RESEARCH DOSSIER
Reinsurance Portfolio Aggregate Limits and Exposure Monitoring
Abstract. Aggregate limits in reinsurance place a contractual ceiling on risk transfer across a defined accumulation of covered losses. A portfolio may exhibit substantial gross insured values, high modeled recoveries under selected catastrophes, apparently adequate occurrence protection, and nonetheless possess little remaining annual reinsurance capacity after earlier events or reserve developments. This monograph analyzes that divergence through contract interpretation, actuarial loss transformations, event aggregation, portfolio exposure data, stochastic dependence, reinstatement mechanics, counterparty recoverability, and supervisory controls. It distinguishes an annual aggregate cap on occurrence recoveries from aggregate excess-of-loss attachment, occurrence limits, reinstated capacity, an insurer's underwriting appetite, and financial-reporting recoverables. The research synthesizes International Association of Insurance Supervisors standards, European Solvency II provisions, Lloyd's catastrophe reporting and syndicate accounts, United States statutory reinsurance schedules, IFRS 17 explanatory material, and actuarial and legal scholarship. An expressly hypothetical four-event ledger demonstrates a layer attaching at 20 with an 80 occurrence limit and a 160 annual recovery cap; event-level eligible losses total 245, while contractual paid or accrued recoveries reach only 160, leaving 85 of otherwise layer-eligible loss uncovered by the exhausted annual cap. A distinct aggregate stop-loss calculation illustrates why coincident numerical results imply neither equivalent protection nor identical exhaustion paths. Original analytical synthesis establishes the need to reconcile six measurement states—exposure inventory, modeled potential recovery, contractually eligible recovery, recognized utilization, legally available reinstatement, and collectible recovery—before decisions about further underwriting or capital can be credible. These are analytical distinctions and audit requirements, not an asserted new theory. Conclusions are bounded by contractual variation, model limitations, jurisdiction-specific regulation, a selected rather than exhaustive evidence base, and the absence of access to proprietary treaty or portfolio data.
Keywords: Reinsurance; portfolio aggregate limits; annual aggregate limit; aggregate excess of loss; occurrence excess of loss; reinstatement; reinstatement premium; exhaustion; attachment point; aggregate deductible; treaty utilization; catastrophe accumulation; exposure monitoring; model uncertainty; claims development; net retention; credit risk; counterparty recoverables; catastrophe scenarios; stochastic dependence; loss occurrence definition; Lloyd's; IAIS; Solvency II; NAIC Schedule F; IFRS 17.
Introduction: Scientific Object and Method
The scientific object is the utilization and continuing availability of a finite contractual aggregate of reinsurance protection when a portfolio generates multiple losses or correlated scenarios over time. Reinsurance transfers selected insurance risk under a separate agreement; once that agreement imposes a cumulative ceiling, knowledge of the current gross portfolio and a nominal treaty occurrence limit alone cannot determine future net exposure. An event occurring after the cap has been consumed can leave the cedent with nearly the whole otherwise transferable layer loss. The critical observations therefore include the contract's covered unit, inception and expiry, the legal grouping of losses, booked and anticipated ceded recoveries, permitted reinstatements, and the contingent exposure still capable of attaching.
The IAIS standards on reinsurance recognize that the ceding insurer acquires credit, operational, and basis risks while transferring underwriting risk. ICP 13.2 discusses controls over implementation and documentation of reinsurance programs, including the management of aggregations and recoveries. ICP 16 identifies risk interdependencies, risk appetite, limits, data and model controls, scenario analysis, and effective group-wide oversight. These standards supply institutional questions; a treaty's actual monetary entitlement is determined by its contract and governing law. Supervisory recognition and contractual recovery are further distinct questions.
