Tools chosen: kalshi_direct kalshi_related polymarket_related claude_news wikipedia code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-tail seismic risk question with a ~10 year window. Key drivers are historical base rates for M8+ in California, current USGS forecasts (UCERF3), and the Kalshi market price as anchor. Cross-market comparisons may exist.

# Event Will there be an M8.0+ earthquake in California before Dec 31, 2035? # Outcomes to forecast - **Yes** (M8.0+ earthquake with epicenter in California or territorial waters before 2035) - **No** # Kalshi market anchor **KXEARTHQUAKECALIFORNIA-35: 30% YES** (⚠️ this is anomalously high — see below) - 7-day change: +4pp | 30-day change: **−23pp** (major recent decline from ~53%+ range) - 58-day price range: 22%–65% — extremely wide, suggesting thin/volatile market - Avg daily volume: 42 contracts (low liquidity; prices are not reliable consensus) - Related: KXEARTHQUAKECALIFORNIA-28 (before 2028) = **18%** YES > **Critical note:** The 30% price is ~10–15× the seismological base rate estimate (~2–3%). Low volume (42 contracts/day) and a 43pp price range over 58 days indicate this market is highly noise-driven, not an efficient consensus. The −23pp drop in 30 days suggests a prior price spike (possibly from news/activity) that has partially corrected. # Sub-question answers 1. **Kalshi market price for M8.0+ CA before 2035?** — 30% YES as of current data; range 22–65% over 58 days; 7-day +4pp, 30-day −23pp. [Kalshi direct] 2. **Historical base rate of M8.0+ in California?** — No instrumentally confirmed M8.0+ has occurred in California. Largest historical events: 1857 Fort Tejon (M7.9), 1906 San Francisco (M7.8–7.9). True M8.0+ paleoseismic recurrence estimated ~400–500 years. [USGS, Wikipedia] 3. **UCERF3 forecast for M8.0+ over ~10 years?** — UCERF3 gives ~7% over 30 years (central); Poisson scaling to 10 years yields **~2.4%**; BPT time-dependent adjustment raises it to ~2.75%. Full uncertainty range: ~1.7%–3.5% for 10-year window. [UCERF3/CGS, code_execution] 4. **Related cross-market signals?** — KXEARTHQUAKECALIFORNIA-28 at 18% (before 2028, ~3 years closer deadline); Japan M8+ before 2030 at 58% (Japan is far more seismically active). No Polymarket markets found. [Kalshi related] 5. **Recent seismic stress/activity?** — Southern San Andreas last ruptured late 17th century (~300+ years ago), flagged as most overdue segment. UCERF3 notes multi-fault ruptures increase M8+ potential. 2023 USGS NSHM update released. No specific recent major precursor events identified. [USGS, NBC News] # Key facts (high-confidence, factual) 1. [UCERF3/CGS] Official 30-year M8.0+ CA probability: ~7% central (range: 0–32%) 2. [Code execution] Poisson 10-year scaling: **2.39%**; BPT-adjusted: **2.75%** 3. [USGS/Wikipedia] No confirmed M8.0+ in California's instrumental or well-documented historical record 4. [Wikipedia] 1857 Fort Tejon = M7.9; 1906 San Francisco = M7.8–7.9 (both sub-threshold) 5. [USGS] Southern San Andreas fault segment overdue; last ruptured ~late 1600s 6. [Kalshi] Before-2028 market at 18% implies ~12pp incremental probability 2028–2035 per market 7. [Code execution] UCERF3 sensitivity range for 10yr: 1.70% (5%/30yr) to 3.45% (10%/30yr) # Cross-market signals - **Kalshi KXEARTHQUAKECALIFORNIA-28** (before 2028): 18% — implies market assigns ~12pp to the 2028–2035 window alone, internally inconsistent with scientific priors - **Kalshi KXEARTHQUAKEJAPAN-30** (M8+ Japan before 2030): 58% — Japan is ~5–10× more seismically active at M8+; useful as upper bound comparator - **Polymarket:** No relevant markets found # Analyst opinions and speculation - UCERF3 authors: increased M8+ probability driven by multi-fault rupture scenarios [CGS] - Media characterizes southern San Andreas as "most dangerous" and "overdue" [NBC News, Newsweek] - Scientific consensus: "Big One" scenario more likely ~M7.8, not necessarily reaching M8.0 threshold - Kalshi price spike to 65% then crash to 22–30% suggests possible news-driven speculation without seismological basis # Directional lean per outcome - **Yes (M8.0+):** *Supporting:* Southern San Andreas stress accumulation; UCERF3 allows max ~32% over 30yr; market at 30%. *Opposing:* No M8.0+ in recorded CA history; UCERF3 central 10yr estimate only ~2.4%; low market liquidity inflates price; before-2028 at 18% is itself likely overpriced - **No:** *Supporting:* Strong seismological base rate (~97–98% Poisson); 10-year window is short; threshold of M8.0 has never been met in CA history; market thinly traded and likely overpriced # Gaps / unknowns - Whether the 30% Kalshi price reflects genuine informed trading or illiquid noise - Whether the 2023 NSHM update materially revises UCERF3's M8+ estimates - Exact current stress state on specific fault segments (post-2023 model) - Whether "territorial waters" could include Cascadia-adjacent offshore events # Calibration anchors - **Kalshi current YES price: 30%** (LOW CONFIDENCE anchor due to thin volume, 43pp range) - **UCERF3 Poisson central (10yr): ~2.4%** (HIGH CONFIDENCE scientific anchor) - **BPT time-dependent (10yr): ~2.75%** - **Recommended blended scientific estimate: ~2.5%** - **Market vs. model gap: ~12×** — market likely overpriced by ~10–27pp; lean strongly toward No (~97%)

