Tools chosen: kalshi_direct kalshi_related polymarket_related wikipedia claude_news code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a rare-event seismic forecast resolving by end of 2028, so the market price is the primary anchor and historical base rates are the key empirical input. California has never recorded an M8.0+ in the modern instrumental record, so I need to anchor on USGS/seismology base rates and compute the probability over a ~3-year window.

# Event Will there be an M8.0+ magnitude earthquake in California before Dec 31, 2028? # Outcomes to forecast - **Yes** (M8.0+ quake with epicenter in CA or territorial waters before 2028) - **No** --- # Kalshi market anchor **Current YES price: 16.00%** — This is the primary consensus anchor. - 7-day change: **+4.00%** (recent upward move) - 30-day change: **-1.00%** (roughly flat over month) - 30-day range: 5.00%–18.00%; avg daily volume: ~80 contracts - Related: 2035 version trades at **26.00%** (7-yr window); implies ~10pp premium for 7 extra years --- # Sub-question answers 1. **Kalshi market price for M8.0+ CA before 2028?** — Current YES price is **16.00%**, up +4pp in 7 days, near the top of its 30-day range. [kalshi_direct] 2. **Has CA ever recorded M8.0+?** — No confirmed M8.0+ in California's instrumental record (~125 years). Largest are the **1857 Fort Tejon (~M7.9)** and **1906 San Francisco (~M7.9)**; neither reached M8.0. [Wikipedia] 3. **USGS UCERF3 annual/multi-year probability?** — UCERF3 estimates **~7% over 30 years** for M8.0+ in CA (vs. 4.7% in prior UCERF2). This implies a raw annual rate of ~0.23%/yr, or roughly **~0.7–1.5% for a 3-year window** under Poisson assumptions. [UCERF3/conservation.ca.gov] 4. **Recent seismic warnings or elevated forecasts?** — 2024 was a record year for modest quakes in SoCal (15 M4.0+ sequences, highest in 65 years). Late 2025 San Ramon swarm (21 tremors, M2.9–4.0). However, USGS experts explicitly state these swarms are **not reliable M8+ precursors** and increase likelihood "only by a very small amount." [columbia.com, rollingout.com, pressdemocrat.com] 5. **Implied probability over ~3-year window from recurrence data?** — Poisson model with USGS recurrence intervals (150–300 years) yields **~1.0%–1.98%**, with uniform-prior mean **~1.38%** and Bayesian posterior (adjusted for 0 events in 125 instrumental years) **~0.92%**. Aggressive upper bound (R=100 yr): ~3.0%. [code_execution] --- # Key facts (high-confidence, factual) 1. [UCERF3] 7% probability of M8.0+ in CA over any 30-year window; Poisson-implied 3-yr rate ≈ 0.7–1.5% 2. [Wikipedia] No confirmed M8.0+ earthquake in California's ~125-year instrumental record 3. [Wikipedia] Both 1857 Fort Tejon and 1906 SF quakes are estimated M7.9 — just below threshold 4. [UCERF3] Southern San Andreas Fault is overdue (~300+ years since last great rupture), modestly elevating near-term risk 5. [code_execution] Bayesian update (geologic prior + 0 obs/125 yrs) yields posterior mean rate ~1/325 yr, P(3yr) ≈ 0.92% 6. [UCERF3] Multi-fault rupture scenarios (new in UCERF3) could produce M8.0+ events more readily than single-fault models --- # Cross-market signals - **Kalshi 2035 market**: 26.00% YES for M8.0+ before 2035 (~10-yr window). Implies ~10pp premium over 7 additional years — roughly consistent with ~1.3%/yr incremental probability - **Kalshi Japan 2030**: 58.00% for M8.0+ Japan before 2030 — confirms Japan is a much higher-risk seismic environment vs. California - **Polymarket**: No matching markets found --- # Analyst opinions and speculation - USGS geophysicist Annemarie Baltay: swarms "not considered reliable indicators of an impending major quake" [rollingout.com] - Scientists: past California swarms "have not been followed by major quakes in same areas" [pressdemocrat.com] - Seismologists note Southern San Andreas is "late in its seismic cycle" — tectonic strain accumulating [conservation.ca.gov] - The Kalshi 16% price is **~10–15x above** the scientifically derived base rate of 1–2%, suggesting significant market mispricing or speculative/fear-driven trading --- # Directional lean per outcome - **Yes (16% Kalshi)**: Southern SAF late in cycle; UCERF3 multi-fault scenarios; recent elevated seismicity; ~3-yr window not negligible - **No (84% Kalshi)**: Base rate ~1–2% from Poisson/Bayesian models; no M8.0+ in 125 years of CA record; expert consensus that swarms aren't precursors; 16% Kalshi price appears ~8–10x overpriced vs. science --- # Gaps / unknowns - No real-time USGS operational forecast elevating M8+ probability - Exact fault-specific time-dependent probability not fully resolved (UCERF3 is time-independent) - Whether thin market (80 contracts/day) is driving speculative vs. informed pricing --- # Calibration anchors - **Kalshi current YES: 16.00%** (primary anchor — but likely overpriced vs. science) - **UCERF3-derived 3-yr base rate: ~0.9–2.0%** - **Historical precedent**: 0 M8.0+ in 125 years of CA instrumental record - **Comparable**: Japan M8.0+ before 2030 at 58% — reflects ~10× higher seismicity rate