The inquiry answers five related questions. Which types of aggregate ceiling govern recoveries, and how do they interact with per-occurrence terms? Which data are necessary to calculate utilization and remaining cover at any valuation date? How should correlated events, reserve deterioration, and reinstatement provisions alter a future-loss distribution? How do counterparty default and legal uncertainty qualify nominally ceded amounts? What constitutes defensible evidence that the remaining portfolio lies within underwriting and capital risk tolerances?
Method combines close reading of documented supervisory provisions, selected market reporting, legal scholarship on aggregation, actuarial aggregate-loss literature, contract algebra, an independently recalculable hypothetical ledger, and operational reconciliation design. The examples are constructed, not estimated from an insurer dataset. Cross-system comparisons identify the scope of rules without assuming that international guidance is self-executing law or that one jurisdiction's capital treatment governs all markets. The dossier makes analytical findings by deductive interpretation and transparent arithmetic; it advances no newly named doctrine or model.
Concepts and Historical Development
Insurance accumulation predates modern catastrophe simulation: an underwriter may hold many nominally separate risks that respond to a single storm, earthquake, fire, pandemic, supply-chain disruption, or liability occurrence. Reinsurance similarly faces a two-sided aggregation problem. The cedent asks how much coverage remains against its portfolio; a reinsurer asks how much a single triggering event will obligate it to pay across multiple cedents, layers, and retrocessions. A treaty portfolio becomes economically concentrated when many claims can access the same finite capacity, when several contracts respond to one cause, or when many obligations are owed by the same counterparty.
Traditional proportional reinsurance shares losses according to agreed participation, while nonproportional cover responds after a retention or attachment threshold. The growth of excess-of-loss contracts sharpened the importance of distinguishing a single loss from an event or an aggregation of related claims. Oliver D. William (2021), in Reinsurance and the Law of Aggregation, demonstrates why deductible and limit calculations depend upon the wording that groups individual losses into a reinsurance occurrence. His analysis is primarily grounded in English law and therefore informs contract interpretation rather than offering a universal legal definition.
The emergence of standardized catastrophe event sets, stochastic loss modeling, and regularly reconciled exposure databases changes the frequency at which aggregate risk can be observed. Lloyd's 2025 disaster scenario specification specifies event footprints, market insured-loss assumptions, and reporting conditions for defined tests. Such scenarios support disciplined comparisons but cannot exhaust all physically or legally possible accumulations. The Lloyd's 2025 capital instructions discuss aggregate exceedance probability, final net loss, model completeness, and sensitivity, providing direct evidence that quantitative accumulations, model limitations, and capital oversight are studied together in market practice.
A useful historical interpretation is that the annual aggregate limit is simultaneously a contract term and a changing operating state. Its nominal face amount is fixed by the bargain, while its available amount evolves with events, reserve revisions, legal classification, restoration provisions, and settlements. An effective monitoring system must reconcile these two temporal descriptions.
Contract Taxonomy and Limit Architecture
At least six different quantities are often described in commercial conversation as an aggregate limit. First, a primary insurance policy can cap its own annual liability. Second, a cedent can set an internal gross or net underwriting accumulation tolerance. Third, a reinsurance treaty can impose an occurrence limit, the most it pays for one defined occurrence. Fourth, a treaty may cap the aggregate recoveries attributable to all covered occurrences during the treaty period. Fifth, a stop-loss agreement can attach only when the cedent's accumulated subject losses exceed an annual aggregate deductible and can then itself carry a separate cover cap. Sixth, the reinsurer may impose a portfolio-level acceptance or retrocession accumulation limit across multiple contracts. Each amount belongs to a different legal or risk-management unit.
For an occurrence excess-of-loss cover described as “80 excess of 20,” a qualifying event with subject loss 150 yields a preliminary recovery of 80, retaining 70 with the cedent. If annual aggregate recoveries are capped at 160, two or more large covered occurrences may together consume that amount, after which otherwise eligible subsequent layer losses fall to the cedent unless a further contractual restoration is available. This language is schematic. Real treaties specify subject premium or losses, insured shares, loss adjustment expense, currencies, exclusions, loss occurrence, attachment bases, claims cooperation, inuring reinsurance, reinstatement premium, and the relevant accounting or risk attachment period.
