RESEARCH DOSSIER
Earthquake Shaking Intensity Parametric Insurance
Abstract. Earthquake shaking intensity parametric insurance is contractual risk transfer in which a qualifying seismic event and a predefined measure of local shaking, or a specified transformation of seismic evidence into modeled loss, determine a scheduled financial payment. This study examines the physical foundations of the measurement, the contractual construction of the index, the distribution of payouts and residual financial need, and the relationship between observation integrity and financial effectiveness. It differentiates source magnitude, peak ground acceleration, peak ground velocity, spectral acceleration, macroseismic and instrumental intensity, and modeled losses. Official seismological documentation, a California specimen policy, a Japanese corporate case, Caribbean sovereign event reports, a Mexican capital-market transaction, and accessible scientific research abstracts establish a comparative evidence base. Original analytical calculations compare binary and graded payouts under explicit hypothetical event rates and demonstrate that equal expected payments can produce substantially different distributions of residual need. The investigation identifies the respective roles of geographic representation, vulnerability, payment timing, contractual eligibility, and seismic uncertainty. It concludes that the scientific adequacy of a shaking-intensity index depends on its reproducibility and its relationship to the defined financial objective, including the time at which resources become available. The study is a comparative documentary and analytical synthesis, not a population-level estimate of product effectiveness, an independently replicated proprietary catastrophe model, or a tariff for actual insurance.
Keywords: Earthquake insurance; parametric insurance; shaking intensity; seismic hazard; ground motion; ShakeMap; peak ground acceleration; peak ground velocity; spectral acceleration; basis risk; event aggregation; catastrophe modeling; disaster liquidity; insurance-linked securities.
Introduction
An earthquake is a physical rupture, a spatial field of motion, a potential cause of damage, and a financial shock whose consequences develop over time. Insurance connects the last of these processes with a contractual obligation. In shaking-intensity parametric insurance, the terms of that obligation are established before the earthquake through a seismic index and an agreed calculation. The research object includes both direct local-intensity indices and specified modeled-loss indices. Source magnitude and event-location instruments enter as comparators because they supply alternative ways of representing earthquake risk.
Financial purpose determines the meaning of protection. A household may need relocation expenses, food replacement, or a deductible; an industrial company may need to pay wages and restore operations; a public authority may need funds for emergency services. The same physical event can therefore support several distinct and economically meaningful indices. The research question is which characteristics of the earthquake, its exposed assets, the observation system, and the contract make the resulting payment relevant to the intended financial objective.
The parent scientific field is Parametric and Index-Based Insurance. This dossier concentrates on the specific relationship between earthquake ground motion, contractual thresholds, and financial protection. Broader questions of seismic engineering, general catastrophe-bond valuation, tsunami modeling, and legal classification enter where they have a direct bearing on that relationship.
The investigation uses purposive documentary selection by evidentiary function. Official physical and technical descriptions establish the measurement architecture; a specimen policy identifies operative written conditions; institutional event and transaction records establish disclosed decisions; accessible original research abstracts identify empirical and methodological findings within their stated limits. The study does not claim exhaustive systematic-review coverage. Three primary PDF documents yielded extracted text, but screenshot retrieval failed; no conclusion depends on an unverified figure. Full texts of several academic papers were inaccessible, so statements derived from them are confined to their available abstracts. Actual archival event maps, proprietary model inputs, independently linked payment and loss records, and complete transaction wordings were not obtained. The numerical analysis is an original illustrative calculation with declared artificial inputs.
The argument proceeds from physical measurement to contractual construction, financial performance, and comparative interpretation. It separates estimation error from genuine differences in vulnerability, the legal conditions governing payment, and the relationship between payment and timely financing.
Scientific Foundations and Historical Development
Source magnitude and local ground motion
Magnitude represents the size of an earthquake. Moment magnitude, commonly denoted Mw, is based on seismic moment, which depends on rigidity, rupture area, and average displacement. The shaking experienced at a particular location instead arises from propagation of the seismic waves and their interaction with local geology. Distance, depth, fault orientation, rupture directivity, geological structure, and surface materials affect the field.
The USGS account of magnitude and shaking intensity provides contrasting examples of local geological conditions and different earthquake depths. In financial terms, a shallow earthquake near a facility may create greater local motion than a larger but distant event. A major rupture can also affect a considerable number of sites simultaneously while individual locations experience different levels of shaking. Source magnitude helps identify the common event; local motion supplies a more spatially specific index.
Estimates of magnitude and hypocenter can change as observations and calculations improve. A revised estimate does not create a new physical earthquake. An insurance instrument consequently requires an event identity, a version rule, and an occurrence definition in addition to its magnitude or motion threshold. A magnitude condition used to screen events for reporting is a separate question from a magnitude condition used to establish entitlement.
Acceleration, velocity, and oscillator response
Peak ground acceleration (PGA) is the maximum absolute acceleration under a defined record and component convention. It may be expressed as a fraction of standard gravity, as percent g, or in SI or centimeter units. Standard gravity is 9.80665 meters per second squared. Hence 0.2 g is approximately 1.96 meters per second squared; 20 percent g is 0.2 g. Confusing 0.2 g with 0.2 percent g introduces a factor of one hundred into the calculations.
Peak ground velocity (PGV) identifies the largest absolute velocity under the specified recording and processing conventions. It is frequently expressed in centimeters per second. Ground velocity can be estimated through integration of acceleration with appropriate treatment of filtering and baseline errors. Velocity cannot be inferred by multiplying a peak acceleration by an arbitrary duration. Even for elementary sinusoidal motion of displacement amplitude A and angular frequency omega, the peak velocity is omega A and the peak acceleration is omega squared A. Equal acceleration at two frequencies therefore need not mean equal velocity.
Spectral acceleration characterizes the response of a damped oscillator at a stated natural period and damping ratio. Pseudo-spectral acceleration is calculated from maximum relative displacement multiplied by squared natural angular frequency. A structure's characteristic period affects how it responds to the input ground motion. PGA, PGV, and spectral accelerations at distinct periods can consequently have different predictive value for short buildings, tall structures, industrial machinery, and extended infrastructure.
