Senate odds move ten policy names, not the party baskets: +7.1% per 10 points
Control of the Senate is the open question of the 3 November vote. This page measures how much of it the market prices, in which names, and what the broad macro assets make of it.Is the midterm priced, where, and through which policy?
Senate odds move ten policy names, not the party baskets: +7.1% per 10 points
Is the midterm priced, where, and through which policy?- The market prices the midterms through the Senate, and only in a handful of policy names: tested (said 14 Oct 2026).→ Exhibit 3
- The Democratic and Republican policy baskets do not move with Senate odds; the ten names the data confirm move +7.1% per 10 points.→ Exhibit 5
- Humana falls 2.2% for every ten-point rise in Democratic Senate odds and Green Thumb rises 3.6%.→ Exhibit 7
- Zero may be the right price: the ex-ante beta is 3.2 standard errors from zero and the 15.6-point September jump carries most of the evidence.→ Exhibit 1
- In six past votes the moved +2.3% per ten points of and lost 5.1% when the vote went as priced.→ Exhibit 10
| What would change this view | Trigger | Today | Decided by |
|---|---|---|---|
| The pricing claim | either Senate basket stops moving with the odds over any 15 sessions: the ex-ante beta at or below zero, or the tuned beta below zero | ex-ante +2.4 · tuned +5.0 % per 10 pts (2 Nov 2026, 15 sessions). The ex-ante and tuned betas are on the 26 and 10 names left by the revision of 25 Oct 2026 | each session's close of odds and prices |
| The sizing claim | Democratic Senate odds above 70% or below 50% | 75% (2 Nov 2026)crossed | the venues' daily close; the vote of 3 Nov 2026 |
| The history claim | the tuned basket's election-week return falls outside the band the fit to past votes predicts | pending · the abnormal return for 4–11 Nov 2026 needs the factor returns for those sessions; the French library runs to 30 Sep 2026. The band is ±6.8% around the fit's prediction at a Senate surprise of +25.4 pointspending | the sessions of 4–11 Nov 2026 |
| The macro claim | on election night gold or the moves more than one typical day in the direction the result implies | 4 Nov 2026: gold +3.7%, the 10-year breakeven −4bp. On this page's betas the result points gold down and the breakeven down. One typical day is 1.6% and 2bpcrossed | the session of 4 Nov 2026 |
The reasoning behind each point
- VerdictThe market prices the midterms through the Senate, and only in a handful of policy names: tested (said 14 Oct 2026).
As written on 14 Oct 2026: “The market prices the 2026 midterms through control of the Senate, and only in a handful of policy names. The party, donor and sell-side baskets carry no Senate beta; the names a Senate vote would change do. The broad macro assets carry none that the data can see.” The verdict is “tested”: a trigger of the falsifier table has crossed. The sizing claim is wrong: the Senate is no longer the open question. The macro claim is wrong: the broad assets do carry a Senate beta the daily data could not see. Democrats won control of both chambers on 3 Nov 2026, and both venues settled that outcome on 4 Nov 2026. The scorecard sets what the log said against what has happened since.
- AnswerThe Democratic and Republican policy baskets do not move with Senate odds; the ten names the data confirm move +7.1% per 10 points.
The party baskets measure +0.5% and −1.0% per 10 points of Democratic Senate odds, their pair +1.6%. The ex-ante basket, long the winners of a Senate vote and short the losers, moves +3.2% per 10 points (t 3.2). The broad macro assets barely move: Senate odds explain at most 5% of any asset's daily swings, the oil shock up to 85%.
- ForHumana falls 2.2% for every ten-point rise in Democratic Senate odds and Green Thumb rises 3.6%.
Bank of America moves −2.3% per 10 points. In the September jump of 15.6 points the tuned basket returned +9.8% and the ex-ante basket +1.5%, after removing the market, Treasuries and oil. In the October fall of 3.1 points the tuned basket returned −4.4% and the ex-ante basket +1.9%.
- AgainstZero may be the right price: the ex-ante beta is 3.2 standard errors from zero and the 15.6-point September jump carries most of the evidence.
With a veto and no two-thirds majority, neither outcome enacts its agenda. The tuned basket was chosen on the data, half of its names on these same betas. The 11–23 Sep window came with the FOMC and the CLARITY vote inside it; without 14–18 Sep the ex-ante beta is +4.0% and the tuned +7.5%.
- HistoryIn six past votes the tuned basket moved +2.3% per ten points of Senate surprise and lost 5.1% when the vote went as priced.
The slope has t 5.9 across the five midterms and the Georgia runoff, with an exact permutation p of 0.02. The ex-ante basket moved +0.7% per ten points of surprise and lost 0.6% in weeks that went as priced. The 2026 odds beta of the tuned basket, +1.1% after pulling toward the average name, sits near the historical slope.
Next: Thematic impact — the rows this theme will feed
Exhibit 2 · The view and the claim as written
The market prices the midterms through the Senate, and only in a handful of policy names — tested
The market prices the 2026 midterms through control of the Senate, and only in a handful of policy names. The party, donor and sell-side baskets carry no Senate beta; the names a Senate vote would change do. The broad macro assets carry none that the data can see.
A Democratic House is about 90% priced and the Senate, near 62%, is the open question. Every single name with a lever that differs between the Democratic sweep and the split Congress, long the winners and short the losers, moved +2.4% per 10 points of Democratic Senate odds from 28 May to 12 Oct; the eleven names whose measured beta agrees with the lever moved +8.0%. The conventional Democrat-minus-Republican policy pair measured 0.0%. Across the six midterm and Senate votes since 2006 the tuned basket moved +2.4% per 10 points of Senate surprise in the first week and lost 5.3% in weeks that went as priced.
The figures in the claim are the research note's own. The Detail fold sets them beside this page's, with the reason they differ.
Exhibit 3 · What the view expected against what has happened, test by test
The view is tested: the word follows a rule set out in the Appendix, from the log and the tests above, never by hand.
