When the lead moved after election night in the 2026 LA mayoral primary, some called it fraud. A decade of official, precinct-level data says otherwise, and shows exactly why.
It feels wrong, but it rests on a hidden assumption: that the last ballots counted should look like the first ones. In California, they structurally can't. Start with one number from the 2026 LA mayoral primary:
of all ballots were counted after election night. That isn't a recount, it's the normal, scheduled California canvass.
Confidence in the count starts with confidence in the rolls. So before the data, here is the front door: who is eligible in California, how people get registered, how that registration is checked against government records, and the question that comes up most, when do you actually have to show ID?
At the state's registration site or on a paper form. You provide your California driver license or ID number, or the last four digits of your Social Security number.
Under California Motor Voter, eligible people completing a DMV transaction are registered automatically unless they opt out, with identity already documented by the DMV.
In the 14 days up to and including Election Day, you can register and vote at a vote center or county office. That ballot is provisional until eligibility is confirmed.
16 and 17 year olds can pre-register, and their registration activates automatically the day they turn 18.
Every new registration runs through California's single statewide voter database, called VoteCal. The identifier you provide, a California driver license or ID number, or the last four digits of your Social Security number, is matched against Department of Motor Vehicles and Social Security Administration records, as federal law requires. If it cannot be matched, you can still register, but you may be asked to show identification the first time you vote.
Because VoteCal is one centralized system rather than thousands of separate lists, it is also what keeps the rolls clean over time. It removes duplicate registrations when someone re-registers, updates records when people move, and cancels registrations against state death records and felony-imprisonment status from the Department of Corrections. That ongoing list maintenance is why the registered count tracks the eligible population instead of drifting above it.
There is no general voter-ID-at-the-polls requirement, because your identity is verified when you register, not each time you vote. The one exception comes from federal law: if you registered online or by mail, are voting for the first time in a federal election, and did not provide an ID number that could be verified, you may be asked to show identification that first time. Everyone else is already verified in the system.
To register in California you must be a United States citizen, a resident of California, and 18 or older on Election Day. You must not currently be serving a state or federal prison term for a felony, and must not have been found mentally incompetent to vote by a court.
Voting rights are restored automatically once a prison term is complete. People on parole, probation, or post-release community supervision can register and vote, with no waiting period and no application required.
VoteCal continuously reconciles registrations against other official records: DMV and Social Security data to confirm identity, U.S. Postal Service and inter-county data to catch moves, California Department of Public Health death records to cancel deceased voters, and Department of Corrections data on felony imprisonment. Duplicate registrations created when a voter re-registers are merged so a person appears once.
This is why a registration total can rise with population yet never exceed the number of eligible people: the same database that adds new voters is constantly removing those who have moved away, died, or registered twice.
California issues AB 60 driver's licenses to undocumented residents, but each one is stamped "FEDERAL LIMITS APPLY," and the voter pipeline is built to exclude them. An AB 60 applicant is never shown the option to register, and their information is not transmitted to the Secretary of State, even on a change-of-address transaction. A driver's license and a voter registration are two separate tracks, and AB 60 holders are blocked from the second one.
Citizenship is required because the law requires it: federal statute makes voting by a non-citizen a felony punishable by prison and deportation (18 U.S.C. §611), and California's Constitution independently limits voting to U.S. citizens. It is not a state policy choice.
One honest point on documentation: to register you provide a California DL/ID number or the last four digits of your Social Security number, which are matched against DMV and Social Security records, and you attest to citizenship under penalty of perjury. California, like most states and federal law, does not require documentary proof such as a birth certificate; verification is sworn attestation plus database matching, backed by the felony exposure and the cross-checks above. Whether to require documentary proof is exactly what is on California's November 2026 voter-ID ballot measure.
This is the part people find suspicious, and it is actually a federal protection. The National Voter Registration Act forbids officials from quickly removing someone just because data suggests they moved. The process is deliberately slow so eligible voters are not wrongly purged: the county gets a signal (returned mail, postal change-of-address, or California's ERIC cross-state report), mails a forwardable notice, and if there is no response moves the record to "inactive." The voter can only be fully removed after staying inactive through two consecutive federal general elections, roughly four years.