1. Both forecasters heavily discount the Kalshi 16% price as "speculative/fear-driven," but the brief notes the 2035 market trades at 26% — implying ~10pp for 7 extra years, which cross-validates a roughly 1.3%/yr incremental rate consistent with science. Neither forecaster adequately interrogates why informed market participants (even in a thin market) would sustain 16% for weeks given such a large alleged mispricing. 2. Both forecasters treat the UCERF3 Poisson model as nearly definitive, but UCERF3 itself is time-independent and explicitly notes that time-dependent models for an "overdue" Southern San Andreas would yield higher near-term probabilities — a scenario neither forecaster quantifies or properly weights. 3. The brief flags that UCERF3 introduced multi-fault rupture scenarios (new vs. UCERF2) that "could produce M8.0+ events more readily than single-fault models," raising the 30-yr estimate from 4.7% to 7%. Neither forecaster discusses whether this structural change in methodology implies the Poisson 3-yr upper bound (~3%) is a floor rather than a ceiling. 4. Both forecasters anchor strongly on the 125-year instrumental record showing zero M8.0+ events, but the brief itself notes the 1857 and 1906 quakes were estimated M7.9 — just below threshold — suggesting the record provides weak evidence against M8.0+ occurrence rather than strong evidence, since measurement uncertainty alone could shift those estimates above 8.0. 5. Neither forecaster addresses the thin market caveat (~80 contracts/day) with appropriate nuance: a low-volume market could be systematically mispriced in either direction, meaning the 16% anchor deserves less weight as a consensus signal than a deeper market would warrant, yet both forecasters treat partial discounting of it as sufficient analysis. 6. Forecaster 1 at 4% and Forecaster 2 at 5.5% differ meaningfully (~38% relative gap) despite using nearly identical evidence, yet neither explains what specific factor drives the disagreement — suggesting both confidence levels are poorly calibrated relative to the genuine uncertainty range implied by the evidence (roughly 1–5%).

6% (65% confidence)

Yes: 6% No: 94%

Scientific base rates (UCERF3 Poisson/Bayesian) put a 3-year M8.0+ probability at ~1-2%, with no confirmed M8.0+ in 125 years. However, the critique fairly notes time-dependent strain on the overdue Southern San Andreas, new multi-fault rupture scenarios, and measurement uncertainty (1857/1906 near M7.9) all push slightly upward. The Kalshi 16% appears speculative for a thin market but shouldn't be fully dismissed. I settle modestly above the science-based rate and both forecasters, but far below the market price.

Total pipeline time: 143.3s