The selection of one peak measure compresses a long motion history. Duration, repeated cycles, permanent ground deformation, loading direction, and frequency composition may be consequential for damage. A peak statistic provides a practical contractual coordinate, but its financial interpretation requires a choice of the target exposure. A simple measure can be entirely suitable for a small cost-of-disruption benefit even when a complex engineering model would be needed to estimate structural replacement costs.
Measurement conventions themselves require explicit definition. Two providers can report a quantity called PGA using different horizontal components, site conditions, or aggregation protocols. A written numerical threshold has meaning within its specified units and conventions. Substitution of a seemingly similar data product should be assessed as a substantive contractual and actuarial change.
Macroseismic and instrumental intensity
Macroseismic intensity describes local effects of earthquake shaking in ordered categories. Modified Mercalli intensity (MMI) is used in the United States. Instrumental estimates can relate observed waveforms to intensity; traditional assignments can also rely on reports of experienced shaking and effects. An estimate of an ordinal intensity level does not imply a linear relationship between the numerical labels and either energy or monetary loss.
Worden, Wald, and Rhoades (2012) developed probabilistic ground-motion relationships using California intensity reports and instrumental motion measures. Their abstract identifies PGV as having the smallest fitting errors among the individual measures studied and reports improvement from magnitude and distance terms. That result supports the use of statistical conversion and the recognition of residual uncertainty. It does not establish a universal monetary conversion or prove that PGV is the best insurance trigger for all portfolios.
Japan's national seismic intensity scale uses instrumental observations and ordered categories from zero through seven, with five and six subdivided. The official explanatory notes distinguish the recorded intensity from descriptions of phenomena potentially associated with it. Conditions can vary across a city and within a building; damage depends in part on the structure and the character and duration of motion. A standardized public category can make a policy's terms easier to communicate. Its interpretation still depends on a specified reporting point or area and the financial exposure.
Care is needed when validating an intensity index derived partly from reports of earthquake effects. If a damage-related observation contributed to the calculation of the intensity and is then reused in an ostensibly independent observed-loss measure, the resulting agreement may be inflated through information leakage. A faithful historical evaluation reconstructs the information that would have been available at the time specified in the policy.
Hazard, exposure, vulnerability, and financial need
Hazard concerns the possible ground motions and their frequency. Exposure concerns the location and amount of objects or economic activities that may suffer consequences. Vulnerability concerns their conditional distribution of damage under those motions. Financial need concerns the resources required to absorb or recover from the consequences at a particular time and under a particular arrangement of insurance, capital, and reserves.
The OpenQuake risk calculation workflows document the combination of modeled ground motions, exposure, fragility, and vulnerability to estimate loss. A schematic direct physical loss is the sum over exposed assets i of V_i D_i, where V_i denotes asset value and D_i is a random damage ratio conditional on local motion and other characteristics. Indirect consequences, financing costs, business interruption, public responsibilities, and deductibles require additional modeling or an alternative definition of the financial target.
The four dimensions need separate data. Accurate local motion can coexist with a poor estimate of vulnerability; a well-calibrated direct physical damage model can coexist with a poor estimate of emergency liquidity. An insurer may use a hazard measure as a contract index without asserting that it determines the full economic loss.
Development of the scientific problem
The development visible in the reviewed evidence brings together public digital seismic mapping, probabilistic catastrophe analysis, and contractually specified event-driven financing. The USGS ShakeMap program provides near-real-time representations and event archives. Research comparing location and station-based triggers and subsequent work on spatial calibration extends the use of this information toward insurance and capital-market instruments.
The history of technical capability is distinct from evidence of financial success. The capacity to produce a more detailed map does not establish that contracts became more affordable, generated less basis risk in the market, or reduced disaster losses. Those propositions require appropriate outcome data and comparisons. The present historical account concerns the development of available mechanisms and the scientific problems they make tractable.
Primary Instruments and Shaking Maps
The digital observation chain
A shaking map is a spatial estimate constructed from observations and admissible modeling. A grid value between stations should not be represented as a direct local measurement. The value at an insured address may depend on a model, measurements at another location, site corrections, the version of the event source, and the extraction procedure. Contractual certainty requires the identity of the estimate as well as its numerical value.
The USGS ShakeMap product documentation identifies event-source information, map status, actual and scenario event categories, processing timestamps, geographic bounds, grid files, uncertainty files, and station information. These properties provide the foundation for an independently reviewable selection of event products. A calculation should accept the stipulated source and event category and reject tests, scenarios, canceled products, and incompatible versions.
More observations can lead to a more accurate estimate after an initial map is issued. Such improvement may change an index that crosses a financial threshold. A policy should therefore specify whether a particular version, a prescribed observation time, or a provider-certified realization binds the parties. A chosen finality rule creates a reproducible determination; it also creates a tradeoff between timeliness and information quality.
The operational evidence record should retain event ID, provider, event origin time, map process time, retrieval time, status, raw-file hash, spatial reference, grid dimensions, units, component convention, extraction method, rounding, and calculated entitlement. When an observation is revised, the original evidence and result must remain retrievable. Later scientific information is not automatically admissible for a payment finalized under an earlier version rule.
Spatial construction
A point-based trigger samples a grid at an agreed coordinate. Nearest-neighbor sampling, bilinear interpolation, or maximum within a small buffer may return different values from the same grid. The spatial rule must be agreed before the event. A site lying on a cell boundary or beyond the grid extent should be part of the design test set. An unavailable value should remain distinguishable from zero.
A station-based trigger can simplify verification by using one or more instruments. This architecture also inherits the distance between the stations and exposure sites, geological differences, potential instrument failure, and the convention for replacing missing observations. The possibility of outage can itself depend on severe motion. Consequently, a missing-data model based solely on ordinary operating conditions can be misleading precisely during the events that matter most for payment.