An aggregate excess-of-loss or stop-loss treaty is different. If a year of subject loss totals 435, aggregate attachment is 200, and maximum cover 160, the simplified recovery is 160. That mathematical answer may resemble a capped occurrence program at year-end, although the two arrangements attach to entirely different loss paths. Contract word choice has economic consequences even when a particular year's ultimate recoveries coincide.
A true treaty aggregate limit is bounded by the contract's scheduled period and scope. It does not automatically reset with a financial year, underwriting year, policy renewal, reinsurance anniversary, commutation, or accounting close. Nor does the attachment of a second catastrophe necessarily trigger a fresh annual cap. The proper period follows the policy's event, risk-attaching, losses-occurring, claims-made, or otherwise expressly specified coverage basis. The first editorial question for every ledger is therefore which losses belong in the same covered accumulation.
Event Definitions, Hours Clauses, and Aggregation
The construction of a covered loss occurrence is a substantive legal input to actuarial calculations. One storm can generate thousands of direct claims; one chain of industrial failures can generate multiple bodily injury and property losses; one causal mechanism can operate across countries or over months. A treaty may group losses by one event, one occurrence, a common original cause, or an hours clause containing time and possibly geographic constraints. These mechanisms need not agree. William's event and cause analysis describes the difficulties produced by combined clauses and divergent judicial interpretations. The selected work studies English reinsurance law and is not a substitute for interpreting a different governing contract.
For a treaty with an 80 limit excess of 20, two individually treated events of subject loss 60 each yield recoveries of 40 plus 40, or 80. If those same losses satisfy an agreed single-occurrence clause, the pooled loss is 120 and the occurrence recovery is only 80; the numerical result is identical in this example. Change both losses to 90 and separate treatment produces 70 plus 70, or 140 before the annual cap, while treating them as one 180 loss produces 80. Thus the grouping convention changes attachment and limit usage, potentially by more than the impact of an ordinary pricing adjustment.
An hours clause is not a universal physical classification of a disaster. The relevant record must retain the clause text, its time-selection mechanism, event footprint and attribution evidence, adjudicated or assumed groupings, and a versioned rationale for any change. A sensitivity analysis can compare plausible groupings but must never book a more favorable recovery merely because a model assigned a different event identification number.
Aggregation also operates at several scales. A direct insurance occurrence may be a single insured event; the reinsurance occurrence may cover several direct losses; a catastrophe portfolio simulation may encompass numerous distinct events in one stochastic year. If these scales are merged in a database, per-occurrence limits, event caps, annual aggregates, and portfolio risk measures will be applied to the wrong set of losses.
Exposure Inventory and Data Lineage
A workable inventory begins with insured locations, interest and coverage, not with a single gross sum insured. A minimum normalized record carries a policy identifier, risk and location identifiers, insured entity, peril and coverage tags, inception and expiry, geography and geocoding confidence, insured value, sublimit and deductible, insurer share, underwriting currency, occupancy or industry attributes, catastrophe model version, and source-system timestamp. Business interruption and contingent business interruption require relevant dependency links. Aggregating policy limits without location, peril, policy period, and coverage conditions is incapable of reproducing a catastrophe recovery.
The treaty side needs a separate structured register: treaty ID and amendment version; subscribing participants and reinsurer shares; inception, expiry and attachment convention; covered business; occurrence definition; retention; occurrence limit; annual aggregate deductible or cap; reinstatement number and charge basis; inuring reinsurance; exclusions; territorial and legal terms; collateral and counterparty information; currency; and the date when placement and documentation are legally effective. ICP 13 implementation controls emphasize documented terms, reinsurance strategy, timely material reporting, and links between facultative placements and aggregation management.