So someone who left years ago can absolutely still show a registration, almost always already flagged inactive. The fastest way it clears is when they register in their new state, which signals California to cancel. Crucially, a stale registration is not a vote. It sits inert. To become fraud, someone would have to physically obtain that person's ballot and forge their signature past verification, one prosecutable felony traceable to that record. The lag inflates the inactive list, not the count of real votes, which is why the rolls still never exceed the eligible population.
Whether to add a broader voter-ID requirement is itself a live question in California: a voter-ID ballot measure qualified for the November 2026 ballot. However that vote goes, the point stands that the rule is being decided in the open, by voters, not changed in secret.
This is the backdrop for everything else. Since California adopted universal vote-by-mail, roughly three-quarters to four-fifths of all votes in every election arrive by mail, and mail is counted last. A large batch of ballots tallied in the days after Election Day is not unusual here. It is simply how most people vote. Hover any point.
Share of votes cast by mail in the marquee LA County contest each cycle. Every bar sits in the same high band, because mail is the normal way Californians now vote.
It does not only rise. In 2024, mail dipped to 72% from 80% in 2022, because in-person voting rebounded (1.04M in-person ballots in 2024 vs 0.90M in 2020) as the pandemic receded and Republicans pushed Election-Day turnout. An honest dataset shows the method mix moving in both directions.
A count isn't one electorate revealed slowly. It's a sequence of different sub-electorates, counted in a legally-fixed order, and those groups vote differently, so the running margin must move.
Ballots returned early, plus election-day voters. Tend to be older, higher-propensity, more consistent voters.
Dropped off on election day, postmarked within the 7-day window, or needing a signature cure. Skew younger and more progressive.
But why does the late group lean Democratic? Because since 2020 the choice of method became partisan. This chart connects each contest's leading Democrat and leading Republican by the share of their own votes that came by mail.
The cause is behavioral, and it is no secret. It traces directly to what each party told its own voters. In 2020, President Trump repeatedly attacked mail voting, calling it "RIGGED" and predicting the "most INACCURATE & FRAUDULENT Election in history," and urged supporters to vote in person instead. Fox hosts and other conservative media echoed it. The message stuck: by 2025, Pew found an 83% to 32% partisan split on whether mail voting should even be allowed, a 51-point gap that barely existed before 2020.
So Democrats vote by mail and Republicans vote in person, and in-person is counted first while mail is counted last. The early count leans Republican, the late count leans Democratic. Nothing about the ballots changes, only which voters' ballots are being counted at each moment. The trend is reversible too: in 2024 the Republican Party ran a "Bank Your Vote" campaign urging its own voters to vote early and by mail. As that messaging shifts, the blue shift will shrink with it, exactly what you'd expect from a behavioral cause and not from fraud.
California counts for weeks by design, not by failure: it accepts late-but-valid mail, gives voters a window to fix signature problems, and hand-checks paper before certifying. So the total keeps climbing for days after Election Day, on a schedule announced in advance. Nearly 900,000 ballots, late-arriving mail, provisional, and signature-cure ballots, were counted in the week after June 2. Hover any point.
The daily batches are lumpy, the biggest single day (June 8, +192,832) was 2.5× the smallest. That looks irregular, but it just reflects mail-processing and signature-cure timing. The total only ever rises, and each batch's size is published before it's added.
Each card is a real LA County contest, from its certified Statement of Vote. The top bar shows who led among in-person voters (counted first); the bottom bar shows who led the vote-by-mail ballots (counted last). For partisan races it is the leading Democrat against the leading Republican. Watch how often the leader changes between the two bars.
In the 2022 mayoral primary, Caruso led the in-person vote and Bass won the mail. In the 2026 Governor primary the Republican, Steve Hilton, led in-person while the Democrat, Xavier Becerra, won on mail. Same mechanism, opposite directions. The party that wins on mail is whichever one votes that way, and mail is always counted last.
Strip out every mail ballot and you get roughly the election-night picture. Here is each race's margin with in-person votes only, then again once all mail is counted. The swing between the two dots is the mail electorate, predictable in advance from the known mail mix, not a surprise injected later.
In 2022, Newsom and Dahle were essentially tied among in-person voters (+0.1), yet Newsom finished +35.6, because mail leaned heavily Democratic. And the 2022 Mayor's race actually changed hands: Caruso led the in-person count, Bass won once mail was tallied. That's not votes appearing, it's a different, later-counted electorate. Every swing points toward the mail-leaning candidate, in proportion to the mail gap. Never random, never unbounded.
