How the AJC and media partners conducted the petition study

Opponents of Atlanta’s planned public safety training started gather out side of the City Hall where they will present their petition with 100,000 signatures to the city of Atlanta on Monday, Sept. 10, 2023. 
Miguel Martinez /

Credit: Miguel Martinez

Credit: Miguel Martinez

Opponents of Atlanta’s planned public safety training started gather out side of the City Hall where they will present their petition with 100,000 signatures to the city of Atlanta on Monday, Sept. 10, 2023. Miguel Martinez /

The Vote to Stop Cop City Coalition is seeking a city-wide vote on whether Atlanta should proceed with the Atlanta Public Safety Training Center currently under construction in DeKalb County. In order to prompt a referendum, petition signers must have been eligible to vote in the November 2021 mayor’s election and still be registered voters in the city of Atlanta.

To be successful, the coalition must collect 58,231 signatures from such people, according to the city clerk.

Petitioners collected and presented 16 boxes of signatures pages to the city clerk’s office on Sept. 11, saying the boxes contained approximately twice as many signatures as required. The city clerk’s office posted PDFs of the signature pages online, organized by box, but has not yet started verifying signatures pending an appellate court ruling.

Media partners at The Atlanta Journal-Constitution, The Associated Press, Georgia Public Broadcasting and WABE spent nearly two months examining and analyzing petitions, counting the total number of entries in each box and determined 108,509 had been submitted — about 8,000 fewer than petitioners estimated. The partners randomly selected 1,000 entries by box and page, and closely examined those samples.

To ensure that each entry was randomly selected, the partners utilized a proportional sample from each of the 16 boxes. That method allowed the partners to obtain a representative sample of names to examine.

The news organizations did not attempt to verify signatures. Instead, they sought to verify if a person was an eligible signer – someone who currently lives in Atlanta, is registered to vote and was registered in Atlanta the last citywide municipal election in November 2021. To reach the threshold of 15% of active, registered 2021 voters, 53.6% of the petition signatures must be valid.

A current state voter file was used to determine if the petition signer is listed as living within the City of Atlanta and a voter file from December 7, 2021 was used to determine if a person was registered at an address within the city at the time of the last municipal election. The media partners also used Fulton County property records as well as multiple sources of public records to determine where a petition signer lived.

City officials have not indicated what lengths they will take to verify names.

The media partners then assigned a value of “Valid”, “Invalid” or “Undetermined” to each signer:

  • VALID: If a signer is currently registered to vote at the City of Atlanta address signed on the petition and was also registered at a City of Atlanta address in 2021.
  • INVALID If a signer was not currently registered to vote at a City of Atlanta address or was not in 2021; if the signer’s name or address was illegible; if the listed address was invalid or unconfirmable; or if the listed address was outside the city of Atlanta.
  • UNDETERMINED: If the name of the signer could be found at a City of Atlanta address in 2021 and 2023, but not confirmed at the address in the petition. For example, if a person named John Smith signed at 123 Peachtree St NW, but none of the 10 registered voters named John Smith live or lived at 123 Peachtree Street.

In the sample of 1,000 signers, the partners found 475 signers who were eligible, 473 who were not and 52 that could not be determined. When “valid”, “invalid” and “undetermined” are considered separately, the study found that 48% of signers were eligible, with a margin of error of +/- 3%. In this scenario, the sample predicts the total population would include between 48,250 and 54,800 valid signatures. Petitioners would fall approximately 3,400 signatures short.

When petition organizers fall short in a referendum, typically a “curing process” is used to allow validation of questionable signatures after petition organizers are notified. In an email, Atlanta’s municipal clerk emeritus Foris Webb III said that he would not respond to hypothetical questions about the curing process, but indicated that many of the scenarios posed by the media partners “have been contemplated in preparation of the verification process.” Some of those questions included: What happens if a person lists their 2021 address, but has a different current address in registration records; and will the city do any research beyond what’s in the voter rolls and the petitions?

Given that information, the partners also conducted an analysis where the undetermined are cured to be valid in a hypothetical verification. In this scenario, the sample predicts the total would include between 53,900 and 60,400 valid signatures, possibly having enough to secure a referendum. However, the margin of error indicates it could go either way.

The same holds true if the City cures only half of the “undetermined” signers in the data. The margin of error makes it too close to make a determination.

The partners shared the methods and findings with Trey Hood, director of the Survey Research Center at the University of Georgia’s School of Public & International Affairs. The partners also consulted with UGA’s Statistical Consulting Center, who reviewed the methodology.

Data reporters Stephanie Lamm, Justin Price and Rahul Deshpande contributed to the data entry and analysis.