Justin Levitt before Texas House of Representatives
Justin Levitt, Counsel
Brennan Center for Justice at NYU School of Law
Texas House of Representatives, Elections Committee
Voter Fraud and Restrictive ID Requirements
April 6, 2009
On behalf of the Brennan Center for Justice at New York University School of Law, I thank the Committee for holding this hearing - and for providing the opportunity to discuss some of the facts and myths about voter fraud, and the effects that some of these myths are having on eligible American citizens.
My name is Justin Levitt, and I am counsel at the Brennan Center. The Brennan Center is a non-partisan public policy and legal advocacy organization that focuses on fundamental issues of democracy and justice. Among other things, we seek to ensure fair and accurate voting procedures and systems and to promote policies that maximize citizen enfranchisement and participation in elections. We have done extensive work on a range of issues relating to election administration, including work to remove unnecessary barriers to voter registration; to make voting machines more secure, reliable, usable, and accessible; and to expand access to the franchise and ensure that elections are conducted with integrity. Our work on these topics has included the publication of studies and reports; assistance to federal and state administrative and legislative bodies with responsibility over elections; and, when necessary, participation in litigation to compel states to comply with their obligations under federal law and the Constitution.
We have paid particular attention in recent years to claims of voter fraud, and to policies purporting to protect against fraud. We have collected allegations of fraud cited by state and federal courts, bipartisan federal commissions, political parties, state and local election officials, authors, journalists, and bloggers. We have analyzed these allegations at length, to distinguish those which are supported from those which have been debunked; furthermore, we have created and published a methodology for investigating future claims, to separate the legitimate from the mistaken or overblown. We have published a monograph reflecting our analysis, entitled "The Truth About Voter Fraud," which compiles for the first time the recurring methodological flaws behind the allegations of widespread voter fraud that are frequently cited but often unsupported. We have similarly examined claims of voter fraud in amicus briefs filed with courts around the country, including cases at the appellate level and with the Supreme Court.
We have also reviewed, in detail, the effect of policies and laws that contribute to the disenfranchisement of eligible citizens. We attempt to bring reliable data to bear on the effort to assess the nature and magnitude of the impact of new election rules, particularly those with the potential to burden eligible citizens' efforts to exercise their right to vote. In helping to quantify the impact of these rules, we have sponsored surveys and sophisticated statistical analyses; we have collected affidavits and anecdotes; and we have conducted in-depth review of voter registration forms and voter registration rolls, line by line.
In my testimony today, I will share some of our findings. Our research suggests that the incidence of fraud by those impersonating others at the polls is strikingly rare. Yet we have seen restrictions proposed to address this perceived or invented threat, often supported by stories about election fraud or abnormalities that the restrictions would not actually prevent. Further empirical research shows that the problems caused by some of these restrictions are more serious than the problems they allegedly resolve. As we stated in The Truth About Voter Fraud, "[t]he voter fraud phantom drives policy that disenfranchises actual legitimate voters, without a corresponding actual benefit."
•I. The Myth of In-Person Voter Fraud
In particular, this Committee has been asked to consider the merits of a policy requiring voters to show particular identification documents at the polls. Others today will speak at length about the impact of such policies on real, eligible citizens. In contrast, I will focus on the evidence - or, rather, lack of evidence - of any problem that could justify such an impact.
Even in theory, the policy before this Committee could provide some protection against only one form of fraud: someone arriving in person at the polls, and pretending to be someone else. As explained in more detail below, we conclude that the incidence of actual in-person impersonation fraud is extraordinarily rare. Though it does occur, there are only a handful of recent accounts, even fewer of which have been substantiated. During this same period, hundreds of millions of ballots have been cast. In the past few years in particular, enormous resources have been deployed to seek fraud out; if in-person impersonation fraud existed to any substantial degree, it would have been found. The most notable significance of the incidents that have surfaced, however, is how rare they appear to be.
We arrive at our conclusion through a focus on evidence: extensive research of reports, citations, and claims of fraud, in popular and scholarly publications, and in documents provided to and produced by public and private investigations. We have prioritized more recent claims, and particularly claims purporting to reveal in-person impersonation fraud. Our review and analysis spans thousands of accounts, including every single assertion of fraud in the most comprehensive collection of claims of in-person impersonation fraud to date: the citations presented to the Supreme Court in the Crawford v. Marion County Election Board case.
