Skip Navigation

Investigator’s Guide to Voter Fraud

In the aftermath of a close election, some are quick to blame voter fraud for the results. Allegations of fraud, however, are often greatly exaggerated on closer examination, the claims amount to a great deal of smoke without much fire.

Published: October 31, 2006


In the aftermath of a close election, some are quick to blame voter fraud for the results. Allegations of fraud, however, are often greatly exaggerated on closer examination, the claims amount to a great deal of smoke without much fire. In an effort to focus the upcoming election on the facts, we offer this guide to common flaws in claims of voter fraud. We urge reporters, analysts, advocates, and scholars interested in the election process to ask the following questions when confronting a claim that voter fraud has occurred.

Is the alleged problem explained by undue reliance on faulty lists or a flawed list-matching process?

Claims of voter fraud are often premised on attempts to compare lists of voters to lists of people ineligible to vote. However, the process of matching information from list to list is full of pitfalls. The following questions will reveal some of the more common ways in which fraud claims based on matching may be bogus.

Are the underlying lists accurate?
Large databases are vulnerable to human error and other inaccuracies. For example, the Social Security Administrations Master Death Index, often used to identify voters who are allegedly deceased, is known to have an error rate of more than 3%.[1] Such errors are compounded over time: the leading expert on list-matching for the U.S. Bureau of the Census estimates that in a large California employment database, [o]ver a period of twenty years, the records [associated] with each individual can expect to contain at least two errors where the [Social Security Number] has been mis-keyed or transcribed improperly.[2]

Were the underlying lists compiled for a different purpose?
Databases that were not compiled for election or identification purposes may contain information that proves misleading in the voting context. In St. Louis in 2000, based on city property records, voters were alleged to have registered from vacant lots. But the property records apparently classified a multi-parcel address as vacant even if only one of the parcels was vacant; further investigation by local reporters revealed that the supposedly vacant lots contained valid residences.[3]

Were the underlying lists missing important information (e.g. date of birth)?
Those searching for fraud often compare identifying information on lists to allege that two entries represent the same person. In New Jersey in 2005, several thousand voters were alleged to have committed fraud based on such a comparison. But records in one county a county disproportionately represented in the claims of fraudulent votes were missing birthdates entirely. Some voters were listed only with month and year of birth. And others were listed with a birthdate of January 1, 1880 which was most likely a system default for missing information.[4] Given the missing information, it was unwarranted to conclude that two individuals sharing the same name alone represented the same person.

Did the comparison process accept a partial match of any field?
Comparing information accurately on two different lists requires precision. In contrast, loose matching criteria produce what are known as false positives: records that appear to correspond to the same individual, but in fact do not. In Florida in 2000, in assembling a set of voters to be purged, a vendor found matches in the first name if the first four letters were the same on two different lists, and matches in the last name if 80% of the letters were the same.[5] The final set of voters to be purged, of course, contained the names of many individuals whose records had been falsely matched.

Did the match process take middle initials or suffixes (e.g. Jr., Sr., III) into account?
The same Florida procedure described above also failed to account for different suffixes and middle initials, so that Rev. Willie D. Whiting, Jr., was confused with Willie J. Whiting.[6] Allegations in New Jersey in 2005 showed the same problem: James A. Smith and James G. Smith were presumed to be the same person, as were J. T. Kearns and J. T. Kearns, Jr.[7] And in New Hampshire, 22 pairs of people who shared the same first and last names were flagged for possible double-voting; in fact, all of the flagged voters have different middle names.[8]

Did the match process take the birthdate problem into account?
Even given an exact match, two entries with the same name and birthdate may not represent the same individual. Statistics students are often surprised to discover that in a group of 23 people, it is more likely than not that two will share the same month and day of birth; in a group of 150, two will probably share the same birthdate. In any group of significant size, statistics tell us that there will be enough people with the same name that two will be born on the same day.[9] Therefore, it should not be surprising that Kathleen Sullivan was most likely listed twice on the 2004 New Jersey voter rolls not because one woman drove 161 miles to cast a second ballot, but because two women named Kathleen Sullivan share the same birthdate.[10]

Is the alleged problem explained by basic recordkeeping errors?

