Data Saves Americapowered by dad and his robots.

Data & sources · Data as of November 19, 2026

The receipts

Everything here comes from public records: Federal Election Commission filings, official election results and the Bureau of Labor Statistics. This page says how the pieces were joined, what was left out, and where the numbers are weakest.

Sources

How it was built, and caveats

  • A race is one general election for one House or Senate seat. Each keeps its top two finishers. The “better-funded side” is the one with more money behind it; races where both sides had exactly the same money, and races not yet called, are left out of the win rates.
  • Correlation is not cause. Money flows to candidates who are already likely to win, so the share of races won by the bigger spender overstates what spending buys. The research page covers how economists try to separate the two.
  • Prices for 2026 use the average of the 10 months published so far.
  • Campaign spending is each candidate's total disbursements for the two-year cycle, net of transfers between the candidate's own committees. It covers the primary and the general election together; the FEC does not split them. Senate figures cover the two-year cycle, not the full six-year term.
  • Outside spending is independent expenditures reported to the FEC: money spent by super PACs, party committees and other groups to support or oppose a named candidate without coordinating with a campaign. It does not include issue ads that avoid naming a vote, party coordinated spending, or most nonprofit 'dark money' that is never reported to the FEC.
  • Outside money counted for a race is general-election, runoff and special-election spending only (primary spending is left out). Money 'behind' a candidate is spending that supports that candidate plus spending that opposes the other top-two finisher.
  • Single independent-expenditure lines above $50,000,000 are dropped as hoax or typo filings (31 lines, listed in this file). Amended filings replace the reports they amend.
  • Races for 2000 and 2002 use the winner and vote-share fields in the FEC's candidate file, which lists only candidates who registered with the FEC (those who raised or spent $5,000 or more). In those two years a race with only a winner listed means the opponent filed nothing or there was no opponent. From 2004 on, results come from official results files that list every candidate on the ballot.
  • A candidate with no FEC record raised and spent under the $5,000 registration threshold; their spending is counted as zero in the win-rate figures and shown as 'no FEC report'.
  • The FEC's candidate file leaves out or under-reports some real campaigns: committees later converted into PACs, and campaign committees not linked to a candidate id. For a top-two finisher with less than $50,000 on file, the pipeline looks up the campaign committee's own totals (FEC committee summary file, 2008 on) and uses them; those candidates are marked with the committee's name.
  • Donald Trump's 2024 campaign committee was converted to a leadership PAC (Never Surrender, Inc.) in 2025 and dropped out of the FEC's candidate summary file; its 2023–24 totals are read from the FEC committee record.
  • House special elections are left out. Senate special elections held on the regular election day are included and marked.
  • 2026 figures are preliminary. Results are unofficial (snapshot of 2026-11-13 (unofficial results; counting ongoing in several states)); only Senate races and the House seats that flipped or were closely contested are included. Campaign reports mostly run through 09/30/2026; post-election reports are due December 3, 2026. Outside spending comes from 24- and 48-hour notices.

Coverage by election year

YearHouse racesSenate racesFinishers with no FEC recordMatched by nameFrom committee recordsOutside spending keptNot matched to a race
200043534000——
200243434000——
20044353478150——
2006435337580——
2008435358842——
20104353753101$387M$14.9M
2012435337090$1.31B$15.9M
2014435368331$781M$26.6M
201643534111134$1.72B$9.7M
20184353559161$1.32B$59.0M
20204353556112$3.37B$40.5M
20224353648233$2.27B$36.5M
20244333552128$4.46B$21.9M
20268135001$4.13B$428M

“No FEC record” counts top-two finishers who never registered a campaign committee, nearly all of them token opponents in safe seats. “Matched by name” are finishers whose FEC id in the results file was missing or wrong. “Not matched to a race” is general-election outside spending in House and Senate contests that named a candidate who was not a top-two finisher (in 2026, mostly House races not yet in this dataset).