About this exhibit
Senate odds have risen 28 points since May, most of it in one fortnight
Where do the odds stand, and when did they move?Exhibit 1 · Market odds of Democratic control of the Senate and the House
The two venues on the day the page opened, as quoted
- The venues on the day the page opened. At midday on 13 Oct 2026 the two venues' Senate contracts had traded $86k (Polymarket) and $6k (Kalshi) over 24 hours and differed by 2.2 points (Polymarket 63.5, Kalshi 61.3 constructed). (research note: Midterm policy baskets (14 Oct 2026) · pricing when the note was written (DeFi Rate, 13 Oct 12:33 ET) · 13 Oct 2026)
About this exhibit
Exhibit 4 · Policy levers under each election outcome, and the stocks each reaches
| Lever | Decided by | Democratic sweep · 60% | Democratic House, Republican Senate · 29% | Republican sweep · 10% | Names it reaches |
|---|---|---|---|---|---|
| Fed Board, bank regulators, SEC, CFTC, judges | Senate confirmation | Stalls | Continues | Continues | BAC, SYF, COF, AXP, COIN, HOOD |
| Oversight and subpoenas | Committee chairs | Both chambers | One chamber only | None | HUM, GEO, CXW, defence |
| Medicare Advantage oversight; CMS nominees | Senate Finance and HELP | Tightens | Loosens | Loosens | HUM |
| 10% credit-card APR cap (Sanders–Hawley; the President asked for one in January) | Law | Highest risk | Low | Low | SYF, COF, AXP, BAC |
| Cannabis banking and rescheduling | Senate floor time | Floor time | No | No | GTBIF, CURLF |
| Clean-energy credits; EPA and FERC nominees | Senate | Credits defended, nominees blocked | Nominees confirmed | Credits at risk | ENPH, FSLR, NEE |
| Permanent ACA enhanced credits | Law; needs the President's signature | Leverage via appropriations only | Unlikely | No | OSCR, CNC, MOH, hospitals |
| CLARITY Act (cloture failed 49–50 on 15 Sep) | 60 Senate votes | Rewritten around consumer protection | Stalled | Revived | COIN, HOOD |
| Reconciliation 3.0: the $350bn defence request; House Republicans eyed about $67bn | Both chambers, simple majority | Dead from 3 Jan | Dead from 3 Jan | Live | LMT, LHX, NOC, KTOS, AVAV |
| Further Medicaid cuts | Reconciliation | Blocked | Blocked | Possible | Hospitals, Medicaid managed care |
| FY27 appropriations (stopgap runs to 11 Dec) | Both chambers | Shutdown leverage | Shutdown leverage | Smooth | Contractors, defence |
| Tariffs: Section 301 at 10–12.5% since 24 Jul, no time limit; Section 232 | Executive; removal needs a two-thirds override | Unchanged | Unchanged | Unchanged | Importers: election-neutral |
Where each lever comes from, as quoted
- Lever sources. Defense One (9 Jun 2026) on McConnell and defence reconciliation; Hoodline (14 Jul) on the enforcement funding; bit.com on the CLARITY cloture vote of 15 Sep; CREFC (8 Sep) on the stopgap to 11 Dec; NLIHC (15 Jun) and Morgan Lewis (7 Jul) on the housing ban; WRVO (12 Jan) on the card APR cap; political.org (16 Jul) and CRFB on Reconciliation 3.0 ($95bn, no offsets; more than $100bn over ten years with interest); Brookings on Fed Board terms (no Board term ends in 2027; the chair runs to May 2030); DLA Piper (24 Feb) and Kroll on the post-IEEPA tariff authorities. (research note: Page plan midterms theme (14 Oct 2026) §9 and research note: Midterm macro assets (14 Oct 2026) sources · levers · 14 Oct 2026)
About this exhibit
The party and donor baskets do not move with Senate odds, the names a Senate vote changes do
Which baskets carry Senate exposure?Exhibit 5 · Each basket's move per ten-point rise in Democratic Senate odds
A4e and A4t are the ex-ante and tuned Senate baskets. The A4t legs are the tuned basket's long and short sides, each against SPY.
About this exhibit
Exhibit 6 · What each of the 16 baskets holds, name by name
| Group | Names | Why these names |
|---|---|---|
| A1 Democratic policy · 12 namesBeneficiaries of Democratic policy, grouped by the lever that would move them · eToro midterm playbook; JPM 1 Oct note | ||
| Hospitals | HCA THC UHS CYH | any Democratic chamber blocks Medicaid cuts |
| Medicaid and ACA insurers | CNC MOH OSCR | Medicaid and ACA credit protection |
| Clean energy | NEE FSLR ENPH | tax-credit and permitting protection |
| Environmental services | CLH | agency funding (JPM) |
| Childcare | BFAM | Democratic childcare agenda |
| A2 Republican policy · 15 namesBeneficiaries of Republican policy, grouped by lever · eToro midterm playbook; JPM 1 Oct note | ||
| Defence, reconciliation-funded | LMT LHX NOC KTOS AVAV | the $350bn reconciliation request |
| Crypto legislation | COIN HOOD | CLARITY Act |
| Oil, gas, LNG | DVN WMB LNG | leasing, pipelines, LNG permits |
| Immigration enforcement | GEO CXW | detention contracts |
| Medicare Advantage | HUM | MA rates and oversight |
| AI power | BE | JPM red-wall pick |
| Bank deregulation | BAC | capital and supervision rules |
| A3 Democratic minus Republican · 27 namesThe spread of the two policy baskets | ||
| Long A1, short A2 | equal weight each side | the conventional election pair |
| A4e Senate basket, ex-ante · 26 namesFilter 1 only: a lever that differs between the Democratic sweep and the split gives the sign; no look at the data; equal weight; single names (XLF and KRE left out). ELV left out by the author's decision of 25 Oct 2026. · this page | ||
| Long, every name with a Senate lever that says up | GTBIF CURLF ENPH FSLR NEE OSCR CNC MOH HCA THC UHS CYH | cannabis floor time; clean-energy credits and nominees; ACA-credit leverage |
| Short, every name with a Senate lever that says down | HUM BAC SYF COF AXP COIN HOOD GEO CXW LMT LHX NOC KTOS AVAV | MA oversight; regulator confirmations; the card APR cap; CLARITY and SEC/CFTC seats; oversight of enforcement and defence |
| A4t Senate basket, tuned · 10 namesFilter 1 and filter 2: the measured beta must agree with the lever's sign; FSLR and NEE kept at zero beta as liquid ballast; UHS and CXW left out on lever quality. ELV left out by the author's decision of 25 Oct 2026. · this page | ||
| Long | GTBIF CURLF ENPH FSLR NEE | cannabis floor time; clean-energy credits and nominees |
| Short | HUM BAC SYF AXP COF | Medicare Advantage oversight; regulator confirmations; the card APR cap |
The lists and the donor data, as quoted
- JPMorgan's outcome lists. Blue wave: BKR, VMI, FLUT, CNC, MOH, WU, CRM, NEE. Gridlock: SHW, ORN, MGM, GILD, CRWV, ORCL, META (SPCX not in the sample). Red wall: DVN, LMT, LHX, BE, BAC (VEEV not in the sample). (JPMorgan note of 1 Oct 2026 (Lakos-Bujas) as reported by StockTwits on 5 Oct 2026 (research note: Midterm policy baskets (14 Oct 2026), baskets C1–C3) · baskets and results · 5 Oct 2026)
- Corporate money, 2026 cycle. Share of party money to Republicans in January 2026: Valero 100%, UPS 67%, Boeing 64%, General Dynamics 61%, Home Depot 61%, Lockheed Martin 59%, Elevance 58%; Honeywell 55% to Democrats, Charter 49%, Northrop 47%. Super PAC money: GEO $1m and Reynolds $5m to MAGA Inc.; Green Thumb and Curaleaf $2.5m each to America First Agriculture Action; Applied Digital $100k to Republican Forward; DraftKings $14.5m and FanDuel $7.5m to Win for America. (Quiver Quantitative (29 Jan 2026) and Public Citizen (27 Aug 2026), as summarised in research note: Midterm policy baskets (14 Oct 2026) (baskets B1–B2) · baskets and results · 27 Aug 2026)
- Employee giving, 2024 cycle. Alphabet $2.16m to Harris against about $54k to Trump; Meta $835k against $25k; Apple $861k against $44k; Microsoft $1.1m against $88k; Amazon $1.0m against $116k. (research note: Midterm policy baskets (14 Oct 2026) (basket B3) · baskets and results · 30 Sept 2024)
About this exhibit
Choosing the ten names on the data lifts the beta from +3.2% to +7.1% at a cost of 41% since 27 May
What does the say, what does the add, and how were its names chosen?Exhibit 7 · Each name's move per ten points of Senate odds, by basket leg
| The two baskets side by side | Ex-ante, 26 names | Tuned, 10 names |
|---|---|---|
| Senate move per 10 points | +3.2% (t 3.2) | +7.1% (t 3.0) |
| Without 14–18 Sept | +4.0% (t 3.8) | +7.5% (t 2.8) |
| Long leg | +2.2% (t 2.3) | +5.7% (t 2.7) |
| Short leg | −1.0% (t −1.1) | −1.4% (t −1.2) |
| Share of daily swings the odds explain | 10% | 23% |
| How far each moves with SPY · with IEF | −0.29 · 0.30 | 0.61 · 0.20 |
| Long minus short, since 27 May | −6.1% | −41.1% |
| Over 11–23 Sep, when the odds moved +15.6 points | +1.5% | +9.8% |
| Over 5–12 Oct, when the odds moved −3.1 points | +1.9% | −4.4% |
| Past votes, first week: move per 10 points of surprise | +0.7% (t 1.6) | +2.3% (t 5.9) |
| Past votes, first week: return when the vote went as priced | −0.6% | −5.1% |
| Past votes, by day 20: move per 10 points of surprise | +0.3% | +3.2% |
| How the names were chosen | the lever alone, before looking at the data | the names whose measured move agrees with the lever |
About this exhibit
The followed the odds in 3 windows of 4, the in 3
Does either basket follow the odds day by day, and which test is clean?Exhibit 8 · Democratic Senate odds and both Senate baskets, day by day
From 14 Oct 2026 each session tests the baskets: cleanly for the ex-ante basket, which never saw the data, and partly for the tuned one.
