RESEARCH DOSSIER
Catastrophe Bond Sponsor Strategy and Layer Selection
Abstract. Catastrophe bond sponsorship is a portfolio-financing decision in which a risk-bearing institution chooses which contingent losses to transfer to capital-market investors, where to locate the protection in a broader risk-financing structure, what event definition should determine payment, and how to sustain protection through successive underwriting and catastrophe seasons. The economic object is the sponsor's net loss and funding distribution across states and time, rather than the amount of bond principal issued. This monograph examines the strategy through primary multilateral transaction records, international supervisory standards, legal rules on insurance special-purpose vehicles, and peer-reviewed studies of catastrophe-linked securities. It derives occurrence-layer and aggregate-layer payoffs, differentiates contractual attachment from hazard intensity thresholds, incorporates parametric basis risk and payment delay, and evaluates the joint effect of retentions, traditional reinsurance, capital relief, funding liquidity and renewal exposure. An explicitly hypothetical distribution demonstrates that alternative attachments can deliver substantially different expected payments while providing equal protection in the most severe modeled state. Documentary comparisons of Mexico's 2024 multiclass coverage, Chile's 2023 bond-and-swap placement, and Jamaica's 2025 payout and 2026 replacement show distinct sponsor solutions to peril segmentation, market-capacity constraints, and funding continuity. The principal result is an integrated analytical interpretation: the economically appropriate layer is conditional on the sponsor's loss distribution, funding deadlines, feasible trigger, existing protection tower, solvency treatment, transaction costs, and multi-year objectives. A cat bond's successful issuance therefore provides evidence of completed risk transfer, while protection adequacy remains a separate question requiring portfolio-specific testing.
Keywords: Catastrophe bonds; insurance-linked securities; sponsor strategy; attachment point; exhaustion point; occurrence excess of loss; aggregate protection; catastrophe reinsurance; parametric basis risk; modeled expected loss; tail risk; risk financing; collateral; capital relief; sovereign disaster finance; multi-period renewal.
Introduction
The sponsor of a catastrophe bond purchases contingent payment capacity from investors who can lose specified principal when a defined catastrophe condition occurs. In a conventional collateralized insurance-linked securities arrangement, the sponsor's risk-transfer contract is associated with a special-purpose vehicle or comparable issuance structure; investors supply risk capital, receive a coupon funded by collateral earnings and a risk premium, and bear principal reductions subject to the documents. The World Bank's capital-at-risk explanation sets out this core arrangement, while the NAIC's insurance-linked securities review describes its role in insurer risk transfer. Economic incidence depends on the actual indemnity, index or parametric trigger and the recourse and collateral structure; the word “bond” alone specifies neither the protection boundary nor the sponsor's recoverability.
The sponsor decision has at least five connected dimensions: the insured peril and geography; the attachment and exhaustion points relative to own losses; the loss definition or physical trigger; the interaction with existing retained, reinsured and funded layers; and the coverage term, annual or aggregate limits, and renewal alternatives. A nominally high layer can be valuable when a rare payment prevents capital distress, whereas a lower layer can reduce expected claims volatility or release cash more often. A cheap premium may reflect limited coverage or basis risk. High modeled recovery can coexist with a long claims-development period. The IAIS Insurance Core Principles expressly integrate reinsurance and other risk transfer into risk management and capital assessment.
The monograph addresses three questions. Which losses does a sponsor need to finance after accounting for operating cash, capital constraints and existing reinsurance? Which layer and trigger best match those losses under uncertainty, including mismatches between model output and realized recovery? How should an instrument with a specified multi-year term be evaluated when portfolio composition, reinsurance prices, investor capacity and catastrophe conditions evolve? The analysis answers them through source-critical synthesis, contractual arithmetic, risk-distribution comparison and actual transaction examples. No proprietary exposure file, individual bond offering memorandum, issuer-level capital model or causal econometric study is claimed. Where an illustration uses invented losses, probabilities or prices, the figures are deliberately hypothetical and have no empirical calibration.
Definitions, Source Record, and Scientific Boundary
The sponsor's contingent asset and investors' contingent loss
A catastrophe bond is an instrument that links investors' interest or principal to an independently specified catastrophic event, loss index or insured loss. In the usual fully collateralized structure, investor principal provides resources against a defined contractual liability. It is crucial to distinguish the risk-bearing sponsor, the formal issuer, a fronting insurer or reinsurer, the collateral custodian and a calculation agent. The World Bank notes program demonstrates a structure in which the World Bank itself issues capital-at-risk notes and transfers a sovereign client's risk to investors under a corresponding risk-transfer agreement. The ultimate client and the security issuer are institutionally distinct.
The scientific unit here is a mapping from event states and contractual conditions to sponsor cash receipts. Let the sponsor's gross catastrophe loss in an event be L, and let a qualifying loss measure be X. For an indemnity instrument, X can approximate eligible sponsor loss after adjustments defined in the contract. For an industry-loss or parametric instrument, X can instead be a third-party index or a physical measurement. The eligible event definition, region, peril, loss-occurrence dates, hours clause, covered exposure and annual aggregate structure determine the event population on which the mapping is evaluated. Two bonds with the same stated limit can therefore protect very different subsets of the sponsor's loss distribution.
Define attachment A, limit C and exhaustion E=A+C in the same monetary loss units. The idealized occurrence-layer recovery for a qualifying indemnity loss X is P(X)=min{C,max(0,X−A)}. It equals zero below A, rises dollar-for-dollar in the layer, and is capped at C above E. If X equals eligible gross loss L, the sponsor's retained loss immediately after the bond is R(L)=L−P(L), before other cover, premiums, taxes and claims expenses. If the trigger is parametric, the actual transfer is P(Z), a function of event measurement Z, and net loss is L−P(Z). These are different random variables. Attachment in monetary loss space must not be confused with the wind-pressure, magnitude, geographic or index threshold of a parametric contract. The practical linkage between them is a modeled calibration, not an identity.
The basic instrument can include occurrence limits, annual aggregate deductibles, franchise terms, reinstatement provisions, multi-event principal reductions, event definition disputes and post-event collateral release conditions. Such features change the cash-flow function. The seminal Cox and Pedersen valuation analysis explains why catastrophe security valuation belongs to incomplete markets; it cannot generally be obtained by replicating a catastrophe payoff with ordinary traded assets.
What is being optimized
Sponsor strategy is not reducible to minimizing expected premium per dollar of nominal limit. An insurer values reduced insolvency probability, lower tail capital demand, stabilization of earnings, capacity to write business and settlement reliability; a sovereign may value emergency liquidity and fiscal continuity. The Cummins and Weiss synthesis locates securitized transfer within wider insurance–capital-market convergence and the search for complementary capacity. The sponsor's relevant objective may be formulated as minimizing total economic cost subject to survival, liquidity and risk-appetite constraints, or maximizing value under specified constraints. These conventional formulations organize the analysis; this work establishes no new optimization theory.
