Six lawful administrative steps create a pathway from housing instability to lost registration. The pathway exists. How often it ends in actual disenfranchisement — versus voters who would not have voted regardless — is the harder question, and one this piece marks honestly.
Pick any large Georgia metro. Roughly 40% of households rent. Most are registered to vote at their current address, the address on the lease, on the utility bill, on the driver's license. Civic standing is anchored to housing.
Housing is the variable. Civic standing is the dependent.
Georgia's eviction process begins with a dispossessory filing. In Fulton County alone, filings have averaged over 60,000 per year through 2024–25. Many filings result in negotiated outcomes; a portion proceed to writ of possession.
The household receives a summons. The clock starts. The county recorder logs the filing. The civic record is, for now, unchanged.
Some move to another rental in the same county. Some move across the county line to take advantage of lower rent. Some leave the metro entirely. A meaningful fraction enter unstable housing — couch-surfing, motels, vehicles. The address on the voter file is now stale.
Update-your-registration is a one-form task. It is also one of the most commonly deferred administrative tasks in the modern American renter experience.
Voter rolls are maintained in part through mailed verification. A confirmation card is sent to addresses flagged for change. If the card comes back undeliverable, the voter is moved to "inactive" status. If the voter doesn't respond and doesn't vote in two consecutive federal elections, they are removed entirely.
The mechanism is lawful, the timeline is statutory, the consequence is silent. Most affected voters do not know it has happened.
If they show up at their old polling place, they are offered a provisional ballot, which may or may not be counted depending on the county and the timing of their move. If they go to the new polling place, they are not on the rolls. If they don't know either polling place is now wrong, they don't go at all.
The pathway from "registered voter who recently moved" to "counted vote" is narrow. The structural failure rate is high.
List maintenance happens on a statutory cycle. The voter, now invisible in election data, will not appear on the next county turnout report, the next demographic crosstab, the next political-targeting model. They become structurally non-participating — not by choice, but by administration.
The eviction filed in November becomes a registration gap by August. By the following primary, the voter is administratively a non-voter. The chain holds across cycles.
The mechanism is statewide. The throughput varies by county — driven by eviction-filing volume, rental tenure rates, registrar staffing, and the share of voters in high-mobility ZIP codes.
Highest absolute eviction filings in the state. Highest absolute count of voters moved to inactive status post-filing.
Adjacent to Fulton; significant net inflow of displaced renters who haven't yet re-registered.
Receiving counties for outer-metro displacement. Registration lag often runs multiple years.
Significant inflow from Fulton/DeKalb. Registration update rates lower than the receiving population's voting age.
Same funnel, smaller absolute numbers, similar rate. The mechanism doesn't require metro scale.
Lower filing volume but lower data quality, the funnel here is observed less because it's recorded less, not because it doesn't happen.
Pick a lens. The room reconfigures. Same fact, different argument, different chart. Press A/B/C/D.
The chain is documented at the population level. The individual causal link — this specific eviction caused this specific non-vote — requires record-linkage work that is doable, not done.
Magistrate court filings are public; Eviction Lab and academic researchers maintain longitudinal datasets for major Georgia counties.
The NVRA list-maintenance process and its Georgia implementation are statutory and public.
Linking specific evictions to specific inactivations requires matched-record work across magistrate and registrar data. Sample studies suggest the linkage rate is meaningful; full enumeration is FOIA-able.
Each county's funnel rate (filings → inactivations → purges) can be computed with current public data. We will publish a 159-county table in the companion deep dive.
The system removes the registration. Whether the voter would have voted, and for whom, is unknowable. The civic harm is removal, not specific electoral counterfactual.
Same-day registration, address-on-file flexibility, and other reform proposals are documentable but outside this descriptive piece. The reform case starts with the descriptive picture.
A reasonable critic — particularly an election administrator — would argue: each of the six steps exists for a defensible administrative reason. Voter-roll maintenance is required by NVRA. Address verification protects against impersonation. The "funnel" framing implies intent where there is procedure.
The critic is partly right. None of the six steps were designed to produce non-voters. But the cumulative effect is measurable, and procedure-without-intent still produces disparate impact under standard civil-rights tests. The episode's argument is structural, not motivational.
If you accept the steps as individually lawful (we do), the question becomes: is the cumulative outcome acceptable? That is a policy question this episode does not pretend to settle.