About this exhibit
Exhibit 9 · Senate odds change and basket returns over each dated period
| Window | Odds, points | Ex-ante, long minus short | Ex-ante long leg | Ex-ante short leg | Tuned, long minus short | Tuned long leg | Tuned short leg | SPY |
|---|---|---|---|---|---|---|---|---|
| 10–12 Aug | +5.4 | +0.1% | +1.4% | +1.3% | −0.6% | +0.8% | +1.5% | −0.1% |
| 11–23 Sep | +15.6 | +1.5% | −1.3% | −2.8% | +9.8% | +0.9% | −8.9% | +0.7% |
| 18–23 Sep | +6.1 | +3.1% | −0.5% | −3.6% | +6.1% | +1.6% | −4.5% | +0.8% |
| 5–12 Oct | −3.1 | +1.9% | +1.6% | −0.3% | −4.4% | −1.6% | +2.8% | +0.4% |
| since 27 May | +27.5 | −6.1% | −4.5% | +1.6% | −41.1% | −31.6% | +9.5% | +0.7% |
About this exhibit
Past votes moved both baskets with the surprise, only the tuned one paid an election-week cost
What did the last eleven elections do to the two baskets?Exhibit 10 · Each basket's return against the Senate surprise, past votes
About this exhibit
Members of Congress own tech, not the Senate: the Democratic members' fund beat SPY by 0.1% with no Senate beta
Who is positioned, and how are they doing against the market?Exhibit 11 · Return over SPY of funds copying each party's members
| Fund | Over SPY since 27 May | After removing the market, Treasuries and oil | Daily edge, bp | A year at that pace | Senate move per 10 points | Moves with SPY |
|---|---|---|---|---|---|---|
| Democratic members' trades (NANC) over SPY | +0.1% | −0.1% | −0.3 t −0.1 | −0.7% | ±0.0% t 0.0 | 1.23 |
| Republican members' trades (GOP) over SPY | −4.0% | −4.4% | −2.7 t −0.5 | −6.7% | −0.3% t −0.7 | 1.10 |
| The gap between the two funds | +4.1% | +4.3% | +2.4 t 0.4 | +6.0% | +0.3% t 0.9 | 0.12 |
What the two funds hold and what research finds, as quoted
- The congressional ETFs. NANC (Unusual Whales Subversive Democratic Trading ETF) replicates disclosed holdings of Democratic members and their spouses, about half in tech and communications, about 85% of names in the S&P 500 (CFRA); GOP (Republican members; KRUZ until 25 Mar 2025) tilts to financials, industrials, energy and crypto. Disclosures arrive up to 45 days late, in dollar ranges. (Tidal Financial fund pages; The Daily Upside, 25 Mar 2025 (research note: Page plan midterms theme (14 Oct 2026) §10) · §10 basket E · 14 Oct 2026)
- Senators' own trades. Belmont, Sacerdote, Van Hoek and Walrath (NBER w26975; Journal of Public Economics 2022): senators' purchases from 2012 to March 2020 underperform matched stocks by 11, 28 and 17 basis points at one, three and six months; no committee-specific skill. (research note: Page plan midterms theme (14 Oct 2026) · §10 evidence on members' own performance · 14 Oct 2026)
About this exhibit
Few macro assets move with Senate odds: the odds explain at most 5% of any asset's daily swings, the oil shock up to 85%
What does the vote mean for rates, the dollar, gold, oil and stocks, once the oil shock is taken out?Exhibit 12 · How much each asset moves when Democratic Senate odds rise 10 points
About this exhibit
Four outcomes priced daily and decided on 3 Nov 2026
What are the scenarios, how likely, and what decides them?Exhibit 13 · Market odds of each control outcome for Congress, and what each changes
Quoted 14 Oct 2026: a traded price, not a forecast.
About this exhibit
Exhibit 14 · What would change the view, and the dates that decide it
| Status | Test | Latest reading | Threshold or date | If crossed | Threshold source |
|---|---|---|---|---|---|
| closing | Either Senate basket stops moving with the odds over any 15 sessions: the ex-ante beta at or below zero, or the tuned beta below zero | ex-ante +2.4 · tuned +5.0 % per 10 pts (2 Nov 2026, 15 sessions). The ex-ante and tuned betas are on the 26 and 10 names left by the revision of 25 Oct 2026 | ≤ 0.0 or < 0.0 % per 10 pts | The pricing claim is wrong: the names a Senate vote would change no longer move with the odds. | research note: Page plan midterms theme (14 Oct 2026) · §11 the outline v3: falsifier (ex-ante beta ≤ 0 over 15 sessions); §6 status rule · 14 Oct 2026 |
| crossed | Democratic Senate odds above 70% or below 50% | 75% (2 Nov 2026) | > 70 or < 50% | The sizing claim is wrong: the Senate is no longer the open question. | research note: Page plan midterms theme (14 Oct 2026) · §3 falsifier table, the sizing claim; §6 status rule · 14 Oct 2026 |
| pending | The tuned basket's election-week return falls outside the band the fit to past votes predicts | pending · the abnormal return for 4–11 Nov 2026 needs the factor returns for those sessions; the French library runs to 30 Sep 2026. The band is ±6.8% around the fit's prediction at a Senate surprise of +25.4 points | the first week after the vote, sessions 4–11 Nov 2026 | The history claim is wrong: the six past votes do not describe this one. | research note: Page plan midterms theme (14 Oct 2026) · §3 falsifier table, the history claim; §6 lifecycle B · 14 Oct 2026 |
| crossed | On election night gold or the 10-year breakeven moves more than one typical day in the direction the Senate result implies | 4 Nov 2026: gold +3.7%, the 10-year breakeven −4bp. On this page's betas the result points gold down and the breakeven down. One typical day is 1.6% and 2bp | the session of 4 Nov 2026 | The macro claim is wrong: the broad assets do carry a Senate beta the daily data could not see. | research note: Midterm macro assets (14 Oct 2026) · §6 challenges to this read; §7 on the page (the fifth falsifier row) · 14 Oct 2026 |
- 9 Dec 2026FOMC meeting with projections, 8–9 December.