Most Allegations of Fraud Do Not Involve In-Person Impersonation Fraud
In order to assess the incidence of in-person impersonation fraud, it is first necessary to cut through a large amount of noise. Upon surveying the news after many, if not most, elections, it is easy to find allegations of election fraud of one sort or another. The vast majority of these allegations do not involve any sort of fraud that an ID bill could prevent.
Many of the apparent allegations of fraud are not actually allegations of fraud at all. Sometimes "voter fraud" stops at the headline: the story claims to involve fraud, but the substance of the story does not support the boldface nine-word summary. Another set of these reports present, not allegations that fraud has occurred, but speculation that fraud might occur in the future. Still other reports show individuals crying foul, but the facts describe straightforward administrative errors. Such errors are not to be ignored - but they are also not to be confused with fraud.
Some of these post-election reports actually do present worrisome allegations of fraud - but not in-person impersonation fraud. Instead, they allege schemes involving fraudulent absentee ballots; or absentee voters who have been coerced; or conspiracies to buy votes; or efforts to tamper with ballots or machines or counting systems. There are reports of fraudulent registration forms, though here too, the actual incidence of such reports has been overblown, and individual allegations usually involve rogue workers hoping to cheat nonprofit organizations out of an honest effort to register real citizens. We are aware of no recent substantiated case in which such registration fraud has resulted in a fraudulent vote.
These forms of fraud exist, at the margins, and they should be both prevented and punished. They are real, legitimate concerns. And requiring identification at the polls does absolutely nothing to address them.
Many Allegations of Fraud Are Also Plagued By Recurring Methodological Errors
One subset of the reports of alleged fraud is particularly toxic because, on first blush, these reports appear to entail the trappings of science, and therefore seem more conclusive than they actually are. Allegations of this sort rely on efforts to match computerized voter rolls either to other computerized voter rolls or to computerized lists of ineligible individuals, such as registries of the deceased; such efforts usually involve an attempt to match names and dates of birth from one list to another. A variant involves the attempt to screen voter rolls against property or zoning records. The results are then usually trumpeted as conclusive evidence of so many dead voters, or double voters, or voters rendered ineligible because of conviction, or voters from vacant lots.
There are common methodological concerns with such matching efforts, including both the quality of the underlying lists and the particular protocol used in matching list to list. We have reviewed these errors at length in several publications, including The Truth About Voter Fraud. Even without an error, however, it is nevertheless a mistake to draw final conclusions from these sorts of matching exercises. As Professor Michael McDonald and I have demonstrated, elementary statistics confirms that in any substantial pool, it is quite common to find two different individuals who share the same name and date of birth. When comparing one list of millions of voters to another list of millions of ineligible individuals, it should not be surprising to find hundreds of perfect "matches" that actually represent different individuals, known to record-linkage experts as "false positives." The incidence of such matches reveals statistics at work, not fraud.
Moreover, the rare cases when matching efforts do actually reveal electoral misconduct rarely involve in-person impersonation fraud. If they do show double voters or citizens intentionally voting despite a disenfranchising conviction, the wrongdoers are caught precisely because they vote using their real name. Those few individuals voting in the name of a deceased citizen find it far easier to do so by absentee ballot. Requiring identification at the polls, again, does not solve the problem.
In sum, we have reviewed many past accounts of fraud, and track contemporary accounts as they arise. Our research confirms that there are hundreds of reports of alleged fraud, in thousands of elections, with millions of ballots cast. Yet after wading through the false and irrelevant reports categorized above, only a handful of reports remain that even allege, much less substantiate, instances of in-person impersonation fraud.
Notorious Allegations of In-Person Impersonation Fraud Have Been Debunked
Even fewer of these allegations stand up to real scrutiny. Indeed, careful investigation has more often than not debunked, not confirmed, allegations of in-person impersonation fraud. One notorious and recurring example is a 2000 investigative report in the Atlanta Journal-Constitution, claiming that "the actual number of ballots cast by the dead" was "5,412 in the past 20 years." The article has been favorably cited by an Assistant U.S. Attorney General, a Governor, a state Secretary of State, and several state Attorneys General, among others.
This article did not, however, actually reveal 5,412 ballots cast by the dead, much less 5,412 instances of in-person impersonation fraud. Instead, it revealed 5,412 matches of Social Security death records to voting records. And it further revealed that these matches are flawed. The reporter acknowledged that death records contain errors, listing people as dead who are actually alive, but apparently did not investigate how many of the 5,412 identified ballots suffered from this error. The reporter also acknowledged that voter records contain errors, reflecting data entry mistakes and those who sign the wrong line of a pollbook, but apparently could not or did not investigate how many of the 5,412 identified ballots suffered from this error. The reporter neither acknowledged nor apparently accounted for the statistical likelihood that a record of John Smith dying and a record of John Smith voting might in fact reflect different "John Smith"s with the same date of birth. Finally, the reporter did not indicate how many of these 5,412 ballots were cast in person, rather than absentee.
