Georgia early voting data suggests that proposed restrictions in Senate runoff will disproportionately impact African American and Hispanic American votes

The Center for New Data
10 min readDec 15, 2020


Oct. 12 was the first day for advance voting in Georgia, and people showed up by the hundreds to cast their ballot early at the Bell Auditorium in Augusta. (MICHAEL HOLAHAN, THE AUGUSTA CHRONICLE)

Originally written on December 8, 2020

December 9 update: Cobb County has announced a reversal of previously announced poll closures, “responding to concerns” over its initial plan to reduce the number of early voting locations to five sites for the entirety of the advance voting period ahead of the Jan. 5 runoffs, down from 11 that were operational ahead of November’s general election.

In October and November of this year, an estimated two-thirds of the eligible American voting population cast a ballot in an unprecedented and historic turnout for the 2020 General Election. Over 100 million of these Americans voted prior to election day, with roughly one-third choosing to vote through in-person early voting. In the context of the ongoing COVID-19 pandemic and mixed messaging from public officials, having a multitude of accessible voting options was more crucial than ever in this election. In the weeks leading up to and following the election however, activists and lawmakers warned of increasingly brazen voter suppression tactics, threatening the equal participation of racial minority voters in the democratic process.

In advance of early voting commencing in Georgia’s runoff election, the Center for New Data’s Observing Democracy program analyzed two large commercially-available geolocation datasets — along with voter files and census records — to report on polling wait times and voting disparities for early voters in Georgia during the 2020 General Election. Findings from this analysis, in the context of announced early voting polling closures for the runoff by Georgia officials, indicate that African American and Hispanic American voters will be disproportionately affected by proposed restrictions.

Why did the Observing Democracy analysis choose to focus on Georgia?

After many decades as a reliably red-leaning state, the public watched as Georgia turned into a key battleground in the General Election this year. Many attribute the tight margins in the presidential race to changing demographics in the state, particularly driven by the Atlanta metro region, which tends to lean blue. The narrow Democratic victories in Georgia have also been credited to the work of Stacey Abrams, an African American woman, former gubernatorial candidate, and the founder of Fair Fight, who many believe lost her 2018 race due to disenfranchisement tactics aimed at African American voters. All of these factors made Georgia a critical determinant in the results of the 2020 election, as well as a ripe target for voter suppression efforts.

Not only did general election results in Georgia help determine a Biden presidential win, they also determined the need for a runoff election for control over two Senate seats. Georgia election law requires a candidate to win a majority of votes (50%+1) in order to be elected to office. If no candidate wins above that threshold of votes, a second runoff election is held between the top two senate candidates of the general election.

This year in Georgia, in a rare fate of luck, both sitting senators are up for election — and the outcome election will determine which political party will control the U.S. Senate. Currently, Republicans hold a 50–48 margin, and if able to retain one or both seats, will maintain control of the Senate. Democrats will be required to win both runoff elections in order to tie any Senate votes, with newly elected Vice President Kamala Harris expected to break any Senate ties favorably for Democrats. With the stakes of the runoff election apparent, voting rights advocates have called for increasing actions to fight voter suppression in the state. The Georgia Early Vote analysis by the Center for New Data addresses these concerns in a non-partisan effort to work towards transparency, integrity, and fairness of all elections.

What did early voting patterns look like for Georgians in the General Election?

An analysis of over 70 million points of individual-level geolocation data undertaken by the Center for New Data found that statewide, the average measured vote time in Georgia was 44 minutes, with over 2.5 million Georgians casting ballots through early voting according to the U.S. Elections Project. These vote times were calculated as total time on-site at a polling location, including wait-times plus the time to sign-in with poll station clerks and to cast a ballot.

This 44 minute average is significant, especially in contrast to best practices laid out by the bipartisan Presidential Commission on Election Administration (PCEA), created in 2013 under President Obama. The PCEA’s final report asserted that “No citizen should have to wait more than 30 minutes to vote; jurisdictions can solve the problem of long lines through a combination of planning, including use of the tools noted in this Report, and the efficient allocation of resources”.

Statewide, the 10 counties with the longest average vote times ranged from Pickens County (82 minutes) to Newton County (75 minutes). Looking at the Atlanta region specifically, the average vote times for the 9 counties and 6 million people that comprise the Atlanta metro area was as follows:

  • Coweta County — 74 minutes
  • Douglas County — 69 minutes
  • Cobb County — 50 minutes
  • Clayton County — 49 minutes
  • DeKalb County — 47 minutes
  • Fulton County — 44 minutes
  • Henry County — 43 minutes
  • Gwinnett County — 42 minutes
  • Fayette County — 42 minutes
This graph visualizes these vote times by county using the 75th percentile of calculated voting. Atlanta metro counties are shown in yellow in order to distinguish from counties in the rest of the state shown in blue.

Also of significance, are the 10% of all Georgia voters who chose to vote on the weekend. Weekend voting options were especially utilized by African American and Hispanic American voters. Observing Democracy’s analysis showed that in weekend voting across Georgia’s 159 counties, African Americans voted at higher rates than Whites in 107 counties and Hispanic Americans voted at higher rates than Whites in 100 counties.

The graphs below depict the fifty counties with the most in-person early votes cast in descending order. The dots mark the difference between the share of votes cast by African American and Hispanic American voters respectively, and White voters. As shown, both African American and Hispanic American voters significantly cast their ballot on the weekends compared to White voters.

What impact would the proposed voting restrictions have on the Senate Runoff races?