Filings thrown out

Anyone can file an independent-expenditure report with the FEC, and a few people file hoaxes claiming billions. Any single line above $50 million was dropped. The largest:

YearReported spenderNamed candidateClaimed amountFEC filing
2022COMMITTEE 300Akhlaghy, Nader$10.00B1577711
2022THE COMMITTEE OF 300Akhlaghy, Nader$10.00B1606351
2024THE COURT OF DIVINE JUSTICEAkhlaghy, Nader$9.98B1778759
2024THE COMMITTEE OF 300Akhlaghy, Nader$9.78B1766664
2022NADER AKHLAGHYAkhlaghy, Nader$9.70B1577002
2024THE COMMITTEE OF 300Akhlaghy, Nader$9.00B1766323
2026TANKINGBettis, Shawn$9.00B1957556
2026Warren BuffetBettis, Shawn$9.00B1957531
2026SHAWN BETTISBettis, Shawn$8.00B1957562
2024THE COMMITTEE OF 300Akhlaghy, Nader$7.96B1798067
2024Republican Emo GirlBounnharat, Amandra$6.35B1763978
2022NADER AKHLAGHYAkhlaghy, Nader$1.60B1553374

Largest outside spenders by year

2026

  1. SLF PAC: $351M
  2. HMP: $295M
  3. WinSenate: $282M
  4. No Going Back PAC Inc.: $274M
  5. CLF: $223M
  6. Texas PAC: $182M
  7. Americans for Prosperity Action, Inc. (AFP Action) DBA CVA Action and DBA LIBRE Action: $107M
  8. Defend American Jobs: $106M

2024

  1. FF PAC: $497M
  2. Make America Great Again Inc.: $377M
  3. WinSenate: $311M
  4. Congressional Leadership Fund: $217M
  5. Senate Leadership Fund: $211M
  6. HMP: $196M
  7. America PAC: $172M
  8. Americans for Prosperity Action, Inc. (AFP Action) DBA CVA Action and DBA LIBRE Action: $152M

2022

  1. Senate Leadership Fund: $246M
  2. Congressional Leadership Fund: $227M
  3. SMP: $160M
  4. House Majority PAC: $145M
  5. DCCC: $96.4M
  6. NRCC: $87.1M
  7. Club For Growth Action: $70.7M
  8. Georgia Honor: $60.6M

2020

  1. Senate Leadership Fund: $294M
  2. SMP: $230M
  3. Congressional Leadership Fund: $143M
  4. House Majority PAC: $139M
  5. Priorities USA Action: $139M
  6. America First Action, Inc.: $134M
  7. NRSC: $121M
  8. DCCC: $104M

Independent expenditures reported to the FEC, all federal races, primaries included. Names as filed.

