The Probability Outlook

Universal Asset Owners · The Probability Desk

Our researched odds on the slow-moving risks that decide long-horizon returns — set side by side with what markets actually price, plus the decade-scale questions no market prices at all.

Updated 2026-08-19 · rebuilt every morning from 6,646 live market contracts and 11 Federal Reserve data series.

What this is

The Probability Desk scores the big, slow-moving risks that decide long-horizon returns — shipping chokepoints, the dollar’s reserve role, pension selling, AI capex, climate and more. For each one we publish our own researched estimate.

How to read it

The Desk is our governed estimate. The market is the nearest tradable contract. A red gap means we see more risk than the market is pricing; green means less; no market means nobody prices it — yet a universal owner still carries it.

Where it fits

One of three daily Desk views: Command Center (what’s moving now), Scenario Register (the ten standing risks), and the Outlook (this page).

Today in one minute

▸

Inflation is doing the damage. Stocks and bonds are moving together (correlation +0.51), so the classic 60/40 hedge is not protecting portfolios right now.

▸

The widest disagreement today: insurers retreat and property values reprice. The Desk says 39%; the nearest traded contracts say 6%.

▸

2 of the Desk's ten standing risks are not priced by any market. If you hold them — and a universal owner does — you are the market.

▸

From The Long Horizon: the US dollar's share of allocated global FX reserves falls below 50% by 2040 — the Desk's estimate is 38% (by 2040).

Where the Desk differs from market prices

Red gap: the Desk sees more risk than the market is pricing. Green: less. Not priced means no liquid market contract exists anywhere — the market has no view. Open The evidence on any row for the real-economy signals (energy, metals, storms, trade, fiscal cash flows) and the traded contracts behind it.