An economically useful cost expression is K=Premium+IssueCosts+OperationalCosts+ExpectedResidualLoss+CapitalCost+LiquidityShortfallCost−RecognizedTransferBenefit, with each term evaluated over a common horizon and with care against double counting. Capital cost and residual loss must be defined consistently with the insurer's valuation method: a risk-adjusted cost of capital already reflecting tail risk cannot simply be added to an independent duplicate charge for the same tail. The sponsor also distinguishes annual contractual premium from a financed coupon: the latter combines investment collateral return, funding margin and risk margin under the actual notes. Transaction disclosures provide exact components only for specific issues.
Hierarchy and limitations of documentary evidence
Primary sources establish their own propositions: issued transaction terms, amount and maturity in an issuer announcement; governing requirements in a regulatory instrument; payout determinations in an official transaction notice; and storm data definitions in the relevant meteorological archive. They do not by themselves establish an issuer's unreported true expected loss, optimal attachment, realized all-in cost or portfolio-wide counterfactual. Peer-reviewed papers furnish models and empirical evidence within their samples, with different access constraints. Braun's primary-market pricing study used cat-bond tranches issued from June 1997 through December 2012 and found expected loss to be the most important spread determinant alongside territory, sponsor, reinsurance cycle and comparable corporate-bond spreads. That historical empirical result informs, but does not mechanically predict, an individual 2026 price.
A geographic or regulatory claim requires its own source. The IAIS ICP 13 standard is an international supervisory benchmark; it is not a substitute for national legislation. For EU-regulated undertakings, Solvency II's special-purpose-vehicle provision identifies an authorization framework, while Delegated Regulation Article 326 specifies criteria concerning full funding and liquidity. The exact jurisdiction, instrument vintage, eligibility of risk mitigation and net solvency benefit must therefore be assessed separately.
The distinct research boundary
The research object is the sponsor's integrated choice of perils, geography, layer, trigger and term. Standalone catastrophe-bond issuance mechanics, securitization history, investor secondary-market liquidity, collateral waterfalls and risk-spread estimation are adjacent areas. They are treated here when needed to explain the sponsor's economic decision, while the independent question remains: under which exposure and funding conditions does a particular layer actually improve the sponsor's position? This boundary also prevents a model demonstration from being mistaken for a proprietary transaction recommendation.
Insurer Portfolio Risk Financing Objectives
Loss, liquidity and capital are different dimensions
Gross event loss is typically generated by the conjunction of hazard intensity, exposed insured values, vulnerability and contractual insurance coverage. Insurer gross loss is further affected by deductibles, policy limits, business interruption, claims inflation, loss adjustment and allocation rules. A sponsor should first model its gross and net distributions across events, years and legal entities, not merely select a quoted market attachment. Geographic concentration matters because an index calibrated to diversified or regional loss can track a concentrated book poorly. In a classic simulation of Florida insurers, Cummins, Lalonde and Phillips (2004) found material differences in index hedge effectiveness by insurer size and subregional design; a single statewide index was comparatively effective mainly for the largest companies in their study.
An insurer's liquidity requirement concerns the timing of claim payments and incoming recoveries. A bond with strong ultimate indemnity recovery can still provide delayed cash if a covered-loss calculation takes years, while a parametric security can pay rapidly when a measurement threshold is confirmed, albeit with possible basis risk. The sponsor's liquidity map therefore records the time-dependent net cash shortfall Q(t)=ClaimsPaid(t)−PremiumsReceived(t)−LiquidAssets(t)−RecoveriesReceived(t)−OtherFunding(t). A decision to transfer a layer should include the probability and severity of Q(t)>0 over operationally important deadlines. This is separate from the distribution of total loss at final settlement.
Solvency concerns the sponsor's capacity to absorb unforeseen loss and maintain regulatory, rating or internal capital thresholds. The IAIS enterprise-risk-management framework connects risk appetite, capital adequacy and the own risk and solvency assessment. Economic capital relief depends on the reduction in the capital metric for net losses and the supervisor's recognition of the transaction, rather than the bond's face value. A transfer of $300 million at an exposure layer that never contributes to the relevant capital quantile may bring little incremental capital relief; the same limit may be decisive when it interrupts the net-loss states driving the solvency constraint.
The shape of the catastrophe loss tail
For a loss variable L and deterministic indemnity recovery P(L), expected recovery is E[P(L)], while the impact on tail risk depends on the joint distribution of L and P. An approximate expected annual loss metric can be estimated from a validated stochastic event set: E[P]=sum over events of annual occurrence frequency times modeled layer payout, with adjustment for clustering, multiple events and contract aggregate provisions. If a Monte Carlo event-year catalog contains correlated storms or earthquakes, events must be aggregated into contract years before annual attachment and exhaustion calculations. Treating all event records as independent one-event years biases aggregate treaty evaluation.
A sponsor can examine occurrence exceedance probability (OEP), aggregate exceedance probability (AEP), value-at-risk and tail value-at-risk, with precise event-year and confidence conventions. For tail value-at-risk at level alpha in a continuous distribution, TVaR_alpha(L)=(1/(1−alpha)) times the integral from alpha to 1 of the loss quantile. Discrete distributions require consistent treatment of probability mass at the quantile. Catastrophe model estimates also have epistemic uncertainty from incomplete hazard records, vulnerability curves, climate assumptions and policy data. A modeled 1-in-200-year loss is a model quantile, not a scheduled physical event, and its numerical value is sensitive to assumptions and model vintage.
A hypothetical loss distribution
Consider a fictional sponsor with four mutually exclusive annual gross catastrophe loss states, in millions of U.S. dollars. The states are $150 million with probability 0.80, $450 million with probability 0.12, $700 million with probability 0.06, and $1.1 billion with probability 0.02. The probabilities sum to one; they are invented to make the arithmetic inspectable and do not describe an actual insurer or catastrophe catalog. The sponsor is comparing an occurrence layer attaching at $400 million and exhausting at $800 million, with a $400 million limit, against one attaching at $600 million and exhausting at $1 billion, also with a $400 million limit.
| Hypothetical annual gross loss (USD million) | Probability | Recovery: 400 xs 400 | Recovery: 400 xs 600 |
|---|---|---|---|
| 150 | 0.80 | 0 | 0 |
| 450 | 0.12 | 50 | 0 |
| 700 | 0.06 | 300 | 100 |
| 1,100 | 0.02 | 400 | 400 |
Here “400 xs 400” denotes a $400 million limit in excess of a $400 million attachment, and “400 xs 600” denotes the same limit in excess of $600 million. The lower layer has expected payment 0.12×50+0.06×300+0.02×400=$32 million per hypothetical year. The higher layer has 0.06×100+0.02×400=$14 million. The two provide the same $400 million payment in the most severe enumerated state. Accordingly, equal protection at the worst modeled state can accompany a $18 million difference in expected recovery and substantial differences at intermediate losses. This arithmetic warns against comparing equal limits only at an extreme return-period label.