The written pages

  • The rules. Every event was checked on 2026-11-19 against a source fetched that day. Statutes before 2014: enactment dates, citations and key provisions were read from the text of the Statutes at Large page images on govinfo.gov. Supreme Court cases: decision dates, holdings and justice lineups were read from the syllabus or opinion text on Cornell LII (and, for NRSC v. FEC, the supremecourt.gov docket). FEC rules, limits and agency history: fec.gov pages and FEC advisory opinion PDFs. U.S. Code sections: Cornell LII. FEC quorum: CRS Report R45160, an FEC court filing and a Nov. 17, 2026 news report. Vote counts were tallied by hand from the lineups and are explained in each event's notes. 'confidence: verified' means the date, citation, vote and holding were confirmed this way; context and 'why it happened' sentences drawn from standard history rather than a fetched source are flagged in notes. Sentences in 'implication' are plain-language interpretation.
    Known gaps (12)
    • Veto overrides for Smith-Connally (1943) and Taft-Hartley (1947) are commonly reported but were not confirmed from a fetched primary source and are left out of the summaries.
    • The year the presidential checkoff rose from $1 to $3 (commonly given as 1993) was not confirmed.
    • Reporter citations for Carey v. FEC and Bluman v. FEC, and the exact day of the 2018 district court ruling in CREW v. FEC, were not confirmed; FEC summaries were used instead.
    • The en banc vote count in the Sixth Circuit's 2024 NRSC decision was not confirmed.
    • Maine super PAC case (Dinner Table Action v. Schneider): status is from advocacy-group case pages and search summaries; the First Circuit docket was not checked. No ruling was found as of Nov. 19, 2026.
    • End Citizens United PAC v. FEC (D.C. Cir. en banc): no decision was found, but the court's docket was not checked directly.
    • FEC quorum: nothing was found on Senate action after the Nov. 17, 2026 committee hearing. Re-check before publication; confirmation of both nominees would restore a four-member quorum.
    • Absence of new federal legislation in 2024-2026 rests on the FEC legislation page and web searches, not a complete review of public laws.
    • Hawaii and Montana 2026 measures are sourced to news reports only; bill numbers, final vote totals and any litigation were not confirmed.
    • Not covered: Executive Order 14215 (Feb. 2025) on White House supervision of independent agencies and its application to the FEC; IRS changes to donor reporting (Schedule B) for 501(c)(4) groups; Doe v. Reed (2010); FEC v. National Conservative PAC (1985); EMILY's List v. FEC (2009); United States v. Classic (1941), which undercut Newberry; state-level limits generally.
    • Whether national banks may make independent expenditures after Citizens United was not confirmed.
    • The legality of President Trump's removal of FEC Commissioner Weintraub has not been decided by any court; she has said she regards it as invalid but, per a Nov. 2026 news report, has not sued.
  • Nonprofits & donor privacy. Five research passes run on 2026-11-19 (entity rules; history to 1963; courts and legislation since 1907; IRS and government custody of data; numbers). Each fact was checked against a page fetched that day with WebFetch or curl: U.S. Code and CFR (law.cornell.edu, ecfr.gov), irs.gov, fec.gov, Federal Register, govinfo.gov, Supreme Court opinions at Cornell LII and the Court's own docket pages, CourtListener/RECAP court filings, TIGTA, the Church Committee report (S. Rep. 94-755, Book III), Senate and House roll-call pages, CRS reports, IRS Data Book spreadsheets, ProPublica Nonprofit Explorer (Form 990 extracts), Giving USA, the DAF Research Collaborative, the Brennan Center, and news outlets for events. url_checks records what each URL returned. Records marked 'verified' may still carry specific sub-claims that could not be confirmed; those are spelled out in each record's notes and should be read before publishing. The lead compiler then re-fetched a sample of the highest-risk and most recent claims (see lead_spot_check fields and meta.lead_spot_checks) and corrected one date (the FEC's Carey statement is October 6, 2011, not October 5).
    Known gaps (13)
    • Nixon enemies list: the account that IRS Commissioner Johnnie Walters and Treasury Secretary George Shultz declined to act on the list is marked unverified (obituaries and the Joint Committee's December 1973 report could not be fetched). Sources also differ on the size of the list handed to Walters.
    • Special Service Staff: the Church Committee gives 11,458 files and rounds to about 8,000 individuals and 3,000 organizations; the widely cited split of 8,585 and 2,873 sums to 11,458 but was not confirmed from the Joint Committee staff report.
    • Kennedy-era Ideological Organizations Project: the Reuther memorandum's date and role were not confirmed; the Church Committee dates the IRS directive to November 30, 1961 and does not mention the memo. Christian Echoes v. United States (10th Cir. 1972) is unverified.