Risk	The Desk	The market	Gap (Desk − market)
Insurers retreat and property values reprice Insurance retreat to collateral repricing — uninsurability bleeds into property value The evidence Real-economy signals Live tropical cyclones: 1 — 1 active system(s): Lala (HU). None currently in the Atlantic. [NOAA NHC, live] Traded contracts [Polymarket] Will a hurricane make landfall in Georgia by November 30, 2026? — 6% [Polymarket] Will a hurricane make landfall in Maine by November 30, 2026? — 3% [Polymarket] Will no Atlantic hurricane form in August 2026? — 18% (direction-inverted) seasonal cat contracts proxy ONE season; the Desk prices the structural retreat	39%	6% median of 3 contracts · closest: “Will a hurricane make landfall in Georgia by November 30, 2026?” (6%)	+34pp
Water and food stress destabilizes sovereigns Water & food-system stress as a sovereign-stability factor The evidence Real-economy signals Wheat (IMF global price): $229/t — Wheat up 38% year-on-year — food-price stress transmits into sovereign stability with a lag. [FRED PWHEAMTUSDM (IMF), 2026-07-01] Traded contracts [Polymarket] Will Virginia reach D4 (Exceptional Drought) by August 31, 2026? — 6% thin markets; mostly unpriced by the crowd	37%	6% nearest contract: “Will Virginia reach D4 (Exceptional Drought) by August 31, 2026?” (6%)	+32pp
Minerals and grid capacity cap the energy transition Transition-mineral & grid-interconnection bottleneck caps electrification / AI The evidence Real-economy signals US natural gas (Henry Hub): $2.79/MMBtu — Spot gas up 1% over ~1 month — the marginal cost of grid power. [EIA, 2026-08-11] Copper (IMF global price): $13,543/t — Copper up 39% year-on-year — the single best price signal of electrification strain. [FRED PCOPPUSDM (IMF), 2026-07-01] Traded contracts [Kalshi] When will insulated copper cable face Section 232 restrictions? — Before Jan 1, 2027 — 14% commodity/grid price contracts proxy the bottleneck, not the multi-year constraint	43%	14% nearest contract: “When will insulated copper cable face Section 232 restrictions? — Before Jan 1, ” (14%)	+28pp
The dollar's reserve role erodes Reserve fragmentation — erosion of the dollar's exorbitant privilege The evidence Traded contracts [Polymarket] Will Gold have the best performance in 2026? — 18% [Polymarket] Will Gold (XAUUSD) hit (HIGH) $4,700 in August? — 3% [Polymarket] Will a country leave BRICS in 2026? — 11% gold-level contracts are the crowd's only liquid expression of reserve diversification	31%	11% median of 3 contracts · closest: “Will a country leave BRICS in 2026?” (11%)	+20pp
Government debt forces the central bank's hand Sovereign-debt sustainability & fiscal dominance The evidence Real-economy signals 10-year term premium: +0.84pp — Investors demand +0.84pp extra to hold long Treasuries — the market's running vote on fiscal credibility (was negative most of 2016-2021). [FRED THREEFYTP10 (NY Fed ACM), 2026-08-14] US federal interest payments: $1.25T/yr — The federal interest bill runs $1.25 trillion annualized, up 7% year-on-year — the cash-flow engine of fiscal dominance. [FRED A091RC1Q027SBEA (BEA), 2026-04-01] Traded contracts [Polymarket] Government shutdown by October 1? — 16% near-term fiscal-event contracts vs the Desk's fiscal-dominance trajectory	36%	16% nearest contract: “Government shutdown by October 1?” (16%)	+20pp
Big pension systems turn into net sellers Pension-system inversion — major retirement systems turn net sellers The evidence Traded contracts [Kalshi] Will Social Security’s retirement trust fund be depleted? — Before Jan 1, 2033 — 18% US trust-fund contracts vs the Desk's global net-seller question	36%	18% nearest contract: “Will Social Security’s retirement trust fund be depleted? — Before Jan 1, 2033” (18%)	+18pp
A major shipping-chokepoint disruption Chokepoint concentration as a standing factor (Hormuz + Taiwan + Malacca + Panama) The evidence Real-economy signals Brent crude: $93/bbl — Brent up 14% over ~1 month — the fastest-reading gauge of chokepoint stress (Desk tripwire: $110). [FRED DCOILBRENTEU, 2026-08-11] Traded contracts [Polymarket] Will China invade Taiwan by end of 2026? — 4% [Polymarket] Strait of Hormuz traffic returns to normal by August 31? — 99% (direction-inverted) [Polymarket] Strait of Hormuz traffic returns to normal by December 31? — 66% (direction-inverted) 6–18m geopolitical contracts vs the Desk's standing 3-year factor	53%	66% median of 3 contracts · closest: “Strait of Hormuz traffic returns to normal by December 31?” (66%)	-12pp
An AI capital-spending air pocket AI data-center capex air-pocket transmits to power & private credit The evidence Traded contracts [Polymarket] AI bubble burst in 2026? — 12% bust-sentiment contracts; thin and noisy	17%	12% nearest contract: “AI bubble burst in 2026?” (12%)	+5pp
The stock-bond hedge stops working Stock-bond correlation regime break — the 60/40 / LDI hedge fails	36%	no market prices this	no market
Aging populations suppress growth and rates Demographic deflation — long-run real-rate suppression ('Japanification')	32%	no market prices this	no market

The Desk = the Probability Desk’s governed estimate (sourced base rates, live signals, public calibration record — methodology). The market = median of the nearest tradable contracts on Polymarket and Kalshi; contracts are short-horizon proxies and are quoted verbatim so you can judge the fit yourself.

What markets already price

Measured from market prices, not opinion — the baseline the Desk’s view is judged against.

14%

US recession within 12 months

What the Treasury yield curve implies (NY Fed method).

Calm

Equity insurance cost

Near-term crash protection is cheaper than long-term — options markets are calm.

Credit stress percentile (10 years)

High-yield spreads at 2.70% — corporate credit is priced for calm.

2.30%

10-year inflation expectations

Long-run expectations sit +0.33pp from the 2% anchor.

The Long Horizon

Markets price the next 18 months. A sovereign portfolio holds for 30 years. These are the Desk’s odds on the questions that horizon actually owns — anchored to the historical record, reviewed quarterly, never silently rewritten.

