Cited, verified accountability journalism.

Crosscheck
3 of 5 False

Fact-check · 5 claims

Are the 2026 Midterm Results Statistically Implausible?

1 Unproven1 Misleading3 False

Photo: Michael Barera / CC BY-SA 4.0 · source

The viral ‘0.15%’ fraud claim contains a textbook statistical error: wave elections aren’t independent coin flips. No auditor or court has found systematic fraud.

By ·

Background: The November 3, 2026 results

Democrats flipped both chambers of Congress in the November 3, 2026 midterms. As of November 12, Democrats had won approximately 255 House seats to Republicans’ 169, with NBC News projecting a final Democratic majority near 258 seats, a net gain of roughly 42 seats from the 213 Democrats held after the 2024 elections.[1] In the Senate, Democrats flipped at least eight Republican-held seats, including in Kansas, Texas, South Carolina, Mississippi, and Maine, reaching at least 55 seats.[2]

Several indicators had pointed to a large Democratic night for months. Trump’s approval rating sat at approximately 38% in Nate Silver’s tracking average as of November 9, with disapproval above 59%, a net rating of roughly −21 points.[3] Pre-election polling had pointed toward Democratic gains across battleground races. Exit surveys found nearly 90% of voters cited cost of living as a deciding factor, and roughly a third ranked democratic governance itself among the top two concerns, the second-highest share on record.[4]

On election night, Trump congratulated Democrats and indicated he would work with the incoming Congress. By November 9, he reversed course. His Truth Social post read:

“Does anybody really believe that this ‘Election’ was honest. How could the Polls be so wrong??? Some early voting ‘dumps’ were so large that it was impossible to win, no matter how many people went to vote. It’s all about the machines, which are far more important than the voter! President DJT”[5]

The post cites three things: polling accuracy, early-vote batch sizes, and voting machines. It names no county, contest, machine vendor, or specific tally discrepancy.[5]

Claim 1: The 2026 midterm Democratic wins are statistically implausible and therefore illegitimate/fraudulent

Claim 1
“The 2026 midterm Democratic wins are statistically implausible and therefore illegitimate/fraudulent” — as stated in and amplified by Trump’s November 9 Truth Social post
False

Frame-lock check: The post asks “how could the results be this large?” — but that is the wrong question for a fraud determination. The right question is: is there specific, documented evidence of fraud affecting enough ballots to change the outcome? Statistical analysis can flag a result as unusual; it cannot, standing alone, establish that votes were manipulated. Courts have held this standard consistently. The claim as circulated conflates improbability with illegitimacy, a logical gap the analysis never closes.

The claim rests on two legs: (a) a specific viral statistical analysis whose methodology is demonstrably flawed, and (b) the premise that statistical anomalies, even if real, constitute evidence of fraud sufficient to override certified results. Both legs fail.

The viral “0.15%” calculation — what’s wrong with it

The analysis was produced by a social media account called “Truth Is Freedom,” amplified by Gateway Pundit’s Jim Hoft, and then shared by Trump on Truth Social.[6] Its central finding: Republican Senate candidates received lower vote percentages than Trump’s 2024 showing in 29 of 30 comparable races. Using a claimed “historical” baseline of 70%— meaning 70% of the time, any given Republican Senate candidate will underperform the prior presidential standard — the analyst calculated the probability of this outcome under a binomial model as approximately 1 in 649, or 0.15%.[6]

There are two separate errors in this method.

Error 1: The assumption of independence. A binomial model treats each event as independent, like flipping a coin 30 times. Senate races in a national midterm are not independent. Every contest is shaped by the same national conditions: the incumbent president’s approval, the economic environment, and the dominant news cycle. When a wave forms, it pushes nearly all candidates in the same direction at the same time. Treating correlated events as independent dramatically understates the probability of near-uniform outcomes. Statisticians call this the ecological correlation problem.[7]

Error 2: The 70% baseline is calibrated to average election years, not wave years. The analyst chose 70% as the historical rate at which a Senate candidate underperforms the prior presidential number. That figure may be defensible for typical years. In a confirmed wave election (with a president at 38% approval, a contested war, and an affordability crisis cited by nearly nine in ten voters), the rate of Senate candidates underperforming the prior presidential standard climbs sharply. Under a 90% per-race probability (plausible for a wave year), the probability of 29 or more of 30 cases rises to 18.4%. Under 95%, it reaches 55.4%. These are not fraud-level anomalies; they are expected outcomes.[7]