The lower layer's second moment of payout is 0.12×50²+0.06×300²+0.02×400²=$8,900 million-squared, yielding variance 8,900−32²=7,876 and standard deviation about $88.75 million. The higher layer's second moment is 0.06×100²+0.02×400²=$3,800 million-squared, yielding variance 3,800−14²=3,604 and standard deviation about $60.03 million. These are payout variances, not the variance of the retained loss or the capital benefit. To identify the latter, each state's net loss must be recomputed and the relevant risk measure applied.
If a hypothetical quoted all-in annual economic cost were $67 million for the lower layer and $39 million for the higher layer, those offers could be compared only after identifying the sponsor's net-loss risk appetite, financing benefit, counterparty and term provisions. The premiums are invented, not drawn from a market benchmark; their purpose is to show why expected payment, price and net economic value are separate quantities. Different investors may price the two layers with different tail loads and correlation concerns. The Braun pricing evidence supports testing modeled expected loss and market factors rather than equating price with expected loss.
Retention as a strategic choice
Higher attachment preserves more working-layer risk and can reserve investor capacity for capital-threatening losses. Lower attachment reduces losses in more states, potentially stabilizing annual results, but uses capacity at relatively frequent attachment probabilities and may attract larger expected-loss and risk charges. The appropriate retention is influenced by the sponsor's liquid reserves, business mix, growth plans, parent support, reinsurance renewal position, rating sensitivity and the cost of carrying capital. Risk reduction at one modeled percentile should be compared with the entire residual distribution, especially where catastrophe years can contain multiple events or correlated perils.
A useful sponsor decision sheet therefore shows, for every candidate: gross loss OEP/AEP curves, net curves after existing covers, incremental bond payout, expected payout, one-year and multiyear exhaustion probability, retained tail measures, capital relief subject to recognition, expected and stressed cash-arrival dates, and all-in transaction costs. This is an analytical decomposition of existing tools rather than an invented universal scoring model.
Layer Tower Integration and Coverage Selection
Placement within a reinsurance tower
A catastrophe risk program is generally a set of complementary loss-absorbing claims. The sponsor may retain losses up to a working retention, buy occurrence excess-of-loss reinsurance in successive layers, purchase aggregate protection for frequency accumulation, and place a catastrophe bond above, beside or instead of part of that program. In many programs the financial exposure of a layer and the legal allocation of ceded claims differ because each contract has its own occurrence definitions, exclusions, reinstatements, annual limits and inuring clauses. The IAIS guidance on reinsurance programs requires effective management of the full program, including associated credit, operational and basis risks. A bond term sheet must be evaluated in the actual tower, not in isolation.
For a stylized tower with nonoverlapping limits C_j attaching at A_j, define payment P_j(L)=min{C_j,max(0,L−A_j)}. If contracts protect disjoint layers and respond to the same gross loss, total recovery is P_total(L)=sum_j P_j(L). When layers overlap or one contract inures to the benefit of another, the eligible loss of contract j must be defined after applicable prior recoveries, and a simple sum may overstate protection. An indemnity bond whose attachment is measured on net retained loss after another treaty cannot be substituted into a formula using gross loss without rewriting the attachment variable. Contract order is therefore a mathematical input.
For an illustrative single event of $900 million, suppose the sponsor retains the first $200 million; a conventional reinsurance treaty covers $300 million excess of $200 million; and a bond covers $400 million excess of $500 million. The three layers allocate $200 million to retention, $300 million to reinsurance and $400 million to investors, leaving zero gross loss above the combined $900 million boundary in this example. At $1.1 billion, $200 million remains above the top of the assumed tower and returns to the sponsor unless covered elsewhere. The same bond nominal can deliver different effective protection if conventional reinsurance is subject to a disputed hours clause, coverage exclusion or counterparty nonpayment, or if the bond's eligible loss is defined net of ceded recoveries.
A tower can contain gap risk. A layer ending at $450 million and another beginning at $500 million leave a $50 million unprotected band, whether intentionally retained or inadvertently created. It can contain overlap risk, where two contracts appear to protect the same segment but their clauses govern whether full simultaneous recovery is possible. It can contain exhaustion risk: the first event consumes principal or annual aggregate capacity before a later loss. It can contain settlement timing risk: one contract pays after parametric confirmation, another after developed claims are final. In every case, the sponsor's integrated net financial position, rather than a picture of stacked nominal rectangles, determines the value of the arrangement.
The marginal value of a layer
In the idealized occurrence-payoff function, holding the limit C fixed and shifting attachment A upward weakly reduces the payout for every fixed loss L. Where differentiable, the marginal decrease in payout from a small upward attachment shift is approximately the probability that L lies between A and A+C, adjusted for the fact that the exhaustion point shifts as well. More precisely, for a continuous L distribution and C constant, dE[P]/dA = −Pr(A<L<A+C). This follows from differentiation under the integral for the capped excess-loss payoff, ignoring atoms at boundaries. The identity describes expected ceded loss; it does not establish an optimal point. The premium schedule, capital effect, liquidity requirement and correlation with the existing tower all change when A changes.
If the sponsor instead holds exhaustion E fixed while moving attachment A upward, then C=E−A declines. The resulting change differs from a fixed-limit comparison, because the maximum recovery itself shrinks. Terms such as “higher attachment” or “more remote layer” are economically incomplete without specifying whether exhaustion or limit is being held constant. An accurate comparison must also state the monetary denomination, insured-loss definition and annual period.
The expected value of incremental recovery from adding layer k to an existing tower is E[P_k], but its marginal capital value is a change in a selected net-loss risk measure: DeltaCapital = Capital(L−P_existing)−Capital(L−P_existing−P_k), with the risk measure and capital conversion explicitly declared. If the bond pays on an index Z, replace P_k(L) by P_k(Z), preserving the modeled joint distribution of Z, L and the other recoveries. Capital relief may be small or even absent under a metric that falls at a loss percentile below the candidate attachment. By contrast, a layer that intersects the capital-defining quantile can have high economic value despite low expected payout.
Occurrence versus aggregate structure
An occurrence contract responds to eligible losses from a specified event as defined by its hours and geographic clauses. An annual aggregate contract responds to accumulated eligible losses, potentially subject to per-event deductibles, annual aggregate attachment and exhaustion, reset or reinstatement provisions. Suppose two independent qualifying storms cause eligible losses of $300 million each. A $400 million occurrence attachment pays zero for both storms if calculated separately. A $400 million annual aggregate attachment with a $400 million limit may pay $200 million, assuming the contract aggregates the same two losses without additional conditions. The protection has moved from severity to frequency accumulation even though the stated attachment and limit are numerically identical.
The sponsor should therefore model event-year realizations rather than only a single-event exceedance curve when considering an aggregate bond. The annual aggregate index, definition of qualifying event, how losses are reported, the treatment of noncatastrophe losses, whether principal recovers after a reduction, and collateral release following late reporting can change the economic outcome. A single annualized expected-loss number conceals these differences. Expected loss is also sensitive to seasonality: a three-year note covering three wind seasons with specific start and end dates can differ from a note lasting three full calendar years.