38%

The US dollar's share of allocated global FX reserves falls below 50% by 2040

by 2040 · on the daily register at 31%

The evidence

History says: USD reserve share has declined from ~71% (2000) to ~58% (2025), roughly -0.5pp/yr; a fall below 50% by 2040 requires the trend to persist without acceleration (IMF COFER (Currency Composition of Official Foreign Exchange Reserves), quarterly)
Why it matters now: Reserve diversification into gold and non-traditional currencies accelerated after 2022 sanctions; central-bank gold buying is at multi-decade highs.
Latest research review (2026-08-13): Deep-research review 2026-08-13: COFER USD share ROSE to 57.13% in 2026Q1 (IMF data brief, 1-Jul-2026) from a revised 56.42% in 2025Q4 — a higher base than at the June review, and the IMF attributes about half the rise to dollar appreciation, i.e. valuation, which does not compound. Reaching <50% by end-2040 now needs 0.48pp/yr against a realized ~0.52pp/yr (OMFIF, 15-Jul-2026, which projects ~52% by 2036). The prior note's '0.87pp/yr' could NOT be verified on the post-Nov-2025 COFER vintage — treat as stale. Central-bank gold, the main diversification channel, slowed to 345t in H1-2026, the weakest first half since 2022 (WGC, 30-Jul-2026). Euro fell to 20.03%; RMB 1.99%.
What would change our mind: IMF COFER (quarterly, late Mar/Jun/Sep/Dec): two CONSECUTIVE quarters print USD below 55.50% (currently 57.13%) — tightened from the old single-quarter <55% trigger because single prints are valuation noise and revise by up to 0.35pp; IMF COFER (quarterly): 'other currencies' resumes rising and exceeds 7.00% in two consecutive quarters (currently 6.18% and falling after seven quarters of increases) — replaces the old RMB >3.5% trigger, which at 1.99% is too far out of the money to inform; WGC Gold Demand Trends full-year edition (late Jan/early Feb): central-bank net purchases above 1,000t for 2026 or 2027 raises the probability; below 700t lowers it (H1-2026 annualises below the lower band).

33%

Stock-bond correlation averages POSITIVE across the 2030s — the diversification regime of 2000-2020 does not return

decade of the 2030s · on the daily register at 36%

The evidence

History says: US stock-bond correlation was positive for most of the 1970s-1990s (inflation-dominant regimes) and negative 2000-2020 (demand-shock regimes); regimes have historically persisted for decades once established (Long-run return datasets (Shiller; AQR 'A Century of Stock-Bond Correlations'))
Why it matters now: Every 60/40 portfolio and LDI hedge built since 2000 embeds the negative-correlation assumption; supply-driven inflation breaks it.
Latest research review (2026-06-14): Most recent multi-year window is positive: AQR (Apr 2026) puts the Bloomberg US Aggregate at a +0.26 equity beta over the 5yr to Feb 2026; BlackRock (Jul 31 2025) calls the positive 12-mo S&P/Agg correlation 'persistent' and says the regime 'may endure' on sticky inflation, fiscal deficits ($38.6T debt, CBO 2026) and a rising term premium. But PIMCO (2025/26) notes the 90-day reading has reverted toward zero/slightly negative as inflation moderates. AQR/RBA show correlation regimes persist for decades, yet the current positive regime is only ~4yrs old and a full-decade positive average is demanding — one disinflation or flight-to-quality episode flips the sign. Net: modest uplift.
What would change our mind: Raise toward 45% if 36-mo rolling S&P 500 / US Treasury correlation stays >= +0.2 through end-2028; Cut toward 20% if core PCE holds <= 2.2% for 8+ consecutive quarters AND rolling 1-yr correlation turns negative by end-2027; Cut toward 22% if the 10-yr term premium (NY Fed ACM) falls back below +0.25% for a sustained year.

50%

G7 public retirement systems in aggregate become persistent net SELLERS of financial assets by 2035