Even the viral analysis’s own cited R-squared value of 0.87 — meaning the 2026 results correlate strongly with the 2024 Trump vote by geography — does not indicate manipulation. High geographic correlation between consecutive elections is a feature of partisan sorting, not a signature of machine interference.[6]

No credible electoral statistician or election administration expert has endorsed the 0.15% figure. The yournews.com article that publicized the finding noted expert methodological objections but chose not to resolve them: “The reported one-in-649 probability does not, by itself, demonstrate that votes were manipulated.”[6]

On polling accuracy: Polls did underestimate Democratic margins. The national House generic ballot polling average showed Democrats leading by 8.9 points going into Election Day, implying about 54.5% of the two-party House vote. Democrats won approximately a 13-point margin, about 56.5% of the two-party vote, a polling miss of roughly 2 percentage points on the two-party share.[8] Individual races ran wider: Texas Senate polls showed Paxton ahead by 1–7 points; he lost by double digits. These polling errors have an established, benign explanation: systematic under-sampling of high-turnout wave conditions. CFR reported that “polls collectively underestimated support for Democratic congressional candidates by a larger margin than they underestimated support for Trump in 2016 or 2024.”[7] The same pattern appeared in 2006 and 2018. DCReport’s post-election analysis called 2026 polling “its most reassuring performance since 2008,” noting polls had pointed in the right direction across all major races.[9]

On “early voting dumps”: Large batches of early votes appearing at a single update are routine. Early ballots cast over days or weeks are frequently uploaded in one consolidated report on election night once counting systems are ready. Election officials in Tarrant County, Texas, for example, uploaded 559,784 early votes as a single update when tabulation opened, not because votes were added improperly, but because ballot-reporting timestamps reflect when data becomes visible, not when ballots were cast.[10] This practice predates electronic voting. It is documented in state election law. Confusion about it has been weaponized to suggest fraud in multiple elections dating back to 2020.[10]

No evidence of machine manipulation: Trump’s post says voting machines “are far more important than the voter,” implying they determined the outcome. Trump named no machine vendor, no specific county, and no mechanism of manipulation. No election official, federal agency, or accredited post-election audit has found evidence of machine-based vote switching in the 2026 election. Machines used in U.S. elections produce paper ballot records; discrepancies between paper and machine totals would appear in routine audits.[11]

Strongest version of the opposing argument

The wave’s magnitude was historically large. The final House popular-vote margin of roughly 13 points for Democrats represents only the seventh time since WWII that a midterm featured a swing exceeding 10 points. Some polling misses in individual races (Texas Senate) were very large. These facts are worth scrutinizing, and routine audits in multiple states are ongoing. The proper response to statistical questions worth investigating is targeted, transparent audit, not a blanket fraud allegation covering all 435 House and 33 Senate contests simultaneously, without a single named precinct, machine, or ballot discrepancy.

Scoring: 2 out of 5. This counter-argument raises legitimate monitoring points but does not directly rebut the statistical error at the core of the fraud claim. It is a reason to audit specific races through proper channels, not a reason to call the election fraudulent. Verdict holds: False.

Claim 2: Officials should audit the vote and not certify the 2026 results until verified

Claim 2 (ticked)
“Officials should audit the vote and not certify the 2026 results until verified”
Misleading

This claim conflates two distinct processes: post-election audit and election certification. It implies that withholding certification until an audit is complete is a legally available and justified option. Neither the law nor the courts support that framing.

Audits and certification are parallel, not sequential. Every state with paper ballots conducts a post-election audit of some kind: risk-limiting audits, hand-count checks of a sample of ballots, or machine-verification procedures. These run on a timetable set by state law and do not delay certification. Certification is governed by statutory deadlines; most states certify within 30–45 days of Election Day regardless of ongoing audit activity.[11]

Legal challenges to withhold certification have not succeeded. Two certification challenges are pending as of November 13. In Texas, the Potter County Republican Party filed a federal lawsuit on October 27, 2026, seeking to require a statewide hand count of all federal ballots before certification, alleging Chinese-manufactured components in voting machines could allow manipulation. The Potter County Elections Administrator filed a motion to dismiss on November 9; the case remains in early procedural stages.[12] In Alaska, the Trump Justice Department filed pre-election suit to block certification of a contested Senate result. Neither has obtained a court order blocking certification.[13]

Courts have consistently rejected certification holds based solely on statistical claims. Following the 2020 presidential election, more than 60 federal and state court challenges, including several that relied on statistical anomaly arguments, were dismissed. The legal standard for staying or overturning a certified election is specific, documented evidence of fraud or error affecting enough ballots to change the outcome, not a statistical inference that results are improbable.[14]

The claim that officials should not certify until an audit is complete is thus misleading on two levels: it implies audits aren’t already happening (they are), and it implies a legal path exists to withhold certification indefinitely on statistical grounds (courts have consistently rejected this).

