80% of Americans support voter ID. 28% support the SAVE Act. The gap is in the details — and the data shows exactly what getting it right looks like.
The same pollster asked about voter ID, proof of citizenship, and the SAVE Act by name — in the same week. The results show why implementation matters.
Any honest framework starts with documented scale. The audits that have actually looked for non-citizen federal voting are consistent:
| Jurisdiction | Scope | Finding | Source |
|---|---|---|---|
| Utah | 2M+ registered voters audited (Apr 2025–Jan 2026) | 1 non-citizen registration, 0 votes cast | BPC / Utah SOS |
| Georgia | 8.2M ballots cast | 9 non-citizen ballots — all caught by existing systems | Georgia SOS |
| Kansas | Full voter roll | 0.002% non-citizen rate; 30,000+ eligible citizens blocked | BPC / Kansas SOS |
| Travis County, TX | SAVE-flagged voters | 25% had already proved citizenship at registration | ProPublica / Texas Tribune |
| Virginia | Voter roll purge | 94% of those removed were U.S. citizens | VoteRiders, 2025 |
| Heritage Foundation | 20-year national database | Fewer than 100 confirmed non-citizen voting cases nationally | Heritage Foundation |
The difference is not whether to verify — it's whether the implementation is designed to work.
| Provision | SAVE Act (House-passed, Feb. 2026) | EISA Framework |
|---|---|---|
| Cost to voter | Full document cost ($25–$60+) | Free — federal program pays vital records offices directly |
| Implementation timeline | Immediate upon enactment | Minimum two federal election cycles with readiness certification |
| Verification method | Document presentation at registration desk | SSA/DMV database match; documents as fallback only |
| Name mismatch (69M women) | Marriage certificate (implementation undefined) | SSA name history database — no bridge document required |
| Tribal IDs | Limited (KIC card only) | All federally recognized tribal IDs with photo + expiration |
| Student IDs | Excluded | State public university IDs with photo + expiration + enrollment verification |
| Official liability | Criminal exposure regardless of intent | Negligence standard; safe harbor for good-faith compliance |
| Mail ballot secrecy | ID copy required with ballot | ID verified at registration; signature verification at submission |
| Roll removal | No notification requirement | Written notice + 90-day cure period required |
| Moving to new state or jurisdiction | Must re-verify from scratch | Registry lookup — no new documents required |
| One-time verification | No — re-verify at each registration event | Yes — verify once, registry carries it forward permanently |
| Constitutional footing | Weaker than Crawford (no free ID) | Consistent with Crawford v. Marion County (2008) |
| 80% consensus served? | No — 28% support once specifics known | Yes — achieves the stated goal without documented barriers |
Each provision addresses a specific, documented failure mode of rushed or underfunded implementation. Click any provision to expand.
If the federal government imposes a new requirement on a constitutional right, the federal government must eliminate cost as a barrier to exercising that right. This is not peripheral — it is the constitutional threshold.
The Supreme Court upheld Indiana's voter ID law in Crawford v. Marion County (2008) in part because Indiana provided free ID. A federal mandate without free ID is constitutionally weaker than Indiana's state law.
Implementation: Federal program pays vital records offices directly — payment flows government-to-government, no individual reimbursement, no fraud surface.
One document per registration event: Each eligible voter receives reimbursement for one citizenship document per registration event. Once citizenship is confirmed in the Federal verified-citizen registry, no further document is required at any subsequent registration — moving, name change, or otherwise. For most voters, this means one document, one time, ever.
Only one state — Arizona — has experience administering proof-of-citizenship voter registration, and its system has been in litigation since implementation. No other state has the infrastructure to comply on day one.
New Hampshire implemented a proof-of-citizenship requirement in 2025 with no meaningful phase-in. In the elections that followed, 244 eligible voters were turned away — not non-citizens, but citizens who were not notified in time.
HAVA precedent: Congress provided states with implementation timelines and $3.86 billion when it passed HAVA in 2002. It took until 2006 for most states to reach full compliance.
The most accurate and least burdensome citizenship verification happens at the database level — not when a voter presents documents to an election official with no training in document authentication.
HAVA Section 303 already established this principle in 2002: all 50 states verify registrations against DMV and SSA databases. That infrastructure exists and works. The question is whether to strengthen it or replace it with a document-presentation requirement that introduces new failure points.
84% of women who marry change their surname — approximately 69 million American women whose birth certificate does not match their current legal name.
The SSA already maintains a complete name change history linked to Social Security numbers. Every legal name change generates an SSA record update. This data already exists. Using it eliminates the name mismatch problem entirely without requiring chain-of-custody documentation.
Once your citizenship is verified, it should stay verified. A Federal verified-citizen registry records citizenship status linked to your Social Security number. Every time you re-register — moving to a new state, moving within the same state, changing your name — the receiving state queries the registry instead of starting over.