by 2035 · on the daily register at 36%

The evidence

History says: Japan's GPIF and several European pay-as-you-go reserves already flipped to structural drawdown in the 2010s-20s as old-age dependency ratios passed ~35%; the US, UK, Canada cohort crosses comparable thresholds 2030-2040 (OECD Pensions at a Glance; UN World Population Prospects dependency ratios)
Why it matters now: The marginal buyer of duration and equities for 40 years becomes a seller; no prediction market prices it because no contract spans it.
Latest research review (2026-06-10): Deep-research review 2026-06-10: US OASI depletion moved EARLIER to Q4-2032 (2026 Trustees Report, Jun-9-2026) — ~$2.5T of mechanical Treasury redemptions now under way; Japan GPIF in framework drawdown since ~FY2009; Germany's reserve ~1 month of outlays. Counterweights: Canada CPP a net buyer to 2030/2057, France AGIRC-ARRCO in surplus (+EUR1.4bn 2025), and post-depletion the US has nothing left to sell — 'persistent at 2035' hinges on US legislation. Near coin-flip.
What would change our mind: 2027 Trustees Report moves OASI depletion into 2031 or earlier; GPIF AUM falls two consecutive fiscal years net of returns; AGIRC-ARRCO technical result turns negative.

40%

At least 5% of US housing stock is effectively uninsurable in the private market by 2035

by 2035 · on the daily register at 39%

The evidence

History says: California FAIR Plan and Florida Citizens policies-of-last-resort have roughly tripled since 2018; state-backed residual pools already cover several percent of the two largest cat-exposed states (State residual-market enrolment data (CA FAIR Plan, FL Citizens); Swiss Re sigma cat-loss series)
Why it matters now: Insurance is the repricing mechanism for climate risk; when cover withdraws, collateral values follow — directly into pension and SWF real-asset books.
Latest research review (2026-06-14): 5% of ~148M US housing units (Census/FRED 2025) is ~7.4M. CA FAIR Plan hit 668,609 residential policies Q4 2025 (Consumer Watchdog, Jan 2026), +165% since 2021; exposure ~$650B (+289% since FY2020, Kennedys 2025). But FL Citizens shed ~541,000 policies in 2025 to 395,144 entering 2026 (Florida Realtors, Jan 2026) — residual pools CAN shrink fast under reform/reinsurance softening (Artemis). Treasury FIO (Jan 2025) confirms non-renewals concentrate in top climate-risk ZIPs; Swiss Re sigma (2026) logs a 6th straight >$100B cat-loss year. Two biggest pools combined are still <1% of national stock; reaching 5% needs ~7x growth spreading to a third region. Definitional caveat: FAIR/Citizens are still insurance. Plausible 2035 range 2-6%, straddling the line.
What would change our mind: Raise toward 55% if CA FAIR exceeds ~1.0M residential policies before 2030 AND a second non-CA/FL residual pool (TX FAIR or LA Citizens) grows >50% YoY for two consecutive years; Raise if a Treasury FIO / NAIC release reports combined national residual-pool + bare-home share crossing 3% before 2030; Lower toward 25% if FL Citizens stays below 500k through 2027 AND CA FAIR policy growth turns negative for four consecutive quarters post-reform.

22%

A cross-strait military conflict (blockade or invasion) occurs by 2035

by 2035 · on the daily register at 53%

The evidence

History says: Great-power territorial flashpoints with standing military build-ups have escalated to armed conflict in roughly 1 in 4 comparable 10-year windows since 1945 (Korea, Suez, Falklands, Ukraine); deterrence held in the rest (Correlates of War project; desk reference-class assembly (ADR-0103 discipline))
Why it matters now: The crowd prices 2026-contract odds at ~7%; the decade-cumulative number — the one a 30-year portfolio actually carries — is several times that.
Latest research review (2026-06-21): Deep-research review 2026-06-21: DoD 2025 China Military Power Report (Dec-2025) — PLA on track to be CAPABLE of a Taiwan campaign by 2027 (capability, not intent); 600+ nuclear warheads. INDOPACOM's Adm. Paparo (Feb-2025): tempo so high an 'exercise could hide operational warning.' US intelligence (reported Mar-2026): Beijing holds no fixed timeline and views amphibious assault as very hard given likely US intervention. Polymarket invade-by-2027 ~6% (1-yr, invasion-only). CSIS 2023 wargames show invasion likely FAILS at catastrophic PLA cost; Rhodium/Atlantic Council put blockade disruption >$2T; EIU calls direct assault 'very unlikely.' Reference class (post-1945 great-power flashpoints): sustained-conflict escalation ~10-20%/decade, upper end higher once blockade counts. Net: trim slightly below prior.
What would change our mind: PLA conducts a declared, sustained (>72hr) quarantine/inspection regime around Taiwan, or mass amphibious-lift mobilization is visible on commercial satellite; Polymarket/Metaculus near-term invasion contract sustains >20% for 60+ days; Explicit US abandonment signal — a formal reversal of arms-sales commitments or a US-PRC communique constraining support for Taiwan.