- 11 Dec 2026The stopgap funding law expires: shutdown leverage in every branch.
- 3 Jan 2027The 120th Congress is seated; the theme resolves and the 2026 vote joins the event-study sample.
Exhibit 15 · No site model yet prices the asset effects of each outcome
Nothing is projected here. Democratic sweep, Democratic House, Republican Senate, Republican sweep and Republican House, Democratic Senate: no model on the site prices a Senate outcome.
What each path would do to core PCE, unemployment and real GDP is read on Thematic impact, under Outlook.
About this exhibit
Appendix tables and the reconciliation to the note
The donor, sell-side and neutral baskets, the event study by theme, the robustness runs and the note's figures beside this page's
Exhibit 16 · Each other basket's move per ten points of Senate odds
| Basket | Names | Per 10 points | t | Without 14–18 Sept | Moves with SPY | Moves with IEF | Odds explain |
|---|---|---|---|---|---|---|---|
| B1 Republican-leaning donors | 11 | +0.6% | 0.9 | +0.5% t 0.6 | 0.46 | 0.34 | 33% |
| B2 Bipartisan donors | 6 | +0.8% | 0.6 | +2.0% t 2.0 | 0.29 | 1.42 | 18% |
| B3 Democratic-leaning employees | 5 | −0.6% | −0.8 | −0.8% t −0.8 | 1.09 | −0.52 | 39% |
| C1 JPM blue wave | 8 | +0.9% | 0.7 | +2.0% t 2.4 | 0.21 | 0.94 | 14% |
| C2 JPM gridlock | 7 | +1.7% | 1.1 | +1.9% t 1.1 | 1.54 | 0.47 | 55% |
| C3 JPM red wall | 5 | −0.5% | −0.3 | −0.6% t −0.4 | 0.89 | 0.12 | 24% |
| D1 Election-neutral | 11 | −0.2% | −0.3 | −0.3% t −0.3 | 1.23 | −0.46 | 53% |
| E1 Democratic members' trades (NANC) | 1 | ±0.0% | 0.0 | −0.1% t −0.3 | 1.23 | −0.18 | 90% |
| E2 Republican members' trades (GOP) | 1 | −0.3% | −0.7 | −0.5% t −1.0 | 1.10 | −0.34 | 69% |
About this exhibit
Exhibit 17 · Each theme's first-week return per ten points of Senate surprise
| Theme | Midterms and Georgia: slope | t · p | All votes: slope | t · p | This year, per 10 points of odds |
|---|---|---|---|---|---|
| Senate basket, tuned | +2.3% | 5.9 · 0.02 | +1.0% | 1.2 · 0.27 | +7.1 measured, +1.1 pulled toward the average name |
| Clean energy | +0.9% | 1.4 · 0.23 | +0.8% | 2.2 · 0.14 | +5.5 |
| Cannabis (from 2016) | too few votes | – | too few votes to read | – | +3.6 to +8.7 |
| Managed care | −0.6% | −1.4 · 0.23 | −0.7% | −2.9 · 0.05 | HUM −2.2 |
| Medicaid insurers | −1.0% | −1.8 · 0.10 | ±0.0% | −0.1 · 0.95 | +0.4 |
| Card lenders | −0.4% | −1.3 · 0.28 | −0.4% | −2.6 · 0.03 | −1.3 to −0.3 |
| Banks | −0.3% | −1.1 · 0.36 | −0.3% | −1.1 · 0.28 | BAC −2.3 |
| Hospitals | ±0.0% | 0.1 · 0.91 | +0.5% | 1.0 · 0.45 | −1.0 |
| Defence | ±0.0% | −0.1 · 0.94 | −0.3% | −1.6 · 0.25 | +0.3 |
| Energy | −0.1% | −0.3 · 0.94 | +0.1% | 0.7 · 0.42 | −1.6 |
| Democratic policy basket | +0.6% | 1.6 · 0.22 | +0.3% | 0.8 · 0.51 | +0.5 |
| Republican policy basket | ±0.0% | 0.1 · 0.90 | −0.3% | −1.2 · 0.58 | −1.0 |
About this exhibit
Exhibit 18 · Past-vote slope across event windows, basket legs and vote samples
| Basket | Votes | Sessions | Per 10 points | t | p |
|---|---|---|---|---|---|
| Tuned basket | midterm and Georgia, six votes | [0,+1] | +0.9% | 3.7 | 0.02 |
| Tuned basket | midterm and Georgia, six votes | [0,+5] | +2.3% | 5.9 | 0.02 |
| Tuned basket | midterm and Georgia, six votes | [0,+20] | +3.2% | 5.1 | 0.04 |
| Tuned basket | midterm and Georgia, six votes | [+6,+60] | −1.2% | −0.7 | 0.57 |
| Tuned basket | midterm and Georgia, six votes | pre[−20,−1] | +1.5% | 3.6 | 0.02 |
| Tuned basket | all eleven votes, with the other two surprises | [0,+5] | +1.0% | 1.2 | 0.28 |
| Tuned basket without cannabis | midterm and Georgia, six votes | [0,+1] | +0.8% | 2.5 | 0.04 |
| Tuned basket without cannabis | midterm and Georgia, six votes | [0,+5] | +1.3% | 2.4 | 0.04 |
| Tuned basket without cannabis | midterm and Georgia, six votes | [0,+20] | +1.0% | 1.1 | 0.26 |
| Tuned basket without cannabis | midterm and Georgia, six votes | [+6,+60] | −3.4% | −1.6 | 0.23 |
| Tuned basket without cannabis | midterm and Georgia, six votes | pre[−20,−1] | +0.5% | 0.7 | 0.43 |
| Tuned basket without cannabis | all eleven votes, with the other two surprises | [0,+5] | +1.3% | 3.2 | 0.07 |
| Long leg | midterm and Georgia, six votes | [0,+1] | +0.9% | 2.7 | 0.06 |
| Long leg | midterm and Georgia, six votes | [0,+5] | +1.9% | 3.8 | 0.03 |
| Long leg | midterm and Georgia, six votes | [0,+20] | +2.8% | 3.6 | 0.05 |
| Long leg | midterm and Georgia, six votes | [+6,+60] | −0.9% | −0.7 | 0.59 |
| Long leg | midterm and Georgia, six votes | pre[−20,−1] | +1.5% | 2.7 | 0.04 |
| Long leg | all eleven votes, with the other two surprises | [0,+5] | +0.5% | 0.7 | 0.53 |
| Short leg | midterm and Georgia, six votes | [0,+1] | ±0.0% | −0.1 | 0.89 |
| Short leg | midterm and Georgia, six votes | [0,+5] | −0.4% | −2.1 | 0.13 |
| Short leg | midterm and Georgia, six votes | [0,+20] | −0.4% | −2.3 | 0.10 |
| Short leg | midterm and Georgia, six votes | [+6,+60] | +0.3% | 0.7 | 0.48 |
| Short leg | midterm and Georgia, six votes | pre[−20,−1] | −0.1% | −0.1 | 0.94 |
| Short leg | all eleven votes, with the other two surprises | [0,+5] | −0.4% | −3.3 | 0.03 |
About this exhibit
Exhibit 19 · The October research note's headline figures beside the site's
| Figure | The note | This page | Difference | Where the note printed it |
|---|---|---|---|---|
| Ex-ante Senate basket, % per 10 points | +2.41t 1.56 | +3.18t 3.18 | +0.77 | research note: Page plan midterms theme (14 Oct 2026) · §11 numbers behind the change · 14 Oct 2026 |
| Tuned Senate basket, % per 10 points | +7.98t 2.70 | +7.14t 3.04 | −0.84 | research note: Page plan midterms theme (14 Oct 2026) · §11 numbers behind the change · 14 Oct 2026 |
| Past votes, tuned basket: first-week slope per 10 points of surprise | +2.38t 5.81 | +2.29t 5.86 | −0.09 | research note: Page plan midterms theme (14 Oct 2026) · §12 history (6 events) · 14 Oct 2026 |
| Past votes, tuned basket: first-week intercept, % | −5.34t −2.77 | −5.13t −2.78 | +0.21 | research note: Page plan midterms theme (14 Oct 2026) · §12 history (6 events) · 14 Oct 2026 |
| Past votes, ex-ante basket: first-week slope per 10 points of surprise | +0.72t 1.65 | +0.71t 1.63 | −0.01 | research note: Page plan midterms theme (14 Oct 2026) · §12 history (6 events) · 14 Oct 2026 |
