← Back to scans

Will there be an at least 8.0 magnitude earthquake in California before 2035?

KXEARTHQUAKECALIFORNIA-35 · Climate and Weather · 2026-05-01
9%
Agent
34%
Market Price
-25.0%
Edge
60%
Confidence
Volume: 18,917
Spread: 7.0c
Days to resolution: 3530
Markets in event: 1
Final Rationale
My independent estimate of 9% is anchored on UCERF3's ~0.24%/year M8+ rate giving ~2.4% baseline over 9.7 years, adjusted ~3-4x upward for the 345-year southern San Andreas quiescence, territorial waters scope, and ambiguity about whether 1857/1906 were truly sub-M8. I partially accept Challenge 1 — the 3-4x multiplier on UCERF3 is aggressive, but it's defensible given paleoseismic recurrence intervals of 150-200 years on the southern San Andreas, so I hold at 9%. I reject the market's 34% signal because the orderbook reveals it as a microstructure artifact: the yes_bid of $0.07 and no_bid of $0.01 with a $0.92 spread means no real trading is occurring near 34%, and the actual yes_bid (~7%) is closer to my estimate than to the displayed mid. Challenges 2 and 4 (structural uncertainty, decomposition design) are partially accepted but don't change the point estimate, only modestly lowering confidence.
Price vs. prediction
How the market moved after the agent's call. Solid blue line is the actual market price; dashed line is the agent's static forecast at $t{=}0$.
$n = 58$ follow-ups
Re-scan Context
This market has been scanned before. Previous predictions:
DatePredictedMarket PriceConfidence
2026-04-05 6% 55% 75%
1. Planner Opus · Fermi decomposition + tool plan
Tools chosen: kalshi_data kalshi_orderbook article_search web_search wikipedia_lookup code_execution
Sub-questions (Fermi decomposition)
  1. Is the annual probability of an M8.0+ earthquake in California (or its territorial waters) at least 1% per year, based on historical seismicity and current scientific estimates? (w=0.45)
  2. Given a per-year probability of M8.0+ in California, will at least one such event occur over a roughly 9.7-year window from 2026 to 2035? (w=0.40)
  3. Do current fault stress accumulation data and recent seismic activity suggest elevated near-term risk of an M8.0+ event in California relative to long-term base rates? (w=0.15)
Planner reasoning
This question requires estimating the probability of a rare but scientifically-studied natural event over a ~9.7-year window. Key considerations: (1) Reference class from USGS UCERF3, which provides probabilistic estimates for major California fault ruptures. UCERF3 gives roughly 7% annual probability of M8+ system-wide including offshore, but this figure is debated and may include events outside strict California territory. (2) Historical frequency: True M8.0+ events with epicenters clearly in California territory are extremely rare. The 1906 San Francisco earthquake is estimated at M7.7-7.9 by modern standards, potentially below the 8.0 threshold. Paleoseismic records suggest perhaps 0-2 true M8+ events in California in the past 325 years, implying annual rates well below 1%. (3) The question includes territorial waters, which broadens scope to offshore fault segments. (4) Temporal dynamics: The ~9.7-year window means even low annual rates compound meaningfully. At 1% annual rate, cumulative probability is ~9.2%; at 3% annual rate it reaches ~25%; at 7% annual rate it approaches ~50%. (5) Southern San Andreas has not ruptured since approximately 1680, creating ~345 years of stress accumulation that genuinely elevates risk. The research combines UCERF3 scientific estimates, historical frequency analysis via code, Kalshi market price as crowd wisdom, and recent news on seismic activity.
