With the FBI recently raiding the Georgia election election office, we thought it was worth re-visiting our analysis of duplicate voter data in the State from 2024. In 2020, Donald Trump lost the Presidential election in Georgia to Joe Biden by a very slim margin of just over 12,000 votes, forcing multiple reviews of the votes.
(Note: the FBI raid relates to voting data from the 2020 presidential election, and our analysis was done on 2024 voting register data).
To cut to the chase: our analysis of the Georgia voter registration data showed that the state had a very “normal” duplicate voter rate of 0.73%, slightly below the overall average of 0.8% that we found across 7 US states (AR, FL, GA, MI, NC, OH, PA).
The registration rate varied county by county, with a slightly higher duplicate rate for Democrat-registered voters than Republican (0.58% vs 0.43%, only counting party-affiliated duplicates).
Note Georgia does not supply data about whether individuals voted or not, so we were unable to look for potential duplicate voting fraud.
Importantly, the state of Georgia is a member of the Electronic Registration Information Centre (ERIC), which uses advanced entity resolution technology to 1) find potential voter registration duplicates, 2) identify duplicate voter fraud, and 3) find potential but un-registered voters to increase voting participation.
Therefore, Georgians and their fellow-Americans can be confident that, at least as far as the voting registration lists and potential duplicate voting are concerned, it is extremely unlikely that any fraudulent activity occurred.
Georgia is especially interesting when it comes to voter registration data. In 2018, the State attempted to implement an “exact match” regulation, meaning that an individual's voter registration had to exactly match their driver’s license or social security data. Any small variance would have meant that their voting rights were placed in a “pending” status, until corrected.
There were 51,000 pending voter profiles in Georgia in 2018, and the majority of these were African Americans, Latinos, and Asian Americans, suggesting that minorities were being kicked off the voting list simply for having uncommon names that are often misspelled.
In the absence of entity resolution technology that is able to handle the “fuzzy matching” between non-identical duplicate data, exact matching is perhaps your only option, but it does not take into account the reality of real-world data.
For example, would it be fair that José Rodríguez might not be able to vote because his driving licence says Jose Rodrigues due to a historical mis-spelling? As long as his other data matched (date of birth, address etc) then it is obvious to a human that Jose and José are the same person. Entity resolution means that computers can recognise this automatically too.
Of course we can’t make any comment on the in-person voting, nor voting machines in Georgia, but at least we can be confident that the voter registration data is about as clean as it can be.