Peril and geographic segmentation
A single multicountry, multiperil bond can pool differently correlated risks and exploit investor demand for diversification, but its aggregate limit can be eroded by one covered peril before another manifests. Segmented tranches establish different premiums, expected-loss profiles, triggers and limits for different regions or hazards. Segmentation also creates documentation and issuance costs. A decision to split, for example, Atlantic and Pacific hurricane exposure should be tested against the sponsor's actual concentration, loss correlation, investor capacity and catastrophe model uncertainty. Geographic labels in a transaction announcement do not necessarily reveal the treaty's precise covered coordinates, exclusions or event definitions.
Mexico provides a primary-source demonstration of this distinction. The World Bank's 2024 Mexico transaction case study reports four classes totaling $595 million: $225 million low-risk earthquake, $70 million high-risk earthquake, $125 million Atlantic hurricane and $175 million Pacific hurricane. These are actual disclosed classes, not the four hypothetical loss states analyzed earlier. They show that a sovereign sponsor can distinguish risk intensity within the same earthquake peril and separate hurricane geographies. The case study states that earthquake parameters and hurricane structures were refined to improve risk matching and faster payment. It does not disclose enough information to establish that this combination is globally optimal, or to compute Mexico's proprietary distribution of net fiscal loss.
Bonds, reinsurance and simultaneous market access
The economic choice between a cat bond and reinsurance depends on price, duration, size, flexibility, documentation, modeled risk acceptance, credit risk, currency, collateral and settlement timing. Conventional reinsurance can offer customized structures, smaller limits, reinstatements and relationship-based flexibility, while bond issuance can offer multiyear terms, collateralized capacity, transparent marketed risk margins and access to a broader capital base. These are tendencies conditional on the transaction, not guarantees. The World Bank's 2026 product note states that smaller volume targets and non-USD risks often fit traditional reinsurance better, whereas large placements with price uncertainty may benefit from marketing into both markets.
Chile offers an observed joint-placement case. The World Bank's March 2023 Chile announcement reports $630 million of earthquake cover, of which $350 million came from a cat bond and $280 million from catastrophe swaps placed with insurance and reinsurance investors. The transaction was marketed simultaneously to the two markets to access more capacity than either would supply on its own. The published bond terms identify a parametric per-occurrence trigger, three-year period and a risk margin of 4.75% per annum above funding components. The risk margin is a contractual component of that particular placement; it is not the all-in annual economic cost of Chile's total program, nor a market-wide price for similar risk.
The principal sponsor implication is that investor-market access itself can be an option worth preserving, particularly when the traditional reinsurance market is concentrated or its pricing cycle diverges from capital market conditions. The work of Finken and Laux (2009) supports a distinct informational mechanism: appropriately designed index or parametric triggers can affect competition in reinsurance markets by reducing the sensitivity of an instrument to private sponsor loss information, provided basis risk remains tolerable. Their theoretical finding should be read within its assumptions and not equated with a verified reduction in every sponsor's realized total protection cost.
The portfolio decision across layers
The sponsor should compare candidate placement combinations using one exposure model and common accounting of all costs. A lower treaty layer and a higher bond layer may stabilize frequent losses and preserve multiyear high-severity protection; an overlapping placement may intentionally diversify counterparty or trigger failure risk, but it may also create duplicate premium for little additional recoverable loss. A high bond attachment can reduce the price and expected payout while leaving the sponsor sensitive to medium losses that impair earnings or liquidity. A lower attachment may consume more principal and raise renewal risk if it triggers repeatedly.
The final design needs a written bridge from model losses to contract loss definitions. The bridge identifies gross and net loss, per-risk or per-occurrence terms, applicable reinstatements, allocation among subsidiaries, expected and stressed claims development, collateral constraints, insurance or reinsurance tax treatment and solvency recognition. No amount of sophistication in a loss simulation can correct a contract that maps the wrong loss measure to the sponsor's actual balance sheet.
Trigger Suitability and Basis-Risk Tolerances
Trigger families and their economic incidence
The principal trigger families are indemnity, industry-loss index, parametric and modeled-loss mechanisms, with hybrids and specialized variants. An indemnity trigger references eligible losses of the sponsor as defined in the agreement; it can track the sponsor's true insured-loss experience closely but requires claims measurement, audit, possible development and fraud or incentive controls. An industry-loss trigger references an independently reported aggregate measure; it can reduce dependence on the sponsor's own reporting while generating mismatch if its portfolio's geographic or peril mix diverges from the index. A parametric trigger references physical event features, such as earthquake magnitude, hypocentral depth, location or tropical-cyclone central pressure and track; it can be independently checked and paid comparatively quickly, yet actual financial loss can differ markedly from the physical index. Modeled-loss triggers apply an agreed model to observed event parameters and a fixed exposure representation, shifting the basis-risk question toward model specification and exposure-vintage mismatch.
The NAIC discussion of cat bonds provides the market context, while the World Bank capital-at-risk examples show parametric structures used by sovereign clients. These terms require careful distinction: a sponsor can receive quick payment after a qualified physical event even when subsequent total economic damage is uncertain, while a separate damaging event can fall outside the contract's thresholds and produce no payment. The occurrence of a real catastrophe and the occurrence of a contractual trigger are two empirically distinct events.
Measuring basis risk
For an intended indemnity protection target P(L) and an actual index or parametric recovery P(Z), define shortfall B(L,Z)=P(L)−P(Z). Positive B identifies under-recovery relative to the target; negative B identifies over-recovery. A sponsor may evaluate the distribution of B, conditional expected shortfall E[max(B,0) | L is large], the probability of material shortfall, and liquidity-sensitive mismatch at specific post-event deadlines. The joint distribution of L and Z is indispensable. Correlation alone is insufficient: two trigger variables may be highly correlated yet generate different cash at a critical contractual threshold.
Suppose a hypothetical event creates $700 million in eligible insured loss, and the sponsor's desired $400 million excess of $400 million indemnity layer would pay $300 million. If a parametric index produces only $180 million under its contractual payout schedule, the event's under-recovery relative to the indemnity target is $120 million. This is a deliberately invented illustration of basis risk; it is not derived from a disclosed World Bank placement. Conversely, an index can pay when the sponsor's eligible loss is below indemnity attachment, producing negative B. The economic importance of these outcomes depends on whether the program is designed to indemnify insured loss, stabilize an overall balance sheet, or provide an independently defined disaster liquidity benefit.
The Cummins, Lalonde and Phillips basis-risk study found that Florida insurers' index-hedging effectiveness varied materially with firm size and intrastate regional segmentation. The lesson is methodological: an index that tracks a diversified aggregate book may be structurally unsuitable for a concentrated sponsor. An effective test uses the sponsor's location-level exposure, policy terms and settlement obligations together with historical and synthetic event realizations to derive the empirical joint distribution of eligible loss and index recovery.
