In June 2007, a steady voter in Memphis named Effie Washington was a Black woman. But before the 2008 presidential election, Effie's voting record was changed to read "gender: man; race: other". In June 2009 her voting record was changed, to "gender: woman; race: other".
At one time Sharonda Williams* was listed as a Black woman. She was purged from Shelby County (TN) voter rolls with a code indicating she had moved out of the county, even though she lived in Millington, a Memphis suburb in Shelby County. Sharonda was issued a different voter ID number and then listed as "race: other". Sharonda voted at her polling place in the May 2010 primary, selecting a Democratic ballot, which is marked as "cast" in the electronic poll book. But that poll book log never was made public, and the official voter history records report that Sharonda Williams did not vote. In August 2010, Williams is reported as voting Republican.
*There are only two persons on the Shelby County voter list named Sharonda Williams. One has a birthdate in 1978, one in 1981. The information above pertains to the one born in 1981.
HOW DOES IT AFFECT THE RESULT IF VOTER LISTS ARE INCORRECT?
First, let's take a look at general bookkeeping issues before racial identity comes into the mix.
For an election to be valid, the books have to balance. And for an election to be democratic, the public must be able to see and authenticate the following:
1) Who can vote (the voter list)
2) Who did vote (the participating voter list, which is later stored as the voter history)
3) Chain of custody
4) The count
When dead people, duplicates, and persons long departed for other jurisdictions sit on the voter list, they can vote -- or more precisely -- be voted for, by insiders, creating a powerful mechanism for fraud. Though the public can never control the hearts, minds, and bribe-sensitivity of every person with inside access to an election, we can mitigate fraud in a meaningful way through full public transparency, by developing the habit of public examinations of who can vote, who did vote, chain of custody and the count.
WHO CAN VOTE: The public needs to be able to examine voter lists to ascertain whether the people on these lists actually live there, are still alive, and that there is only one of each person on the voter list. Without accurate voter lists and public ability to authenticate them, an election fails the test of transparency and opens itself up for fraudulent results.
WHO DID VOTE: When votes are counted, they must match up to the number of people who voted. You cannot have 100 votes if you have only 90 voters.
In a primary election, you can't have more Democratic votes than Democratic ballots; likewise for Republican votes and ballots. In many states, voters choose a party affiliation when they register; in others, like Tennessee, voters can choose which party ballot to vote on Election Day. Regardless of which method is used, you can't have more Republican votes than Republican voters or ballots. Without an accurate record of who voted, and in primaries, who voted Republican or Democratic, an election fails the test of transparency and opens itself up for fraudulent counts.
Some people may try to tell you that if the voter list is wrong, or the list of who voted or what party ballot they voted happens to be incorrect, this has nothing to do with the vote count itself. Not true.
If you have 280 Republican primary votes and just 240 Republican ballots cast, one or the other of those counts is just flat wrong. Such an election cannot be valid, because it is based on impossible numbers.
If a dog casts a vote, like Duncan the dog did in Washington State's absentee voting system, or a poll worker casts votes on behalf of dead people, as in Shelby County in 2010 (resulting in invalidation of Ophelia Ford’s initial election), or if some people have two votes cast in their names, as happened in Shelby County in the August 2010 contest, the election cannot be valid because it is based on invalid votes.
HOW ALTERING RACIAL IDENTITIES CAN AFFECT THE COUNT
The above information pertains in a generic way to all elections. In locations where a significant number of voters belong to a racial minority, additional factors come into play with voter list accuracy.
Many states use a racial identifier on their voter list, originally set up to protect minority voters. The idea is that by identifying the historically targeted voters in a given jurisdiction, it will ascertainable if they have been targeted for disenfranchisements, such as non-working voting machines, ballot shortages, or procedural "glitches."
In locations with strong political affiliations tied to race, racial identity provides another check and balance against wrong election results. If you know that the Cuban population in a particular location votes heavily Republican, or that the Black population votes 96% Democrat, as voter histories show is the case in Shelby County, any result that noticeably diverges from known racial political preferences may become suspect.
In Shelby County, precincts are racially polarized; some are mostly Black, some almost all White, and some are "in play", meaning they are mixed enough that it's anybody's guess which party will win. If votes were simply flipped so that Republicans got the Democratic votes and vice versa, it would be immediately seen in a county with known racially-related political preferences and concentrated ethnic populations. If you wanted to flip the counts, you’d need to either (1) dilute the reported racial demographic or (2) create a precinct-specific vote flipping method, or both. Both of these factors exist in Shelby County. The subject of this report is racial identity dilution; the precinct-specific mechanism available for vote-flipping will be described later in this report series.