Historical wave elections: comparison set

Compared against: post-World War II midterm wave elections by House popular-vote margin and seat change for the opposition party.

The claim that the 2026 results are unprecedented or historically implausible is not supported by the historical record of midterm elections.

Year Context Opposition party gain (House seats) President approval (approx.)
2026 Iran war, affordability crisis, Trump approval 38% ~+42 (D) 38%
2010 ACA backlash, economic anxiety +64 (R) 44%
1974 Watergate, Nixon resignation +49 (D) 24%
2018 First Trump midterm +41 (D) 42%
1958 Eisenhower recession +48 (D) 52%
2006 Iraq War fatigue +31 (D) 37%
1994 Clinton health care failure +54 (R) 46%

The 2026 House popular-vote margin for Democrats was approximately 13 points, which the Council on Foreign Relations identifies as ranking second among post-WWII midterm elections in popular-vote margin, trailing only the 1974 Watergate result.[7] The Democrats’ 2026 Senate gains, including flipping states like Kansas (first Democratic Senate win since 1932) and Mississippi, are historic in geographic scope. But scope does not equal fraud; 1974 produced comparable House shifts and was never seriously alleged to be fraudulent.

The 90% rule of presidential approval also applies: in elections where the sitting president’s approval falls below 40%, the opposition party’s gains historically exceed 35 seats. Trump entered Election Day at 38%. A gain of 42 seats is statistically consistent with historical norms at that approval level.[16]

What the post implies

Implied claim
Implied: Voting machines controlled or altered vote totals to produce the Democratic result.
Unproven

Trump’s post says “it’s all about the machines.” No specific machine vendor, county, or mechanism is named. No audit has found machine-paper ballot discrepancies. This claim is unproven and unsubstantiated; the burden of specificity lies with the claimant.[11]

Implied claim
Implied: The polling miss is evidence of fraud or pre-planned outcome manipulation.
False

Polls systematically underestimated Democrats by approximately 1.8 percentage points on the national House generic ballot, within the normal range of polling error, and by wider margins in individual races. Under-sampling of high-turnout wave conditions is a recognized methodological challenge with well-documented historical precedent in 2006 and 2018. Polling error is not evidence of election manipulation.[8][9]

Implied claim
Implied: A result this large is historically unprecedented and therefore cannot be legitimate.
False

As the historical comparison above shows, the 2026 House Democratic gains are large but precedented. The 1974 and 2010 midterms produced comparable or larger seat swings. The 2026 result ranks second in post-WWII wave size, not first.[7]

Unrated implication: The post is widely read as a call for Trump’s supporters or sympathetic officials to resist certifying or accepting the results. Whether officials should comply with calls to delay or resist certification is a legal and political question that no fact-check can resolve with a verdict. It is noted here as a contextual implication the post carries.

What this finding changes

The False verdict on the statistical fraud claim matters beyond this individual post for several reasons.

The specific “0.15% probability” calculation is currently circulating widely and informing two pending court cases and one proposed congressional audit. If courts and election officials treat the number as credible without examining its methodology, procedural harm can result even from a mathematically invalid claim. This fact-check documents the specific statistical errors so that decision-makers and the public have a named basis for evaluating the claim’s validity.

The pattern also establishes a template. Trump applied the same logic after his 2020 loss (large polling misses, early-vote reporting patterns, and machine allegations without specific precinct-level evidence), and 60+ courts rejected those claims. The 2026 formulation uses the same three legs on a different election. The legal and institutional response will draw on the same precedents.

The Misleading verdict on the certification claim is consequential in a different direction. Routine post-election audits in key states are legitimate and expected; the fact that certification can proceed alongside them does not mean audits lack value. If specific audits find discrepancies, those findings should be pursued through proper legal channels. The misleading element is specifically the suggestion that certification should be withheld on the basis of the flawed statistical argument above.