For a polygon-based trigger, the maximum represents the strongest qualifying portion of the region; an average describes a different relationship; a weighted integral assigns importance according to exposure or another rationale. A maximum may generate an entitlement for a location outside the strongest part of its polygon. An average may dilute localized damage. Neither result is intrinsically an estimation failure: it may be a direct consequence of the chosen contractual geography.
The effective boundaries, geocoding algorithm, coordinate reference system, and edge convention are part of the definition of the index. Updating an exposure address during the policy period requires a dated, controlled change. The parties must distinguish data errors from genuine pre-event exposure modifications. Privacy protection can coexist with reproducibility by using stable pseudonymous identifiers and preserving a restricted authoritative coordinate record.
Earthquake event and insured occurrence
The catalog describes physical events; the policy describes insured occurrences. Earthquake sequences can contain foreshocks, a mainshock, and numerous aftershocks. A grouping condition may select one event, combine eligible observations within a stated interval, or treat events separately. Each choice affects expected payments, accumulation, exhaustion, and the probability of another entitlement during the same period.
A trigger observation deadline, an earthquake occurrence-aggregation window, and a policyholder reporting deadline serve separate functions. The first determines when observation becomes binding. The second determines which physical events share a contractual entitlement. The third governs the provision of required information. A policy using all three intervals should identify them independently.
Seismic catalogs may revise origin time, merge records, or correct identifiers. A contract administered automatically requires procedures for these changes, particularly for events near inception or expiration. The occurrence definition is also important to the insurer's reinsurance: if the original policies and reinsurance treaty aggregate different physical events, the primary and ceded payout distributions can diverge.
California specimen wording
The publicly accessible California earthquake policy form attributed to Jumpstart and certain Lloyd's underwriters supplies direct contract text. Sections II and V establish eligibility using PGV of at least 30 cm/s at any point in the census block containing the address, with reference to the authoritative USGS map reported at 24 hours after earthquake event time. It requires legal interest, a covered loss, and a loss report. It also describes a 168-hour event grouping period.
The form therefore does not establish an unconditional payment solely for a number. A physical threshold determines one part of eligibility; additional insurance conditions remain material. The territorial mechanism means a qualifying block need not have uniform ground velocity at every address. A payment based on the specified maximum follows the geographic rule even when a different sampling procedure would give a lower value.
Sections II and III appear potentially inconsistent concerning the aggregate limit: one invokes the limit stated in the declarations, while the other describes twice the coverage limit. Declarations and endorsements were not available. No definitive annual cap or legal determination can be inferred. This is a concrete demonstration of the need to review the complete policy rather than extracting a threshold in isolation. The accessible specimen is not verified as the current marketed form or the wording applicable to any individual customer.
A contract containing a fixed scheduled benefit and a requirement to establish covered loss can still differ substantially from an indemnity contract in which the amount of each payable cost is assessed. The relevant questions are which conditions govern eligibility, which calculation governs amount, and which documentation establishes each of them.
Provider independence and settlement control
An independent public measurement does not automatically ensure independent administration. An insurer or calculation agent may still choose event records, map versions, addresses, and eligibility decisions. The parametric disaster insurance overview identifies the importance of specifying the parameter, payment, verifying entity, and contingency arrangements.
A reproducible calculation chain distinguishes observation, computation, review, and release. A map revision after finality should have an explicit status: it may be irrelevant, trigger a correction under stated conditions, or be considered in a defined dispute process. Late correction rules should address scientific revisions separately from data corruption and administrative mistakes.
If a station or provider fails, the contract can designate a secondary source, prescribe a substitute calculation, or extend the determination window. These choices have economic consequences. A more distant replacement station may systematically return different observations; a longer window may harm liquidity; a modeled fallback changes the kind of uncertainty borne by the insured. A meaningful fallback is evaluated before the contract is issued.
Risk Transfer Mechanisms of Ground Motion
Attachment and payout functions
Let X denote the accepted index, A the attachment point, U the exhaustion point, and K the event limit. A binary contract pays K when X is at least A and zero otherwise, conditional on other contractual requirements. A graded linear structure with U greater than A has payment P(X)=K max(0,min(1,(X−A)/(U−A))). A stepped contract assigns predeclared payment bands. The limit, equality convention, units, and rounding are part of the specific function.
A binary function contains a discontinuity of K at attachment. A continuous ramp has slope K/(U−A) between attachment and exhaustion. A change in accepted index delta therefore changes payment by no more than K times absolute delta divided by U−A for the specified ramp. That deterministic property improves local stability relative to a discontinuous function, while changing the distribution of payments.
For a 100,000-dollar limit and linear attachment/exhaustion at 20 and 60 cm/s, the slope is 2,500 dollars per cm/s. A change from 29 to 31 cm/s produces a payment difference of 5,000 dollars. With a binary 30 cm/s attachment at the same limit, the difference is 100,000 dollars. A ramp is numerically stable in this example. Its economic desirability still depends on the target: some enterprises have costs that themselves jump when a shutdown threshold is reached.
An ordinal national intensity scale can use a table of payable amounts or percentages. Treating the ordered category numbers as a linear physical variable imposes an assumption that the data do not necessarily support. Category subdivisions must be preserved in the implementation.
Multi-site exposure and catastrophe dependence
A program covering several facilities may establish individual site schedules and combine payments under a common limit. A shared earthquake creates dependence between those payments. Source geometry and spatial ground-motion residuals affect the distribution of eligible site counts. Assuming independent site payouts can materially understate aggregate exposure; assuming perfect dependence can overstate it.
The OpenQuake event-based approach represents stochastic events and spatially correlated field realizations, enabling assessment of portfolio loss per event. A settlement index may be simpler than the underlying actuarial model. The latter supports pricing and aggregate capital assessment, while the former must be defined sufficiently for repeatable contract administration.
