Tools chosen: kalshi_direct wikipedia claude_news code_execution kalshi_related polymarket_related

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-horizon geophysical base-rate question with essentially no news-driven signal; the answer hinges on published seismic hazard forecasts (UCERF3) and the historical record of large California earthquakes. Primary anchor is the Kalshi price, supplemented by USGS/UCERF3 30-year M8+ probabilities converted to a ~10-year window via Poisson math. Cross-venue checks are low-yield but cheap.

# Current state Resolution requires a USGS-recorded M8.0+ earthquake epicentered in California (or its territorial waters) before 2035-12-31. No such event has ever been recorded in California's ~175-year instrumental/historical record (largest: 1857 Fort Tejon M7.9, 1906 SF M7.9). USGS UCERF3 puts the 30-year (from ~2015 baseline) probability of M8.0+ at ~7%. Kalshi currently prices this contract at 25% YES, far above model-implied base rates (~2-4%). # Event Will a magnitude ≥8.0 earthquake with epicenter in California (or territorial waters) occur before Dec 31, 2035? # Outcomes to forecast - Yes - No # Kalshi market anchor KXEARTHQUAKECALIFORNIA-35 current YES price: **25.00%**. 7-day change: -5pp; 30-day change: +1pp. Range since listing: 7%-65% (high volatility/thin market). Avg daily volume: 129 contracts over 47 data points — thin, likely retail-driven, prone to mispricing. [kalshi_direct] # Sub-question answers 1. **Kalshi price/trading** — Currently 25%, down 5pp over 7 days but +1pp over 30 days; has ranged from 7% to 65% historically, indicating a volatile, low-liquidity market not yet converged. [kalshi_direct] 2. **Largest recorded CA earthquake** — No M8.0+ has ever been recorded in California; largest are 1857 Fort Tejon (M7.9) and 1906 San Francisco (M7.9). [wikipedia] 3. **UCERF3 30-yr probability** — ~7.0% (up from 4.7% in UCERF2), driven mainly by southern San Andreas multi-fault rupture scenarios. [claude_news, CA Dept of Conservation/SCEC] 4. **Poisson probability for ~10-yr window to 2035** — Central estimate ~2.4% (λ≈0.00242/yr from UCERF3 7%/30yr); sensitivity range 1-3.5% across plausible P30 inputs; historical-record-based estimates (Jeffreys prior) converge near 2.8%; conservative rule-of-three upper bound ~15.75%. [code_execution] 5. **Fault capability for M8+** — San Andreas alone has a realistic ceiling near M8.3 (not M9-capable; insufficient length/depth); multi-fault rupture models (San Andreas + San Jacinto + others) are precisely why UCERF3 raised long-term M8+ odds from 4.7% to 7%. [claude_news, Quake Trace, SCEC] 6. **Recent 2024-2025 activity/updates** — No M8+ event; largest recent CA quake was M4.2 near Cloverdale. 2023 USGS National Seismic Hazard Model update incorporated new fault/deformation models but no evidence found of a statewide M8+ probability revision beyond UCERF3's ~7%. [claude_news, USGS, EarthquakeTracker.org] 7. **Resolution definitions** — No explicit rules text found on Mw finality or territorial-waters/Cascadia inclusion; description says "epicenter in California or its territorial waters," suggesting offshore near-coast events count but the Cascadia/Mendocino triple junction proper is largely offshore Oregon/Washington-adjacent, ambiguous edge case not resolved by available research. # Key facts (high-confidence, factual) 1. [wikipedia] No M8.0+ earthquake has ever been instrumentally or historically recorded with epicenter in California; largest are M7.9 (1857, 1906). 2. [claude_news/CA Dept Conservation] UCERF3 30-year M8+ probability ≈7.0%, up from UCERF2's 4.7%, due to multi-fault rupture modeling. 3. [claude_news] San Andreas fault's realistic single-fault ceiling is ~M8.3; standard planning scenario (ShakeOut) uses M7.8. 4. [claude_news] Cascadia subduction zone (capable of M9) only marginally touches far Northern California (Cape Mendocino) but is a distinct hazard system. 5. [code_execution] Poisson-derived 10-year probability from UCERF3 base rate ≈2.4%, with historical-based estimates (Jeffreys) ≈2.8%. 6. [kalshi_direct] Current market YES price 25%, roughly 10x higher than model-based central estimates. # Cross-market signals - Kalshi related: KXEARTHQUAKECALIFORNIA-28 (shorter horizon, before 2028) trades at just 8%, consistent with a much shorter time window; the 2035 market at 25% implies a large jump in probability for the ~7 extra years, inconsistent with linear/Poisson scaling (28-market ~2.7yr window at 8% vs 2035 ~10yr window at 25% — a roughly 3x time increase implying only ~3x probability increase under Poisson, i.e., ~22-24%, actually roughly consistent, though 28-market itself appears overpriced vs ~1% Poisson estimate too). - Japan M8+ before 2030 market prices 40% — different tectonic regime (subduction zone, historically far more M8+ events), not directly comparable but shows category-wide retail tendency to overprice tail earthquake risk. - Polymarket: no matching markets found. # Analyst opinions and speculation - Quake Trace: emphasizes San Andreas is NOT capable of M9, capping realistic CA scenarios near M8.3, and that official planning uses M7.8 as the working "Big One" scenario — implying M8.0+ is the extreme tail even within worst-case rupture models. - claude_news synthesis: "implied probability of M8+ CA earthquake by 2035 remains low (order of a few percent)," explicitly flagging the market may be overpriced relative to USGS-based models. # Directional lean per outcome - **Yes**: Supported by UCERF3 showing elevated (vs. earlier models) 7%/30yr risk from multi-fault ruptures; San Andreas physically capable of M8.0-8.3; long fault system with significant slip deficit since 1857/1906. - **No**: Strongly supported — zero M8.0+ events in ~175 years of record; Poisson/base-rate models give only ~2-4% for the ~10-year window; standard hazard planning scenarios use M7.8, below threshold; current Kalshi price (25%) appears substantially above model-implied probability. # Gaps / unknowns - No official USGS/UCERF sub-period (10-year, 2026-2035) estimate found; all 10-year figures are extrapolated/approximated. - No clarity on whether "territorial waters" includes Mendocino triple junction/Cascadia-adjacent offshore events, which could pull in a higher-magnitude-capable zone. - No confirmation of a UCERF4 or updated statewide M8+ probability post-2023 NSHM. - Market thinness (129 avg daily volume, wide 7-65% historical range) suggests price may reflect limited informed trading rather than calibrated consensus. # Calibration anchors - Kalshi current YES price (anchor): **25%** - Model-based central estimate (UCERF3 Poisson, 10-yr): **~2-4%** - Conservative statistical upper bound (rule-of-three, small-sample): **~16%** - Related Kalshi market (before 2028, shorter horizon): **8%** - No historical precedent of M8.0+ in California ever recorded (175-year base rate = 0/175).