In Shelby County, the 2009 census reported a population that is 52% Black, and voter histories show Black voters choosing Democratic primary ballots 96% of the time. White voters choose Democratic ballots 25-40% of the time, depending on the election. In August 2010, this demographic yielded a surprising White Republican sweep. When 10 candidates filed a lawsuit which called the accuracy of the count into question, its accuracy was justified by Republican election commissioners who stated that the results matched the demographic for citizens who turned out to vote.
But what if the demographic is incorrect?
I examined 89 voter databases created between Sept. 2006 and July 2010.
Between 2006 and 2009, racial identity for Black voters in Shelby County was diluted by 10.6%, with racial identities for 6.1% of White voters also removed. A total of 40,881 Black voters show race changes during this period, with 24,029 White race changes, for a total of 64,881 voters whose race was changed in the voter lists. Most changes removed the racial identifier by converting Black or White to the "Other" category.
Current voter lists have removed tens of thousands of voters who moved away, passed away, were removed for committing felonies, or were purged for other reasons. Remaining on the 600,000-voter list in 2010 are 32,918 voters whose racial identifiers were removed; 21,860 of these were originally Black, and 11,058 were originally White.
Thus, when the media reports, as it recently did for Early Vote action in a Memphis election, that "52% of those who cast votes were Black", that number is not the number of Black voters, but instead, represents the diluted number of people who are now said to be Black.
Comparing the list of names on the Memphis Early Voting Participating Voter list with historic voter lists showing the racial identify by name for who "used to be Black", in fact 62% of voters were knowably Black, 24% White, 14% Other.
In truth, the number in that recent early voting list is probably 70% Black, 27% White, 3% Other.
By examining older voter histories, one can see that the number of "Other" who have been identified as Hispanic, Asian, or American Indian was less than one percent. It is likely that the Hispanic population in Shelby County has grown past one percent by now, and will probably climb towards five percent in the foreseeable future.
Based on historic data, after taking out an estimate for Asian, Hispanic and Indian, we would expect to find two-thirds of this remaining pool to be Black, and one-third White. That would give us not the reported 52% Black, but 70% Black voters in the recent Memphis early voters. (52% currently reported, 10.6 percent "used to be Black", and about 7.4% unidentified other we can assume to be Black based on past history in voter lists).
WHY THESE NUMBERS MATTER
If election results are being justified because the vote count "matches the demographics" but the demographics are off by 10 to 18 percent, the vote count actually does not match the demographics. That would indicate that (1) the political preference for Black voters in Shelby County has changed – but there is little indication that this is the case; or (2) the vote counts may not be reported accurately.
This kind of analysis is important for any location with significant populations of Hispanic, Indian, Asian, or Black voters.
WHAT YOU CAN DO
1. Get copies of the latest voter history list every month. These lists can be read by anyone, if you want to take the time. They are just lists. Really, really long lists, provided in files that end in ".txt", ".csv", or ".tab" – files you can open with Notepad on any computer. You can also open them with Excel 2010 (earlier versions of Excel don’t hold enough records); and you can open them with any database program, like Filemaker or Microsoft Access.
2. Compare the changes to the voter list each month. It's not hard. Any database person can identify all the changes within about five minutes. You can merge the files together by using the voter ID number to connect matching records from different databases, and then use an "IF-THEN" formula to identify changes, like this: "If a = b, 0, 1" (if column a is same as column b, put 0 here, otherwise, put 1 here). Then just sort for the "1s" and you have flagged all records with a change in the field.
3. Hook up with someone in your local political party, or an ex-candidate, and collect archive copies of the old voter lists. To find out if voter lists have had race identifiers diluted in your location, compare racial identities in the old lists with current lists and identify voters who have had their racial identity changed.
4. Using voter histories in primary elections, see if you can identify political preference for each group.
5. If voters are obstructed or disenfranchised, compare with actual racial demographics to see if a minority group was disproportionately harmed (a civil rights violation). If you know there are racially-related political preferences, compare with election results to spot red flags.
I will provide examples of this process with video and in our 2012 Tool Kit. Get started now collecting and archiving a chronological sequence of voter databases. If you have a database guy in your group, show him this article. He’ll know what to do.
HOW AND WHY WERE SHELBY COUNTY RACIAL IDENTITIES DILUTED?
When changes are made in voter records, a transaction log is created. In October 2009, for example, we see race changes made by two employees, Tijuana Barrett and Shirlee Borin. They made entries for thousands of voters indicating that a racial identifier was created or changed.