For a network in which one critical facility can interrupt the entire enterprise, a maximum or critical-site trigger can be meaningful. For distributed assets that contribute jointly to costs, a weighted index or summed site schedule may be appropriate. Site weights should be related to the financial objective rather than convenience alone. A replacement-value portfolio and a business-interruption portfolio can need different weights for the same locations.
Modeled-loss layers
A modeled-loss trigger uses predeclared calculations to transform event parameters and exposure assumptions into an amount Y. The payout may be a proportional layer with attachment A, exhaustion U, and ceded share c: c max(0,min(Y−A,U−A)), with limit c(U−A) unless an additional cap applies. Y is not necessarily the insured's actual loss; it is the output of the specified model.
A model can incorporate the spatial distribution of motion, exposure values, and vulnerability functions. Its correspondence to actual financial need depends on the suitability of those inputs. Unrecorded assets, revised valuations, retrofit programs, incomplete business-interruption variables, and outdated regional data can produce error even with accurate ground-motion observations. Model identity, version, exposure date, currency, and transformation rules are thus contractual elements.
A government emergency-cost model and a model of private structural replacement losses answer different questions. A positive model output does not alone establish the amount of public spending needed; its suitability depends on the financial responsibility the policy intends to cover. The reported monetary output and the amount of contractually ceded protection require separate explanation.
Event rates and actuarial expectations
Let a finite collection of disjoint event scenarios have annual rates lambda_e, with random payout Z_e conditional on each scenario. For an uncapped additive payout process with finite means, the annual expected gross payment is the sum of lambda_e E[Z_e]. For independent Poisson scenario processes and deterministic payouts p_e, annual variance is the sum of lambda_e p_e squared. These simplified results require the stipulated assumptions. Earthquake clustering, contractual aggregation, annual caps, changing exposure, and dependencies can change the results.
If qualifying events follow a Poisson process of annual rate lambda, the probability of at least one such event in t years is 1−exp(−lambda t). An instrument paying K per event without an annual cap has annual expected payout K lambda. If it pays at most once each year, the expectation is K(1−exp(−lambda)). The difference is relevant even in products with apparently similar event-level protection.
A physical return period describes frequency under assumptions; it is not a waiting-time guarantee. A site hazard curve alone does not establish the frequency of eligible contractual occurrences. A proper contract simulator accounts for geographic sampling, event grouping, reporting conditions, aggregate limits, and reinstatement.
The actuarial pure premium reflects expected payments. Expenses, reinsurance, capital requirements, uncertainty, taxes, and distribution can affect the actual premium. Identical expected payouts are not therefore a claim of identical market prices. Severe events can trigger numerous policies simultaneously, making the full distribution and correlation of liabilities important for underwriting.
Original comparative calculation
The following scenario inputs are hypothetical. Financial amounts are thousands of dollars; PGV is measured in cm/s. Rates describe disjoint additive per-event possibilities in an uncapped calculation. Eligibility is assumed satisfied. Product A pays 100 at PGV at least 30. Product B pays 50 at PGV at least 20. Product R pays up to 100 linearly between 20 and 60.
| Scenario | Annual rate | PGV | Financial target F | Payout A | Payout B | Payout R |
|---|---|---|---|---|---|---|
| a | 0.030 | 15 | 10 | 0 | 0 | 0 |
| b | 0.010 | 29 | 75 | 0 | 50 | 22.5 |
| c | 0.008 | 31 | 20 | 100 | 50 | 27.5 |
| d | 0.002 | 60 | 180 | 100 | 50 | 100 |
Scenario b has relatively substantial need despite falling just below A's threshold, while scenario c crosses that threshold with smaller need. This construction represents possible heterogeneity of exposure, not measured earthquake frequencies or losses.
For payout P define shortage S=max(F−P,0) and excess O=max(P−F,0). Then S−O=F−P and S+O=abs(F−P). Weighting each quantity by its scenario rate yields:
| Annual expectation, thousands of dollars | A | B | R |
|---|---|---|---|
| Payout | 1.000 | 1.000 | 0.645 |
| Shortage | 1.210 | 0.810 | 0.985 |
| Excess | 0.640 | 0.240 | 0.060 |
| Absolute target error | 1.850 | 1.050 | 1.045 |
The expected annual target is 1.570. For A, expected payout equals 0.008 times 100 plus 0.002 times 100, or 1.000. For B it equals (0.010+0.008+0.002) times 50, also 1.000. B's shortage is 0.030 times 10 plus 0.010 times 25 plus 0.002 times 130, or 0.810. Its excess is 0.008 times 30, or 0.240.
A and B consequently provide the same average annual gross payments with different allocation across scenario states. B reduces expected absolute target error by 0.800 thousand dollars annually relative to A, but in the severe scenario d it leaves a shortage of 130 thousand rather than 80 thousand dollars. A purchaser primarily concerned with a very severe event could prefer A despite its larger mean error.
The ramp has lower expected payment than both binary designs. Its absolute error is marginally below B's, but its expected shortage is higher and its excess lower. A claim that the ramp is economically superior would require a criterion and a comparable payment budget. The numbers demonstrate a distributional tradeoff rather than general superiority of graded contracts.
As a rate sensitivity test, changing scenario b's annual rate by plus 0.005 changes A's annual expected shortage by 0.375 thousand and B's by 0.125 thousand. It changes B's expected payout by 0.250 thousand, while A's remains unaffected. A real actuarial investigation would vary the joint distribution of physical fields, vulnerability, exposure, and dependencies, not just one artificial event rate.
The example establishes a narrow but important analytical result: equal expected payments leave the distribution of financial protection undetermined. Optimizing mean correspondence alone can sacrifice protection in the most severe states. No new named theory or empirical model is claimed.
Comparative Evidence for Threshold Design
Japan: corporate liquidity and confirmation timing
Swiss Re Corporate Solutions' Japanese QUAKE case study (2021) identifies a manufacturing client whose parametric protection began in 2017 to supplement conventional earthquake insurance. An earthquake on February 13, 2021 met the Shindo 6-lower threshold and the client qualified for a partial payment. The insurer reports payment three days after JMA confirmed the level, and 44 days after the earthquake.
