The same provisional ballot, cast for the same reason, can be counted in one Georgia county and rejected in another. We map the forks and name the competing reads: structural disparate impact vs. lawful procedural variation. Both deserve hearing.
Provisional ballots are cast when a voter's eligibility is in question at the polling place, wrong precinct, ID issue, registration discrepancy, signature mismatch. The ballot is sealed and set aside for post-election adjudication. The voter has done their part.
In a typical Georgia general election, total provisional ballots cast statewide can run into the tens of thousands. Most are cast by voters who believe themselves eligible.
The first check: is this voter registered in Georgia? If yes, in which county? Inactive-status voters who reactivate by updating address may have their ballots counted; those whose registration has been fully purged generally do not. Roughly 15–25% fail at this stage.
This is where the eviction-funnel mechanism meets the ballot adjudication mechanism. Voters who lost registration administratively often discover it here.
Georgia counts most provisionals cast in the correct county but wrong precinct only for the contests common to both precincts. A ballot cast in the wrong precinct for a contest only on the correct precinct's ballot is not counted for that contest. Another meaningful share fails here — often voters who moved within the county but failed to update.
The rule is statutory. The enforcement is uniform on paper, variable in practice.
When a provisional is cast because of an ID issue, the voter has a short statutory window (typically three days) to provide acceptable ID to the county registrar. The fraction that successfully cures is small — outreach varies, the window is short, and many voters never learn their ballot is conditional.
Cure success rates differ measurably between counties with active outreach programs and those without.
Provisional ballot envelopes are signed. Election officials compare the signature to the registration record. Adjudication is subjective; training varies; appeal rights are limited. A small but real share are rejected for signature.
The disparate-impact concern: signature variation correlates with age, language, disability, and several other characteristics that are not the question being asked. This is a known weakness of the procedure.
Across recent Georgia general elections, the share of provisional ballots fully counted has averaged roughly 40%, meaning the majority of provisional ballots cast are not fully counted. The disparity by county is the part that does not get reported. Some counties run cure rates well above 60%. Others run below 25%.
The integrity question is not whether ineligible ballots are counted. It is whether eligible ballots are rejected, county-by-county, by an adjudication process whose pass rate depends on the geography of where the ballot was cast.
County-by-county provisional ballot survival rates from recent Georgia general elections. The geography of which-counted-and-which-not is the part of the election-integrity conversation that has no national constituency for telling.
Active outreach programs; relatively high cure-rate; comparatively high count rate.
Large volume; average cure rate; persistent geographic variability within the county.
Effective procedural infrastructure; less consistent cure outreach.
Bibb, Richmond, Chatham, cure-rate variability is high; outreach depends on board priorities.
Counties with high non-English-language voter share; cure rates often lower; rejection rates higher.
Lower absolute provisional volume; lower data quality; cure infrastructure typically minimal.
Pick a lens. The room reconfigures. Same fact, different argument, different chart. Press A/B/C/D.
The procedure is lawful. The disparities are documented in academic literature and in SoS reports. Whether the disparities constitute discriminatory intent or are an artifact of variable local administration is a legal question, not a documentary one.
GA SoS publishes provisional ballot cast and count totals per cycle.
The five-fork adjudication procedure is established in OCGA and county election board rules.
County-level adjudication outcomes are publicly available but require assembly into a single comparable dataset.
Cure rates require pairing cast-counts with cured-counts; this is doable from county filings, not yet centralized.
Demographic patterns in rejection require survey-based or matched-record analysis; aggregate provisional data alone does not establish it.
Whether the disparities constitute a Voting Rights Act issue or other legal claim is for litigators, not for this piece.