Indeed, the article identified only one individual concretely alleged to have been the victim of in-person impersonation fraud. It cited the case of "[Alan Jay] Mandel, the tobacco shop owner, whose voter certificate was signed at the polls by someone after his death." Repeated the reporter, "[S]omebody definitely signed his name on a voter certificate on Nov. 3, 1998."
This allegation, though amounting to only one concrete allegation of in-person impersonation fraud in approximately twenty million votes over 20 years, would nevertheless be disturbing - if it were true. Further investigation, however, proved the allegation false. The signature (and voter certificate) in question belonged to Alan J. Mandell (with two "l"s), who was very much alive and eligible in 1998, but whose vote was mistakenly recorded in the name of Alan Jay Mandel (with one "l").
Investigation as thorough as the investigation into the vote of Mr. Mandel/Mandell is rare. Nevertheless, when researchers do expend the effort to follow through on initial allegations of in-person impersonation fraud, they often find those allegations to be unwarranted. A 2007 investigation of approximately 100 "dead voters" in Missouri, for example, revealed that every single purported case was properly attributed either to a matching error, a problem in the underlying data, or a clerical error by elections officials or voters. Likewise, after compiling a list of potential "dead voters" in New York state, a Poughkeepsie journalist investigated seven local cases - and found that seven out of seven reflected clerical errors or other mistakes, not fraud. An investigation in Hawaii in 1999, after reviewing precinct pollbooks and calling allegedly deceased citizens, similarly found that not one of 170 potential "dead voters" actually reflected fraud.
Texas had a similar investigation just last year. In a story headlined "Dead Voters Cast Ballots in Dallas County" - a headline later changed to reflect reality - 48 specific votes were called into question, as potentially cast by those impersonating the deceased. The Dallas Morning News chose to investigate further, scanning the poll books and signature rosters that were available for 47 of those votes, and found clerical error, voter mistake, or just confusion on the part of those crying wolf. Careful review revealed that of all the cases investigated, "none involved a fraudulently cast vote."
This is not to say that in-person fraud never, ever happens; there have been a few notorious anecdotes where the allegations do pan out. A 14-year conspiracy in Brooklyn from the late 1960s to the early 1980s, for example, appears to have involved widespread tampering on the part of election officials and their co-conspirators, including impersonation fraud at the polls. Similar allegations surfaced in the recent prosecution of Ike Brown in Mississippi. These were widespread schemes of official corruption by local "bosses," involving wrongdoing by those who are employed to safeguard the process. They should be roundly condemned. But here too, voter identification requirements do little good. Policies aimed at policing voters will not be able to stop insiders from corrupting the system. It is impossible to stop local bosses intent on breaking the law by giving them a new law to break.
In sum, every once in a great while, a report of in-person impersonation fraud, or a report of an attempt at in-person impersonation fraud, appears to be substantiated; even more rarely, there will be a substantiated report of the sort that restrictive identification requirements could possibly prevent. It has been known to happen. What is notable, however, is how rarely it has been known to happen. Americans are struck and killed by lightning more often.
Recent Examples Fit the Pattern: Little to No Evidence of In-Person Impersonation
Three recent and very prominent investigations into voter fraud precisely fit the overall pattern that I have described above: many claims of wrongdoing and irregularity, but few that even allege in-person impersonation fraud, and a tiny portion, if any, that substantiate the allegations. These examples are: the compilation of alleged fraud submitted to the Supreme Court in Indiana's Crawford v. Marion County Election Board case, the investigation by the Milwaukee Police Department of fraud in Milwaukee in 2004, and the recent investigation by the Texas Attorney General of fraud in this state.
First, a few words on Crawford. The Supreme Court's Crawford case is often said to have validated laws requiring photo identification at the polls. It did no such thing. In Crawford, the fractured court rejected the plaintiffs' challenge to the law as overbroad, in light of the limited evidence in the record on the extent of the law's burdens. That is, without solid proof of burden in the record, Indiana's asserted justifications were legally sufficient to sustain the law against the particular facial challenge that was lodged. The Court did not issue a blanket statement declaring restrictive identification laws to be legal. And it certainly did not validate ID laws as a matter of good policy.