Here is the cleanest outside test. Whether a state shows a shift at all, and which way it points, comes down to one rule: when officials are allowed to start processing mail ballots. Same kind of votes, opposite optics.
Florida lets officials open and prepare mail ballots starting weeks before Election Day, so it reports almost the entire result within 30 minutes of polls closing. Little left to count late means almost no shift. Fast, purely by procedure.
Pennsylvania law barred processing any mail ballot until Election Day morning. Trump led by ~15 points that night; it took until Saturday, four days later, for the mail to be counted and Biden to win. The famous "red mirage," caused entirely by the counting rule.
Arizona reports early-returned mail first, so Democrats often lead on election night, then late-arriving and Election-Day ballots pull the margin back toward Republicans. A "red shift." Same mechanic, opposite direction.
For a shifting count to be fraud, the fraud would have to run blue in Pennsylvania, red in Arizona, and barely exist in Florida, all in the same election. It doesn't. The shift's size and direction track one boring variable: when each state lets officials start processing mail. California, like Pennsylvania, counts an enormous late-mail load, so it is slow and it shifts. The only thing that would change is speed, not security. Florida proves you can count fast; the single trade is letting officials begin earlier.
Two fears live here: that registration outpaces real people, and that the partisan tilt is engineered. California's own Reports of Registration answer both. Each bar is statewide registration by party; the line is the share of eligible Californians who are registered.
Registration tracks the eligible population, and it never exceeds it. The rolls grew from 17.7M (73% of eligible) to 23.1M (84.8%) as the state automated and expanded registration. High participation, but a register can't hold more voters than there are eligible people, and it doesn't.
The partisan mix isn't engineered, it's just who's registered. Republican registration drifted from 28.6% (2014) to 25.0% (2026) while Democrats hold a ~45% plurality and No-Party-Preference voters ~23%. Combine that standing majority with the fact that Democrats vote by mail more (counted last), and the "blue shift" is simply the rolls expressing themselves in counting order.
The simplest fraud check is the most direct: compare ballots counted to registered voters. Every precinct's Statement of Vote lists both, so we can. Here is turnout, ballots cast as a share of registration, for all seven contests. Hover any bar.
Countywide, turnout never tops ~76%. It rises and falls exactly with the type of election, sleepy primaries near 31–39%, presidential generals near 66–76%. In no election do countywide ballots come close to exceeding registration.
Zoom into individual precincts and a few do show more ballots than registered voters, up to 12 precincts in 2024, one micro-precinct as high as ~1,750% (a few dozen ballots against a tiny registration base). This is the literal "more votes than voters" claim, and the cause is legal: California's Conditional (same-day) Voter Registration lets eligible people register and vote at a Vote Center on the spot, so their ballots can land in a precinct that hadn't counted them in its closing registration. It's a bookkeeping offset that nets out countywide, not extra votes.
Fraud is targeted: you'd rig the race you care about, not the water board. So here's a test: take 14 unrelated countywide contests from the 2024 ballot, President, U.S. Senate, ten state propositions, two county measures, all voted by the same people, and measure the mail share of each. If the presidential race had been stuffed with mail ballots, its dot would sit apart from the pack.
All 14 contests land between 72.0% and 73.2% mail, a 1.2-point spread, standard deviation 0.37 points. The presidential race is dead center, indistinguishable from a county measure nobody would bother to rig. The mail/in-person split is a fact about how people vote, not what they vote on, so it can't be the signature of a scheme aimed at any one race.
Aggregate totals can't reveal a single forged ballot, so look at the gate every mail ballot passes through. Each envelope signature is compared against the voter's registration record, and invalid ballots are rejected every cycle. In California's 2024 general election:
of mail ballots were rejected, 122,480 of them, for failing verification.
Rejection reasons, 2024. Counties ranged from 0.17% (Amador) to 2.54% (Imperial).
If forged or unverified ballots were being waved through, the rejection rate would collapse toward zero. Instead the gate turns away ~1 in 100 mail ballots every election, mostly for signature mismatches and missed deadlines, the exact checks you'd want. And the law requires officials to contact voters to "cure" a problem, so the gate is strict and fair, not a rubber stamp in either direction.
This is the real question under every fraud claim. Set aside the shift for a moment: could election workers simply alter the totals, quietly discard the other side's ballots, or buy votes wholesale? To change a result, fraud has to be systematic, and a system built like California's is exactly where systematic fraud cannot hide. Take the most common versions, point by point.
