Accurate Data from Georgia Shows Racial Turnout Gap

On July 10, during a Georgia field hearing in the Committee on House Administration, Representative Barry Loudermilk (R-GA) raised questions about a Brennan Center study concerning the growing racial turnout gap in Georgia. The Brennan Center has found no way to justify the claims that Representative Loudermilk made regarding turnout in Georgia’s 2022 general election. We confidently stand by our report, and this resource lays out the multiple data sources and analyses that support our findings.

Representative Loudermilk stated the following:

“[I] also would like to submit something for the record. It was mentioned earlier today as well as in this hearing a Brennan study that made the claim that total turnout between 2022 and 2018 had decreased. I would like to submit for the record the actual Georgia Secretary of State’s data that shows that voter turnout actually did increase between 2018 and 2022. However, white voter participation between 2018 and 2022 actually decreased. Non-white voter participation significantly increased between 2018 and 2022. And that is raw data directly from the Georgia Secretary of State’s office.”

The Brennan Center has not been able to verify or replicate Representative Loudermilk’s claims using data from the Georgia secretary of state’s office, L2 Political, or any other currently available information.

Turnout can be calculated using different approaches, as described by the MIT Election and Science Lab. In our analysis, we used a standard definition: we divided the estimated number of ballots cast by the citizen voting age population (CVAP), which comes from the Census Bureau’s American Community Survey. This definition takes into account factors like population growth. Defining turnout in other ways — such as using the raw number of ballots cast or as a percentage of registered voters, as we show below — is insufficient because it fails to account fully for all people who could potentially participate. While CVAP doesn’t account for ineligible citizens, such as those with felony convictions, it nonetheless provides the best estimate of the number of potential voters. And while other measures such as the “Voting Eligible Population” take factors like disenfranchisement into account, they are available only as statewide measures and therefore can’t be used to estimate the turnout for different racial groups.

Representative Loudermilk was factually incorrect that the Brennan Center claimed overall voter turnout in Georgia declined. The report, Georgia’s Racial Turnout Gap Grew in 2022, explained that “overall turnout didn’t change much from 2018.” Our data shows that overall turnout in Georgia was about 51 percent both in the 2018 and 2022 elections. Moreover, data from the Georgia secretary of state’s office shows that the total number of ballots cast grew by about 15,000 votes, and the citizen voting age population of Georgia grew by more than 300,000 people between 2018 and 2021. The CVAP growth in Georgia varied across racial groups, which meant its effects on turnout rates were also different for these groups. Accordingly, our report noted that nonwhite turnout in Georgia declined, while white turnout increased.

Other aspects of Representative Loudermilk’s statement prove to be untrue as well, regardless of how we measure turnout.

The Brennan Center report is based on well-regarded datasets and is valid. The data are from L2 Political, which uses the official records from the Georgia secretary of state’s office. The estimated turnout figures in our report are also consistent with the findings of other established social science researchers, such as Professor Bernard L. Fraga at Emory University. The Washington Post similarly reported in May that census data suggested the decline in turnout in Georgia’s 2022 midterm elections was more concentrated among Black voters.

If anything, we find that these estimates may actually understate the racial turnout gap. That is because the L2 dataset uses self-reported information and other proprietary data to estimate individuals’ race. If we use only the self-reported data, footnote1_akoeg9g 1 In this approach, we model voters whose race is “unknown,” as well as that of individuals marked “other,” using a standard tool called  Bayesian Improved Surname Geocoding. the resulting turnout gap is even more pronounced.

The Brennan Center also attempted to substantiate Representative Loudermilk’s claims about turnout using alternative datasets and methodologies, as described below. These additional steps only reinforced our report’s conclusion that the turnout gap between white and nonwhite voters in Georgia grew between 2018 and 2022.

Using L2 Self-Reported Data on Race Only: Examining only those voters who self-reported their race in the L2 data, the absolute number of Georgia votes cast by nonwhite voters increased by 10,000 between 2018 and 2022. But taking into account the growth in the nonwhite citizen voting age population during this period, the actual nonwhite turnout rate decreased by about 3 percentage points. Self-reported white voters in Georgia cast over 100,000 more ballots in the 2022 election than in 2018, according to L2. But the state’s white citizen voting age population also changed, increasing the white turnout rate in Georgia by about 2 percentage points. footnote2_0rpxg6w 2 We stress that this is not the best way to calculate turnout; roughly 9.4 percent of voters in the L2 dataset didn’t report their race and are thus excluded from this analysis. Political scientists regularly estimate the races of individuals whose race is unknown, but our results are consistent even if we don’t do so.

Using Raw Voter File Data from the Georgia Secretary of State: To further confirm our findings, we obtained the raw voter file produced by the Georgia secretary of state’s office — the material Representative Loudermilk states that he relied on. This dataset uses voter registration snapshots dated October 31, 2022, and December 6, 2018, merged with the state’s online voter history records. footnote3_71k19bn 3 The numbers from the state’s voter file can differ slightly from L2’s for several reasons. First, later snapshots might not reflect the turnout of voters whose registration was canceled for any number of reasons. Many states conduct voter list maintenance shortly after a federal election, which can shift the composition of the voter file. L2 also removes from its lists individuals who it believes are not eligible to vote, even if their registration has not been formally cancelled by the state. While the different data sources can give slightly different results, the point stands: we have not been able to substantiate Representative Loudermilk’s claim.  When including only self-reports from the raw voter file and excluding voters whose race was unknown, the number of ballots cast by white voters in Georgia increased by roughly 35,000 between 2018 and 2022. The number of ballots cast by nonwhite voters in Georgia dropped by about 60,000 between 2018 and 2022. footnote4_a7k4ma5 4 Similarly, we attempted to calculate turnout from these files as a percentage of registered voters who cast a ballot, even though it is a less reliable way to calculate than calculating turnout among citizens eligible to vote. Even using this less reliable method, white voters’ turnout in Georgia decreased from 62 percent to 58 percent between 2018 and 2022, while for self-reported nonwhite voters, it decreased by much more: from 53 percent to 42 percent.

•  •  •

The Committee on House Administration has held a series of hearings on elections and voting in the last five months, and several members repeatedly raised concerns over the disparity between white and nonwhite voters in voter turnout and ballot access. The data demonstrating changes in the racial voter turnout gap in Georgia adds to the body of evidence supporting action, as it provides a clear measure of the real effects that some existing policies have on interfering with the enjoyment of the franchise. The turnout gap represents thousands of missing ballots, each one a loss of someone’s freedom to vote. And crucially, focusing on high overall voter turnout masks the gap between white and nonwhite voting.