The policy decision, instead, must start with the rationale for such a law. Because the Supreme Court represented such a high-profile forum, it provided the most prominent focal event to date for supporters of an identification law to justify their support by showing their rationale to be real. Crawford was a national stage for those who believe in-person impersonation fraud to be a legitimate concern to present their proof. In the case, the lower courts cited several media accounts that, the courts claimed, reflected reports of in-person impersonation fraud. In the Supreme Court, respondents and amici supporting respondents added citations to more than 250 reports, encompassing decades of elections.
The Brennan Center thoroughly examined each one of these citations. The evidence of in-person impersonation fraud was strikingly sparse. The vast majority of cited reports reflected either allegations that could not possibly be related to in-person impersonation fraud and which an identification law could not possibly fix (e.g., absentee ballot problems, or vote-buying schemes, or ballot tampering), or allegations that did not mention whether the alleged wrongdoing was committed in-person or through more susceptible absentee ballots. In elections encompassing hundreds of millions of votes overall, the cited reports mentioned only eleven votes and two attempted but unsuccessful votes that allegedly involved in-person impersonation fraud. Two of the allegations - and the two unsuccessful attempts - were substantiated. The other nine might just have readily involved the same sorts of clerical errors uncovered in the investigations discussed above.
A recent investigation by the Milwaukee Police Department highlights the same pattern. After the 2004 election, initial media accounts featured front-page allegations of widespread fraud in Wisconsin. On February 26, 2008, the Milwaukee Police Department released a report on that election, with what appears to be a painstaking investigation of the facts, and policy recommendations offered with less care and disavowed by the Milwaukee Police Chief. The department's careful factual investigation primarily revealed administrative mistakes and, occasionally, negligence. It showed that much of what had originally been identified as potential fraud was in fact due to clerical error. It also uncovered several votes by potentially ineligible individuals, including some who were allegedly nonresidents, and some who had allegedly been rendered ineligible due to convictions. The report revealed only one potential vote that might have involved in-person impersonation fraud, with no documentation verifying that the vote in question was actually cast.
Here in Texas, the pattern appears the same. Texas Attorney General Abbott spent two years and $1.4 million dollars to fight voter fraud. His efforts yielded indictments for incidents of wrongdoing (and for actions where wrongdoing was vigorously disputed), but not a single case that identification rules could possibly prevent.
Existing Safeguards Would Disclose In-Person Impersonation Fraud
Some have claimed that the incidence of alleged in-person impersonation fraud is extremely low because in-person impersonation fraud is difficult to detect. This is distinct from the issue of whether in-person impersonation fraud is difficult to prosecute: littering clearly exists, but is difficult to address through the criminal justice system, because the wrongdoer is not easily identified. Here, not only are there virtually no prosecutions of in-person impersonation fraud, but there are even strikingly few reports of potential impersonation. It is as if individuals were complaining about littering, but could find no garbage in the street. For those believing in impersonation at the polls, the answer is that this sort of fraud is simply difficult to detect.
In truth, there are multiple means to discover in-person impersonation fraud, all of which should yield many more reports of such fraud, if it actually occurred with any frequency. An individual seeking to commit in-person impersonation fraud must, at a minimum, present himself at a polling place, sign a pollbook, and swear to his identity and eligibility. There will be eyewitnesses: pollworkers and members of the community, any one of whom may personally know the individual impersonated, and recognize that the would-be voter is someone else. There will be documentary evidence: the pollbook signature can be compared, either at the time of an election or after an election, to the signature of the real voter on a registration form, and the real voter can be contacted to confirm or disavow a signature in the event of a question. There may be a victim: if the voter impersonated is alive but later arrives to vote, the impersonator's attempt will be discovered by the voter. (If the voter impersonated is alive and has already voted, the impersonator's attempt will be discovered by the pollworker; if the voter impersonated is deceased, it will be possible to cross-reference death records with voting records, as described above, and review the actual pollbooks to distinguish error from foul play.) If the impersonation is conducted in an attempt to influence the results of an election, it will have to be orchestrated many times over, increasing the likelihood of detection.
As in all law enforcement, none of these detection mechanisms are perfect. Yet in hundreds of millions of ballots cast, they have yielded only a handful of potential instances of in-person impersonation fraud, precisely during a period when investigating voter fraud was expressly deemed a federal law enforcement priority, and when private entities were equipped and highly motivated to seek, collect, and disseminate such reports. Every year, there are far more reports of UFO sightings. The scarcity of reports of in-person impersonation fraud, in this context, is itself meaningful.
