Case Study 14.2: The 2018 Georgia Governor's Race — Turnout, Suppression, and the Abrams-Kemp Margin

Background

The 2018 Georgia gubernatorial election between Democrat Stacey Abrams and Republican Brian Kemp became one of the most scrutinized American elections of the decade, focused attention not just on its outcome but on what the election revealed about the relationship between voter registration, turnout barriers, and electoral administration.

The race was narrow: Kemp won by approximately 55,000 votes out of 3.9 million cast, a margin of about 1.4 percentage points. Abrams declined to concede, stating explicitly that "the architects of suppression" had denied the will of Georgia's voters. Kemp had served simultaneously as Georgia's Secretary of State — its chief election official — throughout the campaign.

The Mechanics of the Dispute

Several specific practices drew scrutiny:

Exact match registration. Under Georgia's "exact match" policy, voter registrations were held in "pending" status when information on the registration form did not precisely match the state's records in the DMV or Social Security Administration database. Even minor discrepancies — a missing hyphen, a transposed letter, an abbreviation — triggered the hold. By late October 2018, approximately 340,000 registrations had been held under this policy.

Critics noted that the policy disproportionately affected voters of color. The Associated Press reported that 70 percent of pending registrations belonged to Black voters, 8 percent to Hispanic voters, and 3 percent to Asian American voters, while Black voters made up about 32 percent of Georgia's population at the time. The disparity was attributed partly to the use of DMV records, which more frequently contain transcription errors or formatting differences for names from non-English languages.

Voter roll purges. The Kemp-led Secretary of State office had purged approximately 1.4 million voters from the rolls between 2012 and 2018, including more than 340,000 removed in a single July 2017 purge of voters who had been inactive for several elections. Civil rights groups argued that many of these purges violated NVRA provisions requiring notice and cure periods.

Polling place closures. Several counties, particularly in rural areas with high Black populations, had consolidated polling places in the years before 2018. Voting rights groups documented cases of voters driving 10 or more miles to reach their nearest polling place, with waiting times exceeding three hours in some locations.

Analytical Challenges

Establishing that these practices changed the outcome requires answering a difficult causal question: how many votes did Abrams lose due to registration problems, purges, and access barriers? This is methodologically challenging because you are trying to estimate a counterfactual.

Several approaches have been attempted:

Pending registration analysis. If the 340,000 pending registrations had been resolved in voters' favor and those voters had turned out at rates comparable to registered Georgia voters in their demographic group, the analysis suggests they could have cast approximately 50,000–100,000 additional votes. Estimating what share of those votes would have gone to Abrams, based on demographic composition, suggests a potential impact approaching the 55,000-vote margin. However, this calculation requires many assumptions at each step.

Purge analysis. Estimating the effect of purges requires knowing how many purged voters would have voted if retained, at what rate. Voters removed for inactivity have, by definition, low recent voting propensity, complicating this analysis.

Polling place access. Research on polling place distance and wait times suggests meaningful deterrent effects, particularly for voters with transportation constraints. Estimating the specific effect in Georgia requires county-level analysis that is feasible but requires data not always publicly available.

The Broader Context: Administrative Suppression

Political scientists have increasingly focused on what some call "administrative suppression" — the accumulation of individually justifiable administrative decisions (maintaining roll accuracy, implementing registration verification, consolidating underutilized polling places) that in combination produce meaningful disparate effects on specific communities.

The challenge for analysts is that each individual decision has a plausible administrative rationale. The Secretary of State can credibly argue that exact-match improves roll accuracy, that infrequent-voter purges keep rolls current, and that polling place consolidation improves efficiency. The cumulative effect on specific populations is harder to attribute causally to intent.

This is precisely why the Fourth Circuit's analysis in the North Carolina case — noting that the legislature had specifically requested racially disaggregated data before making decisions — mattered. It moved the analysis from the structure of each policy in isolation to the pattern of information-seeking and decision-making.

The Abrams Legacy and Subsequent Developments

Abrams did not simply accept the outcome. She founded Fair Fight Action, a voting rights organization that has invested heavily in voter registration, litigation, and advocacy in Georgia and other states. The organization's registration drives contributed to the Georgia electorate's demographic shift that made Biden's 2020 victory possible.

The Georgia legislature subsequently passed SB 202 in 2021, which — among many provisions — reduced the availability of drop boxes, made it a crime to provide food or water to voters waiting in line, and restructured the State Election Board. Civil rights groups have challenged several provisions; federal courts have issued mixed rulings.

Implications for Turnout Analysis

The 2018 Georgia case is a master class in the interaction between turnout analysis and electoral administration:

  1. The voter file is a political document. Who is on the voter rolls, and who has been removed, reflects administrative decisions that have political consequences. Analysts who treat the voter file as a neutral list of eligible voters miss this.

  2. Turnout models trained on biased data reproduce the bias. A model trained on historical turnout in an environment with systematic suppression will treat the low turnout of suppressed groups as a predictable baseline, not as a problem to be corrected.

  3. Margin analysis requires counterfactual thinking. Assessing the impact of any electoral intervention — including administrative decisions — requires estimating what would have happened in a counterfactual world. This is difficult, contested, and deeply political.

  4. The analyst's choices carry ethical weight. Deciding what to model, how to interpret findings, and what questions to ask about administrative practices is not politically neutral. An analyst who builds a sophisticated turnout model but never examines whether the voter file has been systematically purged is missing a crucial piece of the analysis.

Discussion Questions

  1. Brian Kemp's defenders argued that roll purges and exact match requirements are standard administrative practices necessary to maintain roll accuracy. How would you evaluate this argument in light of the disparate racial impact data?

  2. What would a statistically rigorous analysis of the impact of Georgia's 2018 practices on election outcomes require? What data would you need? What assumptions would be unavoidable?

  3. Fair Fight Action invested heavily in voter registration post-2018. Is registration advocacy a legitimate response to the kind of administrative suppression alleged in Georgia, or does it place an unfair burden on suppressed communities to overcome their own disenfranchisement?

  4. How should campaign analysts like Nadia approach a state where voter roll quality is uncertain due to potentially biased administrative practices? How would this affect her turnout model?