    • Johnson Amendment: Johnson's motive is not in the Congressional Record; the Dougherty / Facts Forum account is the usual explanation, not an established fact.
    • IRS targeting: the '150,000 comments' figure on the 2013 proposed 501(c)(4) rule, the text of the Justice Department's October 23, 2015 and September 8, 2017 letters, and the Justice Department's own October 26, 2017 settlement release were not fetched (contents taken from Roll Call, the Ways and Means Committee, NPR and the Ohio court's order). The circulating '248 conservative vs 29 liberal' figure could not be sourced and is omitted.
    • Littlejohn: plea date and statute not confirmed from a fetched primary page; the 405,427-taxpayer figure is from news reports of an IRS letter to the House Judiciary Committee, not the letter itself; the D.C. Circuit opinion of July 17, 2026 was not read.
    • 2026 developments resting on thin sourcing: the reported Treasury review of named left-leaning nonprofits (only the senators' September 4, 2026 letter was verified); the D.C. Attorney General's Leo and Arabella inquiries (one outlet); proposed IRS rules on discrimination by schools (described from the Harvard Crimson, not located in the Federal Register); whether Treasury has issued the promised Johnson Amendment guidance (none found, which is not proof).
    • State laws: the count of 22 Personal Privacy Protection Act states (September 2026) comes from the advocacy group that promotes the model bill; no neutral tracker could be loaded and the pre-2022 roster rests on search results. November 2026 Michigan Proposal 2 and Montana I-194 results are unverified (possibly uncertified totals).
    • Proposition 8: the restaurant manager's name and employer, the exact launch date of the eightmaps site, and the district court's reasoning in ProtectMarriage.com v. Bowen were not read from primary pages. DISCLOSE Act: the terms of the 2010 NRA carve-out rest on search-result summaries.
    • Early history: when Paine's authorship of Common Sense became known; the Federalist's first-publication day (the Library of Congress page gives only the newspaper); what Congress did in 1836 about abolitionist mail (check 5 Stat. 80, 87); the '175,000 pieces' figure for the 1835 mailing; the size of Arthur Tappan's Oberlin gift (sources give $10,000, $12,000 and $17,000); every specific bounty figure for Arthur Tappan; the 7-1 count in the Amistad case (only Justice Baldwin's dissent is recorded); Virginia's 1956 disclosure statutes.
    • Entity rules: whether candidate committees, party committees and hybrid PACs may give to super PACs was not covered by any source fetched; left/right labels on example organizations are conventional descriptions (only legal form was confirmed); the AFL-CIO's reported suit over the 2026 LM-2 rule was not checked.
    • Numbers: OpenSecrets was read only through Internet Archive captures. OpenSecrets' data page ($170.2 million) and the Brennan Center ($43 million) disagree on 2024 dark money reported directly to the FEC, apparently because of different definitions of a non-disclosing group. The 2024 total-cost figure ($14.8 billion) was read from a table in inflation-adjusted view. Not confirmed: the roughly $1.5 trillion foundation-assets figure (IRS SOI gives $1.33 trillion for tax year 2022), any Democracy Alliance total, ActBlue and WinRed totals, total union political spending from LM-2 reports, and Giving USA's revised 2024 total (about $584 billion is our inference from the stated 5.7% growth).
    • No candidate-level or state-level campaign money is covered; this file is about organizations outside candidate campaigns, plus the committees needed for comparison.
  • What research says. Four research passes on 2026-11-19 (academic literature; historical benchmarks; 2026 primaries and themes; OpenSecrets, Michigan and 2026 Senate generals). Each item was checked that day against a fetched source: Crossref/OpenAlex/publisher or RePEc records for papers; FEC press releases and PDF tables; OpenSecrets elections-hub pages read in a browser; Brookings Vital Statistics CSVs; Brennan Center analyses; news reports (AP via local outlets, NBC, Bridge Michigan, Roll Call, NOTUS, Fortune, Newsweek and others); Wikipedia as a pointer for results and FEC tables; and the pipeline's own local copies of FEC bulk files (weball26 candidate summary; independent_expenditure_2026) for 2026 candidate and outside-spending figures. Nothing was taken from memory. Subagent passes were spot-checked (Crossref records for Levitt 1994 and Kalla-Broockman 2018; Wikipedia for the Maine and Michigan primaries; UDP, MD-5, IL-7 and Ohio totals re-derived from the FEC file). All text is paraphrased.
    Known gaps (34)
    • WebSearch budget for the session (200 calls shared by all research passes) ran out partway through; later verification used direct fetches of known URLs, Wikipedia, the Brennan Center, FEC documents and the pipeline's local FEC bulk files.