In-Person Impersonation Fraud Is Irrational
Instead, a more logical explanation for the extraordinary rarity of reported in-person impersonation fraud is that in-person impersonation fraud is extraordinarily rare. In-person impersonation fraud is an extremely inefficient means to influence an election. For each act of in-person impersonation fraud in a federal election, the perpetrator risks 5 years in prison and a $10,000 fine under federal law, in addition to the ten years and $10,000 that could be assessed under Texas law. In return, the perpetrator gains at most one incremental vote. It is sensible that few individuals believe such a trade-off worthwhile.
•II. The Consequences of the Myth
If perpetrating the myth of in-person impersonation fraud had no consequences, it would likely be of little concern to this Committee. There are consequences, however, and these consequences have been and can be quite serious.
First, "crying wolf" about voter fraud, and particularly in-person impersonation fraud, perpetuates the lack of public confidence in the integrity of the election process. Second, it distracts attention from the real and recurring problems that do commonly undermine our elections, including some of the other forms of fraud or administrative irregularity I have discussed above. Third, it can be used to justify placing undue and improper pressure on impartial prosecutors to bring unwarranted criminal prosecutions.
"Crying wolf" about voter fraud has also been used to justify policies that lead to the disenfranchisement of eligible citizens, including the imposition of overly restrictive voter identification requirements. The argument that is usually made - and we have catalogued many such attempts - is a logical non-sequitur. Advocates will identify various deficiencies with the election process, none of which could be solved by requiring more restrictive identification at the polls. And they will then argue that restrictive identification is therefore necessary, as snake oil tonic to combat ills that it cannot possibly combat.
Claims of Fraud Are Used to Justify Policies that Cause Real Problems
Moreover, not only are restrictive identification requirements poorly tailored to any existing problem of any magnitude, but reliable empirical data demonstrates that they cause problems of their own.
The most prominent restrictive voter identification proposals would require government-issued photo identification as a condition of voting a valid ballot at the polls. Most eligible voting-age citizens have such identification. Studies have repeatedly shown, however, that many eligible voting-age citizens do not.
Furthermore, for those who do not currently have government-issued photo identification, obtaining it can be a burden. In some cases, it takes ID to get ID - for example, a certified birth certificate may be required to obtain government-issued photo identification, but government-issued photo identification may be required to get a certified birth certificate. Even when it is possible to get this underlying documentation, doing so costs time and money, with fees up to $380 for a replacement certificate of naturalization. Bringing the assembled paperwork to the agency that distributes the photo identification costs further time and money, as individuals must travel - without a valid driver's license - to a particular office during government hours. For individuals with disabilities or lower-income or more elderly citizens, these burdens may become particularly acute.
In the two states where government-issued photo identification is required to cast a valid ballot, these laws have already contributed to the disenfranchisement of eligible citizens. In one Indiana county's 2007 elections, for example, 32 voters arriving without acceptable photo identification cast ballots that were not counted, apparently solely because of the voter identification law. Fourteen of these voters had previously voted in at least ten previous elections at these same polling places. More recently, in Georgia's 2008 presidential primary election, 296 voters arriving without acceptable photo identification reportedly cast ballots that were not counted, again apparently solely because of the ID law. Reports from the 2008 general election are currently being assembled. It is impossible to know exactly how many additional individuals without identification arrived at the polls but did not cast the futile provisional ballots, or declined to make the futile trip to the polls. In less than two years, however, there are already reports of far more individuals adversely affected by restrictive photo identification laws than the total cited instances of in-person impersonation fraud reported over the last few decades.
There are also disturbing indications that far more eligible citizens may be affected by such laws in the future. Reliable surveys of registered voters conclude that restrictive identification laws do impact eligible citizens, and that they disproportionately impact minority and elderly populations. Although there is some disagreement over the precise magnitude of the effect, even the studies with more modest results estimate an impact that would reach more than two million registered voters if applied nationwide. Surveys of voting-age citizens - including eligible citizens who have not been engaged in the election process, but whom we should hope to engage - are somewhat less fully developed, but find an even more substantial effect, and similar disproportionate impact.