Minor human error infects many election records: typos and misspellings on voter registration lists, the wrong name marked on a pollbook. Though these errors may not alter the result of any election, they could well create the appearance of fraud where none exists.

Is there any way to be sure that a data entry error didnt lead to false information?
Typographical errors happen. In Washington State in 2006, Marina Petrienko tried to register to vote for the first time, but a county official mis-typed the year of her birth, entering 1976 into the database, instead of the year on her form: 1975. First-time Illinois voters Mike and Sung Kim had been mistakenly registered with Kim as their first names in 2004.[11] And in Milwaukee, Victor Moy was listed on the rolls as living at 8183 W. Thurston Avenue, but actually resides at number 8153.[12] Because such typos may prevent registrations from being externally validated by information in other sources, officials and observers may believe that registrations are fraudulent when they are, in reality, entirely legitimate.

Is there any way to be sure that the correct name was marked in the pollbooks?
Mistakes are made at the polls as well. In a jurisdiction of any significant size, it is unfortunately easy to make an entry in the pollbook next to the wrong voters name. For example, despite having died in 1997, Alan J. Mandel was alleged to have voted in 2000; Alan J. Mandell (two ls), who was very much alive and voting at the time, explained that local election workers simply checked the wrong name off of the list.[13] The same problem may occur when information from a pollbook is entered incorrectly into a countys computer system, as in Milwaukee in 2004.[14] Or voters legitimate voters may make a mistake: a 1994 investigation of fraud allegations in California, for example, revealed that voters accidentally signed the poll books on the wrong lines, next to the names of deceased voters.[15]

Is there a legitimate explanation for the alleged facts?

On occasion, those who claim voter fraud jump to conclusions not justified by the data in front of them. Upon further investigation, the facts often reveal that the alleged misconduct was actually entirely legitimate.

Did an individual actually live at an address zoned for business use?
In most states, voters must register at a residential address; those looking for fraud may flag addresses zoned for business use as an indication of fraudulent activity. Broad zoning restrictions, however, do not account for many less traditional but legitimate residences. Barbara Taylor was among hundreds of Washington voters challenged in 2005 for this reason. While it is true that the address on her registration was the address of a public storage facility, Taylor explained that she is a manager for the company and has lived in an apartment on the site for 12 years.[16]

Did a person register twice but vote only once?
Registering twice or mistakenly leaving an old registration on the rolls is not meaningful evidence of voting twice. In 2004, for example, federal prosecutors charged 23-year-old Wisconsinite Cynthia Alicea with double-voting. Wisconsin allows residents to register on Election Day, which Alicea did. Pollworkers found an error on the form, and asked Alicea to fill out another, which she also did but the first form was never discarded. Although Alicea completed two registration forms, she voted only once. Yet the two registration forms sufficed to take Alicea to trial. She was promptly acquitted.[17]

Did a person listed as deceased die after casting a vote?
Voting from the grave offers salacious headlines, and investigators often attempt to match death records to voter rolls in an attempt to find fraud. Yet in addition to the problems with inaccurate matching identified above, death records may only give the month or year of death concealing citizens who voted before dying, in quite ordinary fashion. In Maryland in 1995, for example, an exhaustive investigation revealed that of 89 alleged deceased voters, none were actually dead at the time the ballot was cast. The federal agent in charge of the investigation said that the nearest they came was when they found one person who had voted then died a week after the election.[18] Similarly, in New Hampshire, postcards were sent to the addresses of citizens who voted in the 2004 general election; one card was returned as undeliverable because the voter died after election day, but before the postcard arrived at her home.[19]

Was a person convicted of a crime not convicted of a disqualifying felony?
Reports of votes by persons with convictions have often fed claims of fraud. Yet in the vast majority of states, most convictions do not affect the defendants voting rights. Wallace McDonald, 64, was purged from the Florida voter rolls in 2000 because of a conviction. Yet Mr. McDonalds crime was not a felony, for which Floridians forfeit voting rights forever but merely a misdemeanor, which should not affect voting rights at all. Indeed, Mr. McDonald had been convicted only of falling asleep on a bench.[20] Similarly, in Washingtons 2004 gubernatorial election, hundreds of citizens were alleged to have voted illegally because of convictions that were actually juvenile dispositions which do not disqualify voters.[21]