Both time intervals matter. The shorter concerns post-confirmation administration; the longer concerns the actual event-to-payment horizon. Under the reported timing convention there is approximately a 41-day difference, but precise confirmation timestamps and the reasons for this interval are unavailable. A statement that the client was paid three days after the earthquake would be incorrect.
The case provides evidence of the disclosed purpose and one reported outcome. It does not provide full policy wording, the actual schedule, the payment amount, actual loss, other available resources, or a recovery counterfactual. It therefore does not establish an average settlement time or an estimate of the product's causal effect on business recovery.
Shindo offers a publicly identifiable intensity scale. The JMA explanation of local effects also establishes reasons why uniform intensity need not imply uniform building damage. A corporation can select a financially meaningful tier without claiming that the tier describes every component of its loss.
United States: velocity and geographic eligibility
The California specimen establishes a different architecture. PGV is the physical metric, the census block is the geographic object, and a 24-hour map is the temporal observation rule. Its requirement of legal interest and covered loss further distinguishes eligibility from payment amount.
The specimen discloses written provisions but no observed payment outcomes. The Japanese case supplies an observed outcome attributed to the insurer but limited contractual detail. A sound comparison preserves these evidentiary differences. Neither source establishes a denominator for claims, a population estimate of basis risk, or a representative time-to-payment distribution.
The two architectures can be compared along the same dimensions: what is measured, what geography represents the exposure, what observation binds the calculation, how payment varies, and which additional conditions apply. PGV and Shindo belong to distinct measurement and institutional contexts; their numerical levels cannot simply be equated.
Haiti: modeled losses and event discrimination
CCRIF SPC's August 14 final briefing (2021a) reports that a magnitude 7.2 earthquake produced a modeled PGA footprint of approximately 0.01 to 0.46 g in Haiti. The final calculation showed 39,953,272 dollars due under the country's earthquake policy. The report describes a model-reporting protocol involving magnitude at least 5.0 and modeled PGA at least 0.01 g within relevant member territory. These conditions initiate reporting; financial attachment remains separate.
The August 15 final briefing (CCRIF SPC, 2021b) concerns a magnitude 5.8 earthquake and modeled acceleration of approximately 0.01 to 0.14 g. It reports modeled losses below the attachment and no payout due; a separate aggregated deductible feature was also not activated. The briefing emphasizes difficulty attributing consequences of the smaller earthquake separately from the preceding event. A nonpayment is not evidence of zero actual loss.
The first briefing includes an event-definition footnote describing grouping within a 25-day interval and 50-kilometer radius and selection by greatest modeled loss. This documents one disclosed occurrence provision. It is insufficient to resolve all event-clustering interpretations or establish that every subsequent event discussed in a separate report fell into that group. The actual full contemporaneous policy and geographic determination were not available.
The comparison shows how qualifying physical observations, modeled loss, contractual attachment, and payment decisions are separate levels of determination. The reported due amount should not be confused with an independently bank-verified payment date or an independently reproduced loss estimate.
Mexico: earthquake criteria and capital-market transfer
The World Bank's April 17, 2024 Mexico catastrophe bond issuance documents three initial classes with a combined amount of 420 million dollars. The two earthquake classes account for 225 million and 70 million dollars, totaling 295 million; the remaining 125 million concerns Atlantic named storms. The disclosed instrument uses per-occurrence parametric criteria based on event location and severity. Principal reductions finance payments through the specified intermediary arrangement.
This evidence establishes the structure of the initial issuance, rather than all subsequent transactions or the precise local shaking trigger of each bond. The press release does not provide complete calculation terms or realized payout data. Its substantive body and transaction date support the year 2024 despite an inconsistent incidental metadata field.
Capital-market certainty requires an unambiguous determination because a qualifying event can reduce investor principal. A sponsor needs correspondence between the event criteria and fiscal exposure. The location-and-severity comparator helps distinguish source-oriented mechanisms from direct local shaking criteria and modeled-loss triggers. The appropriate index depends on the actual transfer terms and portfolio.
Canada and China: scientific calibration studies
Goda (2013) compares station and scenario triggers for a modeled portfolio of 2,000 conventional wood-frame houses in southwestern British Columbia. The accessible abstract reports that the station-intensity approach performed at least as well on aggregate trigger error and that incorrect spatial-correlation models introduced further errors. That conclusion remains specific to the studied portfolio, methods, and evaluation criteria.
Pai, Li, Yang, and Li (2022) examine Bayesian spatial quantile regression in the context of earthquake parametric insurance in Yunnan. The available abstract describes historical loss data from 1992–2019, exposure and vulnerability as spatially latent quantities, and calculated mechanisms and premiums for 16 regions. The authors report an improvement in basis risk relative to their comparator. Numerical effect sizes, diagnostics, actual comparator details, and implementation were not independently inspected.
The studies identify complementary spatial questions: whether the intensity observed at a station or represented by a scenario captures relevant loss, and whether a statistical model can represent regionally varying exposure and vulnerability. Their dependent variables and portfolios differ. Pooling their effects or transferring their thresholds directly would require additional evidence.
Cross-case evidence structure
| Case | Financial object | Disclosed index | Evidentiary boundary |
|---|---|---|---|
| California specimen | Household or business loss | Census-block PGV, 24-hour map | Form text, no claims data |
| Japan case | Corporate supplemental liquidity | JMA Shindo tier | Insurer report, no full terms |
| Haiti | Sovereign earthquake financing | Modeled loss from PGA | Reported model, no independent replication |
| Mexico | Sovereign catastrophe financing | Event location and severity | Issuance disclosure, no complete bond terms |
| British Columbia | Modeled residential portfolio | Station and scenario comparison | Abstract-level evidence |
| Yunnan | Regional economic loss | Spatial statistical calibration | Abstract-level evidence |
The cases were selected for diversity of measurement and financial architecture. They do not constitute a representative international sample. Their contribution is to establish concrete possibilities and expose the distinct kinds of evidence needed for stronger evaluation.