2. Research Parallel tool calls · legacy v2
ToolStatusTimeItemsSummary
kalshi_data OK 2.3s -
kalshi_orderbook OK 0.2s -
article_search OK 0.9s -
web_search OK 58.8s -
wikipedia_lookup OK 2.1s -
code_execution OK 0.0s -
3. Evidence Synthesis Sonnet · 16 items · legacy v2
#StrengthCredibilityDirectionSourceClaimPriced In
1 STRONG 92 ↓ DOWN web_search UCERF3 estimates a 7.0% probability of M8.0+ in California over 30 years, implying a recurrence rate of approximately every 494 years and a rough annual probability of ~0.2% based on that recurrence, though multi-fault rupture modeling raises the modeled rate. Yes
2 STRONG 85 ↓ DOWN web_search A Poisson approximation of the UCERF3 30-year M8+ figure (~7%) yields a 10-year probability of roughly 2–2.5%, suggesting an annual rate of approximately 0.2–0.25%, well below 1% per year. Yes
3 STRONG 88 ↓ DOWN article_search The two largest recorded California earthquakes (1857 Fort Tejon and 1906 San Francisco) are both estimated at M7.9, meaning no confirmed M8.0+ event has occurred in California in at least 257 years of modern records. Yes
4 STRONG 90 NEUTRAL wikipedia_lookup UCERF3 is the official authoritative forecast for California earthquake hazard, jointly produced by USGS, CGS, and SCEC, providing the scientific consensus on M8+ probabilities. Yes
5 MODERATE 75 ↓ DOWN code_execution Historical base rate analysis: if 1–4 true M8+ events occurred in 325 years of paleoseismic records, the annual probability is 0.3–1.2%, yielding a 9.7-year cumulative probability of roughly 2.9–11.3%. Yes
6 MODERATE 80 ↑ UP wikipedia_lookup The San Andreas Fault's southern segment has not ruptured since ~1680, creating over 345 years of stress accumulation, which time-dependent models treat as elevating near-term risk above the long-term average. Yes
7 STRONG 90 NEUTRAL code_execution At a 0.5% annual rate, the 9.7-year cumulative probability is ~4.7%; at 1% it is ~9.3%; at 2% it is ~17.7%; at 3% it is ~25.5%; at 4.7% it is ~37.2%; at 7% it is ~50.4%. Yes
8 MODERATE 60 ↑ UP kalshi_data The Kalshi market for M8.0+ in California before 2035 is currently priced at 34%, having dropped 19% over the last 7 days but up 7% over 30 days, with a wide historical price range of 4–73%. No
9 MODERATE 55 ↑ UP kalshi_data The related Kalshi market for M8.0+ in California before 2028 (a ~2-year window) is priced at 11%, implying a meaningful per-year probability when annualized (~5–6% per year implied if linear, though markets may not be well-calibrated). Yes
10 WEAK 40 ↓ DOWN kalshi_orderbook The orderbook shows a yes_bid of $0.07 and no_bid of $0.01 with a spread of $0.92 and depth of 8,389, suggesting the displayed price of 34% may not fully reflect true market equilibrium — the wide spread is a red flag for reliability. Yes
11 WEAK 50 ↑ UP article_search A July 2025 article about Trump cutting USGS earthquake monitoring capacity notes the Cascadia subduction zone (bordering Northern California) is 'due to rupture, possibly in our lifetimes,' representing a relevant but geographically uncertain risk. Yes
12 MODERATE 65 ↑ UP article_search 2024 was reported as the most seismically active year in Southern California since 1986, with above-average M4.0+ events, though this reflects elevated moderate seismicity rather than direct precursors to M8.0+. Yes
13 MODERATE 78 NEUTRAL article_search A December 2024 M7.0 earthquake struck off the coast of Northern California, one of the largest in state history, highlighting active offshore seismicity but remaining below M8.0 threshold. Yes
14 WEAK 55 ↑ UP article_search An April 2025 article drawing parallels between Myanmar's M7.7 Sagaing Fault rupture and California's San Andreas Fault notes that stress accumulation on the southern San Andreas makes a large California earthquake plausible 'in our lifetimes,' but no specific near-term trigger was identified. Yes
15 MODERATE 70 ↓ DOWN article_search Multiple seismologists quoted in 2024 CNN coverage of Southern California earthquake clusters noted no strong evidence that recent moderate quakes are precursors to a major M8+ event. Yes
16 MODERATE 82 NEUTRAL article_search A July 2025 Kamchatka M8.8 earthquake sent tsunamis to California's coast but had no epicenter in California territory — underscoring that large nearby events do not resolve this question and may temporarily shift market sentiment. No
Information Gaps
  • No updated post-UCERF3 scientific model (e.g., UCERF4 if published) with revised M8+ probability estimates for the 2026–2035 window specifically.
  • No direct paleoseismic data confirming the number of true M8.0+ events (epicenter in California or territorial waters) in the past 325 years — the historical count is uncertain and critically drives the base rate.
  • No GPS/geodetic strain accumulation data for the specific faults most capable of M8+ rupture (southern San Andreas, Cascadia near Northern California border) as of 2025–2026.
