID might or might not match. I don’t necessarily buy the idea that this explains a large portion of the differences, though, since the overwhelming majority of voters presented a DPS-issued ID.
– The timing of the voter file comparison remains unconfirmed. Having heard the 40% number repeated and occasionally mis-characterized, my first instinct was to check for a few records that might have been fixed in the Early Voting process under the assumption that I might be looking at a post-election comparison file – perhaps one used for the current primaries. The request for the file specified that I wanted the one used for the 2013 elections, so I’m operating under the assumption that the Clerk’s office got that right. I don’t have reason to believe otherwise after some spot-checking of folks I know who did make a correction to their voter registration.
– The voter file used for comparison is nearly 100,000 voters short of an exact match reported on the canvass. While I’m still waiting for an official response from the County Clerk, it could be that much of that difference could be due to the fact that there was no DPS match -and hence, those 100,000 voters would have a mismatch. This would raise the overall “shoulda” comparison to 25% of voters required to initial an affidavit. Since the calculated rates of what “shoulda” happened and what actually happened were surprisingly close, it isn’t inconceivable that a “revised-shoulda” of 25% mismatches would bring the efficiency rate calculations more inline with expectations. For instance, the 112% efficiency rate for Champions Forest would end up under 100% and more reflective of what we’d expect to see. It still wouldn’t explain Pasadena (or Tomball, or Bear Creek, or Glen Cheek), but it would go a long way towards making everything else make sense.
If you’re inclined to look at any of the detail, here are two reports to go nuts with:
1. This shows the calculated “shoulda” rate for Substantially Similar names based on a calculation of DPS matches from the voter file.
2. This report shows the daily comparison of the “shoulda” vs actual affidavit collection for each location by day.
I may do some more work to look at the name matches by precinct with the data in hand. I definitely hope to break down some results by Male/Female to show the different impact this has on those categories of voters. And I may do a limited request to look at what happened on Election Day 2013 with regard to the affidavits. But I think the next, most interesting view of this implementation is going to be from the primaries now underway.
Most locations, in 2013, got better as the Early Vote calendar went forward. I suspect that there remains an issue about how well the law can be administered under the busiest days, but that remains a hypothesis and may not have anything more conclusive until November of this year. I would hope to see some improvement in the process for the 2014 primaries. But there’s sure to be some fuel available for any fires once we see whether Dem or GOP voters tend to fall into the Substantially Similar umbrella.