Parties affected

Party Effect Why
Democratic Party Mixed Won both chambers but faces an organized delegitimization campaign that could affect institutional cooperation and transition processes.
Republican Party Mixed Even conservative commentators (Megyn Kelly: “just stop”) and former Trump allies are distancing from the fraud claims after the decisive loss, creating intra-party tension.
State election officials Costs Fielding legal challenges, court orders, and public pressure to delay certification based on a claim their own auditors have not flagged. Texas and Alaska officials face active suits.
Voters Costs Widespread fraud narratives erode confidence in electoral outcomes; AP-NORC found under half of voters “very confident” ballots would be counted accurately before the election.
Federal judiciary Mixed Absorbs post-election litigation; courts have a consistent track record of dismissing statistical-anomaly challenges (60+ cases post-2020), but each case consumes resources.
Analysis: Rows above infer effects from documented reporting; the intra-party Republican split row reflects November 9–12 coverage and may shift.

What to watch

Upcoming signposts

  • Late November 2026: Texas certification deadline approaches. The Potter County lawsuit (filed Oct. 27) seeks to require a statewide hand count before certification. The state’s motion to dismiss is before the court; a ruling is expected before the statutory certification window closes.
  • December 2026: Alaska Senate certification challenge from the DOJ. The first attempt by a presidential administration to block a congressional election certification; outcome will set a significant precedent for executive interference in state election processes.
  • January 3, 2027: New Congress convenes. Democrats take control of both chambers. If certification challenges in any state succeed in delaying a seat’s certification, that seat may be vacant for the opening of the 120th Congress, affecting the official majority tally.
  • Ongoing: State-level post-election audits in Texas, Georgia, and Michigan. If any audit surfaces a material discrepancy between paper ballots and machine totals, that would constitute the type of specific, documentary evidence the courts require — and would warrant separate coverage.

What happened

Democrats won both the House and Senate in the November 3, 2026 midterm elections. It was a large Democratic victory. Democrats gained about 42 seats in the House and flipped at least eight Senate seats, including in Texas, Kansas, Mississippi, and Maine.

On election night, President Trump said he accepted the results. Six days later, he changed course. On November 9, he posted on Truth Social:

“Does anybody really believe that this ‘Election’ was honest. How could the Polls be so wrong??? Some early voting ‘dumps’ were so large that it was impossible to win, no matter how many people went to vote. It’s all about the machines, which are far more important than the voter!”

Trump did not name any county, any specific voting machine, or any single ballot discrepancy. He provided no evidence.

Claim 1: The results are statistically impossible and the election was fraudulent — Verdict: False

The fraud claim depends on a viral statistical analysis. A social media account called “Truth Is Freedom” ran the numbers, and Gateway Pundit’s Jim Hoft amplified them before Trump shared the claim on Truth Social.

The analysis found that in 29 out of 30 Senate races, Republican candidates got a lower share of the vote than Trump did in 2024. The analyst said the odds of that happening by chance were only 1 in 649, or 0.15%. The claim was that this proved the election was rigged.

That math has two specific errors.

Error 1: Treating races as independent coin flips. The calculation uses a binomial model — the same math used to calculate the odds of flipping heads 29 times in a row. That math only works if each event is independent, meaning one flip does not affect the next. Senate races are not independent. They are all shaped by the same national conditions: the president’s approval rating, gas prices, the Iran war, and the economy. When voters are angry at the president, nearly all Senate candidates from his party suffer together. This is called ecological correlation. Treating correlated events as independent coin flips produces a misleadingly small probability number.

Error 2: The 70% baseline is wrong for a wave election. The analyst assumed that in any given race, there is a 70% chance a Senate Republican would underperform Trump. That figure may apply to normal years. But this was not a normal year. Trump’s approval was at 38%. Nearly nine in ten voters cited cost of living as a deciding factor. In a confirmed wave election, the probability that any given Republican underperforms the presidential number is much closer to 90–95%. Under a 90% rate, the probability of 29 of 30 underperforming is 18.4%. Under 95%, it rises to 55.4%. These are not suspicious numbers. They are expected outcomes in a wave year.

On the polling miss

Polls did underestimate Democrats. The national generic ballot polling average predicted Democrats would win the House popular vote by about 8.9 points. Democrats actually won it by about 13 points. That is a miss of roughly 4 points on the margin. Individual races missed wider: in Texas, pre-election polls showed Republican Ken Paxton slightly ahead; he lost by double digits.