| Past votes, ex-ante basket: first-week intercept, % | −0.76t −0.37 | −0.65t −0.32 | +0.11 | research note: Page plan midterms theme (14 Oct 2026) · §12 history (6 events) · 14 Oct 2026 |
| NANC over SPY since 27 May, % | +3.67 | +0.10 | −3.57 | research note: Page plan midterms theme (14 Oct 2026) · §12 Congress (27 May – 12 Oct) · 14 Oct 2026 |
| SPY per 10 points, after removing the oil shock, % | +0.83t 1.03 | −0.73t −0.92 | −1.56 | research note: Midterm macro assets (14 Oct 2026) · §2 betas · 14 Oct 2026 |
| GLD per 10 points, after removing the oil shock, % | −1.07t −0.83 | −1.70t −0.90 | −0.63 | research note: Midterm macro assets (14 Oct 2026) · §2 betas · 14 Oct 2026 |
The note read its odds from DeFi Rate's venue average and its closes from TradingView. This page reads both venues' own price histories and Yahoo Finance closes to the latest session. The note compounded each name's price change over a window, where this page sums each basket's daily log returns. Yields come from the Treasury's par curves rather than FRED. The inputs differ by construction, so the figures sit side by side and are not checks.
About this exhibit
- DeFi Rate's daily history of the Kalshi and Polymarket Senate odds, read 13 Oct 2026 12:33 ET
- closes from TradingView to 12 Oct 2026
- FRED yields
The Appendix
The model, the data, the checks and the sources80 checks, all pass
This page measures what the market prices of the 2026 midterms, in its own data: Democratic control odds from the two venues, closes from Yahoo Finance, yields from the Treasury and the research factors from Kenneth French's library. Only the page's ending reads other pages of this site, the Projections and Transmission pages, and each value it copies from them is checked against the page it comes from.
- OddsWhere the odds come from, venue by venue.
Odds. Polymarket: the Yes token of the Democratic-control market of each chamber and of the four balance-of-power contracts, the CLOB price history at one point a UTC day; the point stamped in the first minute after midnight is read as the close of the day just ended. Kalshi lists no tradable standalone chamber market for 2026, so its chamber odds are sums of its four joint contracts: Senate = DD + RD, the House = DD + DR; the UTC-day close is the last traded price of the last hourly candle ending at or before midnight. The page's Senate series is the two venues' average on every day both trade (average of Polymarket and Kalshi where both trade, else the one that trades); the House series is Kalshi's. On a session the odds are the last daily close on or before the session's date. Every probability on this page is a traded price, printed with its venue and date; no author's odds appear anywhere. On 2 Nov 2026 the venues read 75.5 (Polymarket) and 73.7 (Kalshi) for the Senate.
- ModelHow a basket's Senate move is measured.
Odds betas. Daily log returns from 29 May 2026 to 2 Nov 2026 (109 sessions) on SPY, IEF, USO, the change in Democratic Senate odds (ΔS), its one-day lag and the change in the odds of Democratic control of the House (ΔH); Newey–West standard errors with 3 lags; Senate beta = 10 × (γ₀ + γ₁), in % per 10 points. The lag absorbs the gap between the odds' daily close and the 4 pm equity close. Every result is rerun without 14 Sep 2026–18 Sep 2026 and with returns trimmed at ±8%. The 15-session rolling beta re-estimates the same regression on each window of 15 sessions.
- BasketsHow the two Senate baskets are made.
The two Senate baskets. Ex-ante: every single name in the sample with a lever that differs between the Democratic sweep and the split Congress, long where the lever says up and short where it says down, equal weight, no look at the data; XLF and KRE are left out to keep it single-name. Tuned: the subset whose measured beta agrees with the lever's sign, with FSLR and NEE kept at zero beta as liquid ballast and UHS and CXW left out on lever quality. A basket's daily return is the equal-weighted mean of its names' daily log returns; a window's return is the sum of the basket's daily log returns inside it. From 25 Oct 2026, by the author's decision of 25 Oct 2026 (D-2026-10-25-EL2), ELV is left out of the A4e, A4t and B1 baskets, the funnel, the lever rows, the sub-theme lists, the event study's themes and the event study's tuned basket; the logged rows are unedited.
- ShrinkagePulling single names toward the average name.
Empirical Bayes shrinkage: the prior spread is the cross-section of 69 single-name betas, τ² = var(β) − mean(se²), τ = 1.03% per 10 points; each beta is multiplied by τ² / (τ² + se²). The page calls the result 'pulled toward the average name'.
- Event studyThe last eleven votes.
For each of the 11 votes since 2006 a factor model (Fama–French five factors, momentum, IEF, crude oil) is estimated over sessions −280 to −31 on adjusted closes; abnormal returns are summed over the windows; the sums are regressed across votes on the Senate surprise (outcome minus the eve-of-vote market probability), with exact permutation p-values over 720 orderings for the 6 midterm and Senate votes and 20,000 random orderings for all eleven, where the other two surprises are controls. Day 0 is the first session after election night (6 Jan 2021 for Georgia). The November 2020 Senate was unresolved, so its surprise is the IEM price change from 0.698 to 0.099. Simple daily returns; a daily move of 60% or more is dropped as a bad print; MJ is masked before 26 Dec 2017, when it was a Latin American real-estate ETF. The historical baskets are formed from the names and ETFs that existed at each date: a name enters when it has 120 valid sessions in the estimation window and no gap in the event window. The factors run to 30 Sep 2026.
- PositioningCongress's own portfolios, the flows and the survey.
Congress's own portfolios are measured three ways: the cumulative log return over SPY from the base date; the cumulative return after hedging SPY, IEF and USO at the ETF's own betas; and the intercept of the odds regression as a daily alpha with its HAC t-statistic. ETF flows and the fund-manager survey are quoted with their dates rather than read from a feed.
- MacroThe macro assets in four designs.
