Tools chosen: kalshi_direct kalshi_related polymarket_related claude_news wikipedia code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-horizon geophysical question (10 years) about a rare extreme event. The key drivers are the historical base rate of M8.0+ earthquakes in California and the seismological consensus on maximum credible magnitudes for California faults (notably the San Andreas). The market price anchors consensus, and base-rate math from USGS data is the core driver.

# Current state The market resolves YES if a confirmed M8.0+ earthquake occurs with epicenter in California or its territorial waters before Dec 31, 2035. No such event has occurred; the question is purely forward-looking with ~9.5 years remaining. # Timeline of key events - **1857-01-09** (confirmed): Fort Tejon earthquake ~M7.9 on southern San Andreas — largest historical CA event, sub-M8.0 - **1906-04-18** (confirmed): San Francisco earthquake ~M7.9 on northern San Andreas — also sub-M8.0 - **2008** (confirmed): UCERF2 published; 30-yr M8.0+ probability ~4.7% statewide - **2015** (confirmed): UCERF3 published; 30-yr M8.0+ probability raised to ~7%; M8+ recurrence interval ~494 yrs [Wikipedia/SCEC] - **2025 (present)**: No M8.0+ has ever been instrumentally confirmed in California; southern San Andreas remains >300 yrs since last major rupture --- # Event Will an M8.0+ earthquake occur in California (or its territorial waters) before Dec 31, 2035? # Outcomes to forecast - **Yes** (M8.0+ occurs before 2035) - **No** (no M8.0+ before 2035) # Kalshi market anchor **KXEARTHQUAKECALIFORNIA-35 current YES price: 26%** - 7-day change: 0%; 30-day change: −2% (mild downward drift) - Avg daily volume: 46 contracts (low liquidity) - Price range over 48 days: 24%–65% (wide range suggests early price discovery) - Related: KXEARTHQUAKECALIFORNIA-28 (before 2028) = **16%** [Kalshi-related] # Sub-question answers 1. **Kalshi market price for M8.0+ CA before 2035?** — 26% YES as of current data; 30-day drift −2%. [Kalshi-related] 2. **Has CA ever recorded M8.0+? Max credible magnitude?** — No confirmed M8.0+ in instrumental record; largest are 1857 Fort Tejon and 1906 SF, both ~M7.9. San Andreas capable of up to ~M8.3 per computer models. [earthquakecountry.org, Wikipedia] 3. **USGS M7.0+ and M8.0+ 10-year probability?** — UCERF3 gives ~7% over 30 years for M8.0+; implying ~2.3–2.5% per 10-year window under Poisson assumption. M7.0+ probability is substantially higher (~60% per UCERF3 for 30 yrs). [SCEC/UCERF3] 4. **Historical annual base rate for M8.0+ in CA?** — UCERF3 recurrence ~494 years → ~0.002/yr → ~2% per decade. Gutenberg-Richter scaling (b=1.0, M7+ rate = 1/10yr) implies ~100-yr recurrence → ~9.5% per decade, but this likely overstates. [Wikipedia UCERF3, code_execution] 5. **Recent UCERF3 developments?** — UCERF3 (2015, still current official model) raised M8.0+ probability vs UCERF2 by incorporating multi-fault ruptures. Southern San Andreas is primary candidate, >300 yrs since last major rupture. No UCERF4 published yet. [SCEC, Wikipedia] # Key facts (high-confidence, factual) 1. [UCERF3/SCEC] 30-yr probability of M8.0+ in CA = ~7%; implies ~2.3–2.5% per 10-yr Poisson window 2. [Wikipedia/UCERF3] M8.0+ recurrence interval in CA = ~494 years 3. [earthquakecountry.org] San Andreas max credible magnitude ~8.3; no confirmed M8.0+ in California's recorded history 4. [code_execution] Poisson model across 200–400 yr recurrences yields 2.5%–4.9% for ~10-year window; blended ~3.5% 5. [code_execution] G-R scaling (b=1.0) gives ~9.5% — likely upper bound due to regional magnitude ceiling effects 6. [Kalshi-related] Before-2028 market at 16%; before-2035 at 26% — implied 2028–2035 incremental probability ~10pp per ~7 years # Cross-market signals - **Kalshi (KXEARTHQUAKECALIFORNIA-28, before 2028):** 16% YES — implies market assigns ~10pp for the 2028–2035 window - **Kalshi (KXEARTHQUAKEJAPAN-30, before 2030):** 65% YES — Japan's subduction zone risk is far higher; useful calibration contrast - **Polymarket:** No active markets found [polymarket_related] - **Sportsbook:** None identified # Analyst opinions and speculation - Kalshi market appears to significantly overstate scientific base rate (26% vs ~3–6% scientifically derived), likely due to low liquidity and unsophisticated traders [low volume: 46 contracts/day] - The wide historical price range (24%–65%) suggests the market has not converged to a stable estimate - Scientific consensus (UCERF3) anchors the probability firmly in the 2–7% range for a 10-year window # Directional lean per outcome - **Yes (M8.0+ before 2035):** Supporting — southern San Andreas overdue (>300 yrs), UCERF3 multifault rupture scenarios, G-R scaling upper bound ~9.5%. Opposing — no confirmed M8.0+ in CA history, 494-yr average recurrence, Poisson base rate ~3–5%, 9.5-yr window is short - **No:** Strongly supported by base rates (94–97% of any given decade sees no M8.0+), historical record, and UCERF3 central estimates # Gaps / unknowns - No UCERF4 update yet; UCERF3 (2015) is a decade old - Cascadia subduction zone could theoretically trigger San Andreas cascade, but this is speculative - Resolution criterion: "territorial waters" scope is unclear for offshore faults # Calibration anchors - **Kalshi current YES price: 26%** — appears substantially inflated vs scientific base rate - **UCERF3 Poisson 10-yr estimate: ~2.3–2.5%** (official lower anchor) - **Blended Poisson/G-R model: ~3.5–6%** (reasonable range) - **Precedent:** Japan M8.0+ before 2030 priced at 65% — reflects subduction zone rates ~10× higher than CA strike-slip

