The SAVE Act requires states to verify the citizenship of registered voters using one or more of four approved sources. Here is the relevant text:
A State may meet the requirements of paragraph (3) by establishing a program under which the State identifies individuals who are not United States citizens using information supplied by one or more of the following sources:
(A) The Department of Homeland Security through the Systematic Alien Verification for Entitlements (‘SAVE’) or otherwise.
(B) The Social Security Administration through the Social Security Number Verification Service, or otherwise.
(C) State agencies that supply State identification cards or driver’s licenses where the agency confirms the United States citizenship status of applicants.
(D) Other sources, including databases, which provide confirmation of United States citizenship status.
All four sources rely on data matching — comparing voter registration records against existing government or third-party databases. The question worth examining is who gets flagged in that process even when they are fully eligible to vote.
Source A: DHS / SAVE Database
Naturalized Citizens Whose Records Haven’t Been Updated
Real Example: In 2012, Florida used the SAVE database to identify suspected non-citizens on its voter rolls. The initial purge list flagged nearly 2,700 registered voters. Two of the named plaintiffs in the resulting lawsuit — Karla Arcia, a naturalized citizen from Nicaragua, and Melande Antoine, a naturalized citizen from Haiti — were both wrongly identified as non-citizens despite being fully eligible to vote. More than 80 percent of those flagged were voters of color. The 11th U.S. Circuit Court of Appeals ultimately ruled the purge violated the National Voter Registration Act. (Arcia v. Florida Secretary of State, 772 F.3d 1335 (11th Cir. 2014))
Why It Happens: SAVE was built for benefits verification, not voter roll maintenance. When someone naturalizes, USCIS updates their record — but SAVE is a separate system. DHS also maintains records across multiple components (USCIS, ICE, CBP) that don’t always sync. A SAVE query can return a person’s prior immigration status rather than their current citizenship status, particularly if the record update hasn’t fully propagated.
How Often: Florida’s 2012 experience is the most documented large-scale example. The state eventually abandoned the effort after the legal challenge revealed how many legitimate citizens had been flagged. A 2024 challenge to a similar Alabama program raised the same concerns, with courts noting that naturalized citizens are uniquely and systematically at risk under database matching programs because their records move through more government systems than native-born citizens. (NAACP LDF Letter on Alabama Voter Purge, August 2024)
Legal Permanent Residents Who Naturalized — But Whose DHS Record Still Reflects Prior Status
Real Example: This issue is structural to how DHS maintains records. A person who held a green card and then naturalized has records in multiple DHS systems. The naturalization record is created by USCIS. Older records held by ICE or CBP reflecting prior non-citizen status are not automatically deleted — they coexist. Courts in the Florida litigation noted this explicitly, finding that the matching methodology made it likely that “properly registered citizens who would be required to respond and provide documentation would be primarily newly naturalized citizens.” (United States v. Florida, 870 F. Supp. 2d 1346 (N.D. Fla. 2012))
Why It Happens: Federal immigration record architecture is not built around a single authoritative record per person. Multiple agencies hold overlapping records that reflect different moments in time. Which record surfaces during a SAVE query depends on how the query is structured and which system responds first.
How Often: No federal agency publishes a rate for this specific failure mode. What is documented is that courts have repeatedly found these programs disproportionately burden naturalized citizens, and that “naturalized citizens will always be at risk” under database-matching approaches. (Mi Familia Vota v. Fontes, D. Ariz. 2024)
Name Discrepancies
Real Example: Georgia’s “exact match” voter registration policy required that voter registration data match government database records exactly. Between July 2013 and July 2016, 34,874 voter registration applications were cancelled due to non-matches. Of those, 76.3 percent were submitted by Black, Latino, or Asian-American applicants. In 2018, a separate batch of 51,111 pending registrations showed the same pattern: 80 percent belonged to Black, Latino, or Asian-American voters. Triggers included hyphenated last names, accent marks, minor typos, and name order variations. Georgia eventually abandoned the system after three federal lawsuits. (Campaign Legal Center: Challenging Georgia’s Exact Match Policy)
Why It Happens: American databases store names in First/Middle/Last fields. Many cultures do not use that structure. Vietnamese, Korean, and Chinese naming conventions place the family name first. Hispanic naming conventions frequently include two surnames, sometimes hyphenated, sometimes not. When records are entered across different agencies — a naturalization certificate, a Social Security card, a driver’s license, a voter registration form — clerks make independent judgment calls about which name goes in which field. Those judgments are not consistent. A woman named Maria Garcia-Rodriguez who naturalized under that full hyphenated name but registered to vote as Maria Rodriguez generates a non-match. A Vietnamese voter whose name is entered in family-name-first order on one record and given-name-first order on another generates a non-match. Neither voter made an error.
How Often: This is the most consistently documented failure mode across all database matching programs. The Georgia litigation is the most complete public record of its scale. A 2009 SSA report examining how well voter registration data matched SSA records found a high no-match response rate and concluded the inconsistencies “could hinder the states’ ability to determine whether applicants should be allowed to vote.” (NBC News: Lawsuit — Exact Match System Negatively Impacts Georgia’s Minority Voters)
Source B: Social Security Administration
Citizens Whose SSA Records Predate Citizenship Data Fields
Real Example: SSA records date to the 1930s. The citizenship status field was not a standard part of earlier records and was not retroactively populated for older accounts. A person who obtained their Social Security number in the 1950s or 1960s may have a record that contains no citizenship data at all. A SAVE Act query to SSA for that record returns a blank field — which, depending on how a state interprets it, may be treated the same as a non-citizen flag.