Basis Risk, Error, and Alternative Explanations
Financial target and residual exposure
Let F describe a contract-relevant financial target. It may be immediate uninsured cost, a deductible, an emergency spending requirement, or a selected portion of total loss. Let P be the net payment available at the required time. The distribution of F−P across relevant events and exposure states describes residual need relative to that target.
The NAIC discussion of basis risk identifies mismatch between an indexed payment and experienced loss as a central limitation of parametric products. Monetary shortage and excess measures capture different aspects of that mismatch. A negative residual relative to a narrowly defined research target may be a legitimate payment under broader policy terms. It is not, by itself, a legal finding of overinsurance. Likewise, a positive payment does not establish that the relevant financial need was satisfied.
An evaluation should identify the time horizon and the financial target before computing differences. A modest early benefit can be adequate for a specific liquidity purpose while covering little of a building's ultimate replacement cost. Comparing it to the latter alone can mistake its intended scope.
Sources of error
Physical estimation error arises when the accepted map or instrument value differs from the relevant actual motion. Spatial representation error arises when the index location or area is poorly aligned with an exposure. Financial translation error arises when accurate motion is a weak predictor of the financial target because of vulnerability, assets, dependencies, or recovery conditions. Contractual mismatch arises from payment schedules, limits, grouping, exclusions, or eligibility. Timing mismatch arises when the amount is determined correctly but funds arrive too late for the defined objective.
These are analytical distinctions among known mechanisms. They can interact and cannot always be identified separately from observed payment data. The remedy depends on the origin of error: improved seismic mapping can correct one problem; altered exposure representation or payout design can address another; more reliable administration can address a delay.
Consider an index estimate with latent true value X and conditional normal uncertainty of mean mu and standard deviation sigma in a suitable scale. The conditional probability of X at least A is 1 minus the standard normal cumulative probability at (A−mu)/sigma. This provides a diagnostic measure of uncertainty near a threshold under a stated distributional assumption, not an alternate settlement rule. Positive and skewed ground-motion quantities may require transformation and a different uncertainty distribution.
The use of a deterministic threshold creates a discontinuity relative to perturbed observations. Measurement uncertainty should therefore be examined particularly in the neighborhood of attachment and exhaustion. A product can achieve a low average estimation error and still produce significant errors in liability if observations often occur near a financially material threshold.
Validation denominators
For a predetermined material need threshold f, classify a need event by F at least f and a payment event by P greater than zero. The false-negative rate is the share of need events with no payment. The false-positive rate is the share of events below f that receive a payment. Neither denominator is the entire event catalog. A model that nearly always predicts no payment can look accurate in a catalog dominated by small events and fail to protect high-need cases.
Appropriate quantitative reporting includes conditional error rates, expected shortage and excess, payment-to-target ratios where denominators are meaningful, tail shortage, and the distribution of credited-fund times. The original numerical example demonstrates how a mean improvement can coexist with a worse extreme-state outcome. The target error metric must be chosen based on the financial objective.
Calibration and validation should be separated. Fitting threshold values and evaluating the resulting performance on the same events yields optimistic results. Geographic and temporal holdouts help assess transferability. Closely related events in a sequence should be grouped for evaluation so that training and testing do not contain near-duplicate observations.
Vulnerability and networks
Shaking at a site can be accurately estimated while its losses vary substantially. A modern building and an older structure can respond differently to the same motion. Sensitive machinery can fail in the absence of major structural damage. An enterprise can lose operating capacity through disruptions to electric supply, communications, transportation, staff access, or critical suppliers.
The JMA utility and infrastructure tables describe potential service interruptions and long-period motion effects. Their relevance is mechanistic: local shaking can affect economic function through more than direct building damage. They do not establish uniform outage thresholds or business interruption durations across enterprises and countries.
Liquefaction, landslides, and tsunami require particular attention because their economic consequences can be weakly summarized by a simple shaking statistic. A policy must define the eligibility of resulting losses. Inclusion of other hazards may require independent measurement and validation. A shaking index is not automatically an adequate description of every earthquake-generated financial exposure.
The distribution of vulnerability can shift due to retrofits, building replacement, economic development, and changes in infrastructure. Monetary exposure can increase even when engineering vulnerability declines. Historical losses should therefore be normalized in a way that reflects the intended objective and the exposure at each event date.
Dynamic seismicity and model updates
Franco and colleagues (2020), in their California dynamic-risk experiment, report a retrospective assessment of parametric risk transfer updating. Their abstract reports an average frequency swing of 16% and a maximum of 36% over three-year periods in the investigated setting, and average and maximum reductions in their cover-efficiency measure of 13% and 35% without annual updating. These numbers are attributed to their experiment; the full paper was not accessed and the results are not universal estimates.
Updating the pricing model and changing the binding trigger are separate actions. Scientific knowledge may change estimates for renewal while the existing policy retains its established coverage. Dynamic adjustment during a contract requires authorized and predeclared terms or another legally valid mechanism. Retrospective alteration of a threshold after an event would destroy certainty.
A governance procedure should record the reason for each data or model change, affected terms, effective date, validation evidence, and previous versions. Recalibration should be tested on previously untouched events where possible. Performance changes can otherwise reflect tuning to the very circumstances used to justify revision.
Operationally faithful backtesting
A retrospective study starts with exact contract wording, a dated exposure inventory, and an archived event catalog. Each event receives the observation data that would have been admissible at the contract time. The actual geographic, aggregation, missing-data, and payout rules are then applied. The generated payment is compared with independently collected target data.
Later, more accurate maps can be used for a separate sensitivity exercise. Substitution of a final map for the required earlier version introduces hindsight bias. A complete backtest also includes nonpaying events and missing reports. A missing loss entry must distinguish unknown from zero; otherwise false-negative and severity estimates become unreliable.