Information asymmetry and the cost of alignment
An indemnity contract better tracks eligible sponsor loss but can expose investors to private portfolio information and claim-settlement incentives. An index contract limits some information dependence but transfers mismatch risk back to the sponsor. The Finken and Laux competition analysis shows how information-insensitive triggers may alter the competitive interaction with traditional reinsurers. Separately, Götze and Gürtler's moral-hazard study reports evidence of ex ante moral-hazard susceptibility among indemnity-triggered bonds, particularly in higher-expected-loss instruments, with a different finding for ex post moral hazard in their sample. Those results do not imply that every indemnity sponsor behaves opportunistically; they show why contract incentive design is an economic variable.
The sponsor's decision must weigh expected basis-risk cost against the price, claim-proof burden and incentive cost of indemnity settlement. A mathematically smaller basis-risk measure may be undesirable if its settlement takes years after the date by which the sponsor must pay policyholder claims. A rapid parametric trigger may be preferable for a government facing immediate disaster response costs even if it differs from the eventual measured aggregate damage. The choice is tied to the function being financed and the documentary conditions governing the payment.
Calibration, validation and event data
Parametric calibration begins with spatially resolved insured-loss or fiscal-cost data, event footprints, hazard parameters, explicit data vintages and a contract-consistent simulation. For cyclone triggers, the sponsor should identify whether the referenced source is provisional advisory data, operational best-track data or post-season reanalysis. The National Hurricane Center data archive documents HURDAT2 storm-track parameters and revisions; data series are not interchangeable merely because they contain storm coordinates and intensity. A trigger referencing a specified B-deck snapshot at a defined time can produce a different settlement outcome from a later revised climatological dataset.
The World Bank's Mexico 2024 case study reports finer earthquake grids and a depth condition, along with a move to earlier NHC B-deck data for hurricane payment calculations. The record says the B-deck information relevant to that contract became available about two weeks after the final public advisory, rather than the months sometimes required for later tropical-cyclone reports. The precise payment timetable remains contract dependent. The strategic point is that changing a data source is simultaneously a latency decision, a version-control decision and a potential basis-risk decision.
The testing procedure should include: simulated sensitivity to hazard intensity and coordinates around contractual boundaries; replacement of current exposure data with plausible future portfolio mixes; extreme-event scenarios outside the calibration record; alternative vulnerability assumptions; data outages and revision events; and model-agent disagreement. Conditional underpayment frequency is especially important when the sponsor is attempting to protect a solvency or sovereign emergency-funding tail. It is poor practice to optimize trigger correlation over all ordinary years when the actual objective is payment during a small subset of fiscal or claims-paying crises.
Claims certainty, discretion and documentary control
The sponsor should document who declares an event, how the calculation agent is appointed and replaced, which data are authoritative, when estimates become final, and how corrections, disputes and loss development are handled. For indemnity contracts, audit rights, reporting calendars, information rights and the treatment of allocated loss-adjustment expenses affect both economic and operational value. For parametric contracts, geographic grids, intensity interpolation, instrument thresholds, rounding and cutoff dates are payment terms, not merely modeling assumptions. Collateral release rules can influence the investor's principal availability and the sponsor's ability to recover additional claims following the first covered event.
The independent evidence for a payment remains transaction-specific. The World Bank's November 2025 Jamaica payout notice states that Hurricane Melissa met the predefined path and central-pressure requirements and that the $150 million 2024 catastrophe bond would pay in full, following a third-party event calculation. This substantiates a contractual success for that event and instrument, while the broader adequacy of Jamaica's disaster financing program must still be evaluated against its full damages, fiscal needs and recovery timing.
Multi-Period Program Evaluation
Maturity is not identical to continuous protection
A multi-year cat bond can secure specified investor capacity through several catastrophe seasons and reduce the sponsor's repeated exposure to short-term renewal price movements. However, the risk period, scheduled maturity, principal-reduction provisions, early redemption triggers and extension for unresolved events determine the actual continuity. If a fully covered catastrophe consumes principal before the scheduled maturity, the remaining program may have materially less available protection. The sponsor must plan both scheduled expiry and event-driven exhaustion. An issuance with a four-year date on its cover sheet does not guarantee four years of unchanged capacity after every covered event.
For a hypothetical single-event trigger probability p=0.02 in each of four independent annual periods, the probability of at least one trigger over the horizon is 1−(1−0.02)^4, approximately 7.76%. This calculation is illustrative and assumes identical independent annual hazards, unchanged coverage and a trigger that consumes capacity after the first qualifying event. It is not an empirical probability for any named bond. If multi-event loss processes are clustered, exposures grow, conditions shift or a principal-reduction schedule is partial, the true cumulative distribution requires a more detailed event-year model. The mathematical result still makes the general point that even a remote annual event can be economically material over a multiyear contract term.
A rolling protection program compares an initial issue date, ongoing changes in reinsurance, investor market windows and future maturity concentrations. One way to represent discounted sponsor cost is PV(K)=sum from t=0 to T of K_t/(1+r)^t, with risk-adjustment and discount conventions declared and future contingent payoffs integrated consistently. A single inaugural risk margin cannot determine expected cost over several renewals. A sponsor should stress a hard-market repricing at maturity, unavailability of specific perils, increased attachment demands from investors, and temporary inability to obtain replacement protection following a covered event.
Insurance and capital market cycles
Cat-bond pricing reflects modeled expected loss, demand for disaster risk, investor diversification and portfolio constraints, market-wide conditions and sponsor characteristics. The Braun (2016) empirical article identifies expected loss and the reinsurance cycle among primary-market spread determinants in its historical issuance sample. Götze and Gürtler (2020) find an association of secondary-market premia with sponsor experience, reputation and financial strength, even for instruments meant to isolate natural-catastrophe underwriting risk from sponsor credit quality. This finding cautions against assuming that collateralization removes every market valuation channel related to the sponsor.
A multiyear contract can lock risk capacity while the model-calibrated exposure, eligible policy portfolio and wider capital-market spreads change. Whether this is advantageous depends on the contract's exposure-adjustment mechanisms and the sponsor's opportunity cost. Rising insured values may erode the economic adequacy of a fixed dollar limit. A shrinking or geographically shifted portfolio may leave the issuer paying for an inefficiently located layer. A reinsurance market that softens after issuance may offer cheaper capacity, but terminating a bond may be restricted or costly. Thus, tenure is valuable primarily when its stability matches a sufficiently stable risk-financing need.
Issuance costs and minimum efficient size
Sponsors incur structuring, catastrophe modeling, external legal, calculation-agent, issuance, placement, collateral oversight, reporting and ongoing compliance costs. These can contain significant fixed components. All else equal, a small issue carries a larger fixed-cost burden per dollar of risk capital than a large issue. A larger issuance can, however, raise total premium or exceed the market's capacity for that peril or attachment. The World Bank disaster risk transfer note reports this practical asymmetry between bond and traditional-market transactions. A rational comparison explicitly amortizes costs over the actual risk period, noting that premium and legal risk may continue after an early event.