To move a statewide race you would need to alter hundreds of thousands of ballots across 58 independently run county systems, handled by thousands of temporary workers from both parties, with observers from both parties watching, and not one person leaks it. A conspiracy that large has never held, because it cannot stay secret.
Every vote is a paper ballot that is kept. After the election, a public hand count of randomly drawn paper is checked against the machine totals. Alter the electronic numbers and the hand count no longer matches, in the open, and the paper wins. You would also have to change one race without disturbing the other races on the same ballots, the precise fingerprint the previous section found absent.
The signature is checked on the outer envelope, and the ballot is separated from it before it is ever opened. At the instant a ballot is accepted or rejected, no one knows which candidates it chose. Rejections follow objective rules (signature, deadline), are itemized by reason in public reports, and can be cured by the voter. There is no partisan lever to pull, because the people judging the envelope are blind to the vote inside.
Vote-buying is a felony for both sides (up to 5 years, federal). It is real but tiny: documented cases overturned a city-council seat won by 10 votes, never a statewide race. Two reasons it can't scale. The secret ballot means a buyer can't confirm how anyone actually voted, so to enforce it they must fill out or collect the ballots themselves, the traceable harvesting that gets caught. And the witnesses: swinging California means paying hundreds of thousands of people, each one able to report a crime. That paid voters come forward at all is proof the secrecy collapses at a handful, let alone a million.
Add it up: a blind acceptance process, a retained paper trail, a public hand-count audit, bipartisan staffing and observation, and 58 separate systems that would all have to be defeated at once, in the open. The shifting count needs none of that. It is just mail counted last. One of these explanations requires a flawless silent conspiracy across tens of thousands of people. The other requires only that Democrats and Republicans mail their ballots at different rates. The evidence fits the second.
Every rebuttal below is anchored to the certified data on this page or to California's published procedures.
None of this asks for faith. California's process is built to be independently re-checked, by anyone.
After certification, officials hand-count randomly drawn paper ballots until the statistics confirm the machine totals. It's the actual proof, and it's public.
The precinct-by-precinct, by-method file, the exact data behind this page. Recompute any total yourself.
Ballots issued vs. returned vs. counted vs. cured vs. rejected must balance. Counted can never exceed registration.
Both parties watch the count in person, and every batch's size is reported as it's added to the running total.
The fairest hard question, and you don't have to take anyone's word for it, because the verification is adversarial, public, and reproducible. Observers from both parties have a legal right to watch every stage, so a skipped step means the other side's watchers, the people who most want to catch you, stayed silent. The hand-count audit is done in the open, and the chain-of-custody logs, reconciliation reports, and cast-vote records are released, so anyone can recompute the result from the paper. When violations do happen they get caught and prosecuted, which is how we know about them: a conservative think tank's own database lists more than 1,400 proven cases, every one of them tiny and every one of them found. And when fraud is alleged it is tested in court, under oath, with discovery; after 2020 more than 60 lawsuits, many before Republican- and Trump-appointed judges, found no systemic fraud. A claim that wins online but loses in every courtroom is the tell.
The full chain a vote-by-mail ballot travels in Los Angeles County, start to finish. Tap any stage to open it.
A ballot can be returned three ways: by mail, in an official drop box, or at a vote center. The moment it enters the system its custody is logged. Drop boxes are emptied on a fixed schedule by two-person retrieval teams, never one person alone. The ballots are placed in transport containers that are sealed with numbered, tamper-evident seals, and the seal numbers are written into a chain-of-custody log.
At the processing center, before any container is opened, staff confirm the seal number on the container matches the number recorded in the log when it was sealed. If a seal is broken or a number does not match, that container is quarantined and investigated. Every transfer of custody, from drop box to van to building to vault, is signed for. After the canvass, these logs become public records that anyone can inspect.
Every vote-by-mail ballot arrives inside an envelope the voter has signed. Before the envelope is ever opened, trained staff compare that signature against every signature the county has on file for that voter, which can include the registration card, prior ballot envelopes, and DMV records. The ballot inside stays sealed and secret during this step.
If the first reviewer believes the signature does not match, the envelope is not rejected on one person's judgment. It is escalated to additional, more senior reviewers, and a ballot is only set aside as a non-match after multiple staff agree. California law sets out what counts as a match and explicitly instructs officials to give the voter the benefit of the doubt on stylistic variation. A missing signature, or a signature that cannot be confirmed, both route the ballot into the cure process rather than straight to rejection.