Stochastic event catalogs can supplement observational data for rare tail losses. The event-based hazard workflow provides a documented architecture for combining rupture realizations and ground-motion fields with exposure and vulnerability. Its output remains conditional on model assumptions. An insurer should assess alternative hazard, spatial correlation, vulnerability, and availability models, including malfunctioning observation systems and incorrect exposure coordinates.
A reliable comparison would fix the expected payout budget or another pricing constraint and use the same event and loss set across different index structures. Sampling uncertainty should be reported, including scarcity of extreme events. The original example gives an arithmetic template, not an estimate of real-world basis risk.
Alternative interpretations of case evidence
Institutional reports can highlight events where the product worked. Without a known denominator, such cases do not establish the probability of missed payment. The same selection problem applies to widely publicized negative outcomes. Case evidence can establish terms, observed decisions, and operational problems; statistical performance requires representative or appropriately weighted observations.
Correlation of a payment with economic loss may partly reflect asset concentration, regional wealth, measurement availability, or reporting intensity. Comparing exposure-adjusted losses and financial needs can clarify which relationship the index captures. A ratio of loss to insured value and an absolute emergency funding amount are distinct targets.
Claims about faster recovery require comparative evidence about what would have happened with alternative financing. A reported settlement speed establishes a time interval, not the incremental economic effect of the payment. The reviewed cases establish arrangements and disclosed determinations. They do not independently identify the causal welfare effects of the programs.
Integrated Scientific Synthesis and Research Agenda
Financial purpose and time horizon
The design of a shaking-intensity contract begins with a specified financial target and horizon. A household may need cash shortly after displacement, while a sovereign program may finance emergency procurement. A commercial enterprise may seek a deductible supplement or a reserve to maintain payroll. The definition shapes permissible evidence, the attachment, the payout amount, and the significance of latency.
If C(t) represents cumulative eligible expenditure by time t and R(t) represents other available funding, the unmet need before the parametric payment is max(C(t)−R(t),0). A payment received at time tau reduces that need only for horizons at or after tau. This accounting observation explains why a settlement interval belongs in a quantitative measurement of product performance.
The Japanese reported interval from event to funds illustrates the distinction between observation and confirmation latency and subsequent administrative speed. A future outcome database should record event origin, time of qualifying data, determination, transfer initiation, and credited-funds time individually. Comparisons should use a common start and end definition.
Choosing an index against exposure
A site or station index may be a transparent starting point where one facility dominates the financial target. A geographically weighted footprint or modeled-loss index can be considered for a distributed portfolio. An event-oriented magnitude and location comparator can suit certain capital-market arrangements. None can be ranked independently of the objective, data quality, and terms.
A meaningful calibration compares candidates against a simple benchmark and a common financial objective. A complex algorithm incurs maintenance and interpretability costs. Its improved statistical fit has limited value if the data cannot be maintained reliably or if settlement decisions cannot be independently reproduced. The gain must be assessed out of sample and against the contractual and operational burden.
Exposure used to calibrate a schedule should match the intended loss concept. An enterprise's insured replacement value may be an inadequate weight for its dependence on a facility. A geographic data provider can accurately represent shaking while the chosen weights misrepresent the financial problem. Both parts need independent validation.
Operational resilience
A complete operational process identifies the party responsible for each stage. Observation products are selected under a fixed rule; their spatial and temporal transformations are reproduced; eligibility and the amount are reviewed; the payer records the final instruction. Negative decisions require the same discipline as positive ones. An absence of communication after an earthquake may conceal calculation delay, data unavailability, or a genuine failure to meet contractual conditions.
Test scenarios should include exact threshold equality, values just above and below attachment, conflicting map versions, out-of-grid points, missing stations, a compound event, and a revision of event time near policy expiration. The exposure registry should be tested for duplicate locations and invalid coordinates. Round only at the point and precision stipulated in the policy.
The protocol for unavailable data is part of the product's risk transfer. Replacing a sensor, extending the observation window, or selecting an alternate model can change the payment distribution. Fallbacks should be included in underwriting and independent validation rather than treated as incidental implementation decisions.
Finality rules preserve the value of public and independent observation. They should specify when corrections are allowed, who authorizes them, whether previously disbursed amounts are affected, and how a dispute is resolved. Original determinations and supporting evidence remain part of the audit history.
Legal boundaries and coordination
Parametric instruments may have distinct legal forms depending on applicable jurisdiction and contract. The regulatory overview notes differences in insurance-law requirements and the possible importance of proving loss. The California specimen demonstrates objective event eligibility coupled with requirements of legal interest and covered financial loss. It does not establish the legal treatment of every fixed-payment policy or derivative.
A comparative legal study would require current statutes, regulatory rules, filings, and case law in each jurisdiction. The present document identifies contract questions supported by the reviewed specimen and institutional overview; it does not supply a universal legal classification. Purchaser disclosures should identify payment conditions, scenarios of substantial need without payment, limits, required reports, governing observation data, and correction procedures.
A parametric contract can supplement indemnity, especially where financing retained expenses or early needs is the objective. The parametric-indemnity coordination discussion describes this broad possibility. Actual coordination depends on whether the contracts are independent, whether payments are additional or advance amounts, and which costs are covered. The combination must be assessed under its actual terms.
Disaster finance can include reserves, credit, indemnity coverage, index payments, and contingent public or capital-market resources. Each provides different timing, cost, and risk allocation. A comprehensive financial comparison evaluates resources net of carrying costs and premiums and identifies the residual need under stress scenarios. Nominal coverage limits alone do not measure timely protection.
Original synthesis and research agenda
The scientific contribution is an integrated interpretation of the measurement-to-payment relationship. The reviewed mechanisms establish that a threshold has several components: physical metric, geography, observation version, event definition, eligibility, and monetary schedule. A model-derived monetary quantity can itself be a contractual index. Accurate measurement and correct application of the contract are compatible with significant residual financial need.