For illustration only, suppose a sponsor pays a fictional $12 million fixed issuance cost for a four-year cat bond, compared with $4 million each year for arranging equivalent annually renewed alternative protection. Ignoring discounting, the fixed component would total $12 million versus $16 million. This $4 million apparent saving does not prove the bond is preferable: the annual alternative may vary coverage, permit reinstatement, or avoid event-driven principal exhaustion, while the bond may have different premiums and basis risk. A present-value comparison should show each of these contingent differences and the funding source for transaction costs.
Climate and model drift
Peril models reflect particular data and scientific assumptions. Changes in sea surface temperatures, rainfall extremes, coastal exposure, construction standards, asset inflation and insured values can influence modeled loss distributions, but the change is peril- and geography-specific. A robust sponsor evaluation tests several internally coherent hazard and exposure scenarios rather than applying a universal multiplicative “climate uplift.” Severe convective storm, tropical cyclone, flood and earthquake risks respond to different physical drivers; earthquakes are not treated as climate-driven simply because they are pooled in a catastrophe bond.
Model drift also occurs without physical climate change. A catastrophe-model vendor may revise vulnerability curves, insured policy geocoding or loss reporting; an insurer may enter a new region or retain larger policy limits. A bond's original expected-loss estimate can therefore become stale during the term. The relevant question is whether contract terms still match the sponsor's emerging net risk, including the remaining traditional reinsurance tower and available liquidity. Renewal decisions require documentation of changes in both the hazard model and the portfolio.
Event-driven replacement: Jamaica
Jamaica demonstrates why principal exhaustion and replacement financing belong to the original strategic decision. The World Bank's Hurricane Melissa payout announcement confirms the full $150 million payout under the 2024 bond. Its May 2026 replacement terms disclose a new $200 million parametric per-occurrence instrument, priced on May 18, 2026, settled on May 26 and scheduled to mature on May 23, 2030. The stated annual risk margin is 6.75%, with a separate 0.12% funding margin and a compounded SOFR component. These figures describe the new issue, not the economic return on the exhausted old bond.
The 2026 Jamaica case study places the bond in a policy combining fiscal reserves, contingency borrowing and Caribbean catastrophe risk pooling, and characterizes bonds as protection for the least frequent, most severe events. The event led to a full contractual payout, and Jamaica subsequently increased nominal bond cover by $50 million. The official record supports these sequential facts. It does not identify a single counterfactual optimum or permit attribution of later fiscal outcomes solely to the bond. A source-critical detail matters: the June 2026 World Bank case study dates the new $200 million transaction to “May of 2025” in one paragraph, while its cover, opening summary, 2026 trigger graphic and the World Bank’s contemporaneous May 18, 2026 announcement identify the issue as a 2026 transaction. The year in that case-study paragraph is therefore treated as an apparent editorial error, with transaction dating anchored to the dated May 2026 issuance announcement.
The payout also illustrates why liquidity should be evaluated against actual dates. The World Bank's post-event Jamaica account reports that the bond proceeds were disbursed on December 1, 2025, after the November determination, alongside earlier Caribbean Catastrophe Risk Insurance Facility payments and contingent financing arrangements. The timing demonstrates that payout determination, actual disbursement and emergency cash needs belong to separate schedules. An issuer's language about a rapid trigger is therefore not a substitute for the sponsor's funding-calendar analysis.
Stress testing the renewal decision
A sponsor's decision memo should contrast at least three conditional futures: uneventful expiry into a soft market; a major catastrophe causing principal loss and a hard-market renewal; and exposure growth that moves the desired layer upward while contractual coverage remains fixed. These are scenarios, not probabilistic forecasts unless calibrated. Each requires a projected retained loss distribution, available reinsurance capacity, incremental cost, replacement lead time and possible interruption in protection. Stress testing should also ask whether multiple bonds mature during the same season, whether a swap or reinsurance alternative is available, and whether the sponsor has advance authority and liquidity for a timely replacement issue.
A multi-period program is strongest when each renewal has explicit evidence that the exposure, trigger, attachment, exhaustion and capital rationale still cohere. Investors' willingness to buy the next bond cannot be taken for granted on the day a prior bond is issued. Long-duration capital-market capacity reduces one category of renewal uncertainty while introducing contract rigidity and potentially a new post-event replacement problem.
Integrated Comparative Synthesis and Research Implications
Three observable sponsor decisions
The documentary cases demonstrate distinct functions of catastrophe bond sponsorship without requiring the assumption that their loss profiles are comparable. Mexico's 2024 four-class program divided earthquake intensity categories and Atlantic/Pacific hurricane geographies, indicating an effort to align segmented triggers with different hazard exposures. Chile's 2023 bond-and-swap transaction integrated two placement channels, illustrating the value of market capacity beyond either capital market or reinsurance alone. Jamaica's 2025 payout and 2026 replacement demonstrate the usefulness of a high-severity parametric financing layer and the importance of renewing exhausted protection. These are specific observations of contract design and results, not a ranking of countries' risk-management effectiveness.
| Observed program | Verified transaction fact | Sponsor-strategy interpretation | What evidence does not establish |
|---|---|---|---|
| Mexico, 2024 | Four classes totaling $595 million, separated by peril, region and earthquake risk category | Segmentation can support different hazard and basis-risk calibrations | No publicly proven optimal attachment relative to the entire fiscal loss distribution |
| Chile, 2023 | $350 million cat bond combined with $280 million catastrophe swaps | Parallel access to capital market and reinsurance capacity | No universal cost superiority of dual placement or disclosed whole-program counterfactual |
| Jamaica, 2025–2026 | Full $150 million 2024 bond payout after Melissa; new $200 million bond in May 2026 | High-severity fiscal funding and replacement of exhausted capacity | No demonstration that nominal limit equals total disaster financing need |
The transaction values and dates in this table come from the Mexico 2024 case record, the Chile 2023 terms and the Jamaica replacement announcement. They represent different policy and market environments, and their published sponsor materials supply more information about transaction structures than about complete portfolio counterfactuals.
Layer suitability as a conditional judgment
The integrated analysis yields five defensible propositions. First, the relevant loss distribution is the sponsor's net liability and cash-funding distribution after other risk-transfer arrangements. Second, nominal layer size and attachment have different meanings under occurrence, aggregate and nonindemnity contracts. Third, a trigger's quality is conditional on the sponsor's funding objective and joint event-loss distribution. Fourth, fixed and variable transaction costs must be compared against the marginal capital and liquidity benefit over a consistent horizon. Fifth, the sponsor must account for expiry, principal exhaustion and reinsurance-market access across multiple periods. These are analytic findings obtained by combining established actuarial arithmetic, supervisory requirements and observed transactions; they are not advanced as an original theoretical system or exclusive doctrine.