Polling misses this size are not evidence of fraud. Under-sampling of high-turnout wave conditions is a recognized methodological challenge. The same pattern appeared in 2006 and 2018. Post-election analysts called this the most accurate polling environment since 2008 for directional predictions — meaning polls told you who would win, even if they missed the margin.

On early voting “dumps”

Trump said some early voting batches were too large to overcome. Large batches of early votes appearing at once are normal. Early ballots cast over several days are often uploaded in a single report on election night when the counting system opens. In Tarrant County, Texas, for example, 559,784 early ballots appeared in one update. Those ballots had been cast over days. The timestamp on a vote report is when the data was uploaded to the public, not when the ballot was cast. This practice is documented in state law and predates electronic voting.

On voting machines

Trump’s post says “it’s all about the machines.” He named no specific vendor, county, or mechanism. No post-election audit has found a discrepancy between paper ballots and machine totals in the 2026 election. U.S. voting machines that use paper records can be checked against those records. No such check has found evidence of manipulation.

Claim 2: Officials should not certify the results until they are verified — Verdict: Misleading

This claim mixes two different things: post-election audits and election certification.

Audits already happen. Every state with paper ballots conducts some form of post-election check. These are built into state law and run on a fixed schedule. They do not delay certification.

Certification has its own statutory deadline, usually 30 to 45 days after Election Day. It proceeds on its own timeline alongside audits.

Courts have rejected the argument that certification can be legally withheld because of a statistical claim. After the 2020 election, more than 60 federal and state court cases made similar arguments. All were dismissed. The legal standard for blocking a certified election requires specific documentary evidence — a named ballot discrepancy, a named machine error, something checkable. A probability calculation does not meet that standard.

As of November 13, two certification challenges are pending. In Texas, a Republican group filed suit to require a statewide hand count before certification; the motion to dismiss is before the court. In Alaska, the Trump Justice Department filed a pre-election challenge. Neither has obtained a court order blocking certification.

Historical context: Wave elections this large have happened before

Democrats won about 56% of the two-party House vote in 2026. Columbia University political scientist Andrew Gelman’s analysis ranks this as the second-largest partisan wave in post-World War II House elections. The largest was 1974, the year after Watergate. Republicans gained 64 seats in 2010. Democrats gained 49 in 1974, 48 in 1958, and 41 in 2018. The 2026 gain of about 42 seats is large but not historically unprecedented.

Trump entered Election Day with a 38% approval rating. Historical data shows that when a president’s approval falls below 40%, the opposing party typically gains more than 35 House seats. A gain of 42 seats is consistent with what that data would predict.

What to watch

The Texas certification lawsuit is expected to be ruled on before the state’s certification deadline in late November. The Alaska DOJ challenge is also pending. If either court issues an order blocking certification, that would be the first time a congressional election certification has been successfully challenged in the modern era. The new Congress convenes January 3, 2027. Any seat whose certification is delayed past that date would be vacant at the start of the session.

Sources

  1. U.S. House Elections 2026 Live Results: Democrats Win Control
  2. Where Democrats Won Big in the 2026 Midterms: Inside the Most Decisive Election Results
  3. Donald Trump Approval Rating, November 2026
  4. What Drove the Historic Blue Wave in the 2026 Midterms
  5. Trump Truth Social post, November 9, 2026
  6. New Analysis Claims 2026 Midterm Voting Pattern Had Just 0.15% Probability
  7. The 2026 Election by the Numbers
  8. After an Extraordinary Wave, What Comes Next?
  9. 2026 Midterm Polls Got the Big Picture Right
  10. Trump Questions the US Midterms: Why Do Early Votes Show Up in Big Batches?
  11. Midterms 2026: Voting Problems Experts Feared Largely Haven’t Materialized
  12. Texas Election Certification Challenge
  13. Alaska U.S. Senate Election Certification Challenge (DOJ)
  14. Republican Reactions to Donald Trump’s Claims of 2020 Election Fraud
  15. The Midterm Election Blue Wave Was Historic. Have We Seen One Like It Before?
  16. Midterm Election Patterns: Why the President’s Party Usually Loses
  17. A Look at Wave Elections Since 1918 and the Thresholds Needed for a Wave Election in 2026

Get the biweekly digest

New Crosscheck fact-checks, every other Monday. Sourced, never spun.