1. Both forecasts anchor almost entirely on the UCERF3 Poisson extrapolation (~2-4%) and barely engage with the rule-of-three upper bound (~16%) or the fact that 0/175 historical base rate with small-sample Bayesian priors (Jeffreys ~2.8%) could understate risk given the short/biased historical record (only ~175 years vs. multi-century recurrence intervals implied by slip-deficit arguments); neither forecast meaningfully grapples with why the "conservative" 16% estimate should be discounted so heavily. 2. Neither forecaster addresses the cross-market inconsistency flagged in the brief: the 2028 market prices 8% for a 2.7-year window while the 2035 market prices 25% for a 10-year window—implying an odd super-linear scaling under Poisson assumptions that neither Poisson estimate (~1% and ~2.4%) fits, suggesting either both markets are irrationally priced or the resolution/definition (territorial waters, Cascadia edge cases) materially differs between them in ways not modeled. 3. The "territorial waters" ambiguity regarding the Mendocino triple junction/Cascadia zone (capable of M9) is flagged as an unresolved gap in the brief, yet both forecasts treat it as a minor tail adjustment rather than a potentially significant probability mass—if Cascadia-adjacent offshore events count, the effective probability could be meaningfully higher than the San Andreas-only estimate. 4. Both forecasters converge to very similar point estimates (4% and 7%) despite claiming independent reasoning, and neither explores why their confidence levels (0.76 vs 0.68) differ so much given nearly identical evidence use—this convergence without engaging tail scenarios (e.g., an unmodeled fault, historical record too short to capture rare M8+ events) suggests insufficient exploration of distributional uncertainty. 5. Neither forecast explicitly reconciles why the market (25%) has stayed anomalously high and volatile (ranging 7-65%) over time rather than converging toward the ~2-4% "true" model estimate if the market were simply irrational noise—persistent divergence could indicate the market/traders possess information (e.g., updated NSHM 2023 fault models) not fully captured in the brief's gaps section, which explicitly notes "no confirmation of a UCERF4 or updated statewide M8+ probability post-2023 NSHM." 6. Both forecasts treat the standard ShakeOut M7.8 planning scenario as strong evidence for a hard ceiling near M8.0-8.3, but this scenario was chosen for infrastructure planning purposes (conservative/likely scenario), not as a statement of maximum physical possibility—overreliance on this heuristic may anchor both forecasts too low relative to genuine tail risk from complex multi-fault ruptures.

5% (72% confidence)

Yes: 5% No: 95%

The scientific anchor is clear: UCERF3 puts M8.0+ at ~7% per 30 years, which Poisson-scales to ~2.4% over the ~10-year window, and a Jeffreys-prior treatment of the 0/175-year historical record converges near 2.8%. California's largest recorded events (1857, 1906) were M7.9, and even multi-fault rupture models cap realistic scenarios near M8.3, so M8.0+ is genuinely an extreme tail rather than a modal 'Big One' outcome. The 25% Kalshi price is ~10x model estimates and comes from a thin (129 contracts/day), wildly volatile (7-65%) market; long-dated contracts also structurally inflate YES prices because NO holders must lock capital for a decade, which plausibly explains the persistent divergence the critique flags without implying private information. I take the critique's points on record-length bias, unmodeled faults, and the Mendocino/territorial-waters edge case seriously, nudging above the pure Poisson central estimate toward 5%, but the rule-of-three ~16% bound is a conservative statistical ceiling, not a best estimate, and does not justify anything close to the market price.

Total pipeline time: 106.3s