1. Both forecasters heavily discount the Kalshi anchor (26%) without adequately justifying why the market is so far above scientific estimates. The brief notes a price range of 24%–65% over 48 days with low liquidity, but the current 26% price is already near the bottom of that range—suggesting the market has partially converged downward, not that it remains at an unsophisticated high. Simply dismissing it as "inflated" may be overconfident given that the brief itself flags gaps (no UCERF4, territorial waters ambiguity, multi-fault scenarios). 2. Both forecasters anchor primarily on the UCERF3 Poisson 30-year estimate (~7%) scaled to 10 years, but the brief explicitly notes UCERF3 raises probability by incorporating multi-fault ruptures. The southern San Andreas is >300 years past its last major rupture, and UCERF3's recurrence model is not purely Poisson—conditional (time-dependent) probabilities for an overdue fault segment can be meaningfully higher than the raw Poisson rate, yet neither forecaster accounts for this distinction. 3. Forecast 1 (8%) appears internally inconsistent: it claims to "allow some upside given the overdue southern San Andreas and G-R upper bounds" but only moves from the ~3.5–6% blended estimate to 8%—a modest adjustment that doesn't reflect the ~9.5% G-R upper bound the brief presents or the magnitude of the overdue-fault argument. 4. Neither forecaster addresses the territorial waters ambiguity flagged in the brief. An M8.0+ on an offshore fault segment (e.g., near the Cascadia interface or offshore southern California) could qualify, and this expands the effective fault area beyond the San Andreas, potentially adding non-trivial probability that both forecasters ignore. 5. The before-2028 vs. before-2035 cross-market signal (16% vs. 26%) implies ~10 percentage points for a 7-year window (2028–2035), which annualizes to roughly 1.5% per year—substantially above the UCERF3 Poisson rate of ~0.23% per year. Neither forecaster engages with why this implied rate is so elevated or whether it contains information about market beliefs in near-term elevated risk. 6. Both forecasters treat the absence of historical M8.0+ in the instrumental record as strong evidence against Yes, but the instrumental record spans only ~125 years against a ~494-year recurrence interval—meaning we'd expect roughly 0.25 events in the record, so zero observations is barely distinguishable from the base rate and provides weak Bayesian updating; neither forecaster explicitly corrects for this reference-class error.

9% (50% confidence)

Yes: 9% No: 91%

Scientific base rates (UCERF3 ~2.3-2.5% Poisson per decade, blended ~3.5-6%) strongly favor No. However, the critique raises valid points: time-dependent conditional probability for the overdue southern San Andreas, multi-fault rupture scenarios, territorial-waters/offshore fault scope, and the weak Bayesian power of a short instrumental record all push slightly upward. The 26% Kalshi price is likely inflated by low liquidity, but fully dismissing it to ~6% is overconfident. I settle at ~9%, between the scientific base rate and the market.

Total pipeline time: 164.6s