The same odds regression on SPY, IEF, TLT, TIP, UUP, GLD, SLV, USO, copper, IBIT, HYG, EEM and on the Treasury's 2-, 5- and 10-year par yields, the 10-year real yield and the 10-year breakeven, with the Brent return as the single control: in this sample Brent loads at −1.30 on SPY with 83% of its variance orthogonal to equities and copper, so the oil move is supply-dominated and the raw return is the Hormuz proxy. Four designs: raw; Brent-controlled; Brent-controlled without FOMC decision days (17 Jun, 29 Jul, 16 Sep), CPI days (10 Jun, 14 Jul, 12 Aug, 11 Sep) and jobs days (5 Jun, 2 Jul, 7 Aug, 4 Sep, 2 Oct) and the FOMC week of 14 Sep 2026–18 Sep 2026; and Rigobon's heteroskedasticity estimator with the 33 days of |Δodds| ≥ 1 point outside those exclusions as the high-variance regime, bootstrapped 1000 times (16–84% band). Variance shares are the R² of each asset on Brent alone and on the odds alone. The smallest detectable beta is twice the HAC standard error. The oil assets (USO) are run without the Brent control.
- DataPrices, yields, factors and the dated log.
Prices are Yahoo Finance's split- and distribution-adjusted closes, completed sessions only (a bar is admitted only if its session had closed when the file was downloaded). Yields are the Treasury's daily par yield curve and par real yield curve; the breakeven is par less real. The Fama–French five factors and momentum are the library's daily files. The eve-of-vote prices, the levers, the baskets, the funnel's verdicts and every quotation are rows of the page's dated file with their sources, written on 14 Oct 2026; the pipeline never revises them.
- StatusHow a test's status is set.
crossed — the threshold condition holds on the latest observation; closing — not crossed, the series has printed since the page opened (data to 13 October 2026), and the gap to the threshold is half or less of the gap on that date; not crossed — otherwise; pending — the test is dated in the future (the vote). Colours flag attention, not good or bad news for the book.
- LifecycleHow the theme opens, closes and resolves.
Holding; under pressure if Democratic Senate odds cross 70% or 50%, or a rolling beta is closing on its threshold; tested if the ex-ante 15-session beta falls to zero or the tuned one flips sign; resolved on 3 Jan 2027 when the 120th Congress is seated, with the outcome logged and the 2026 vote added to the event-study sample.
- VerdictHow the verdict word is set.
The verdict word is set by rule on every update: resolved — the log records the thesis as resolved; tested — a trigger of the falsifier table has crossed and the log has not yet closed the theme; under pressure — the log marks the theme closing, or a falsifier trigger is closing; holding — otherwise. The words of the verdict restate a passage of the logged claim, which this update checks is in the log.
- ProbabilitiesEvery probability is a traded price.
Every probability on this page is market-implied: a traded price from Polymarket or Kalshi, printed with its venue and date. The scenario set's logged revision is the two venues' average of the four joint contracts on its date, normalised to 100; the daily odds live in the trackers. No author's odds appear anywhere.
- ReconciliationThe note's figures beside this page's.
The October note's headline figures are printed beside this page's in the reconciliation fold. They are not checks: the note read DeFi Rate's venue average and TradingView closes to 12 Oct; this page reads the venues' own histories and Yahoo Finance closes to the latest session.
- LicencesWhat is shown, and on what terms.
Yahoo Finance prices are shown with attribution under the site's decision of 21 Sep 2026; the venues' prices, the Treasury's curves and the library's files carry no licence restriction; ETF flows and the survey are quoted rather than read from a feed.
- DocumentsWhere the view was argued.
- research note: Midterm policy baskets (14 Oct 2026) · summary, method, baskets and results, event study, flows and positioning · 14 Oct 2026
- research note: Midterm macro assets (14 Oct 2026) · §0 summary; §2 betas; §3 windows; §4 history; §5 channels · 14 Oct 2026
- research note: Page plan midterms theme (14 Oct 2026) · §11 reorganised outline; §12 prototype v3; §13 macro assets; §14 phrasing · 14 Oct 2026
- the site's specification for this page · 14 Oct 2026
How the numbers are made
polymarket_history + kalshi_candles ──▶ daily closes on UTC days; Kalshi chamber odds as sums of its joint contracts; the Senate series as the venues' average where both trade; the last close on or before each NYSE session ──▶ the odds on sessions yahoo_closes ──▶ daily log returns in percent for every instrument; equal-weighted basket returns, long leg less short leg, for the sixteen baskets of the dated file ──▶ the daily returns and baskets the odds on sessions + the daily returns and baskets ──▶ the odds regression on SPY, IEF, USO, ΔS, its lag and ΔH with Newey–West errors (3 lags); Senate beta = 10(γ0+γ1); rerun without the FOMC week and with returns trimmed at ±8%; the rolling 15-session beta ──▶ the betas and the rolling betas the betas and the rolling betas ──▶ empirical-Bayes shrinkage from the single-name cross-section (τ² = var(β) − mean(se²)); the funnel's live betas beside the dated verdicts; the record (hedged return, predicted move, R²) and the window sums ──▶ the baskets, the record and the windows yahoo_closes + french_factors + dated_file ──▶ the event study: the historical baskets formed from the names that existed at each vote; FF5 + momentum + IEF + oil estimated over sessions −280 to −31; abnormal returns summed over windows; cross-vote slopes with exact and random permutation p-values ──▶ the past votes the odds on sessions + the daily returns and baskets ──▶ the congressional ETFs as outperformance: cumulative log return over SPY, the hedged residual, the regression's intercept as a daily alpha ──▶ Congress's own portfolios the odds on sessions + yahoo_closes + Treasury's daily par yield curve ──▶ the macro set in four designs — raw, with Brent as the oil control, without release days and the FOMC week, Rigobon on the high-variance odds days — with variance shares, lead-lag tests, indexed levels, the windows and the six past votes ──▶ the macro assets the baskets, the record and the windows + the past votes + Congress's own portfolios + the macro assets + dated_file ──▶ trackers, triggers, tiles, takeaways, the falsifier and every section title generated from the numbers; the dated file's scenario set, timeline and quotations attached; the reproduction checks (validate, check, publish) ──▶ the data the page reads
Sources
- publicPolymarket, Gamma and CLOB APIs: the 2026 Senate and lower-chamber control markets and the four balance-of-power contracts, daily price history
Source: Polymarket · https://gamma-api.polymarket.com/markets
- publicKalshi, trade API v2: the four KXBALANCEPOWERCOMBO contracts, hourly and daily candlesticks
Source: Kalshi · https://api.elections.kalshi.com/trade-api/v2/series/KXBALANCEPOWERCOMBO
- licensedYahoo Finance chart API: adjusted daily closes of the 81 single names and ETFs, the congressional ETFs, the macro set and the event-study proxies (since listing)
Source: Yahoo Finance · https://query2.finance.yahoo.com/v8/finance/chart/
- publicU.S. Department of the Treasury: daily par yield curve (2-, 5-, 10-year) and daily par real yield curve (5-, 10-year); breakeven = par − real