Did a person convicted of a disqualifying felony have her voting rights legally restored?
Even in Florida, where persons with felony convictions lose their voting rights permanently, not every person convicted of a felony is ineligible to vote. Reverend Willie Dixon, 70, was purged from the Florida voter rolls in 2000 because of a felony conviction but Reverend Dixon had been pardoned for his crime and his voting rights had been restored.[22] In most other states, persons with convictions automatically regain the franchise after release from either incarceration, probation, or parole. Allegations of fraud that look to convictions without accounting for the restoration of voting rights will often miss the mark.

Does the person alleging fraud propose a policy to solve the problem, and is the proposed policy the best solution to the problem alleged?

Allegations of fraud may serve the interests of those who favor particular changes to existing election law. Some of these proposals are ostensibly justified by the need to protect against fraud, but are actually poorly tailored to serve that purpose. Investigators should carefully probe the fit between those fraud allegations that bear scrutiny and any policies proposed to address voter fraud.

Does the proposed policy require photo identification?
Alleged fraud is most frequently used to justify proposals requiring photo identification of voters, including the controversial new Federal Election Integrity Act of 2006 passed by the House of Representatives in September.[23] In 2006 alone, a quick search of news archives showed that at least 50 newspaper op-eds and editorials and at least 272 other published articles attempted to make a connection between the need for photo identification and the need to reduce voter fraud.

Does the proposed policy actually solve the problem?
Some of those who support photo identification requirements indiscriminately cite examples of fraud, whether the cited anecdotes can be remedied by photo identification or not. In Wisconsin in 2005, for example, supporters of a restrictive identification requirement pointed repeatedly to voting by allegedly ineligible persons with convictions even though requiring restrictive ID would not prevent voting by persons who are rendered ineligible by a conviction.[24]

Does it leave a substantial loophole (e.g. absentee ballots)?
In 2005, the Georgia legislature passed a law requiring voters to show photo identification at the polls. A 1996 county vote-buying scheme was cited as justification, despite the fact that the 1996 scheme involved absentee ballot fraud, which the new photo-identification law would do nothing to prevent.[25] Indeed, by exempting absentee ballots entirely, the Georgia photo-ID law left open the most commonly cited vehicle for the occasional acts of voter fraud that have been proven.[26]

Would the alleged problem be solved by proper implementation of existing federal law?
Many of the more commonly proffered allegations of voter fraud involve concerns that should be resolved by existing federal law, once new statewide voter registration databases have been fully implemented. The Help America Vote Act of 2002 (HAVA) imposes an identification requirement on first-time voters who have registered by mail, as long as their registration information has not been externally validated.[27] It also requires states to purge voters confirmed as deceased from the rolls, and to coordinate voter registration lists with computerized sources regarding the voting rights status of persons who have been convicted of felonies.[28] In addition to the other list maintenance required by HAVA, these three simple requirements should eventually eliminate any significant threat caused by the most commonly cited sources of alleged voter fraud.

Does the proposed policy create more problems than it solves?
Legitimate cases of fraud that could be addressed by a photo identification requirement are proven to occur approximately as often as Americans are struck and killed by lightning.[29] Given the frequency of the problem, proposed solutions may be more harmful than helpful. Restrictive photo identification requirements, for example, will likely have an impact that far exceeds the negligible rate of voter fraud. Up to 10% of the voting-age population does not have state-issued photo identification.[30]This rate is disproportionately higher among minorities, low-income populations, youth, and the elderly. A recent Wisconsin study, for example, found that 78% of black men aged 18–24 had no valid drivers license.[31]


