The Effect of Voter ID Laws on Turnout

The NYT has a short piece on this new wave of reformist legislation that, save for one state, has been universally passed by Republican legislatures and signed into law by Republican governors.  The NYT piece summarized a new study released yesterday by the Brennan Center for Justice at NYU that suggests around 5 million potential voters would face additional burdens because of the new legislation.  In the spirit of balance, the NYT briefly mentions a Heritage Foundation report, recently (August 2011) authored by Hans von Spakovsky & Alex Ingram, itself a direct critique of a previous (2006) Brennan Center publication entitled Citizens Without Proof.  I’ve read both, and while the Brennan Center report is not perfect, it’s largely compelling, and certainly more rigorous than the hack job released in the name of the Heritage Foundation.  It is here where one can mine pure comic gold.

Such genius begins with the opening sentence.

The primary argument against voter identification requirements—that many Americans lack proper identification and would therefore be prevented from voting—is not supported by credible studies of voter turnout rates.

This, simply stated, is untrue.  To at once paraphrase and directly quote the Heritage piece, their “study suffers from sloppy—or perhaps purposefully misrepresented—” knowledge and / or understanding of the extant political science literature.  A quick lit review came up with several citations, including empirical articles by Alvarez et al. (2007), Barreto et al. (2007), Vercellotti and Anderson (2006); a review of the extant literature by Hershey (2009) (published in PS, of course), and an interesting piece by Robert S. Erikson and Lorraine C. Minnite (2009) published in the Election Law Journal, which is a solid consideration of the practicalities and limitations that we social scientists face in observing marginal effects within the limits of data availability and research design.  (Indeed, I’ll be using this article in my methods class this year).  Of these cites, all agree that theoretically, whenever the costs of voting are increased, additional burdens to participation will result, and this should decrease the probability that any given individual casts a ballot.  It would be foolish to argue otherwise (although the Heritage piece certainly attempts to do so, albeit with an argumentative logic that wouldn’t pass muster on a high school debating team).

Not surprisingly, empirical results are mixed, but only to the extent that a handful of null results coexist with findings supportive of the hypothesis that voter ID laws indeed reduce turnout.  That theory suggests a certain outcome does not mean we will be able to observe that outcome, especially when the effect is at the margins.  The best explanation for this can be found in the Erikson & Minnite article:

Yet we see the existing science regarding vote suppression as incomplete and inconclusive. This is not because of any reason to doubt the suppression effect but rather because the data that have been analyzed to date do not allow a conclusive test.  What can be done to boost the empirical analysis of the problem? Additional elections and additional states enforcing strict voter ID laws will provide more and better data.

Thankfully, a bunch of states are now happy to supply us social scientists with the “more and better data” we crave.

Alvarez et al. find no effect with aggregate data, but using CPS data, do “find that registered voters with low levels of educational attainment or lower levels of income of all racial/ethnic groups are less likely to vote the more restrictive the voter identification regime.”  This is broadly consistent with both the Barreto et al. paper and the Vercellotti & Anderson paper.  Indeed, the only (unpublished) paper I’ve found that argues the opposite is Milyo 2007, but it suffers from two fatal methodological limitations.  First, it’s a comparison of county-level turnout rates in the 2002 and 2006 Congressional elections in Indiana; the method employed is limited to bi-variate cross tabs which use an interactive term (ethnicity or partisanship multiplied by the presence or absence of the voter ID law, itself effective from only 2006).  By my read, it’s not possible to move beyond estimating variance in turnout across the counties between 2002 and 2006; any independent effect of the voter ID law is impossible to isolate with this design.  However, the author over-reaches in attributing a rise in minority turnout as resulting from Voter ID laws based on county-level aggregate data.  In other words, a classic ecological fallacy, a fact pointed out by Alvarez et al.  (who, it should be noted in fairness, also have a strong methodological critique against Vercellotti & Anderson, though the findings of the former ultimately confirmed those of the latter). Of course, this doesn’t stop von Spakovsky from suggesting that as A) turnout in Indiana 2006 was 2% higher than Indiana 2002, B) 2006 featured the voter ID law, ergo C) the voter ID law clearly serves to increase turnout.

Finally, the extant literature goes deeper than testing the simple hypothesis and finds “systematic biases in the application of the identification statute across racial lines.”  (Atkeson et al. 2010).  In other words, if your skin is brown, you’re more likely to get carded when trying to cast a ballot.

The authors of the older (2006)  Brennan Center report swiftly replied to the Heritage hack job, but I add a few points here.  von Spakovsky writes:

Election turnout data also reveal that significant numbers of Americans who may be eligible still do not vote for various personal reasons that have nothing to do with registration or voting rules and regulations. In fact, the largest group of nonvoters “are more affluent, better educated, and more involved in their communities or volunteer groups” than voters. They are just not interested in voting.

Bullshit.

“Election turnout data” reveal no such thing; in fact this is precisely contrary to one of the core tenets underlying our understanding of turnout.  von Spakovsky couldn’t be further from the truth even if he were explicitly trying.  The source cited for the passage above is a book by two Northwestern journalism professors, Jack C. Doppelt, Ellen Shearer, entitled Nonvoters: America’s No-Shows, published by Sage in 1999.  The authors use a survey of >3000 to identify around 1000 “likely non-voters” who are then asked 64 questions.  A further 30 are selected for the sort of in-depth, qualitative interviews that while illustrative, are quite uncommon in the turnout literature.  This wasn’t an examination of the determinants of turnout; rather it was an attempt to understand and classify those who don’t vote.

Wait!  There’s more:

legitimate studies of voter turnout rates in states with identification requirements demonstrate that such laws do not disenfranchise voters; indeed, Americans overwhelmingly support such requirements that increase the reliability and trustworthiness of our election system.

The presence or absence of support for such laws is irrelevant in trying to determine whether or not they create an additional burden on the act of voting.  Fallacious reasoning such as this is common throughout the von Spakovsky reports; in fact this is one of the less egregious examples.  The list of  “legitimate studies” is footnoted as being available in an earlier Heritage report (itself comical as the August report critiques the Brennan Center for including “Buried Footnotes . . . The study is further undermined by several footnotes buried in the report.”)  In this earlier report, von Spakovsky does provide something of a humble lit review.  In considering this literature search, we must remember that he is on record in his August report as valuing an unbiased approach to evidence gathering (e.g. he critiques Citizens Without Proof for “ignoring evidence”).  Hence, I can only conclude that in compiling his exhaustive list of “legitimate studies”, all of which coincidentally supports his policy preference, he remains ignorant of the broader body of political science literature on the subject, much if not most of which disagrees with him.

Ironically, the very Brennan Center he assails as having a selectivity bias regarding the inclusion of evidence maintains an exhaustive literature review on the effect of voter ID, including several that von Spakovsky includes in his own (e.g. Lott 2006, Milyo 2007, Mycoff et al. 2009).

Ultimately, von Spakovsky argues that “by eschewing many of the traditional scientific methods of data collection and analysis, the authors of the Brennan Center study appear to have pursued results that advance a particular political agenda rather than the truth about voter identification.”  While I would never dare suggest that the Heritage Foundation exists to “advance a particular political agenda”, it’s clear to me that at least this one author has only the fuzziest notion of what constitutes “traditional scientific methods of data collection and analysis”, let alone logical rigor in reasoned argumentation.

My recommendation?  Get a better unpaid intern, preferably one who knows about and understands these things.