By reviewing before-and-after voter lists, we see that Barrett was changing race "A" (Asian) to "Other", and Shirlee Borin was adding race "other" for new registration cards and changing race to "other" with voters who had moved to a new address.
At some point in 2009, under the administration of Republican-appointed election supervisor Richard Holden, all voters identified as "H" (Hispanic), "I" (American Indian) and "A" (Asian) were changed to "O" (Other). These racial categories have, in the past, represented a miniscule portion of the Shelby County voter list, but the Hispanic component is growing and its actual representation on the voter lists is now not being tracked at all. Fifteen voters who had been identified as "C" (Colored) were changed to "Other", instead of changing them to "Black".
One reason racial identifiers are eroding is that the Tennessee voter registration card has changed. It used to contain a check box for race, but now has a blank and the word "optional" over it. When someone moves or changes marital status, or corrects information, or a duplicate is removed, this tends to correspond with removal of racial identifier. If they fill out a new registration card and leave racial identifier blank, Shelby County staff has no choice but to match that blank space by changing race to "other".
Routine updates are resulting in loss of racial demographic information. My examination of the records shows no targeting, no disproportionate alterations for any precinct. It does not appear that anything illegal is going on.
This issue doesn't need to be illegal to be important, however.
You simply can't use racial demographics as corroboration for any election result if the lists are not accurate. Organizations involved in protecting minority rights should take steps to "un-dilute" the information, at least for internal use, to best protect their voting blocs.
BIGGER INACCURACIES
Several hundred Shelby County voter records had issues that are hard to justify as routine voter update-caused changes in race/gender identity. Consider the following:
- Ronald Lacey, of Collierville, was listed in June 2007 as a White male. As of Oct. 2008, Ronald Lacey was shown as a Black female. In 2010, he was listed as "Other, male."
- Eunice Cathey was shown in June 2007 as Black, female, but after September 2009, is being shown as White, male.
- Lisa Truesby was reported as White, female in 2007; Other, female, in 2008; and White, male in 2009.
- Meredith Stewart was shown as White, female in 2007; then Black, male in 2008, then Black, female in 2009.
Altogether there are over 500 records that have bounced back and forth in both race and gender. These are hard to explain.
To support the important election protection work by Black Box Voting:
Black Box Voting
You can discuss this article here
At one time Sharonda Williams* was listed as a Black woman. She was purged from Shelby County (TN) voter rolls with a code indicating she had moved out of the county, even though she lived in Millington, a Memphis suburb in Shelby County. Sharonda was issued a different voter ID number and then listed as "race: other". Sharonda voted at her polling place in the May 2010 primary, selecting a Democratic ballot, which is marked as "cast" in the electronic poll book. But that poll book log never was made public, and the official voter history records report that Sharonda Williams did not vote. In August 2010, Williams is reported as voting Republican.
*There are only two persons on the Shelby County voter list named Sharonda Williams. One has a birthdate in 1978, one in 1981. The information above pertains to the one born in 1981.
HOW DOES IT AFFECT THE RESULT IF VOTER LISTS ARE INCORRECT?
First, let's take a look at general bookkeeping issues before racial identity comes into the mix.
For an election to be valid, the books have to balance. And for an election to be democratic, the public must be able to see and authenticate the following:
1) Who can vote (the voter list)
2) Who did vote (the participating voter list, which is later stored as the voter history)
3) Chain of custody
4) The count
When dead people, duplicates, and persons long departed for other jurisdictions sit on the voter list, they can vote -- or more precisely -- be voted for, by insiders, creating a powerful mechanism for fraud. Though the public can never control the hearts, minds, and bribe-sensitivity of every person with inside access to an election, we can mitigate fraud in a meaningful way through full public transparency, by developing the habit of public examinations of who can vote, who did vote, chain of custody and the count.
WHO CAN VOTE: The public needs to be able to examine voter lists to ascertain whether the people on these lists actually live there, are still alive, and that there is only one of each person on the voter list. Without accurate voter lists and public ability to authenticate them, an election fails the test of transparency and opens itself up for fraudulent results.
WHO DID VOTE: When votes are counted, they must match up to the number of people who voted. You cannot have 100 votes if you have only 90 voters.
In a primary election, you can't have more Democratic votes than Democratic ballots; likewise for Republican votes and ballots. In many states, voters choose a party affiliation when they register; in others, like Tennessee, voters can choose which party ballot to vote on Election Day. Regardless of which method is used, you can't have more Republican votes than Republican voters or ballots. Without an accurate record of who voted, and in primaries, who voted Republican or Democratic, an election fails the test of transparency and opens itself up for fraudulent counts.