What if someone just writes an "X"? Because the check measures consistency with your own record and not penmanship, an X only passes if your registration signature is also that mark. A regular signer whose envelope suddenly shows an X is a non-match, so it goes to cure, not to the count. Voters who genuinely cannot sign may use a witnessed mark, an accessibility accommodation with its own check. The failure mode here is being too strict and curing a real voter, never quietly accepting an unverifiable or forged one.
If a signature is missing or does not match, the law requires the county to contact the voter and give them a chance to fix it. This is called curing. The county reaches out by mail, and where possible by phone, email, and text, and sends the voter a signature verification statement to sign and return.
The voter has until roughly 28 days after Election Day to cure, which is one of the reasons the count runs for weeks. This step exists to protect real voters from losing their vote over a sloppy or changed signature. It is also why a low rejection rate is a sign of a working system, not a leaky one: many ballots that would otherwise be rejected are recovered because the voter confirmed them.
Only after a signature is verified is the envelope opened. The process is deliberately designed so that the ballot is separated from the envelope that carries the voter's name, which preserves the secret ballot. Once the ballot is detached from any identifying envelope, there is no way to tie a specific marked ballot back to a specific person.
Ballots are flattened and sorted into batches for scanning. Batches are tracked by count, so the number of ballots extracted is reconciled against the number of verified envelopes at every step.
Some ballots cannot be read by a scanner: they are torn, stained, hand-marked in an ambiguous way, or printed on a replacement sheet. These are handled by a bipartisan board. When a damaged ballot must be remade onto a clean card so a machine can read it, two staff from different parties do it together, the new card is marked as a duplicate with a serial number linking it to the original, and the original is preserved.
When voter intent is unclear, for example a circled name instead of a filled bubble, the board applies the state's published voter-intent standards. The original ballots are never destroyed, so any adjudication can be reviewed later.
Counting is done on voting systems that are tested and certified by the state, and that are not connected to the internet. Before an election, each system goes through public logic-and-accuracy testing, where a known set of test ballots is run through and the output is checked against the expected result. The test is repeated after the election.
Crucially, the machine is not the final authority. Every ballot is a piece of paper that the voter could verify, and the paper is retained. If a machine count and a hand count of the same paper ever disagreed, the paper would win. That is what makes the audit in step 8 meaningful.
Some people vote provisionally, for example if their eligibility needs to be checked, or if they registered and voted on the same day under California's Conditional Voter Registration. These ballots are set aside, not counted on the spot, and only added after staff research the voter's record to confirm they were eligible and did not vote elsewhere.
This research takes time, which is another reason the count grows for weeks, and it is also the honest explanation for why a few small precincts can show more ballots than their listed registration: a same-day registrant's ballot lands in a precinct whose closing registration number did not yet include them.
Before results are certified, the books must balance. Officials reconcile the number of ballots issued, returned, accepted, cured, rejected, and counted, precinct by precinct, and the totals have to add up. Counted ballots can never exceed the ballots returned, which can never exceed the people registered.
California then requires a post-election check of the paper. Traditionally this is a 1 percent manual tally, a public hand count of ballots from randomly selected precincts compared against the machine totals. The state is moving toward risk-limiting audits, which hand-count randomly drawn ballots until the result is statistically confirmed to a high confidence. Either way, humans count paper to confirm the machines, in public, before anything is official.
Honesty about limits is what separates analysis from propaganda. So, plainly:
The shift is real enough to flip a close lead, Caruso led the 2022 in-person count and lost. Treat election-night numbers as a first partial, not a result. That's a feature of counting late mail, not a bug.
Mail ballots are rejected more than in-person ones, and young and first-time voters are hit hardest (3%+). That's a genuine access problem worth fixing, it just isn't fraud.
Isolated individual fraud happens and gets prosecuted. What the data rules out is systematic, outcome-changing fraud, not the existence of any single bad actor.
The analysis centers on Los Angeles County, about a quarter of California, and its marquee contests. A strong, representative sample, but not a recount of all 58 counties.
What it does establish: the shifting count, the late mail, the partisan tilt, the turnout, the registration. Every pattern critics point to has a measured, documented, non-fraud explanation, consistent across a decade and across races nobody would rig. The burden of a fraud claim is to explain all of that better. None has.