[1] Preventing Identity Theft by Terrorists: Joint Hearing on SSNs and Identity Theft Before the Subcomm. on Oversight and Investigations of the H. Comm. on Financial Services, and the Subcomm. on Social Security of the H. Comm. on Ways and Means, 107th Cong. (2001) (statement of Marc Rotenberg, Executive Director, Electronic Privacy Information Center), available at
[2] WILLIAM E. WINKLER, QUALITY OF VERY LARGE DATABASES (Bureau of the Census Stat. Res. Div., Statistical Research Report Series No. RR2001/04, 2001).
[4] Brennan Center for Justice at NYU School of Law & Michael McDonald, Preliminary Analysis of the September 15, 2005 Report Submitted to the New Jersey Attorney General by the New Jersey Republican Party 6–7 (2005) [hereinafter New Jersey Fraud Analysis].
[5] Greg Palast, The Wrong Way to Fix the Vote, WASH. POST, June 10, 2001, at B1.
[6] Id.
[7] New Jersey Fraud Analysis, supra note 4, at 2.
[8] Memorandum from Bud Fitch, Deputy Atty Gen., N.H. Dept of Justice, to Robert Boyce, Chairman, N.H. Sen. Internal Affairs Comm., et al. (Apr. 6, 2006),
[9] Michael P. McDonald & Justin Levitt, Seeing Double Voting 11 (Aug. 30, 2006) (unpublished manuscript, presented at the 2006 American Political Science Association Conference).
[10] Id.
[11] Burt Constable, 1st-time Voters Join Immigrant Turnout, CHI. DAILY HERALD, Nov. 5, 2004, at 1.
[12] See e.g., Greg. J. Borowski, GOP Fails to Get 5,619 Names Removed From Voting Lists, MILWAUKEE J. SENTINEL, Oct. 29, 2004, at A1.
[13] Greg Palast, The Wrong Way to Fix the Vote, WASH. POST, June 10, 2001, at B1; see also Marcia Myers, Election Theft Ruled Out, BALT. SUN, Aug. 24, 1995, at 1A.
[14] Greg J. Borowski, Nothing Points to Fraud in 9 Double Voting Cases, MILWAUKEE J. SENTINEL, Aug. 22, 2005.
[15] Lou Cannon, Editorial, Huffingtons Claims of a Stolen Election Preposterous, SEATTLE POST-INTELLIGENCER, Jan. 23, 1995, at A7.
[16] Gregory Roberts, GOP Challenges Rights of Hundreds of Voters, SEATTLE POST-INTELLIGENCER, Nov. 5, 2005, at A1.
[17] Gregory Stanford, Editorial, Election Fraud Witch Hunt Disillusions Young Voter, MILWAUKEE J. SENTINEL, Jan. 8, 2006, at J4.
[18] Marcia Myers, Election Theft Ruled Out, BALT. SUN, Aug. 24, 1995, at 1A.
[19] Memorandum from Bud Fitch, supra note 8.
[20] Greg Palast, Ex-con Game: How Floridas Felon Voter-Purge was Itself Felonious, HARPERS MAG., Mar. 1, 2002, at 48.
[21] David Postman, GOPs Felon List May Be Way Off, SEATTLE TIMES, Mar. 17, 2005.
[22] Id.
[23] See, e.g., 152 CONG. REC. H6768 (daily ed. Sept. 20, 2006) (statement of Rep. Norwood); see also H6772 (statement of Rep. Burton); H6775 (statement of Rep. Royce); H6776 (statement of Rep. Ehlers).
[24] Gregory Stanford, Editorial, Voting Irregularities: State GOP Should Apologize, MILWAUKEE J. SENTINEL, Aug. 24, 2005, at A14.
[25] Jay Bookman, Editorial, Rationale for Voter ID Law Fraudulent, ATLANTA J.-CONST., Oct. 20, 2005, at 17A.
[26] Brennan Center for Justice at NYU School of Law & Spencer Overton, Response to the Report of the 2005 Commission on Federal Election Reform 8 (2005) [hereinafter Carter-Baker Dissent]
[27] See 42 U.S.C.  15483(b).
[28] See 42 U.S.C.  15483(a)(2), (4).
[29] Carter-Baker Dissent, supra note 26, at 10.
[30] Id. at 3.
[31] John Pawasarat, The Driver License Status of the Voting Age Population in Wisconsin 1, 11 (2005),

Last Updated October 2006