Some seek to justify these restrictive laws, despite their demonstrated impact on many American citizens, and despite the fact that they do not correct an existing problem with recurring in-person impersonation, by claiming that they will at least increase public confidence in the election process. Even if the unfounded fears of the many were sufficient justification to burden the constitutional rights of the few, however, a careful study cited in the Harvard Law Review casts serious doubt on the validity of such assertions. The data show no support for the notion that requiring identification will increase voter confidence; the study found no statistically significant correlation between the rate at which citizens were asked to produce photo ID and their perception that either voter fraud generally, or voter impersonation in particular, exists. Photo identification laws do not, in short, appear to make citizens feel more secure about their elections.
There is No Reason to Believe that Existing Protections in Texas are Ineffective
Calls for increased restrictions should cause particular skepticism in Texas, which already has protections in place that adequately balance this Committee's apparent concerns: protecting against the unlikely possibility of in-person impersonation fraud without disenfranchising eligible voters. That is, even if in-person impersonation fraud were a significant concern, Texas would not be particularly vulnerable. Protections already exist against impersonation fraud at the polling place.
First, there is the system provided by the federal Help America Vote Act: any new voter who registers by mail must either have information on their voter registration form verified against motor vehicle or Social Security data, or present documentary proof of identity, from a broad list calculated by Congress to minimize the chance that an eligible voter will be unduly turned away.
Then, once a voter has been registered, Texas law requires the registrar to deliver a registration certificate; successful delivery confirms the voter's name and address. The voter must sign the certificate, with a signature that matches his registration form. At the polls, after again signing the roster, voters must either show this same registration certificate, or sign an affidavit and present an identification document - again, from the same broad list designed by Congress to minimize disenfranchisement.
All of the available evidence indicates that these existing safeguards work. Eligible Texans vote; people impersonating those eligible Texans do not - at least, not at the polls or without insider assistance. There appears to be little reason to break that which is not currently broken.
* * *
Given the amount of speculation and misinformation in the public sphere concerning in-person impersonation fraud, and restrictions ostensibly intended to address such fraud, we thank the Committee for holding this hearing. This represents a welcome effort to ensure that the serious policy debate around election reform remains grounded in the facts.
The available empirical research shows that although in-person impersonation fraud is an occurrence of extraordinary rarity, it has been used to justify policies that appear to offer little benefit and impose substantial cost. The existing safeguards and deterrents have been successful in preventing in-person impersonation fraud to any significant degree; further measures are not only unnecessary, but risk compromising the integrity of our elections to the extent that they shut out eligible citizens.
In contrast, there remain serious concerns about aspects of our election process where existing safeguards are not sufficient, and where remedies would not risk harm to eligible voters - including, for example, threats to the security of our voting systems, and concerns with the transparency of purges of the voter rolls. We still face substantial challenges in ensuring that all eligible citizens are able to exercise the franchise effectively, in Texas and beyond, and we look forward to the Committee's efforts to achieve this common goal.
Thank you very much.
 Justin Levitt, The Truth About Voter Fraud (2007), available at http://www.brennancenter.org/content/resource/truthaboutvoterfraud/.
 Brief of the Brennan Center for Justice et al. as Amici Curiae Supporting Petitioners, Crawford v. Marion County Election Board, 128 S. Ct. 1610 (2008) ("Brennan Center Brief"); Brief of Brennan Center for Justice at NYU School of Law as Amicus Curiae in Support of Plaintiffs-Appellants and Reversal, Crawford v. Marion County Election Board, 472 F.3d 949 (7th Cir. 2007); Brief Amicus Curiae of the Brennan Center for Justice at NYU School of Law in Support of Plaintiffs/Appellants and Reversal, Gonzalez v. Arizona, 485 F.3d 1041 (9th Cir. 2007); Brief of Brennan Center for Justice at NYU School of Law as Amicus Curiae in Support of Plaintiffs/Appellees and Affirmance, ACLU of N.M. v. Santillanes, 546 F.3d 1313 (10th Cir. 2008); Brief of Brennan Center for Justice at N.Y.U. School of Law as Amicus Curiae in Support of Plaintiffs-Appellants and Reversal, Common Cause/Georgia v. Billups, 554 F. 3d 1340 (11th Cir. 2009); Brief of Brennan Center for Justice at N.Y.U. Law School as Amicus Curiae in Support of Plaintiffs-Appellees and Affirmance, Common Cause/Georgia v. Cox, Case No. 05-15784-G (11th Cir. Jan. 14, 2006); Brief of Brennan Center for Justice at NYU School of Law as Amicus Curiae in Support of Plaintiffs' Motion for Summary Judgment, ACLU of N.M. v. Santillanes, 506 F. Supp. 2d 598 (D.N.M. 2007).