Maybe the deepest version of the worry isn't any single fraud claim. It's this: if millions of people don't trust the count, the system has a problem worth solving, whatever the data shows. That's fair. Confidence isn't a luxury in a democracy; an outcome everyone accepts is the entire point. But look at what the changes demanded loudest would actually do. Ending mail voting and requiring strict ID or documentary proof of citizenship would remove millions of real, eligible voters to stop a fraud the evidence says is not changing outcomes. Start with who gets removed.
Californians voted by mail in 2024, about 80% of all votes. Forcing them back to a polling place falls hardest on the people for whom in-person voting is hardest: rural voters far from a site, people with disabilities, the elderly, caregivers, shift workers, and military and overseas voters.
voting-age Americans, about 1 in 10, lack a current government photo ID, and the gap is not even: 25% of Black Americans, 18% of those over 65, plus large shares of students and low-income voters. Nearly 500,000 have no car and live 10+ miles from an ID office open more than two days a week.
citizens can't quickly produce a passport or birth certificate to register. It lands especially on the 34% of women whose current legal name no longer matches the birth certificate they were issued, after a marriage.
Set that next to the fraud it claims to stop. Documented non-citizen voting in California: 6 ballots, caught and disclosed. "More votes than voters": a same-day-registration bookkeeping offset. The cure removes voters by the million; the disease is measured in single digits. When a remedy disenfranchises thousands of real citizens for every fake vote it could theoretically block, it is not protecting the election. It is shrinking it.
Stealing a count is hard, public, and catchable, as this whole page shows. Changing the rules is neither. A faction that fears losing does not need to fake votes; it can manufacture doubt, then use "restoring confidence" as cover to remove the other side's voters or seize control of how ballots are counted. The false claim and the proposed cure are the same move, and it is self-reinforcing: a lie lowers trust, low trust is cited to justify restrictions, the restrictions are challenged, and the challenge is spun as proof that critics "don't want integrity." Manufactured distrust becomes a ratchet that only tightens.
This is the textbook path of democratic backsliding. Elections keep happening, but the terms are quietly rigged: who may register, who can cast a ballot, who administers the count. The vote survives as theater while the choice is hollowed out, all under the banner of integrity. The tell is simple. A genuine reform expands verified participation, speed, and transparency. A suppressive one shrinks the electorate to solve a problem the evidence says barely exists, and it shrinks it unevenly, by race, age, and income. When the people most affected are not fraudsters but the elderly, students, married women, and voters without a car, the fraud was never the point.
If confidence really is the goal, there is a whole set of reforms that build it without taking the vote from a single eligible person. These are the ones worth fighting for.
Let officials process mail before Election Day, the way Florida does. It kills the days-long "red mirage" that breeds suspicion, at zero security cost. The single biggest win for confidence, and California and Pennsylvania could adopt it tomorrow.
Expand risk-limiting audits and publicize them, so the public hand-count that confirms the machines is something people can watch and verify, not just be told happened.
Livestreamed processing, public chain-of-custody logs, and released cast-vote records let anyone recompute the result. Transparency turns "trust me" into "check me."
When officials and media explain before Election Day that the lead will move and why, the shift becomes expected instead of shocking, and misinformation has less to feed on.
Every one of these adds confidence and costs no one their vote. That is the line between hardening an election and hollowing it out: the cure for a slow, opaque count is speed and transparency, and the cure for a false perception is accurate information. This page is the second half of that.
Every chart here is computed from Los Angeles County Registrar-Recorder/County Clerk Statement of Vote files, the certified, precinct-level public record, and the County's dated post-election ballot-count releases. The in-person vs. vote-by-mail split comes straight from each contest's POLLING PLACE and VBM PORTION precinct rows. Download the parsed data:
We flag deviations in both directions, 2024's dip in mail share, the shifting precinct definitions, and the handful of precincts that top 100% turnout. An honest read names what looks odd regardless of which narrative it might seem to help; every anomaly here has a documented, non-fraud explanation. The registration and turnout figures behind those checks are in the downloads above.
Sources
Note: 2026 figures are preliminary (pre-certification, ~July 10, 2026). Contests compared: Governor (2022 general; 2026 primary), President (2020, 2024 general), LA Mayor (2022, 2026 primary). 2016 SOV files are no longer hosted by the County.