Source: U.S. Department of the Treasury · https://home.treasury.gov/resource-center/data-chart-center/interest-rates/
- publicKenneth R. French Data Library: Fama–French five factors and the momentum factor, daily
Source: Kenneth R. French Data Library · https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ftp
- publicEve-of-vote market prices, 2006–2024: Snowberg, Wolfers and Zitzewitz (NBER WP 12751), the Iowa Electronic Markets, Intrade and PredictIt as reported by Ray Fair, The Block, Casino.org, davemanuel.com, Forbes.com, Asharq Al-Awsat, ElectionBettingOdds.com, and Polymarket's price history — each row with its URL in the dated file
Sources as listed per vote · https://iemweb.biz.uiowa.edu/iem_market_info/
- computed hereThe October note and its page plan (US midterm policy baskets; the macro assets; the lever map, the baskets and the funnel; quotations of the BofA survey via Finvaulta, ETF flows read from TradingView, JPMorgan's lists via StockTwits, Quiver Quantitative, Public Citizen, OpenSecrets via Reuters)
The site's own research, Macro Review
Pages this one copies from
| Page | Title | Updated | Data to | Arrays | Values | Source checks |
|---|---|---|---|---|---|---|
| Range and scenarios | Projections | 2026-11-17 | 16 Nov 2026 | 0 | 2 | 80 pass |
| Transmission | Transmission | 2026-11-17 | Oct 2026 | 0 | 0 | 67 pass |
Reproduction checks
| Check | Published | Computed | Tolerance | Result |
|---|---|---|---|---|
| The log's dated revisions: 1 applied on 17 Nov 2026 to a copy of the log (ELV is dropped from the elections theme — desk decision of 2026-10-25 (D-2026-10-25-EL2)); each is dated on or before the build, cites a decision in specs/decisions.yaml in force on that day and superseded by none dated on or before it, every row it names is in the log, and the logged rows are unedited (not applied: revision of 2026-10-24: D-2026-10-24-EL1 superseded by D-2026-10-25-EL2 from 2026-10-25; not found: none) | 0 | 0 | 0 | pass |
| Every basket's Senate beta (18 baskets and legs) is reproduced by an independent numpy least-squares fit, |Δ| ≤ 1e-09 (% per 10 points; baskets off) | 0 | 0 | 0 | pass |
| Every basket's standard error (18) is reproduced by the Newey–West formula in numpy (3 lags, Bartlett weights, no small-sample correction), |Δ| ≤ 1e-09 (baskets off) | 0 | 0 | 0 | pass |
| Each basket's daily return is the equal-weighted mean of its names' daily log returns, long leg less short leg (14 baskets; identity, 1e-09) | 0 | 0 | 0 | pass |
| Long leg less short leg equals the pair, day by day, for both Senate baskets and the policy pair (identity, max |Δ|, %) | 0 | 0 | 0 | pass |
| Shrinkage prior: τ² = var(β) − mean(se²) over the 69 single names (7.519 − 6.468); each basket's factor equals τ² / (τ² + se²) (max |Δ|) | 0 | 0 | 0 | pass |
| The funnel: every Long and Short verdict's measured beta carries the sign its lever implies (8 names; disagreements) | 0 | 0 | 0 | pass |
| The funnel's names are the ex-ante basket's and the two sector ETFs it leaves out: as logged, 29 names and 27, and as revised on 17 Nov 2026, 28 and 26 (names that differ) | 0 | 0 | 0 | pass |
| The tuned basket's names are the funnel's Long, Long-ballast and Short verdicts: as logged, eleven, and as revised on 17 Nov 2026, ten (names that differ) | 0 | 0 | 0 | pass |
| The record: each session's predicted move equals γ0·ΔS_t + γ1·ΔS_{t−1} from the published coefficients (tuned basket; identity, max |Δ|, %) | 0 | 0 | 0 | pass |
| The record: R² of the tuned basket equals 1 − SSR/SST recomputed from the residuals | 0.229 | 0.229 | 0 | pass |
| The record: R² of the ex-ante basket equals 1 − SSR/SST recomputed from the residuals | 0.1 | 0.1 | 0 | pass |
| The odds chart's marks (the FOMC decision and the CLARITY vote) are rows of the log's timeline (marks missing) | 2 | 2 | 0 | pass |
| Each window's basket return equals the sum of the basket's daily log returns inside it, both baskets, five windows (identity, max |Δ|, %) | 0 | 0 | 0 | pass |
| The windows table's odds moves equal the differences of the Senate odds series between the window's first and last named days (max |Δ|, points) | 0 | 0 | 0 | pass |
| The odds windows' moves equal the windows table's (max |Δ|, points) | 0 | 0 | 0 | pass |
| The odds: both venues' files are present, each with a daily close on at least 95% of the days since the base date (venues short of it) | 0 | 0 | 0 | pass |
| The odds: on a day only one venue trades (6 such days since the base date) the Senate series equals that venue (max |Δ|, points) | 0 | 0 | 0 | pass |
| The odds: the page's Senate series equals the two venues' average on every day both trade (max |Δ|, points) | 0 | 0 | 0 | pass |
| The odds: Kalshi's chamber odds are the sums of its joint contracts (Senate = DD + RD; the other chamber = DD + DR; max |Δ|, points) | 0 | 0 | 0 | pass |
| After the vote: every one of the four contracts on both venues is settled (Polymarket closed and resolved; Kalshi finalized; contracts not settled) | 0 | 0 | 0 | pass |
| After the vote: each venue settles exactly one of the four outcomes Yes, and both settle the same one (violations) | 0 | 0 | 0 | pass |
| After the vote: the outcome both venues settled Yes is the branch the log records (dem_sweep; violations) | 0 | 0 | 0 | pass |
| The logged scenario set (14 Oct 2026) equals the two venues' average of the four joint contracts at the close of 13 Oct 2026, normalised to 100 by the largest-remainder rule (branches that differ) | 0 | 0 | 0 | pass |
| Kalshi and Polymarket differ by no more than the 2.2-point spread the log quotes at midday 13 Oct, on the close of 2 Nov 2026 (the spread is 1.8 points; excess over the quoted bound) | 0 | 0 | 0 | pass |
| The odds tile's value equals the Senate series' latest observation | 74.6 | 74.6 | 0 | pass |
| The sample runs from 29 May 2026 (the first session with a lagged odds change after the base date) to 2 Nov 2026, the last NYSE session on or before the venues' last close (109 sessions; the trackers' dates are the sample's; violations) | 0 | 0 | 0 | pass |
| Event study: each vote's day 0 is the first NYSE session after election day, as the log dates it (votes off) | 0 | 0 | 0 | pass |
| Event study: the November 2020 Senate surprise is the IEM price change the log states (0.099 − 0.698) | -0.599 | -0.599 | 0 | pass |
| Event study (tuned basket): every window's abnormal return is the sum of the daily abnormal returns inside it, and the path at day +5 equals the first-week sum (11 votes; violations) | 0 | 0 | 0 | pass |
| Event study (tuned basket): the six-event first-week slope per 10 points is reproduced by numpy polyfit | 2.292 | 2.292 | 0 | pass |
| Event study (tuned basket): the exact permutation p equals the count over 720 orderings and lies in [0, 1] | 0.022 | 0.022 | 0 | pass |
| Event study (ex-ante basket): every window's abnormal return is the sum of the daily abnormal returns inside it, and the path at day +5 equals the first-week sum (11 votes; violations) | 0 | 0 | 0 | pass |
| Event study (ex-ante basket): the six-event first-week slope per 10 points is reproduced by numpy polyfit | 0.708 | 0.708 | 0 | pass |
| Event study (ex-ante basket): the exact permutation p equals the count over 720 orderings and lies in [0, 1] | 0.188 | 0.188 | 0 | pass |
| Event study: every permutation p in the theme and robustness tables lies in [0, 1] (values outside) | 0 | 0 | 0 | pass |