A lower-attachment bond can be defensible when frequent midseverity losses threaten earnings or timely payments, the premium is affordable relative to the reduction in those outcomes, and the collateralized coverage coordinates with existing treaties. A high-attachment bond can be defensible when rare extreme losses dominate capital or sovereign fiscal risk and the sponsor can absorb medium losses with retentions, reinsurance, reserves or contingent credit. A blended strategy is defensible where two markets jointly provide capacity, diversified settlement mechanisms or greater resilience to renewal shocks. All three judgments depend on quantified sponsor-specific evidence.
Distinguishing three kinds of uncertainty
Aleatory catastrophe uncertainty concerns future event occurrence, intensity and location. Epistemic model uncertainty concerns the hazard, exposure and vulnerability representations used to estimate financial loss. Contractual and operational uncertainty concerns whether and when a specified event is reported, qualified, calculated and paid. These uncertainties interact but require different evidence. More simulation trials can reduce Monte Carlo sampling error without repairing misspecified contract language or improving geocoding. More precise storm data can improve an event calculation while leaving a financial-loss mismatch unresolved. Fully funded collateral can reduce a counterparty-credit channel while leaving event-definition, model and liquidity risks intact.
The Solvency II full-funding requirements address a key part of the contractual capacity to pay; IAIS ICP 13 requires broader risk management of transferred and newly assumed risks. Market documentation and academic studies likewise show that pricing and incentives are multidimensional. The Galeotti, Gürtler and Winkelvos study compares premium models and finds predictive rankings sensitive to whether financial-crisis observations are included, reinforcing the case for scenario and model sensitivity rather than an unqualified single price formula.
Legal, regulatory and governance implications
A board or risk committee selecting a catastrophe bond layer should retain a documented statement of risk appetite, the exposure model and its version, the tested alternatives, the expected and stress-state recoveries, trigger verification procedures, collateral and counterparty arrangements, the effect on solvency and liquidity, and proposed renewal and exhaustion responses. Governance is important because the transaction can span an underwriting cycle and involve multiple affiliated risk bearers. The IAIS enterprise-risk-management principles and ICP 13 internal-control requirements supply international supervisory support for integrated risk management, while national and regional rules determine legal enforceability and specific regulatory recognition.
In the European Union, a proposed program must separately satisfy applicable supervisory approval and the relevant qualitative conditions for recognizing reinsurance or SPV risk mitigation. The EIOPA rulebook on risk mitigation sets out the Delegated Regulation's Article 211 requirements, while Directive Article 211 governs authorization of special-purpose vehicles. These provisions have separate legal roles despite the coincident article numbering. U.S. insurance statutory accounting and investment rules may apply differently according to issuer and investor status; the NAIC investment analysis manual defines catastrophe-linked securities for an insurer-investor classification purpose, which should not be mistaken for a sponsor's automatic capital credit. The governing law, domicile and supervisory practice require precise transaction-by-transaction analysis.
Evidence limitations and the research frontier
Public issuer announcements and multilateral case studies are strong evidence of named amounts, dates, disclosed trigger classes and official payment determinations. They are generally weak evidence for counterfactual pricing, proprietary loss distributions, full legal waterfalls and economically optimal sponsor behavior. The principal peer-reviewed studies considered here use different dates, samples and theoretical assumptions; they cannot directly be pooled into an estimate of present-day sponsor benefit. Scientific generalization requires original insurer-level exposure and loss records, event-set simulations, term sheets, independent estimates of market pricing, and longitudinal recovery and renewal observations.
Several specific empirical questions follow. How does the distribution of capital benefits across candidate layers change when economic and regulatory capital measures are evaluated on the same net loss model? Does faster parametric settlement consistently outweigh basis-risk cost for sovereign funding objectives when both are valued at actual disbursement dates? How does an issuer's portfolio concentration alter the joint price and hedge effectiveness of an industry-loss trigger? What is the distribution of replacement-cost shocks after early principal exhaustion? Which layer choices prove robust across changes in catastrophe-model versions? These are open research questions requiring new evidence; this monograph does not claim to have statistically resolved them.
Limits of transportability
The sovereign case records cannot be transferred to commercial insurer capital management by changing only the sponsor label. A government may have tax revenues, contingent development finance, donor support and broad economic reconstruction obligations, whereas a licensed insurer has policyholder claims, local solvency regulations and an underwriting portfolio. Similarly, a diversified global reinsurer, a regional mutual and a single-peril specialist may have different dependence on indexes and market placements. The strategic logic is shared at the level of contingent payments and risk financing, while the legally and economically relevant loss variables differ.
The most useful contribution of the comparative analysis is a disciplined separation of observable contractual terms from inferred strategic purpose and unobserved optimality. A published size is an issuance fact. A published payout is a fulfillment fact. A modeled expected loss is a conditional estimate. A claim of optimal layer selection requires a specified objective, feasible set, calibrated exposures, cost assumptions, constraints and a reproducible comparison. This hierarchy allows strong scientific conclusions without inventing unobserved sponsor data.
Conclusion
Catastrophe bond sponsor strategy is the selection of contingent capital-market protection within the sponsor's entire insurance, reinsurance, capital and liquidity system. The layer's attachment, limit, exhaustion, trigger, peril and geography determine the states in which payment becomes available. Its nominal amount alone describes neither the sponsor's net catastrophe exposure nor the economic value of the transfer.
The analysis establishes that attachment selection is conditional on the shape of the loss distribution, the sponsor's working retention, the position of conventional reinsurance, the chosen capital risk measure and the dates when liquidity is required. Occurrence and aggregate contracts with identical monetary labels can produce different payments. Indemnity, industry-loss, parametric and modeled-loss triggers transfer different combinations of claim-alignment, information and settlement risk. Basis risk must be assessed in the joint distribution of sponsor loss and contractual payment, especially during the extreme states the sponsor seeks to finance. A bond can provide strong high-severity protection and still leave large medium-loss or post-event funding gaps.
The hypothetical calculation shows how two equally sized limits can provide identical recovery in the most severe enumerated state while differing sharply in expected payout and their responses to medium catastrophes. Documentary evidence from Mexico supports segmented peril and hazard treatment, from Chile the use of simultaneous bond-and-swap placement, and from Jamaica the importance of parametric payment and replacement after exhaustion. These cases illustrate observed design choices and independently confirmed transaction events; they do not prove one universal optimal layer.
Over multiple years, the sponsor must plan for principal reduction, changing exposures, catastrophe-model revision, market cycles, issuance expenses and renewal failure. Sound governance requires an explicit reconciliation of insured or fiscal losses, eligible contractual losses, modeled recovery, cash timing and any supervisory recognition. The appropriate choice is the feasible package that meets the sponsor's stated risk-financing objective under tested assumptions and documented constraints. The scientific result is an original comparative and mathematical synthesis of established knowledge, with its empirical boundaries expressly identified. No novel theoretical system is asserted.