30%

A G7 central bank is forced into explicit yield-curve control or debt-monetization for FISCAL reasons by 2035

by 2035 · on the daily register at 36%

The evidence

History says: Japan ran explicit YCC 2016-2024 at ~260% gross debt/GDP; the US, UK, France and Italy are converging on the debt-service share of revenue at which fiscal pressure historically dominates monetary policy (financial-repression episodes 1945-1980) (BIS debt statistics; Reinhart & Sbrancia, 'The Liquidation of Government Debt')
Why it matters now: Fiscal dominance is the single largest unhedgeable exposure in a sovereign-bond-heavy portfolio.
Latest research review (2026-06-21): Deep-research review 2026-06-21: IMF Fiscal Monitor (Apr-2026) — Japan debt/GDP ~204%, US 126% (->142% by 2031), France/Italy >100%. CBO (Feb-2026): US net interest 19% of revenues in 2026 -> 37% by 2056; $1.0T -> $2.1T by 2036. 30y JGB hit a record 3.88% (Jan-2026); Fitch cut France to A+ (Sep-2025), S&P to A+ (Oct-2025). Reference class (Reinhart & Sbrancia): post-WWII financial repression was the norm, liquidating ~3-5% GDP/yr. BUT the bar is EXPLICIT + FISCAL: the BoJ ENDED YCC (Mar-2024) and is hiking/tapering — walking away from caps; BoE/ECB backstops (2022 LDI, TPI) are framed as monetary/financial-stability, dodging the bar. Direction of travel is adverse; the strict conjunction caps it. Trim slightly below prior.
What would change our mind: Any G7 central bank announces an explicit long-yield ceiling/target or unlimited purchases citing a debt-service/fiscal rationale (not inflation); US net interest crosses 25% of federal revenues (CBO actuals) before 2032, OR 30y JGB/UST sustains >5-6% with the central bank resuming uncapped buying; ECB activates TPI or France-specific support on stated debt-sustainability/fiscal grounds rather than monetary transmission.

30%

US trend productivity growth sustains above 2.5%/yr for 5+ consecutive years before 2035 (the AI dividend materializes at macro scale)

before 2035 · on the daily register at 17%

The evidence

History says: In the postwar US record only one five-year span cleared 2.5% sustained — the 1995-2004 IT diffusion wave; general-purpose technologies have taken 10-20 years from invention to measured productivity (BLS multifactor productivity series; Brynjolfsson 'productivity J-curve' literature)
Why it matters now: Whether the AI capex boom earns its cost of capital is the central asset-allocation question of the decade; the macro data, not the narrative, will answer it.
Latest research review (2026-06-28): Deep-research review 2026-06-28: BLS Productivity & Costs Q1-2026 (May-7-2026) — nonfarm productivity +0.3% annualized Q1 but +2.8% y/y; current cycle (Q4-2019->Q1-2026) 2.1% annualized. BLS TFP (Mar-19-2026): 2025 labor productivity 2.2% (down from 3.0% in 2024), TFP +0.8%. Only postwar >2.5% five-year run = the 1995-2004 IT wave (~2.7%; NY Fed/SF Fed). Too-low: Goldman (2025) GenAI scenario +1.5pp/decade, McKinsey (2023) $17-26T. Too-high: Acemoglu (NBER w32487, 2024) AI TFP ~0.06%/yr; MIT GenAI Divide (Aug-2025) 95% of pilots zero return; BofA (Dec-2025) 2026 capex air-pocket (~$725B). Net: hold near 0.30, recommended 0.28 on 2025 deceleration (within 3pp -> desk_p unchanged).
What would change our mind: Raise toward 0.45 if BLS nonfarm productivity prints >=2.5% y/y for 4+ consecutive quarters through end-2027; Raise if BLS annual TFP growth re-accelerates above 1.5% in the 2026 or 2027 release (vs 0.8% in 2025); Lower toward 0.20 if 2026-27 hyperscaler capex cuts >20% YoY (BofA 'air pocket' realized) OR repeat MIT/Census evidence of <10% enterprise AI ROI.