| Event study: the tuned basket's first-week robustness row equals the main six-event fit (slope per 10 points) | 2.292 | 2.292 | 0 | pass |
| Election-week test: the scoring path reproduces the 2022 midterm's first-week abnormal return in the six-vote table (%) | 12.492 | 12.492 | 0 | pass |
| Congress ETF NANC: the daily alpha equals the regression's intercept, reproduced by numpy (bp a day) | -0.258 | -0.258 | 0 | pass |
| Congress ETF GOP: the daily alpha equals the regression's intercept, reproduced by numpy (bp a day) | -2.656 | -2.656 | 0 | pass |
| Congress ETF spread: the daily alpha equals the regression's intercept, reproduced by numpy (bp a day) | 2.399 | 2.399 | 0 | pass |
| Congress ETFs: the return over SPY equals the difference of the two cumulative log returns since the base date (identity, max |Δ|, %) | 0 | 0 | 0 | pass |
| Congress ETFs: the spread's path equals NANC's excess less GOP's, day by day (identity, max |Δ|, %) | 0 | 0 | 0 | pass |
| Macro assets: every beta and standard error in the raw, oil-control and ex-release designs (17 assets × 3) is reproduced by numpy (|Δ| ≤ 1e-09; values off) | 0 | 0 | 0 | pass |
| Macro assets: each asset's R² on Brent alone and on the odds alone is recomputed by numpy (1e-9; values off) | 0 | 0 | 0 | pass |
| Macro assets: each asset's typical day equals the sample standard deviation of its daily move (values off) | 0 | 0 | 0 | pass |
| Macro assets: the smallest move the data could detect equals two standard errors of the oil-control design (values off) | 0 | 0 | 0 | pass |
| Macro assets: the oil-control design is not run on the oil assets themselves (oily assets with a Brent coefficient) | 0 | 0 | 0 | pass |
| Macro levels: each indexed series equals 100 on the base date (max |Δ|) | 0 | 0 | 0 | pass |
| Macro history: six midterm and Senate votes, each with its first-day, first-week and twenty-session moves (votes) | 6 | 6 | 0 | pass |
| Macro history: the SPY first-week slope per 10 points of surprise is reproduced by numpy polyfit | 0.25 | 0.25 | 0 | pass |
| The vote's result: the log records it for the balance-of-power set (dem_sweep), a branch of the set, with both venues' records and a source (violations) | 0 | 0 | 0 | pass |
| The vote's result: Kalshi's last Senate and House prices after the vote (4 Nov 2026) are those the logged branch pays (Senate 100, House 100; max abs difference, points) | 0 | 0 | 1 | pass |
| The macro claim after the vote: gold's and the breakeven's moves on 4 Nov 2026 are the closes' changes from 3 Nov 2026 (GLD close ratio, the Treasury par 10-year less the real 10-year; abs difference) | 0 | 0 | 0 | pass |
| Every date the tiles, takeaways, titles and falsifier print is a log row, a window bound, the base date or a data date (11 dates; stray: none) | 0 | 0 | 0 | pass |
| What each path would move: a branch Scenarios does not project, or a theme branch not on its set, carries no number here (numbers on such rows) | 0 | 0 | 0 | pass |
| What each path would move: every projected row copies the ten-year, the S&P 500 and the mortgage move of the one Scenarios branch it names (rows missing one) | 0 | 0 | 0 | pass |
| Verdict: the passage the verdict restates is in the logged claim (thesis.claim; passages not found) | 0 | 0 | 0 | pass |
| Verdict: the head is one takeaway under the grammar (≤ 25 words, ≤ 2 figures beside a date; faults: none) | 0 | 0 | 0 | pass |
| Scorecard: no judgement odds are printed — a set without a market-implied or model basis for every branch says its branches or its base case in words (rows printing a percentage) | 0 | 0 | 0 | pass |
| Headline: one conclusion of at most 14 words, not a question, with one or two figures ('Senate odds move ten policy names, not the party baskets: +7.1% per 10 points'; faults: none) | 0 | 0 | 0 | pass |
| Thesis.status is one of active, closing, resolved (now 'active'; violations) | 0 | 0 | 0 | pass |
| Active theme: no resolved block is carried before the status is resolved (fields present) | 0 | 0 | 0 | pass |
| Active theme: no closing_tracker is named before the status is closing (violations) | 0 | 0 | 0 | pass |
| Scenario set 'balance_of_power': probabilities cover every branch and sum to 100 at each of 1 dated revisions (revisions that do not) | 0 | 0 | 0 | pass |
| Scenario set 'balance_of_power': every revision carries a date, a reason and a source document with its date (revisions missing one) | 0 | 0 | 0 | pass |
| Scenario set 'balance_of_power': revisions are in date order (out of order) | 0 | 0 | 0 | pass |
| Scenario set 'balance_of_power': every branch states its basis (market-implied, model or evidence log) (branches without) | 0 | 0 | 0 | pass |
| All 6 triggers cite the issue document, section and date their threshold comes from (triggers without) | 0 | 0 | 0 | pass |
| Every data trigger has a numeric threshold (triggers without) | 0 | 0 | 0 | pass |
| All 16 timeline events carry a date, a statement and a source (events without) | 0 | 0 | 0 | pass |
| All 0 exhibit references map to a published panel, a panel declared pending, or none (unmapped) | 0 | 0 | 0 | pass |
| The Range and scenarios page: the source page is published with all its checks passing (dated 17 Nov 2026, 80 checks) | 1 | 1 | 0 | pass |
| The Range and scenarios page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Transmission page: the source page is published with all its checks passing (dated 17 Nov 2026, 67 checks) | 1 | 1 | 0 | pass |
| The Transmission page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Range and scenarios page: every copied single value equals the source page (2 values; mismatches) | 0 | 0 | 0 | pass |
| Every read point names the exhibit it is read from (`ex`; 5 points; without one: none) | 0 | 0 | 0 | pass |
| Every read point's exhibit code names a card of this page (themes.elections-…; naming another page: none) | 0 | 0 | 0 | pass |
| Decisions: every desk decision this page's specification names is in the decisions file (D-2026-10-24-EL1, D-2026-10-25-EL2; ids missing) | 0 | 0 | 0 | pass |
- Polymarket, Gamma and CLOB APIs: the 2026 Senate and lower-chamber control markets and the four balance-of-power contracts, daily price history — Source: Polymarket
- Kalshi, trade API v2: the four KXBALANCEPOWERCOMBO contracts, hourly and daily candlesticks — Source: Kalshi
- Yahoo Finance chart API: adjusted daily closes of the 81 single names and ETFs, the congressional ETFs, the macro set and the event-study proxies (since listing) — Source: Yahoo Finance
- U.S. Department of the Treasury: daily par yield curve (2-, 5-, 10-year) and daily par real yield curve (5-, 10-year); breakeven = par − real — Source: U.S. Department of the Treasury
- Kenneth R. French Data Library: Fama–French five factors and the momentum factor, daily — Source: Kenneth R. French Data Library
- Eve-of-vote market prices, 2006–2024: Snowberg, Wolfers and Zitzewitz (NBER WP 12751), the Iowa Electronic Markets, Intrade and PredictIt as reported by Ray Fair, The Block, Casino.org, davemanuel.com, Forbes.com, Asharq Al-Awsat, ElectionBettingOdds.com, and Polymarket's price history — each row with its URL in the dated file — Sources as listed per vote
- The October note and its page plan (US midterm policy baskets; the macro assets; the lever map, the baskets and the funnel; quotations of the BofA survey via Finvaulta, ETF flows read from TradingView, JPMorgan's lists via StockTwits, Quiver Quantitative, Public Citizen, OpenSecrets via Reuters) — The site's own research, Macro Review