    • OpenSecrets: only the elections-hub pages (reelection rates, cost of election, election trends 2022, incumbent advantage 2024, stats at a glance 2024) could be read, in a browser. Its 'Winning vs. Spending' page no longer exists; its outside-spending, race-profile and self-funder pages hit a bot-verification wall or failed to load, and web.archive.org was unreachable. So: no OpenSecrets top-spender-wins percentage for 2002, 2016 or 2024 (and only a House figure for 2018; the 2012 House and 2022 figures were seen only in search excerpts); no OpenSecrets outside-spending-by-cycle table (the FEC independent-expenditure series is the verified substitute); no OpenSecrets race totals for the 2024 Ohio, Pennsylvania and Montana Senate races (candidate-committee totals are from FEC tables; race totals for Ohio and Montana are Wikipedia-only; Pennsylvania outside spending not found); no OpenSecrets self-funder win-rate series.
    • 2026 outside-spending-by-side figures labelled 'FEC IE file' were computed here from the FEC bulk file and are approximate (see meta.ie_method). Year-end FEC reports are due January 2027; all 2026 totals are preliminary and candidate coverage dates vary (mostly 10/14/2026).
    • FEC API (DEMO_KEY) returned HTTP 429 all day, so per-report (pre-primary) candidate figures come from Wikipedia's FEC tables or news reports rather than FEC directly.
    • Minnesota general: Michele Tafoya's campaign total could not be confirmed (FEC summary file shows an implausible $46.6K).
    • Alaska Senate: winner (Peltola) taken from Wikipedia on 2026-11-19; the repository's results file still had it uncalled.
    • Literature: 45 of 49 entries are verified, mostly at abstract level (Crossref/OpenAlex/publisher/RePEc); full text was read for only three. Not fully verified: Stratmann 2005 (conclusions paywalled), Branham-Soroka-Wlezien 2017 (no abstract; headline percentages not confirmed), Steen 2006 (no win rates found), Hall 2014 working paper (known only from the author's blog summary). Levitt's per-$100,000 effect size, Gerber et al. 2011's decay timeline and Gilens & Page's coefficients were not confirmed and are deliberately left out of the findings.
    • Benchmarks pass: A3: no fetched source gives the OpenSecrets top-spender win rate for 2016 or 2024, or the Senate figure for 2018. The 2012 House figure (93.8%) and 2022 figures (over 94% House, 88% Senate) were seen only in search-result excerpts; the pages themselves would not load (UNT web archive returned empty, ZNetwork and Bloomberg Government returned 403).
    • Benchmarks pass: A3 conflicts to resolve against OpenSecrets' own table: 2004 Senate (88% CRP vs about 94% AP), 2006 (93%/67% preliminary vs 94%/73% later), 2008 Senate (94% preliminary vs 86% later), 2020 (89.2%/71.9% on Nov 24 vs 90.4%/73% on Nov 6). Election-week numbers are consistently different from later restatements.
    • Benchmarks pass: A4: OpenSecrets outside-spending totals for 2010, 2012, 2014 and 2016, and a final 2024 total, were not found in any fetched article. 2020 ($3.3B) and 2022 final ($2.3B) were seen only in search excerpts. The FEC independent-expenditure series is the verified substitute.
    • Benchmarks pass: A2: outside-group spending by side for the 2020 presidential race was not verified. The 2016 outside figures are an Oct 31, 2016 snapshot, not final.
    • Benchmarks pass: A2: FEC Presidential Table 2 (through Dec 31, 2024) prints Trump disbursements of $255.2M and cash on hand of $119.7M against receipts of $474.9M; these do not reconcile, so Trump's 2024 campaign spending is unverified. The FEC API (DEMO_KEY) was rate-limited (HTTP 429) for the whole session, so candidate totals could not be cross-checked there.
    • Benchmarks pass: A5: outside spending by side and race totals for Pennsylvania 2024 were not found; the Ohio $483.4M total, the $40M Fairshake figure and the Montana $315M / $50M / $487-per-voter figures were read only on Wikipedia, with underlying outlets not confirmed. No AdImpact or OpenSecrets race page was fetched.
    • Benchmarks pass: Part B: Linda McMahon's 2012 self-funding total is unverified (Wikipedia shows only $15.7M, apparently partial). Steve Forbes's 1996 self-funding amount, J.B. Pritzker's 2022 self-funding amount, Right to Rise's final spending, Michael Huffington's personal share of his $28M, and Rick Scott's final 2010 personal total (reports range $60M-$78M) are not pinned down. Vote margins for O'Rourke 2018, Gideon 2020, Fiorina 2010 and outcomes for Allred/Hovde 2024 and the four extra 2020 Senate challengers were not separately fetched.
    • Benchmarks pass: Part B: no cycle-level OpenSecrets statistic on self-funder win rates (for example, share of $1M+ self-funders who won) could be fetched. Fairshake's general-election win count was not found.