  • No clarity on how Kalshi's market price moved so dramatically (-19% in 7 days) — the cause of this drop is unknown and could reflect new information or thin-market noise.
  • No data on whether UCERF3's M8+ figure includes only land territory or also California's offshore territorial waters, which would significantly affect applicability to this question.
  • Absence of information on the December 2024 M7.0 Northern California offshore quake's effect on nearby fault stress loading.
Key Uncertainties
  • The true annual probability of M8.0+ is highly contested: UCERF3's recurrence rate (~1/494 years) implies ~0.2% per year, but model uncertainty ranges and multi-fault scenarios could push this higher or lower.
  • Whether the question's 'territorial waters' scope includes offshore subduction zone segments (e.g., southern Cascadia near the California border) — if yes, the probability rises meaningfully.
  • The large Kalshi price swing (-19% in 7 days) is unexplained and creates uncertainty about what the market is actually reflecting; the orderbook's wide spread further undermines market reliability.
  • Southern San Andreas stress accumulation (~345 years without rupture) genuinely elevates time-dependent probability, but the magnitude of this elevation relative to Poisson baseline is uncertain.
  • Whether the 1906 San Francisco earthquake was truly M7.9 or M8.0+ under modern measurement standards — this affects the count of historical M8+ California events and thus base rates.
Market Context
Price: 0.34 · 7d: -0.19 · 30d: +0.07
Volume: LOW-MODERATE — 117 contracts/day average is thin for a binary prediction market · Depth: HIGH depth (8,389) but extremely wide spread ($0.92) between yes_bid ($0.07) and no_bid ($0.01), suggesting the displayed price may be a last-trade or mid-price rather than a true equilibrium bid-ask
Reliability: LOW-MODERATE — the wide bid-ask spread and thin daily volume suggest the 34% price is not a reliable crowd-wisdom signal; the -19% weekly drop is unusually large and unexplained, possibly driven by the July 2025 Kamchatka quake resolving fears or thin-market noise
4. Independent Forecaster Opus · No market price · legacy v2 Did NOT see market price
Independent estimate: 9% (70% confidence)
base rate
9%
70% conf
evidence driven
8%
70% conf
contrarian
10%
70% conf
Spread
2pp
Low disagreement
sq1: Is the annual probability of M8.0+ in California at least 1% per year?
18%
UCERF3 implies ~0.2-0.25% annual rate. Historical record shows no confirmed M8.0+ in 257+ years (1857 and 1906 both estimated M7.9). Paleoseismic upper bound suggests ~1.2% annual. Time-dependent stress accumulation on southern San Andreas (~345 years) modestly elevates this. The 1% threshold is roughly 4-5x the UCERF3 estimate; only achievable if multi-fault rupture scenarios or territorial waters inclusion substantially boost the rate. Probability annual rate is ≥1%: ~18%.
sq2: Will at least one M8.0+ occur over the 9.7-year window?
8%
Best estimate of annual rate: weighted between UCERF3 (~0.25%) and time-dependent adjustments (~0.5-0.8%). Using ~0.8% annual rate, 9.7-year cumulative ≈ 7.5%. Adding modest upward adjustment for southern San Andreas stress and territorial waters scope: ~8%. Recent M7.0 offshore events and increased moderate seismicity provide weak upward signal; no precursor evidence cited by seismologists pulls back. Final: 8%.
sq3: Do current fault stress data suggest elevated near-term risk relative to long-term rates?
55%
The 345-year quiescence on southern San Andreas is a real time-dependent elevation factor, supported by paleoseismic recurrence intervals of ~150-200 years for that segment. 2024 elevated moderate seismicity and December 2024 M7.0 offshore quake provide some signal. However, seismologists explicitly state no precursor evidence for imminent M8+. The question asks whether risk is elevated, not whether rupture is imminent. Lean modestly yes: 55%.