 Levitt, supra note 1, at 6.
 See, e.g., Megan Matteucci, Riverdale Suit Alleges Election Fraud, Atlanta J.-Const., Nov. 20, 2007; Sandy Coleman, Randolph Petition Claims Voter Fraud, Boston Globe, May 21, 2006, at 5.
 See, e.g., Hoosier Voter Lists Loaded With Bogus Names, South Bend Tribune (Ind.), Nov. 6, 2000, at C4. Such speculation, for example, may be driven by poor maintenance of the registration rolls, which proper implementation of the National Voter Registration Act of 1993 and the Help America Vote Act of 2002 should correct. See 42 U.S.C. §§ 1973gg-6; 15483(a)(2), (4).
 See, e.g., Michelle Hillen, Recount of Runoff Reverses 1st Result, Ark. Democrat-Gazette, June 20, 2006.
 See, e.g., Pabey v. Pastrick, 816 N.E.2d 1138, 1144-46 (Ind. 2004); In re The Matter of the Protest of Election Returns and Absentee Ballots in the November 4, 1997 Election for the City of Miami, Florida, 707 So.2d 1170 (Fla. Ct. App. 1998).
 See, e.g., Anastasia Hendrix, City Workers: We Were Told To Vote, Work for Newsom, S.F. Chronicle, Jan. 15, 2004; Matthew Purdy, 5 Bronx School Officials Are Indicted in Absentee Ballot Fraud, N.Y. Times, Apr. 25, 1996.
 See, e.g., Beth Musgrave, Three Sentenced in Bath Vote Fraud, Lexington Herald-Leader, Sept. 25, 2007; Nicklaus Lovelady, Investigation Into Vote Fraud in Benton County Nets 14th Arrest, Miss. Clarion-Ledger, Aug. 31, 2007; Tom Searls, Six To Learn Fate in Lincoln Vote Buying Case, Charleston Gazette (W.Va.), May 3, 2006, at 1C; Michael E. Ruane, FBI's Sham Candidate Crawled Under W. Va's Political Rock, Wash. Post, Dec. 2, 2005, at A1.
 See, e.g., John M. Glionna, S.F., State Wade Into Vote Count Controversy, L.A. Times, Nov. 21, 2001.
 See, e.g., Todd C. Frankel, 8 Charged in StL Voter Fraud, St. Louis Post-Dispatch, Dec. 21, 2007; Keith Ervin, Felony Charges Filed Against 7 in State's Biggest Case of Voter-Registration Fraud, Seattle Times, July 26, 2007; Carlos Campos, Bogus Voter Forms Pop Up in Fulton, Atlanta J.-Const., Oct. 21, 2004.
 See, e.g., David W. Chen, Among Voters in New Jersey, G.O.P. Sees Dead People, N.Y. Times, Sept. 16, 2005, at B5.
 See, e.g., Russ Buettner, Exposed: Scandal of Double Voters, Daily News (N.Y.), Aug. 22, 2004.
 See, e.g., Scott Hiaasen et al., Felon Purge Sacrificed Innocent Voters, Palm Beach Post, May 27, 2001, at A1.
 See, e.g., Greg J. Borowski, GOP Demands IDs of 37,000 in City, Milwaukee J. Sentinel, Oct. 30, 2004.
 Levitt, supra note 1; see also, e.g., Brennan Center for Justice, Investigator's Guide to "Voter Fraud" (2006), at http://www.brennancenter.org/content/resource/investigators_guide_to_voter_fraud/.
 Michael P. McDonald & Justin Levitt, Seeing Double Voting: An Extension of the Birthday Problem, 7 Elect. L.J. 111 (2008).
 Jingle Davis, Even Death Can't Stop Some Voters, Atlanta J.-Const., Nov. 6, 2000.
 Letter from William E. Moschella, U.S. Assistant Attorney General, to Christopher S. Bond, U.S. Senator (Oct. 7, 2005), at http://www.usdoj.gov/crt/voting/misc/ga_id_bond_ltr.htm.
 Darryl Fears & Jonathan Weisman, Georgia Law Requiring Voters to Show Photo ID Is Thrown Out, Wash. Post, Sept. 20, 2006, at A6.
 Brief of State Respondents, at 2, Crawford v. Marion County Election Board, 128 S. Ct. 1610 (2008).