Why It Happens: Legacy database architecture. Fields added to a system decades after initial record creation are only populated when there is a triggering event — a record update, a benefits application, a name change. For people whose SSA record was created and never substantially updated, the citizenship field simply does not exist.
How Often: SSA has acknowledged data gaps in older records. There are approximately 50 million Americans over age 65. The precise share of those with unpopulated citizenship fields in SSA records is not publicly reported.
Citizens Born Abroad to American Parents
Real Example: A child born in Germany to a U.S. military family acquires citizenship at birth under INA Section 301. Their Social Security number is issued based on a Consular Report of Birth Abroad rather than a U.S. birth certificate. The SSA record creation pathway for this population differs from domestically born citizens, and the citizenship confirmation field may be populated inconsistently — or not at all.
Why It Happens: The process by which citizenship is recorded in SSA’s system was designed primarily around domestic birth records. Foreign-born citizens who acquired citizenship at birth follow a different administrative pathway, and the resulting records are not always structured identically.
How Often: Approximately 3 million Americans are estimated to be living abroad at any given time, with additional military and diplomatic families who have returned. The false-flag rate for this population under SSA queries is not publicly documented.
Source C: State DMV Records
Licenses Obtained Before States Tracked Citizenship Status
Real Example: A person who obtained a driver’s license in California in 1995 and renewed it several times since. California’s compliance with the REAL ID Act — which requires states to verify citizenship or lawful presence — was delayed for years due to the state’s legal challenges to the law. Records created before states implemented REAL ID-compliant systems were not always retroactively updated. That person’s DMV record may have no citizenship field, or one marked “not verified,” rather than a confirmed citizen status.
Why It Happens: States upgraded their DMV systems at different times and under different circumstances. REAL ID compliance was phased in unevenly across the country. Pre-upgrade records were generally not retroactively corrected because there was no triggering event to force an update.
How Often: This is highly state-dependent. States that resisted or delayed REAL ID compliance — including California, which fought the law for years — have larger pools of records with no citizenship data. California’s DMV manages records for approximately 27 million licensed drivers. The share of those records with no citizenship field populated is not publicly reported.
Licenses Obtained Using Alternative Documents
Real Example: Several states permit applicants to obtain a standard or limited driver’s license using documents other than a U.S. birth certificate or passport. In those cases, the DMV record may reflect only that a license was issued — not whether the applicant was a citizen or a lawful non-citizen. Someone who was a non-citizen at the time they obtained their license, later naturalized, and never updated their DMV record would have a record that reflects their prior status.
Why It Happens: DMV records are designed to track driving eligibility, not citizenship status. The citizenship field — where it exists — reflects what was presented at the time of the original application. It is not automatically updated when someone’s immigration status changes.
How Often: Specific figures are not publicly available. The issue is most prevalent in states with large immigrant populations and a history of alternative-document licensing.
Source D: “Other Sources, Including Databases”
The Open Category
Source D is the most difficult to analyze with precision because the law does not define it. The language — “other sources, including databases, which provide confirmation of United States citizenship status” — sets no standard for what qualifies. Any database a state deems credible potentially qualifies.
Real Example: The Interstate Crosscheck Program, operated by Kansas under then-Secretary of State Kris Kobach, is the most documented example of what unregulated database matching produces. Crosscheck compared voter registration records across participating states to identify potential double registrants. In 2017, it analyzed 98 million records and flagged 7.2 million as “potential duplicate registrants.” Research by Harvard, Stanford, the University of Pennsylvania, Yale Law School, and Microsoft Research found the false positive rate was 99.5 percent. Less than four people were charged with double voting out of those 7.2 million flags — and not a single flagging led to a conviction. The program was suspended in 2019 as part of an ACLU lawsuit settlement. (ACLU of Kansas v. Kobach)
Why It Happens: Crosscheck matched records on first name, last name, and birthdate only. It did not require matching Social Security numbers. Common surnames — disproportionately shared among Black, Hispanic, and Asian Americans — generated enormous numbers of false matches. According to the program’s own flagged lists, African Americans were overrepresented by 45 percent, Hispanic voters by 24 percent, and Asian voters by 31 percent relative to their share of the voter population. (Wikipedia: Interstate Voter Registration Crosscheck Program)
Why It’s Relevant to Source D: Crosscheck was not a citizenship verification program — it was a double-registration program. But its mechanics illustrate what happens when a state uses a loosely defined database with no accuracy standard. Source D imposes no accuracy requirement on whatever databases a state chooses to use.
The Common Thread
Each of the four sources is a data matching system. Data matching has a documented accuracy problem that falls hardest on specific populations:
- Naturalized citizens — whose records move through more government systems than native-born citizens, creating more opportunities for mismatch.
- Voters with non-Western naming conventions — Vietnamese, Korean, Chinese, and many Hispanic naming structures conflict with how American databases store names in First/Middle/Last fields.
- Older voters — whose records in SSA and DMV systems predate the fields used to confirm citizenship.
- Citizens born abroad — whose records were created through a different administrative pathway than domestically born citizens.
The SAVE Act does not require states to use all four sources. It requires them to use at least one. It does not set an accuracy standard for any of them.
Learn More: