Georgia's "Model" Election Management System
By Ellen Theisen*
"Georgia has a model system for the deployment and management of elections technology, which combines the resources of its Secretary of State (SOS), its University system, and its county election officials."
Georgia's "Model" Election Management System
Unfortunately, Georgia's "model system" appears to be incapable of electronically tracking election turnout data by precinct. I discovered this by accident while I was trying to confirm Georgia's 0.4% presidential undervote rate as reported by Charles Stewart III in a February report of the Caltech/MIT Voting Integrity Project. I attempted to obtain the number of ballots cast in each precinct in Georgia, and that's what led me to discover the limitations of Georgia's computerized voter registration system. Here's how it happened.
Around the middle of March, I called the Georgia Secretary of State's office and left a message for Frances Jackson, asking where I could locate a file showing the turnout, by precinct, for the 2004 general election. Her assistant called me back (I neglected to get his name) and pointed me to the Credit for Voting Reports.
I downloaded a pdf file entitled "Georgia Secretary Of State; Active Voters By Race/Gender; General Election Voting History; Election Date: November 02, 2004" and saw that it showed the "Total Voters" for each precinct in each county, along with the "Total Voters" for each county.
I entered the county turnout totals into an Excel spreadsheet and compared them with the number of presidential votes cast in each county. The data showed 81 counties with more votes than voters, for a total of 22,399 "phantom votes" in the state's counties. In fact, the total turnout for the state added up to 3,285,140, which is 19,344 fewer voters than the 3,384,404 presidential votes reported by Georgia.
Curious about the discrepancy between this data and the turnout data used by Charles Stewart in his report, I called Frances Jackson and asked. She told me that the person I had spoken with before was mistaken when he told me the Voter Credit Report showed the turnout for the election; the report was updated regularly to remove voters who had died, moved, or been convicted of a felony. The file was dated "03/09/05", so I expected it would be updated again around the beginning of April. As of April 29, 2005 it has not been updated.
I asked her if the actual turnout totals for Georgia were available on the website, and she led me back to the SoS home page, through "Other Services," to the "Press Office," and then to the "2004 Press Releases." At the bottom of the November 18, 2004 press release, there was a link which opened an Excel file with the turnout by county and vote totals by county.
I thanked her and downloaded the file, entered the turnout data it into my Excel spreadsheet, and compared the figures with both the Voter Credit figures and the official presidential vote totals. To my surprise, the turnout reported in the Voter Credit report was higher than the actual turnout figures in nineteen of the state's 159 counties. Then I realized that each voter's voting history must follow them if they move to another county.
Consider, for example, County "W". In the four months following the 2004 election, 125 people who had voted in the election moved into County "W" from another Georgia county and registered to vote. The registrar in County "W" entered their new addresses into the state voter registration database, and their voting credit was immediately transferred to County "W". That's 0.45% of the voters in County "W", a higher percentage than the total percentage of undervotes in the state of Georgia
While the data only proves this to be true of 125 voters, it was probably more, since most counties had seen many voters die, move, and be convicted of felonies during those four months, and surely some died in County "W" as well. For example, in County "F" alone, 7074 people who voted in the 2004 election ÷ a little over 2% of the voting population ÷ died, moved out of the county, or were convicted of felonies during the four months following the election. And the county registrar recorded the changed status of all of them in the state voter registration database.
When I spoke with the Director of Elections and Registration in a North Georgia County, he told me that it can take as long as 6 weeks for the registrars to enter the voter credit data from a major election into the state system. So, their dedication to keeping the rolls updated with the deaths, moves, and convictions of voters is remarkable.
Even though I had the authoritative answer from Frances Jackson, I continued poking around on the Georgia SoS website, and I found huge zip files called "Voter History Files" all dated 2/1/05. I downloaded the one entitled "2004.zip", unzipped the 131MB text file, and imported it into Access. The file was a list of voter registration numbers, each with an associated election date and type, and whether or not the voter voted absentee.
By creating some queries and forms, I was able to check the November 2, 2004 "Voter History" figures for each of the 159 counties in Georgia and compare the figures with the presidential totals for the counties. This data indicated that 90 Georgia counties had more votes than voters listed in the Voter History file, for a total of 32,473.
It seemed odd that the number of voters given credit for voting in the 2004 election would have decreased by a total of 32,473 during the three months following the election and then increased by 10,074 during February. The data reported for some counties seemed particularly odd. For example, consider County "B". The actual turnout for that county, as reported by the SoS press release of November 18, was 5,628 voters. Three months later, the Voter History report listed 5,388 voters (240 fewer ÷ 4%) as voting in the November election; and a month after that the Voter Credit once again showed 5,628 voters receiving credit for voting in the election.
I decided to ask Frances Johnson about the Voter History file, but this time, I wanted more information about it and I wanted the information in writing, since I had been given conflicting information last time. So I sent her an open records request with my questions ÷ basically asking the purpose of the Voter History file, the source of the data, and what the data represented. She passed my request on to Cliff Tatum, Assistant Director of the Legal Affairs Unit of the Elections Division.
Four days later, in response to my email asking about the status of my open records request, Mr. Tatum replied in a pdf file saying that since I had asked questions rather than requesting documents, there was nothing for me to inspect, and they could not fulfill my request. He suggested that I contact Ralph Jones at the Voter Registration help desk. He told me that Mr. Jones would be able to answer my questions.
I was delayed in calling Mr. Jones, and a few days later he was kind enough to telephone me in response to a suggestion from Mr. Tatum. Briefly, here is what he told me. Both the Voter History files and the Voter Credit Report are generated from the same state voter registration database. Both reflect the voter credits at the time they are generated. While the Voter History files are updated periodically, the Voter Credit Report is static. Once it is produced, it is never updated.
According to Mr. Jones, both files provide data that is a 97, 98, or 99% accurate indication of the actual turnout. The data is accumulated based on the information entered by the county registrars. He didn't know where they obtained their data. When I asked why the Voter History file might show fewer voters than the Voter Credit Report, he didn't seem to understand my question. He told me that the precinct-by-precinct vote totals data included the number of ballots cast, but I showed him that it didn't. Then he pointed me to the press release, and I commented that it didn't include precinct data. He couldn't point me to precinct turnout data, but he suggested that I call Michael Barnes, who handles the ballots-cast information.
After we talked, I sent him an email, summarizing the information he had given me and asking if I had understood him correctly. Then I sent an open records request to Mr. Tatum, asking for electronic documentation showing the number of ballots cast in each precinct in the 2004 presidential election.
Three days later, when I had not received a response from Mr. Jones, I emailed him again. Thirty-four minutes later, Mr. Tatum emailed me a pdf response to my open records request, along with a message saying he would be out of the office the remainder of the day and the entire following week.
This time his response indicated that he could fulfill my request. I would be welcome to come into the office and inspect the relevant records, but not until he was back from his week's absence. He estimated 27 hours of his staff's time to "prepare the documents" for my inspection and 4 hours for me to inspect them, for a total of $686. 65, which would be billed to me.
The only possible conclusion is that Georgia's "model system for the deployment and management of elections technology" is incapable of quickly and easily producing an electronic document showing the number of ballots cast in each precinct in the November 2004 election.
*Ellen Theisen is the Executive Director of VotersUnite.Org and Director of Information Resources for Vote Trust USA.
"Implementing Voting Systems: The Georgia Method." By Brit J. Williams AND Merle S. King. Communications Of The ACM October 2004/Vol. 47, No. 10. Page 39.
Copyright © 2005 Vote Trust USA