30%

The euro area averages nominal GDP growth below 2%/yr across the 2030s ('Japanification' completes)

decade of the 2030s · on the daily register at 32%

The evidence

History says: Japan averaged ~0.4% nominal growth 1995-2015 after its working-age population peaked; the euro area's working-age population peaked ~2010 and its old-age dependency ratio tracks Japan's with a ~15-year lag (UN World Population Prospects; IMF WEO long-run series)
Why it matters now: A repeat compresses long-run real rates and re-rates every euro-denominated liability stream — the quiet scenario nobody trades.
Latest research review (2026-06-28): Deep-research review 2026-06-28: Eurostat (Mar-6-2026) euro-area real GDP +1.4% in 2025; ECB staff (Jun-12-2026) real 0.8%/1.2%/1.5% (2026-28) and HICP 3.0%/2.3%/2.0% -> ~3.2-3.5% nominal near-term. EC (Autumn-2025) potential growth 1.4%->1.2% by 2027; working-age population peaked ~2010 (~15yr lag to Japan). Too-low: Draghi report (Sep-9-2024) 'existential' stagnation risk; labour force -20% by 2070. Too-high (decisive): the bar is NOMINAL -- if the ECB hits ~2% HICP, even 0-1% real = 2-3% nominal; Germany EUR500bn fund + debt-brake reform (Mar-2025), NATO 3.5% (Jun-2025), migration (foreign-born 12.6% of EU working-age, 2024). A full decade sub-2% nominal needs stagnation AND chronic inflation undershoot. Trim 0.35->0.30.
What would change our mind: Raise if the ECB HICP projection for the 3rd-out year falls below 1.5% in two consecutive quarterly rounds (2026-2028); Raise if Germany's EUR500bn infrastructure fund is <40% disbursed by end-2029 (Bundesbank/EC); Raise if the European Commission euro-area potential-growth estimate falls below 1.0% in any Spring/Autumn forecast.

70%

Combined sovereign-wealth-fund and public-pension AUM exceeds USD 75 trillion by 2035

by 2035 · on the daily register at 36%

The evidence

History says: The pool grew from ~USD 30T (2010) to ~USD 58-62T (2025), ~5%/yr through two drawdowns; USD 75T by 2035 needs only ~2.5%/yr — but pension-system drawdown (see LO-PENSION-NETSELLER) cuts both ways (Global SWF annual rankings; Thinking Ahead Institute Global Pension Assets Study; UAO Registry data)
Why it matters now: The growth of the universal-owner pool itself — who owns the future market — is the meta-question of this publication.
Latest research review (2026-07-12): Deep-research review 2026-07-12: Global SWF Annual Report (Jan-1-2026) — SWFs crossed USD 15T (Dec-2025); combined SWF+public-pension+central-bank AUM ~USD 60T, projected USD 80T by 2030. Thinking Ahead Institute GPAS 2026 (Feb-9-2026): P22 pension assets a record USD 68.3T at YE2025 (+9.6% y/y). From ~USD 60T, reaching USD 75T by 2035 needs only ~2.3%/yr vs ~5%+ realized — Global SWF's own path clears USD 75T mid-decade. Counterweights: the USD 60T includes central banks (double-counting inflates the base), mature-DB decumulation (Netherlands/UK/Japan net sellers), a flat-real return decade, and USD strength deflating ~60% non-USD AUM. Recommended 0.70 (from 0.55).
What would change our mind: Global SWF Annual Report (each Jan): combined SWF+pension+central-bank AUM prints below ~USD 66T by Jan-2030 (trailing CAGR <2%) — downgrade; on-projection >= USD 80T — raise toward 0.85; TAI/WTW Global Pension Assets Study (early Feb): P22 pension assets print below ~USD 70T at YE2028 (stall from USD 68.3T) — demographic drawdown biting, downgrade; Fed Broad Dollar Index sustains >15% appreciation over any 24-month window — flag as a headline-only (non-real) risk to the USD-denominated threshold.