Several counties in Georgia have proposed voting restrictions including operating fewer polling locations, removing Saturday voting options, and closing early voting at 5 P.M., citing resource constraints as the primary driver behind the voting limitations. Unlike in a general election, a weekend early voting day is not mandatory in the Senate Runoffs.

Amongst others, Cobb County, Forsyth County and Hall County all have plans to cut their polling locations down by more than half in the Senate Runoff races, despite all being in the top ten most populous counties in Georgia. Weekend closures and other voting limitations would likely disproportionately burden lower and middle class Georgians with less flexibility around work schedules and other weekday commitments. In addition, voting rights advocates have cautioned that the early voting sites closed, like in Cobb County, where many closures are located in Black and Latinx communities, could have especially harmful impacts for African American and Hispanic American voters.

The Center for New Data’s analysis looked at the 42 early vote polling locations marked for closure as of Dec. 6, 2020, and found that slated closures would potentially disenfranchise more than 223,000 votes. The analysis showed that out of the 10 polling locations in Georgia with the highest estimated portion of voters spending longer than 30 minutes on site, 5 are slated for closure. Compounding this, the analysis also found that voters at polling locations slated to close had wait times averaging 49 minutes, a 12% longer wait time than voters at other polling locations (who spent an average of 43 minutes). These closures would have a consequential negative effect on minority voter turnout based on historical analysis from the general election, which showed that individuals identifying as white were the least likely to cast their vote on weekends (8.6%), compared to those identifying as Asian (13.1%), African American (11.8%) and Hispanic (11.4).

The table below summarizes the Center for New Data’s findings depicting the rates of early weekend voting by racial groups.

How can state officials leverage this data to protect each individual’s right to vote?

To ensure free and fair elections, election officials in counties across Georgia may wish to reconsider their proposed restrictions on voting for the Runoff elections. Per findings from the Center for New Data’s analysis, poll closures and limited voting times pose a significant threat to the voting rights of minority voters, drawing heavy criticism from civil rights groups across the country.

Activist and equal rights organizations, such as the Georgia Branch of the NAACP, the American Civil Liberties Union, and Fair Fight, have pushed for maintaining the same number of polling locations as the general election, extending polling hours past 5 P.M. and providing options for Saturday voting. In a tweet commenting on the Center for New Data’s analysis, Fair Fight founder Stacey Abrams declared “Victories in Georgia and around the country were driven, in no small part, by ensuring access to the ballot. Voter suppression is not done yet — and we’ve got to be vigilant as we head to Jan. 5.”

As the early voting period kicks off on December 14, 2020, election officials are beginning to answer these concerns. In a December 9 news release, Cobb County officials announced major reversals to their original early voting plans, responding to outcry raised by protestors and activists. The Center for New Data’s Observing Democracy analysis played a prominent role in providing relevant data to help make the case for rolling back these restrictions. These new changes include the addition of two more locations for the third and final week of early voting in December.

Additionally, officials announced plans to move a voting site in Powder Springs from the Ward Recreation Center to the Ron Anderson Community Center, an area with a greater share of African American voters. This brings the total number of polling places in Cobb County to 5 for the first two weeks of the early voting period, and then up to 7 for the final week, still markedly lower than the 11 polling locations the county operated in the general election.

How was the Georgia Early Vote analysis undertaken?

The Center for New Data’s methods include a replication and an enhancement of a 2019 paper posted in the National Bureau of Economic Research (NBER) — “Racial Disparities in Voting Wait Times: Evidence from Smartphone Data” — which used geolocation data to examine and document racial disparities in wait times at polling locations. The paper’s original authors, two of whom serve as advisors to the Observing Democracy program, were detailed in their sensitivity analysis and showed results that were robust. To update the analysis for the 2020 early vote, these methods were replicated on a newer dataset.

To perform this analysis, The Center for New Data acquired and processed two large commercially-available geolocation datasets (provided via X-MODE & Veraset), which together comprised over 40 million daily devices, each of which generated an average of more than 150 daily pings. These data feeds were pseudonymized — transformed to no longer carry personally identifiable information — to provide only latitude and longitude indicators, timestamps, a randomly generated device ID and additional metadata such as device and OS characteristics. This enabled data processing methods to map mobility activity over time without revealing personal information about the device owners.

To put this geolocation data in context, The Center for New Data gathered a near-complete set of national early vote and election day polling locations for the 2020 General Election, as well as publicly available voter registration and demographic data. As part of its commitment to transparency, The Center for New Data has provided a full inventory of data sources below:

Data scientists began by transforming the raw data into mobility patterns by aggregating individual latitude and longitudinal pings into “stationary” and “travel” clusters. All datasets used for this analysis are already in widespread use, collected by “opt-in” location settings on mobile devices. The Center for New Data does not collect original data from individuals, nor undertake or support targeting in any way.

From there, stationary clusters within the vicinity of polling locations were identified as potential voters (subject to a set of exclusionary criteria based on mobility patterns and local precinct data). Based on the observed activity around the polling stations, The Center for New Data’s algorithms then measured the duration of time each device spent at a station, leveraging statistical methods and queuing theory to estimate key metrics. The sample was then adjusted using block-revel census data and re-weighted to approximate representativeness of the voting population at the national, state, and local levels.

Algorithms used in the Georgia Early Vote Analysis were tested using South Carolina’s June 9th primary and early voting in Georgia and North Carolina. To calibrate, results were validated against on-the-ground reports of wait times documented by county government agencies where available, as well as ballot return data. The results are believed to be robust, largely due to the extensive sensitivity testing performed in advance of the analysis.



The Center for New Data

A new kind of data nonprofit, born from the pandemic to advance policy research for the public interest. Our programs: Covid Alliance & Observing Democracy.