For most voters this means citizenship verification happens exactly once in their lives. The registry carries the verification forward permanently, unless citizenship status is disputed.
This is the provision that most directly reduces friction for the average voter — and it makes the entire framework more efficient. One verification event funds the registry entry. Every subsequent registration event is a lookup, not a re-verification.
This registry is a voter verification tool — nothing else. It cannot be used for any purpose other than confirming citizenship for voter registration. It is not a general citizenship database. The bill text prohibits any use outside voter registration verification, and the registry is subject to the full protections of the Privacy Act of 1974.
A verification system is only as reliable as the incentive structure facing the people who administer it. When the penalty for accepting a questionable registration is up to five years in prison — and the penalty for rejecting an eligible voter is nothing — officials will reject in all cases of uncertainty. That produces errors in one direction only.
This framework establishes:
The goal is accurate elections, not paralyzed election officials. These are not in conflict.
Mail ballot secrecy: Citizenship verification belongs at registration — not at ballot submission. Requiring an ID copy with a mail ballot connects an identifying document to specific ballot choices, compromising the secret ballot. Oregon has voted entirely by mail since 2000. Military absentee voting dates to the Civil War. Signature verification remains the authentication standard at submission.
Notification before roll removal: Virginia conducted a voter roll purge in which 94% of those removed were U.S. citizens. Roll maintenance without notice and cure produces that outcome.
The framework is designed so that most eligible voters are verified automatically — no document presentation required. Documents are the fallback, not the primary mechanism.
If Congress imposes a new requirement on a constitutional right, the burden of funding falls on the federal government. The HAVA model provides the proven infrastructure: formula grants through the Election Assistance Commission, tied to state implementation plans.
| Program Component | Estimated Cost | Basis |
|---|---|---|
| Free birth certificate program | $750M–$1.2B (one-time) | ~30M voters lacking ready access × avg $25 document cost + administrative overhead |
| Federal verified-citizen registry + SSA integration | $300–$500M (one-time) | Federal IT modernization at comparable scope; registry replaces repeated per-state verification costs |
| State implementation grants (2-cycle phase-in) | $400–$600M | HAVA formula grant model, scaled to new mandate scope |
| Mobile ID/document assistance (rural) | $200–$400M | Comparable mobile DMV programs in rural states |
| Election official training + safe harbor documentation | $50–$100M | HAVA poll worker training program precedent |
| Total estimated federal investment | $1.7B–$2.9B | One-time + phase-in costs; registry eliminates ongoing per-state re-verification costs; ongoing administration via EAC |
This framework takes the counterarguments seriously. Each answer is grounded in the data.
The audits that have actually looked for non-citizen federal voting consistently find it is vanishingly rare: Utah audited 2 million voters and found 1 non-citizen registration and 0 votes. Georgia found 9 non-citizen ballots out of 8.2 million cast. The Heritage Foundation's 20-year national database contains fewer than 100 confirmed cases nationally.
A two-cycle phase-in exists to ensure that eligible citizens are not removed from rolls or turned away while states build the infrastructure to implement verification correctly. The alternative — immediate implementation — has a documented track record: New Hampshire 2025, 244 eligible voters turned away in a single round of local elections. Those were not non-citizens. They were citizens who were not notified in time.
The core trade-off: accepting certain, documented disenfranchisement of eligible citizens in exchange for faster implementation of a system designed to address a problem the data consistently struggles to find is not a trade-off that holds up to scrutiny. If the problem is rare enough that audits can barely find it, a two-cycle phase-in to implement the solution correctly costs very little. The cost of getting it wrong falls entirely on eligible voters.
This framework does not oppose citizenship verification. It accepts the premise and proposes an implementation that works.
But the question's premise reverses the constitutional burden of proof. In the American legal framework, the government bears the burden of demonstrating that a restriction on a constitutional right is justified — citizens do not bear the burden of proving they deserve to exercise a right they already hold.
The Supreme Court addressed this directly in Crawford v. Marion County (2008). The Court upheld Indiana's voter ID law — but the constitutional reasoning depended on the ID being free. Justice Stevens wrote explicitly that if the State required voters to pay a fee to obtain identification, the law would not survive constitutional scrutiny. The government's obligation to provide free ID is not a courtesy — it is the constitutional load-bearing condition that makes the requirement valid at all.
The question is not whether voters should prove innocence. It is whether the government's implementation meets the standard the Court already established. This framework does. An unfunded immediate mandate does not.
Many countries do require identity verification to vote. The comparison is instructive — but it supports this framework, not an unfunded immediate mandate.
Canada, Germany, and most European democracies with strict verification requirements also provide national ID at no cost, maintain automatic voter registration, and fund the infrastructure required to implement the system. They did not impose the document requirement and leave citizens to source their own documents at their own expense.
