Early Voting in the November 2013 Texas Constitutional Amendment Election: Two Views

In the election news cycle yesterday and today, one of the stories has been about how smoothly picture I.D. voting has been going in the first statewide election since the law went into effect.

On the one hand, state and county election officials report that there have been almost no problems at all. Married and divorced women have not been disproportionately turned away at the polls due to identification issues resulting from name changes, and there are no reports of widespread provisional voting or unprepared voters. The media take on this is that the Texas Democratic Party has lost traction and credibility on the issue of picture I.D. voting, arguably because the party oversold the potential that voters would be turned away at the polling places.

The State of Texas and the U.S. Department of Justice might both cite improvements in the adminstration of picture I.D. requirements in Texas for this election – among other things, the State broadly expanded the number of entities authorized to issue photo i.d.s, and the Department of Justice might regard these improvements as attempts by the State to mitigate liabilities for civil rights violations in direct response to DOJ’s civil rights lawsuit over picture I.D.

A number of commentators have been careful to point out that the policy questions about the adoption of picture I.D. requirements for voting are clouded by political rhetoric from both sides of the political spectrum. On the left, there is probably a tendency to overstate the suppressive effect of picture I.D. laws in general. On the right, meanwhile, buzzwords like “security and election integrity” are chosen to justify picture i.d. laws on the basis of purely imaginary waves of election fraud.

Anecdote is poor evidence of systemic problems with elections, and it is unwise to rely on either individual reports of voter problems (“I had to vote provisionally because my name on my picture I.D. doesn’t exactly match my name, therefore picture I.D. suppresses voting.”) or the evident lack of such reports (“County voter registrars report that after two days of early voting, not a single person has called to complain or been turned away from voting due to an i.d. issue.”)

Why is it so easy to fall into the habit of false generalization on this issue? Here are some reasons that come to mind, off the top of my head.

1. Selection bias – The population of voters is self-selected, and is not directly comparable to the population of all people who could vote, or of all people who (from a social utility perspective) should vote or ought to vote. This selection bias is more pronounced in an election as sleepy and low profile as the November 2013 state constitutional amendment election in Texas – the tiny number of people who actually manage to turn out and vote in an “off-season” election like this one are arguably a committed and motivated group when compared to the larger population of all people eligible to register to vote.
2. Sampling bias – Sampling just the complainers is probably not the best way to collect data on election problems. One of the things that struck me as interesting about working in the Elections Division at the Texas Secretary of State’s Office wasn’t how many phone calls we would get (the number ranging anywhere from around a hundred on a slow day to a couple thousand on an election day), but how few calls we got. Neither the roughly 12 million registered Texas voters, nor the roughly 1 million eligible non-registered people of voting age make a habit of calling with election questions or complaints, but we can’t infer that the population is happy just because the per capita number of complaints is low.
3. Modeling bias – The media, election officials, and political party strategists all envision “voting problems” as meaning something, but they don’t necessarily share the same mental impression of what that something is. As widely reported, Texas gubernatorial candidate Wendy Davis had to sign a “same name” affidavit (provided as part of the new name-comparison law) because her name changed as a result of marriage. This could either be reported as a problem (“Requiring a voter to sign an affidavit creates an unacceptable risk of bureaucratic error in the handling of many women’s votes, the danger of interference with those votes, and an institutional discouragement of voting by women.”) or as a non-problem (“Aside from having to complete an affidavit (comparable procedurally with the old “no voter certificate” affidavit), Senator Davis successfully cast a ballot that will be tabulated with all the other valid ballots in this election. Therefore, the name comparison system is working perfectly, Senator Davis’s ballot will be counted, and this so-called problem isn’t a problem.”)

It turns out that actual social science research costs actual money and time and intellectual creativity. When institutions such as the Brennan Center for Justice or the Pew Charitable Trust are trying to figure out if changes in voting procedures are harmful or helpful, they do things like conduct carefully designed polls, build economic and policy models, conduct deep interviews of voters, non-voters, discouraged voters, almost-voters, and election workers, conduct mock elections and user forums, focus groups and behavioral studies, enlist sociologists and statisticians, and attempt valiantly to identify and counter all sources of conscious and unconscious bias that might skew the results.

Good journalists can do the same sort of deep digging, but sadly it happens less and less.

I’m no fan of picture I.D., partly because I think its a product of ignorance, graft, and simple, brutish, mean-spirited bullying. But I also hesitate to say that the ills of photo i.d. will manifest themselves in big, camera-friendly riots and mayhem at the polls. In other words, the suppressive effects of photo i.d. requirements are all the more worrisome because those effects are likely to be invisible – isn’t it at least theoretically possible that some of the reason for the low turnout might well be attributable to the new voting procedures?

[UPDATE: The first version of this entry erroneously reported that Senator Davis voted provisionally. She did not – she completed an affidavit regarding her name change, and otherwise voted normally].