The original calculation demonstrates that equal expected payments at the gross payout level do not determine the allocation of protection among event states. It also distinguishes mean error from severe-state shortage and quantifies local sensitivity to observation revisions. These results are analytical consequences of explicit assumptions and known mathematical operations. No new theory, doctrine, empirical discovery, or scientific priority is claimed.
Further research should obtain archives of exact event products at operative timestamps, complete contract terms and endorsements, independent event-level payment and financial-need records, and applicable jurisdictional authorities. Data collection must include nonpayment events, sensor outages, map revisions, invalid exposure coordinates, and different vulnerability conditions. A standardized record of the reasons for each decision would enable direct comparison of observational, contractual, and financial error.
A future empirical study could separate households, corporate facilities, networked businesses, sovereign programs, and capital-market instruments; validate thresholds within each relevant population; and report uncertainty in estimates from rare severe earthquakes. It would use temporally and geographically separated evaluation sets, exposure histories, and alternative hazard and vulnerability assumptions.
The highest-priority unresolved empirical boundary is the absence of paired independent loss, payment, and measurement records for the reviewed contracts. The accessible documents do not support estimating actual basis-risk distributions or causal recovery effects. Restricted academic work and inaccessible document imagery also constrain the depth of independent scholarly verification. These limitations are disclosed rather than replaced with apparent numerical precision.
Conclusion
Earthquake shaking intensity parametric insurance establishes a contractual relationship between earthquake evidence and a scheduled financial payment. Source magnitude represents earthquake size, whereas PGA, PGV, spectral acceleration, and intensity describe distinct aspects of local motion. A modeled-loss index transforms selected physical, exposure, and vulnerability information into a contractually relevant monetary quantity.
The reviewed documents show distinct approaches. The California wording couples territorial PGV eligibility with covered loss and legal interest. The Japanese corporate example uses a national instrumental-intensity tier to deliver supplemental financing. The Caribbean event reports distinguish physical reporting criteria and modeled loss from financial attachment. The Mexican transaction demonstrates event-oriented criteria in sovereign capital-market transfer. Accessible scholarly findings identify important spatial and temporal risks in calibration.
Scientific adequacy depends on a reproducible measurement, a geography and index relevant to the financial exposure, a schedule that allocates payment to appropriate event states, and a timing rule consistent with the funding objective. The original illustrative calculation demonstrates that two contracts with equal expected gross payment can generate different ordinary and severe-state shortages. Evaluation requires the full distribution of residual financial need, not only agreement with physical intensity or the mean payout.
The credibility of earthquake parametric protection rests on the combined quality of observational evidence, contractual specification, exposure representation, actuarial validation, and settlement administration. Their distinct roles provide the basis for evaluating actual instruments and specifying the additional evidence required for more complete empirical research.
References
U.S. Geological Survey (USGS). (n.d.). Earthquake magnitude, energy release, and shaking intensity. Earthquake Hazards Program. Source record. Retrieved for research November 18, 2026.
U.S. Geological Survey (USGS). (n.d.). ShakeMap. Earthquake Hazards Program. Source record. Retrieved for research November 18, 2026.
U.S. Geological Survey (USGS). (n.d.). ShakeMap product documentation. Product Distribution Layer User Guide. Source record. Retrieved for research November 18, 2026.
Worden, C. B., Wald, D. J., & Rhoades, D. A. (2012). Probabilistic relationships between ground-motion parameters and Modified Mercalli intensity in California. Bulletin of the Seismological Society of America, 102(1), 204–221. DOI: 10.1785/0120110156. Source record. Retrieved for research November 18, 2026.
Japan Meteorological Agency (JMA). (n.d.). Tables explaining the JMA seismic intensity scale. Monitoring of Earthquakes, Tsunamis & Volcanic Activity. Source record. Retrieved for research November 18, 2026.
Global Earthquake Model Foundation (GEM). (n.d.). Overview of the OpenQuake engine. OpenQuake Engine documentation, version 3.26.2 as identified in retrieved page. Source record. Retrieved for research November 18, 2026.
National Association of Insurance Commissioners (NAIC). (n.d.). Parametric disaster insurance. Insurance Topics. Source record. Retrieved for research November 18, 2026.
Jumpstart / Certain Underwriters at Lloyd's. (n.d.). Earthquake insurance policy form (California, JS POL CA). Six-page policy form, sections II–V and VIII. Source record. Retrieved for research November 18, 2026.
Swiss Re Corporate Solutions. (2021, May 21). Customer case study: Parametric QUAKE Japan. Source record. Retrieved for research November 18, 2026.
CCRIF SPC. (2021a, August 22). Haiti—Earthquake; 14 August 2021: Final event briefing. Source record. Retrieved for research November 18, 2026.
CCRIF SPC. (2021b, August 25). Haiti—Earthquake; 15 August 2021: Final event briefing. Source record. Retrieved for research November 18, 2026.
World Bank. (2024, April 17). World Bank issues $420 million in catastrophe bonds for renewed disaster risk protection for Mexico. Press release and transaction summary. Source record. Retrieved for research November 18, 2026.
Goda, K. (2013). Basis risk of earthquake catastrophe bond trigger using scenario-based versus station intensity–based approaches: A case study for southwestern British Columbia. Earthquake Spectra, 29(3), 757–775. DOI: 10.1193/1.4000164. Source record. Retrieved for research November 18, 2026.
Pai, J., Li, Y., Yang, A., & Li, C. (2022). Earthquake parametric insurance with Bayesian spatial quantile regression. Insurance: Mathematics and Economics, 106, 1–12. DOI: 10.1016/j.insmatheco.2022.04.007. Source record. Retrieved for research November 18, 2026.
Franco, G., Guidotti, R., Field, E. H., Milner, K. R., Lee, Y. J., & Stein, R. S. (2020). An exploration of parametric earthquake risk transfer solutions that dynamically adapt to seismicity changes. Conference paper. USGS Publications Warehouse, record 70208800. Source record. Retrieved for research November 18, 2026.