If the standard is other developed democracies, that standard includes free ID, government-funded document retrieval, and phased implementation — all of which this framework provides. An unfunded mandate with no phase-in does not meet that standard.
For most Americans, obtaining ID is not difficult. This framework does not dispute that. The question is whether "most Americans" is the right threshold for restricting a constitutional right.
The data on the populations most affected is specific:
The free ID provision in this framework addresses this directly. If obtaining ID is straightforward for most people, the cost of providing it free to those for whom it is not is modest. The cost of not providing it is disenfranchisement of eligible citizens.
Yes — and that vulnerability must be built into the framework, not ignored. This is a genuine concern that deserves a direct answer.
Database errors exist in every large-scale government system. The DHS SAVE system acknowledged in February 2026 that it cannot make confident citizenship determinations in all cases. A ProPublica/Texas Tribune investigation found DHS rushed the SAVE system before it could verify updated citizenship data. Travis County, Texas found that 25% of voters flagged as potential non-citizens had already proved citizenship when they registered.
This is precisely why this framework includes: a 90-day cure period for flagged registrations before any removal; provisional ballots for voters under review; annual database error audits with corrective procedures; and document presentation as a fallback for inconclusive cases. The answer to database vulnerability is not to abandon database verification — it is to build error-correction into the system.
Provisional ballots are a mitigation, not a solution. They have a documented failure rate that falls disproportionately on the voters least likely to have the documents a new requirement demands.
Provisional ballots require voters to follow up within a narrow cure window — typically a few days — to provide qualifying documentation. Voters who lack the documentation on election day are generally the same voters who will have difficulty obtaining it within the cure window. The New Hampshire data is instructive: of 244 voters turned away in 2025 local elections, the turnaway itself was the problem — not the absence of a provisional process.
Provisional ballots are an appropriate backstop within a fully implemented, properly funded system. They are not a substitute for the phase-in period that allows states to build that system.
The Kansas and Virginia data are worth examining carefully — because they support the case for this framework, not against it.
Kansas implemented proof-of-citizenship voter registration. The result: a 0.002% non-citizen registration rate — and more than 30,000 eligible citizens blocked from registering. Courts struck the law down. The Kansas data does not show the current system failing to catch non-citizens. It shows a poorly designed verification system blocking eligible citizens at a rate thousands of times higher than the problem it was designed to solve.
Virginia conducted a voter roll purge. 94% of those removed were U.S. citizens. That is not evidence that voter rolls are full of non-citizens. It is evidence that roll purges without adequate notice and cure periods remove eligible citizens. Both data points argue for the provisions in this framework — not against them.
Under EISA, no. This is one of the most important distinctions between this framework and a document-based mandate with no portability mechanism.
Once a voter's citizenship is confirmed — either through SSA/DMV database matching or through the document fallback process — that verification is recorded in the Federal verified-citizen registry, linked to their Social Security number. When they re-register in a new state, a new county, or the same jurisdiction after a move, the receiving election authority queries the registry. A confirmed match is sufficient. No new documents required.
The same applies to name changes. When the SSA updates its name records following a marriage, divorce, or court order, the registry updates automatically. A voter who re-registers under their new name does not need to present a marriage certificate or other bridge document — the registry confirms both citizenship and the name change.
The practical result: for most eligible voters, citizenship verification happens exactly once in their lives. Every subsequent registration event — moving across the country, moving across town, updating a name — is a registry lookup, not a re-verification. This is the provision that most directly reduces friction while maintaining the integrity of the verification record.
Yes — and that's precisely the argument for EISA over the SAVE Act.
Utah and Georgia are both REAL ID compliant and already use HAVA §303 database verification and the federal SAVE system for voter registration cross-checks. Their audit results reflect a system that's already filtering non-citizens through existing infrastructure. Utah audited 2M+ voters and found 1 non-citizen registration and 0 votes cast. Georgia found 9 non-citizen ballots out of 8.2M cast.
That infrastructure is exactly what EISA is built on. EISA strengthens and formalizes database-first verification using the same SSA/DMV matching HAVA §303 has required since 2002, adds a federal registry for portability, and creates a document fallback only when databases are inconclusive.
The SAVE Act bypasses that working infrastructure and adds a mandatory document layer on top of it — at the cost of 30,000+ eligible citizens blocked in Kansas and 94% citizen removal rates in Virginia. Those outcomes occurred in states that also had existing verification systems. Adding a document mandate on top of working database verification doesn't improve accuracy. It adds friction that falls entirely on eligible citizens.
The audit results from Utah and Georgia aren't evidence that the honor system works. They're evidence that the database verification system HAVA already requires works. EISA keeps that system and improves it. The SAVE Act adds a layer that the data doesn't support needing.
This framework has been drafted as model legislation in proper Congressional bill format. It amends HAVA and NVRA directly, includes a full findings section sourced to the audit data, and is designed so any Senator or Representative could introduce it.