Some people may try to tell you that if the voter list is wrong, or the list of who voted or what party ballot they voted happens to be incorrect, this has nothing to do with the vote count itself. Not true.
If you have 280 Republican primary votes and just 240 Republican ballots cast, one or the other of those counts is just flat wrong. Such an election cannot be valid, because it is based on impossible numbers.
If a dog casts a vote, like Duncan the dog did in Washington State's absentee voting system, or a poll worker casts votes on behalf of dead people, as in Shelby County in 2010 (resulting in invalidation of Ophelia Ford’s initial election), or if some people have two votes cast in their names, as happened in Shelby County in the August 2010 contest, the election cannot be valid because it is based on invalid votes.
HOW ALTERING RACIAL IDENTITIES CAN AFFECT THE COUNT
The above information pertains in a generic way to all elections. In locations where a significant number of voters belong to a racial minority, additional factors come into play with voter list accuracy.
Many states use a racial identifier on their voter list, originally set up to protect minority voters. The idea is that by identifying the historically targeted voters in a given jurisdiction, it will ascertainable if they have been targeted for disenfranchisements, such as non-working voting machines, ballot shortages, or procedural "glitches."
In locations with strong political affiliations tied to race, racial identity provides another check and balance against wrong election results. If you know that the Cuban population in a particular location votes heavily Republican, or that the Black population votes 96% Democrat, as voter histories show is the case in Shelby County, any result that noticeably diverges from known racial political preferences may become suspect.
In Shelby County, precincts are racially polarized; some are mostly Black, some almost all White, and some are "in play", meaning they are mixed enough that it's anybody's guess which party will win. If votes were simply flipped so that Republicans got the Democratic votes and vice versa, it would be immediately seen in a county with known racially-related political preferences and concentrated ethnic populations. If you wanted to flip the counts, you’d need to either (1) dilute the reported racial demographic or (2) create a precinct-specific vote flipping method, or both. Both of these factors exist in Shelby County. The subject of this report is racial identity dilution; the precinct-specific mechanism available for vote-flipping will be described later in this report series.
In Shelby County, the 2009 census reported a population that is 52% Black, and voter histories show Black voters choosing Democratic primary ballots 96% of the time. White voters choose Democratic ballots 25-40% of the time, depending on the election. In August 2010, this demographic yielded a surprising White Republican sweep. When 10 candidates filed a lawsuit which called the accuracy of the count into question, its accuracy was justified by Republican election commissioners who stated that the results matched the demographic for citizens who turned out to vote.
But what if the demographic is incorrect?
I examined 89 voter databases created between Sept. 2006 and July 2010.
Between 2006 and 2009, racial identity for Black voters in Shelby County was diluted by 10.6%, with racial identities for 6.1% of White voters also removed. A total of 40,881 Black voters show race changes during this period, with 24,029 White race changes, for a total of 64,881 voters whose race was changed in the voter lists. Most changes removed the racial identifier by converting Black or White to the "Other" category.
Current voter lists have removed tens of thousands of voters who moved away, passed away, were removed for committing felonies, or were purged for other reasons. Remaining on the 600,000-voter list in 2010 are 32,918 voters whose racial identifiers were removed; 21,860 of these were originally Black, and 11,058 were originally White.
Thus, when the media reports, as it recently did for Early Vote action in a Memphis election, that "52% of those who cast votes were Black", that number is not the number of Black voters, but instead, represents the diluted number of people who are now said to be Black.
Comparing the list of names on the Memphis Early Voting Participating Voter list with historic voter lists showing the racial identify by name for who "used to be Black", in fact 62% of voters were knowably Black, 24% White, 14% Other.
In truth, the number in that recent early voting list is probably 70% Black, 27% White, 3% Other.
By examining older voter histories, one can see that the number of "Other" who have been identified as Hispanic, Asian, or American Indian was less than one percent. It is likely that the Hispanic population in Shelby County has grown past one percent by now, and will probably climb towards five percent in the foreseeable future.
Based on historic data, after taking out an estimate for Asian, Hispanic and Indian, we would expect to find two-thirds of this remaining pool to be Black, and one-third White. That would give us not the reported 52% Black, but 70% Black voters in the recent Memphis early voters. (52% currently reported, 10.6 percent "used to be Black", and about 7.4% unidentified other we can assume to be Black based on past history in voter lists).
WHY THESE NUMBERS MATTER
If election results are being justified because the vote count "matches the demographics" but the demographics are off by 10 to 18 percent, the vote count actually does not match the demographics. That would indicate that (1) the political preference for Black voters in Shelby County has changed – but there is little indication that this is the case; or (2) the vote counts may not be reported accurately.
This kind of analysis is important for any location with significant populations of Hispanic, Indian, Asian, or Black voters.
WHAT YOU CAN DO
1. Get copies of the latest voter history list every month. These lists can be read by anyone, if you want to take the time. They are just lists. Really, really long lists, provided in files that end in ".txt", ".csv", or ".tab" – files you can open with Notepad on any computer. You can also open them with Excel 2010 (earlier versions of Excel don’t hold enough records); and you can open them with any database program, like Filemaker or Microsoft Access.
2. Compare the changes to the voter list each month. It's not hard. Any database person can identify all the changes within about five minutes. You can merge the files together by using the voter ID number to connect matching records from different databases, and then use an "IF-THEN" formula to identify changes, like this: "If a = b, 0, 1" (if column a is same as column b, put 0 here, otherwise, put 1 here). Then just sort for the "1s" and you have flagged all records with a change in the field.
3. Hook up with someone in your local political party, or an ex-candidate, and collect archive copies of the old voter lists. To find out if voter lists have had race identifiers diluted in your location, compare racial identities in the old lists with current lists and identify voters who have had their racial identity changed.
4. Using voter histories in primary elections, see if you can identify political preference for each group.
5. If voters are obstructed or disenfranchised, compare with actual racial demographics to see if a minority group was disproportionately harmed (a civil rights violation). If you know there are racially-related political preferences, compare with election results to spot red flags.
I will provide examples of this process with video and in our 2012 Tool Kit. Get started now collecting and archiving a chronological sequence of voter databases. If you have a database guy in your group, show him this article. He’ll know what to do.
HOW AND WHY WERE SHELBY COUNTY RACIAL IDENTITIES DILUTED?
When changes are made in voter records, a transaction log is created. In October 2009, for example, we see race changes made by two employees, Tijuana Barrett and Shirlee Borin. They made entries for thousands of voters indicating that a racial identifier was created or changed.
By reviewing before-and-after voter lists, we see that Barrett was changing race "A" (Asian) to "Other", and Shirlee Borin was adding race "other" for new registration cards and changing race to "other" with voters who had moved to a new address.
At some point in 2009, under the administration of Republican-appointed election supervisor Richard Holden, all voters identified as "H" (Hispanic), "I" (American Indian) and "A" (Asian) were changed to "O" (Other). These racial categories have, in the past, represented a miniscule portion of the Shelby County voter list, but the Hispanic component is growing and its actual representation on the voter lists is now not being tracked at all. Fifteen voters who had been identified as "C" (Colored) were changed to "Other", instead of changing them to "Black".
One reason racial identifiers are eroding is that the Tennessee voter registration card has changed. It used to contain a check box for race, but now has a blank and the word "optional" over it. When someone moves or changes marital status, or corrects information, or a duplicate is removed, this tends to correspond with removal of racial identifier. If they fill out a new registration card and leave racial identifier blank, Shelby County staff has no choice but to match that blank space by changing race to "other".
Routine updates are resulting in loss of racial demographic information. My examination of the records shows no targeting, no disproportionate alterations for any precinct. It does not appear that anything illegal is going on.
This issue doesn't need to be illegal to be important, however.
You simply can't use racial demographics as corroboration for any election result if the lists are not accurate. Organizations involved in protecting minority rights should take steps to "un-dilute" the information, at least for internal use, to best protect their voting blocs.
BIGGER INACCURACIES
Several hundred Shelby County voter records had issues that are hard to justify as routine voter update-caused changes in race/gender identity. Consider the following:
- Ronald Lacey, of Collierville, was listed in June 2007 as a White male. As of Oct. 2008, Ronald Lacey was shown as a Black female. In 2010, he was listed as "Other, male."
- Eunice Cathey was shown in June 2007 as Black, female, but after September 2009, is being shown as White, male.
- Lisa Truesby was reported as White, female in 2007; Other, female, in 2008; and White, male in 2009.
- Meredith Stewart was shown as White, female in 2007; then Black, male in 2008, then Black, female in 2009.
Altogether there are over 500 records that have bounced back and forth in both race and gender. These are hard to explain.
To support the important election protection work by Black Box Voting:
Black Box Voting
You can discuss this article here