References
Barrieu, Pauline, and Henri Loubergé. (2009). “Hybrid Cat Bonds.” Journal of Risk and Insurance 76(3): 547–578. https://doi.org/10.1111/j.1539-6975.2009.01312.x
Braun, Alexander. (2016). “Pricing in the Primary Market for Cat Bonds: New Empirical Evidence.” Journal of Risk and Insurance 83(4): 811–847. https://doi.org/10.1111/jori.12067
Cox, Samuel H., and Hal W. Pedersen. (2000). “Catastrophe Risk Bonds.” North American Actuarial Journal 4(4): 56–82. https://doi.org/10.1080/10920277.2000.10595938
Cummins, J. David, David Lalonde, and Richard D. Phillips. (2004). “The Basis Risk of Catastrophic-Loss Index Securities.” Journal of Financial Economics 71(1): 77–111. https://doi.org/10.1016/S0304-405X(03)00172-7
Cummins, J. David, and Mary A. Weiss. (2009). “Convergence of Insurance and Financial Markets: Hybrid and Securitized Risk-Transfer Solutions.” Journal of Risk and Insurance 76(3): 493–545. https://doi.org/10.1111/j.1539-6975.2009.01311.x
European Parliament and Council. (2009, as amended). Directive 2009/138/EC, Article 211, special purpose vehicles, as presented in the EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2298_en
European Commission. (2015, as amended). Commission Delegated Regulation (EU) 2015/35, Article 211, risk-mitigation techniques using reinsurance contracts or special purpose vehicles, as presented in the EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5864_en?prefLang=da
European Commission. (2015, as amended). Commission Delegated Regulation (EU) 2015/35, Article 326, solvency requirements for special purpose vehicles, as presented in the EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2615_en
Finken, Silke, and Christian Laux. (2009). “Catastrophe Bonds and Reinsurance: The Competitive Effect of Information-Insensitive Triggers.” Journal of Risk and Insurance 76(3): 579–605. https://doi.org/10.1111/j.1539-6975.2009.01317.x
Galeotti, Marcello, Marc Gürtler, and Christine Winkelvos. (2013). “Accuracy of Premium Calculation Models for CAT Bonds—An Empirical Analysis.” Journal of Risk and Insurance 80(2): 401–421. https://doi.org/10.1111/j.1539-6975.2012.01489.x
Götze, Tobias, and Marc Gürtler. (2020). “Hard Markets, Hard Times: On the Inefficiency of the CAT Bond Market.” Journal of Corporate Finance 62: 101553. https://doi.org/10.1016/j.jcorpfin.2019.101553
Götze, Tobias, and Marc Gürtler. (2020). “Risk Transfer and Moral Hazard: An Examination on the Market for Insurance-Linked Securities.” Journal of Economic Behavior & Organization 180: 758–777. https://doi.org/10.1016/j.jebo.2019.06.010
International Association of Insurance Supervisors. (2024). Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups, December 2024, including ICP 13 and ICP 16. https://www.iais.org/uploads/2025/01/IAIS-ICPs-and-ComFrame-adopted-in-December-2024.pdf
International Association of Insurance Supervisors. (2024, online). “ICP 13: Reinsurance and Other Forms of Risk Transfer.” ICPs and ComFrame Online Tool. https://www.iais.org/icp-online-tool/13524-icp-13reinsurance-and-other-forms-of-risk-transfer/
International Association of Insurance Supervisors. (2024, online). ICP and ComFrame Online Tool, including enterprise risk management for solvency purposes. https://www.iais.org/icp-online-tool/
National Association of Insurance Commissioners. (2025). “Insurance-Linked Securities.” Insurance Topics, updated September 24, 2025. https://content.naic.org/insurance-topics/insurance-linked-securities
National Association of Insurance Commissioners. (2024). Purposes and Procedures Manual of the NAIC Investment Analysis Office, Part Three, “Catastrophe-Linked Securities.” https://content.naic.org/sites/default/files/publications-ppm-manual.pdf
National Hurricane Center, National Oceanic and Atmospheric Administration. (2026). NHC Data Archive, including the HURDAT2 best-track data and revisions. https://www.nhc.noaa.gov/data/index.php
World Bank. (n.d.). Capital at Risk Notes, IBRD Treasury, program and explanatory examples. https://treasury.worldbank.org/en/about/unit/treasury/ibrd/ibrd-capital-at-risk-notes
World Bank. (n.d.). Capital at Risk Notes, IBRD projects program explanation. https://projects.worldbank.org/en/about/unit/treasury/ibrd/ibrd-capital-at-risk-notes
World Bank. (2023). “World Bank Executes Its Largest Single Country Catastrophe Bond and Swap Transaction to Provide Chile $630 Million in Financial Protection Against Earthquakes.” News release, March 17, 2023. https://www.worldbank.org/en/news/press-release/2023/03/17/world-bank-executes-its-largest-single-country-catastrophe-bond-and-swap-transaction-to-provide-chile-630-million-in-fin
World Bank. (2025). World Bank Catastrophe Bonds Provide Mexico with $595 Million Financial Protection against Hurricanes and Earthquakes. Treasury case study, February 27, 2025, describing the 2024 transaction. https://thedocs.worldbank.org/en/doc/8f7222d60ec54d0b6571131b754319c1-0340012025/original/Case-Study-Mexico-2024-Cat-Bond.pdf
World Bank. (2025). “Hurricane Melissa Triggers 100% Payout of $150 Million World Bank Catastrophe Bond for Jamaica.” News release, November 7, 2025. https://www.worldbank.org/en/news/press-release/2025/11/07/hurricane-melissa-triggers-100-payout-of-150-million-world-bank-catastrophe-bond-for-jamaica?intcid=ecr_hp_sidekick3_en_ext
World Bank. (2025). “How Jamaica’s Preparedness Delivers After a Devastating Hurricane.” Feature, December 17, 2025. https://www.worldbank.org/en/news/feature/2025/12/17/jamaica-preparedness-delivers-after-devastating-hurricane
World Bank. (2026). Disaster Risk Transfer. Treasury product note, updated May 2026. https://thedocs.worldbank.org/en/doc/46c9b740fe83bc96681ff097681d9e6b-0340012024/original/Disaster-Risk-Transfer-Product-Note.pdf
World Bank. (2026). “World Bank Prices Catastrophe Bond Providing Protection for Jamaica Replacing Coverage Triggered by Hurricane Melissa.” News release, May 18, 2026. https://www.worldbank.org/en/news/press-release/2026/05/18/world-bank-prices-catastrophe-bond-for-disaster-protection-for-jamaica
World Bank. (2026). Third Catastrophe Bond for Jamaica: Increased Cover Following Hurricane Melissa. World Bank Group case study, updated June 2026. https://thedocs.worldbank.org/en/doc/f66f98e820b44dcec2070273327df145-0340012026/original/Case-Study-Jamaica-2026-Cat-Bond.pdf