    • Benchmarks pass: Part C: spending totals for California Props 26/27 (2022) not confirmed; New Jersey 2025 governor race spending and result not retrieved (Wikipedia page too long, general-election section not reached); Virginia 2025 has only an early fundraising snapshot; no verified evidence on the electoral effects of Arizona/Maine/Connecticut public financing; Montana or other state contribution-limit facts not researched.
    • Benchmarks pass: Most Part B and Part C dollar figures marked partially_verified were read on Wikipedia only; before publication each should be checked against the outlet Wikipedia cites.
    • Benchmarks pass: WebSearch quota for the session was exhausted about one-third of the way through, so later items could only be pursued by fetching known URLs.
    • 2026 primaries/themes pass: Confidence key: verified = figures and result read today from at least one fetched reputable source or the FEC file; partially_verified = some elements rest on a search-result summary, a single secondary source or Wikipedia; unverified = could not confirm today. Compiled 2026-11-19. Items carry an extra "additional_sources" list beyond the requested fields.
    • 2026 primaries/themes pass: WebSearch budget for the session ran out about one-third of the way through (shared limit). Everything after that came from direct WebFetch of URLs already surfaced, Wikipedia/Ballotpedia pages, Brennan Center series pages and the FEC file. Topics below were therefore not completed.
    • 2026 primaries/themes pass: ActBlue vs WinRed 2026 totals and an explicit small-donor share of 2026 fundraising: not found.
    • 2026 primaries/themes pass: End Citizens United / "No Corporate PAC" pledge count for 2026: not found. Explicit pledge status confirmed only for Peggy Flanagan (and April McClain Delaney's campaign claim); for Talarico, El-Sayed, Platner, Jackson and Osborn only the FEC funding mix was verified.
    • 2026 primaries/themes pass: Final full-cycle spending totals for MAGA Inc., Senate Leadership Fund, Congressional Leadership Fund, Senate Majority PAC and House Majority PAC: not found (post-general and year-end FEC reports not yet reflected in coverage). Only cash-on-hand snapshots and through-September figures are given.
    • 2026 primaries/themes pass: Ranked list of the most expensive 2026 House general-election races and their winners: not found.
    • 2026 primaries/themes pass: MI-13 (Thanedar vs McKinney) and MI-11 primary results: not confirmed. NY-21 general result (Constantino) and CA-4 general result (Eric Jones vs Mike Thompson): not confirmed; CA-11 November winner (Wiener vs Chan) not retrieved.
    • 2026 primaries/themes pass: North Carolina Senate primary vote shares and any NC primary money story beyond NC-4: not retrieved. Tennessee AIPAC-targeted primaries: nothing retrieved.
    • 2026 primaries/themes pass: 2025 New Jersey and Virginia governor races: results verified (Wikipedia) but candidate and outside spending were not retrieved today.
    • 2026 primaries/themes pass: Outside-spending breakdowns by group were not found for: Louisiana Senate primary (other than a $4.1M pro-Letlow super PAC in the runoff), Georgia Senate GOP primary, Kentucky Senate GOP primary (other than ~$8M crypto), Minnesota DFL primary (totals known, groups not named), NY-10, NY-13, MO-1, Massachusetts Senate primary (conflicting figures).
    • 2026 primaries/themes pass: Single-source claims needing confirmation before publication: Latin Times report that AIPAC announced on Oct 21, 2026 it would stop endorsements/direct PAC contributions, and that 30 incoming members explicitly rejected AIPAC; AI super PAC totals of $81.3M (aggregator page returned 403); Musk's $10M to a pro-Nate Morris effort (Wikipedia reference title only); Chakrabarti's ~$9M self-funding (Axios, 403).
    • 2026 primaries/themes pass: Portside article "Knowledge Beat Money in the 2026 Midterms": not found on portside.org front page; not used.
    • 2026 primaries/themes pass: opensecrets.org, web.archive.org, axios.com, michiganadvance.com, scdailygazette.com and fourweekmba.com could not be fetched; Axios figures are quoted only via Roll Call's column or search-result summaries.
    • 2026 primaries/themes pass: FEC weball26 caveats: figures are two-year-cycle-to-date totals as of each committee's latest report (coverage dates vary from mid-2026 to 10/14/2026); for nominees they include general-election activity; candidates with both House and Senate IDs show the same committee twice; Saikat Chakrabarti was not located in the file; Senate incumbents' six-year totals are understated.
    • 2026 primaries/themes pass: Several vote percentages are early or unofficial counts (TX-D primary 52.4% vs 52.8%; MA primary at 81% counted; NY-10 from an unofficial tally; Prop 40 52.1%-52.8% No depending on count date).
    • 2026 primaries/themes pass: WebFetch answers are produced by a summarizing model; figures were cross-checked across sources where possible but long Wikipedia pages were read only in part.