12%

A binding carbon price of at least USD 100/tCO2 covers more than 50% of global emissions by 2040

by 2040 · on the daily register at 43%

The evidence

History says: Carbon pricing covered ~24% of global emissions in 2024-25, with a volume-weighted average price far below USD 100; coverage has grown ~1pp/yr over the last decade (World Bank State and Trends of Carbon Pricing (annual))
Why it matters now: Transition-risk models embedded in TCFD scenarios quietly assume carbon prices the political record does not support — the gap is itself a mispricing.
Latest research review (2026-07-12): Deep-research review 2026-07-12: World Bank State and Trends of Carbon Pricing 2026 (~Jun-2026) — direct pricing covers ~29% of global GHGs (87 instruments), up from ~24% (2023); under 0.5% of emissions priced above USD 100/tCO2; global average price ~USD 21/t. ICAP Status Report 2026 (Apr-2026): 41 ETSs cover 26%. EU ETS ~EUR 80-95, China national ETS ~USD 12-14, California ~USD 30. Coverage grew ~1pp/yr, so linear extrapolation reaches only ~44% by 2040 — short of 50% at ANY price. Too-low: EU ETS toward EUR 100+, ETS2 (2027), CBAM, China cap. Too-high: the punishing USD 100 AND 50% conjunction, US/India unpriced, political backlash. Trim 0.15->0.12.
What would change our mind: World Bank State and Trends reports direct-pricing coverage above 40% before 2032 (currently ~29%) — raise; World Bank reports more than 10% of global emissions priced at >= USD 100/tCO2 (currently <0.5%), OR EU ETS sustains > EUR 100 for a full year (ICE/EEX) — raise; A US federal carbon price OR an India national ETS at >= USD 50/t enters force (ICAP factsheets / official gazette) — raise.

22%

An interstate armed conflict in which water access is a primary driver occurs by 2040

by 2040 · on the daily register at 37%

The evidence

History says: No modern war has been fought PRIMARILY over water, but water has been a contributing driver in multiple conflicts (Indus standoffs, Nile/GERD escalation, Tigris-Euphrates tensions); basin-stress projections put 5+ major shared basins into severe deficit by the 2030s (Pacific Institute Water Conflict Chronology; WRI Aqueduct basin projections)
Why it matters now: Water stress transmits into sovereign credit, food prices, and migration — three channels every long-horizon portfolio carries.
Latest research review (2026-08-13): Deep-research review 2026-08-13 (first full re-underwrite): UCDP/Uppsala (9-Jun-2026, Journal of Peace Research 63:4) records 8 interstate conflicts in 2025 — most since 1946 — but ZERO with water as the coded incompatibility in 80 years. Pacific Institute Water Conflict Chronology (11-Nov-2025) logs 2,750+ events, 420 in 2024 (+18%), yet 63% intrastate. India put the Indus Waters Treaty in abeyance (23-Apr-2025; reaffirmed 3-Jul-2026) and rejected the Hague award of 15-May-2026; Ethiopia inaugurated GERD 9-Sep-2025 with no tripartite deal. Against that, Turkiye–Iraq raised flows under the Apr-2024 framework. The 'primary driver' AND 'interstate' bars are the binding constraint, not water stress.
What would change our mind: Pacific Institute Water Conflict Chronology (annual, November): interstate share of events exceeds 50% in any year, or any entry codes a state-vs-state exchange with 10+ fatalities as water-triggered; UCDP annual release (June): any new interstate dyad enters the dataset whose stated incompatibility references a river, dam, or water allocation rather than territory or government; India moves the Indus Waters Treaty from abeyance to formal withdrawal, or begins western-river storage beyond treaty limits — the act Pakistan has publicly called an act of war.

How to read this page. The Desk is the Probability Desk’s governed estimate — sourced base rates, live data feeds, and a public scoring record (methodology). The market is what short-horizon market contracts (Polymarket, Kalshi) and bond and options markets imply — used strictly as a comparison object, never as a source of our estimates. The Long Horizon questions are decade-scale estimates no market prices. Metaculus is closed to automated access and is therefore not included — we say so rather than approximate it. Part of the Probability Desk system: Command Center · Scenario Register · Scenario Lab. Editorial analysis for long-duration capital — not investment advice.