 Brief of Texas et al. as Amicus Curiae Supporting Respondents, at 8, Crawford v. Marion County Election Board, 128 S. Ct. 1610 (2008).
 Over twenty years and approximately 20 million votes cast, even if all 5,412 ballots had in fact been fraudulent, the overall rate of fraud would have been 0.027%. Secretary of State Cathy Cox, The 2000 Election: A Wake-Up Call For Reform and Change 11 n.3 (2001), available at http://www.sos.state.ga.us/acrobat/elections/2000_election_report.pdf. In reality, many of these ballots are likely attributable to clerical error, data error, or statistical coincidence.
 Indeed, if the voter voted in several elections over the twenty-year span reflected in the voter history, each such mismatch would likely account for several "false positive" ballots. Thus, the 5,412 identified ballots might reflect far fewer matched - or mismatched - voters.
 Davis, supra note 18.
 Id. (emphasis added).
 Cox, supra note 23, at 11 n.3.
 Jingle Davis, State Plans to Update Voter Lists, Atlanta J.-Const., Feb. 10, 2001, at 4H; Cox, supra note 23, at 11 n.3.
 Steve Chamraz, "The Dead List" Backstory . . . , News 4 Daily Briefing, May 7, 2007, at http://www.beloblog.com/KMOV_Blogs/n4idailybriefing/2007/05/the_dead_list_backstory.html.
 John Ferro, Deceased Residents on Statewide Voter List, Poughkeepsie J., Oct. 29, 2006.
 Waite David, Review Turns Up No Signs of Fraud, Honolulu Advertiser, Mar. 16, 1999, at A1.
 Lee Ann O'Neal, Dead Voters Cast Ballots in Dallas County, Texas Watchdog, Oct. 30, 2008.
 Rudolph Bush, Review Finds No Vote Fraud in Dallas County, Dallas Morning News, Dec. 13, 2008.
 Frank Lynn, Boss Tweed Is Gone, But Not His Vote, N.Y. Times, Sept. 9, 1984.
 United States v. Brown, 494 F. Supp. 2d 440, 486 n.73 (S.D. Miss. 2007), aff'd, 2009 WL 485709 (5th Cir. 2009).
 Michael Cass, Poll Worker Indicted in Vote Probe, The Tennessean, Dec. 20, 2007 (discussing one incident in Tennessee in 2007); Madeline Friedman, Anatomy of Voter Fraud, Hoboken Reporter, July 1, 2007 (discussing one attempt in New Jersey in 2007); Brown, 494 F. Supp. 2d at 486 n.73 (S.D. Miss. 2007) (discussing one incident in Mississippi in 2005); Manny Garcia & Tom Dubucq, Unregistered Voters Cast Ballots in Dade: Dead Man's Vote, Scores of Others Were Allowed Illegally, Herald Finds, Miami Herald, Dec. 24, 2000 (discussing one incident in Florida in 2000); Larry J. Sabato & Glenn R. Simpson, Dirty Little Secrets 292 & n.70 (1996) (citing Doug Haaland and Doug Swordstrom, A Report on Election Law Irregularities: California 16th Senate District 10 (1995)) (discussing one attempt in California in 1994); Lynn, supra note 34 (discussing a scheme in Brooklyn primarily in the 1970s).
 Of the reports in note 36, supra, several indicate that election officials were complicit in the scheme, or that the individual in question produced fraudulent photo identification. Cass, supra note 36; Brown, 494 F. Supp. 2d at 486 n.73; Garcia & Dubucq, supra note 36; Lynn, supra note 34.
 Nat'l Oceanic & Atmospheric Admin., Nat'l Weather Service, Lightning Safety: Medical Aspects of Lightning, at http://www.lightningsafety.noaa.gov/medical.htm (last visited Mar. 7, 2008); Ronald L. Holle, Lightning Deaths by State, 1997 to 2006 (2007), at http://www.lightningsafety.noaa.gov/stats/1997-2006_Fatalities+Rates.pdf.
 Crawford v. Marion County Election Board, 128 S. Ct. 1610, 1615, 1622-23 (2008) (Stevens, J.). Part of the difficulty is that the case was a pre-enforcement challenge, brought before Indiana's law was put into effect and therefore without direct evidence of past harm. See generally Justin Levitt, Crawford-More Rhetorical Bark than Legal Bite?, May 2, 2008, at http://www.brennancenter.org/blog/archives/crawford_more_rhetorical_bark_than_legal_bite/.