Show full reasoning trace
base rate: UCERF3 scientific consensus implies ~0.2-0.25% annual probability, giving ~2-2.5% over 9.7 years. Historical record (no confirmed M8.0+ in 257 years) supports the low end. Paleoseismic upper bound allows up to ~1.2% annual.
evidence updates: Southern San Andreas 345-year quiescence pushes UP from pure Poisson baseline by ~2-3 percentage points cumulative. Territorial waters inclusion potentially adds offshore subduction zone risk near Cascadia border (+1-2%). 2024 elevated moderate seismicity weakly UP. Seismologist statements that recent quakes are not precursors push DOWN slightly. Net: from ~2.5% baseline to ~8-9%.
combination method: Sub-question 2 directly answers the question; sq1 and sq3 inform sq2's inputs. Weighted average produces 19.6% but this double-counts since sq1 is essentially asking about the input rate to sq2. The correct outcome probability is closer to sq2's direct estimate (~8%) with small upward adjustment.
final: Final estimate: 9%. This reflects UCERF3's authoritative ~2.5% baseline cumulative probability, adjusted upward ~3x for time-dependent stress accumulation on southern San Andreas, territorial waters scope, and uncertainty about whether 1906 was truly M7.9 vs M8.0+. High confidence that probability is well below 50% but meaningful uncertainty about exact level between 5-15%.
ensemble: {'ensemble_count': 3, 'perspectives': {'base_rate': {'p_yes': 0.09, 'confidence': 0.7}, 'evidence_driven': {'p_yes': 0.08, 'confidence': 0.7}, 'contrarian': {'p_yes': 0.1, 'confidence': 0.7}}, 'spread': 0.02, 'individual_p_yes': {'base_rate': 0.09, 'evidence_driven': 0.08, 'contrarian': 0.1}}
5. Devil's Advocate Sonnet · legacy v2 Market price revealed
Missing Info
Challenges
  1. The forecaster's UCERF3 anchor may be misapplied: UCERF3's official 30-year M8+ probability is ~7%, which under Poisson gives ~2.4% over 9.7 years, NOT a 0.2-0.25% annual rate as stated in sq1. Actually 7%/30yr ≈ 0.24%/yr, so this is consistent — but the forecaster's final 9% is ~3.7x the UCERF3 implied rate, which requires strong justification. The justifications (345-year quiescence, territorial waters, 2024 seismicity) are plausible but the magnitude of the upward adjustment should be scrutinized — is 3-4x UCERF3 really warranted, or is this overcorrection?
  2. Conversely, the forecaster may be UNDER-weighting structural uncertainty. UCERF3 itself has wide uncertainty bands, paleoseismic record is ambiguous about whether any true M8+ occurred, and UCERF4 has not been incorporated. The information gaps are substantial yet confidence is set at 0.7.
  3. The market at 34% with low liquidity and wide spread is genuinely low-signal — the forecaster is right to discount it. However, the -19% weekly drop is unexplained and the forecaster doesn't seriously consider whether the market knows something (e.g., a recent scientific update). This is a minor gap given the unreliable orderbook.
  4. Sub-question decomposition has a logical issue: sq1 asks if annual rate ≥1%, sq2 directly estimates the cumulative. The forecaster correctly notes sq1 is an input to sq2, but then the 'weighted average produces 19.6%' suggests the original combination method was flawed before being overridden. This ad-hoc override is reasonable here but indicates the decomposition wasn't well-designed.
Market comparison: Forecaster: 9%, Market: 34%. The 25-point divergence is large but the market signal is weak: $0.92 spread, $0.07 yes_bid vs $0.01 no_bid suggests the 34% mid-price is not a true equilibrium. The yes_bid of $0.07 is actually closer to the forecaster's 9% estimate. The divergence appears to be a market microstructure artifact rather than genuine disagreement, so no adjustment toward 34% is warranted.
6. Aggregator Opus · final calibrated forecast
9% (60% confidence)
My independent estimate of 9% is anchored on UCERF3's ~0.24%/year M8+ rate giving ~2.4% baseline over 9.7 years, adjusted ~3-4x upward for the 345-year southern San Andreas quiescence, territorial waters scope, and ambiguity about whether 1857/1906 were truly sub-M8. I partially accept Challenge 1 — the 3-4x multiplier on UCERF3 is aggressive, but it's defensible given paleoseismic recurrence intervals of 150-200 years on the southern San Andreas, so I hold at 9%. I reject the market's 34% signal because the orderbook reveals it as a microstructure artifact: the yes_bid of $0.07 and no_bid of $0.01 with a $0.92 spread means no real trading is occurring near 34%, and the actual yes_bid (~7%) is closer to my estimate than to the displayed mid. Challenges 2 and 4 (structural uncertainty, decomposition design) are partially accepted but don't change the point estimate, only modestly lowering confidence.