1. Both forecasters appropriately dismiss the Kalshi anchor but may still be over-anchoring to it: Forecast 1 settles at 6% (roughly 2.5× the scientific central), explicitly citing partial deference to the market despite flagging it as noise-driven; this upward adjustment lacks a stated scientific justification beyond vague uncertainty inflation. 2. Neither forecaster addresses the internal inconsistency of the KXEARTHQUAKECALIFORNIA-28 price (18% for M8.0+ before 2028): if the before-2028 market is itself likely overpriced by the same liquidity/noise dynamics, it cannot serve as a coherent cross-market signal, yet the brief uses it to imply a "12pp incremental" 2028–2035 probability that neither forecaster explicitly challenges or corrects. 3. Both forecasters under-engage with the "territorial waters" scope question flagged in the brief: a Cascadia subduction zone rupture with an epicenter in California's offshore territorial waters could plausibly reach M8.0+, which is a meaningfully different scenario than a San Andreas event and could modestly raise the true scientific base rate above the UCERF3 onshore-focused estimate. 4. Neither forecaster quantifies or explicitly weighs the UCERF3 30-year sensitivity range upper bound (~10%/30yr → ~3.45%/10yr) against the possibility that the 2023 NSHM update materially revised these figures upward; given this is flagged as a known gap, a brief acknowledgment of how it affects their confidence interval would strengthen calibration. 5. Forecast 1's 6% estimate implies roughly a 3.5pp premium above the scientific central (~2.5%) to account for model uncertainty and tail scenarios, but this premium is not decomposed — the brief notes the full UCERF3 uncertainty range runs 1.7%–3.5%, meaning 6% already sits outside the scientific sensitivity range without explanation.

4% (65% confidence)

Yes: 4% No: 96%

Scientific base rate dominates: UCERF3 Poisson 10-year estimate is ~2.4%, BPT-adjusted ~2.75%, with sensitivity ceiling ~3.5%. The Kalshi 30% anchor is unreliable (42 contracts/day, 43pp range). I add a small premium above the central estimate for tail scenarios (Cascadia-adjacent offshore, multi-fault rupture, overdue southern San Andreas, possible NSHM upward revisions), landing at 4% — within the scientific sensitivity envelope.

Total pipeline time: 143.5s