The claims side then reconciles each loss occurrence against both the primary inventory and relevant treaties. Paid claims, case reserves, incurred-but-not-reported provisions, dispute adjustments, loss adjustment expenses, salvage, subrogation, estimated ceded recoveries, cash receipts, and cash settlements require separate fields. Reserves are estimates subject to revision. A cash receipt is evidence of collectibility, but it is not the sole measure of a treaty's consumed limit; unpaid ceded claims may already use capacity under the contract's applicable basis.
The four indispensable provenance questions are: Which underlying items entered a particular scenario? Which treaty wording was applied? Which valuation snapshot produced each amount? Which responsible actor authorized overrides? Without a traceable answer, the apparent precision of a dashboard is misleading. Data completeness, location uncertainty, policy duplication, mismatched currency conversion, unknown risk participation, and claims lag should be visible audit attributes rather than silently absorbed into net loss.
An exposure register is also a temporal system. Writing new risks increases prospective loss concentration without changing prior paid recovery. An adverse reserve development can retroactively affect modeled remaining capacity. A renewal or endorsement can move an exposure between contract periods. For that reason, a reliable monitoring process supports both as-of-date reproducibility and later corrected views, documenting the delta between them.
Ceded-Loss Mathematics
Let a treaty year contain contractually defined covered occurrences indexed by i, with eligible subject loss G_i in a common valuation currency. Let R denote the per-occurrence retention, L the per-occurrence payable limit, and C the annual aggregate maximum recoverable under the treaty. For a simplified positive excess-of-loss layer, preliminary occurrence recovery is
If the treaty's aggregate cap applies to the sum of these preliminary recoveries and no other terms intervene, the year's ultimate reinsurance recovery is
This representation applies only where the policy describes those exact mechanics. It does not independently establish whether loss adjustment expense erodes the layer, which period a loss enters, whether a participating reinsurer has a separate limit, or whether a reinstatement is purchased or automatic. For aggregate excess-of-loss, using annual subject loss S, annual aggregate deductible A, and aggregate cover limit M, a different simplified recovery is
The second expression attaches at a portfolio annual threshold rather than at each occurrence. The two formulas should remain distinct in contracts, pricing models and monitoring reports. Panjer's aggregate-loss overview describes the underlying actuarial problem of deriving loss distributions from claim count and severity, while Reijnen and colleagues' stop-loss study examines the effect of dependence and alternative approximations on aggregate stop-loss premiums.
The pre-cap future payable amount also requires a time index. Let U_t be utilization recognized according to the treaty's contractual basis after events or estimates booked through t, with U_t between zero and C. The current aggregate headroom is H_t=C-U_t, subject to signed changes from reserve revisions, reinstatement terms, and disputed allocation. When eligible next-event preliminary recovery p arrives, the recognized incremental recovery in the simple sequential model is min(p,H_t), and new headroom becomes H_{t+1}=H_t-min(p,H_t). Negative reserve corrections may restore booked headroom under the governing accounting and contract interpretation, but collected cash does not itself replenish an exhausted treaty.
The mathematical cap imposes a bound, not a forecast of when it will be reached. A portfolio risk assessment must calculate a distribution of future preliminary recoveries conditional on current exposure and prior experience. Expected gross loss, expected ceded loss, a probable maximum loss at one return period, and probability of aggregate exhaustion are different statistical functions. The difference becomes critical when two geographically or economically related catastrophes can occur in the same contract year.
Aggregate Exhaustion as a State-Dependent Process
An aggregate cap transforms the recovery from a new event because the payout depends on the preceding sequence. At inception the full headroom may be available. After a first loss, both the remaining aggregate headroom and potentially the immediately usable occurrence capacity are reduced. Where a reinstatement is conditional on notice, payment, additional premium or specified approval, the value and timing of restored protection also depend on that condition. A credible monitor therefore produces a treaty-state ledger rather than a static picture of nominal limits.
Three states require separate treatment. A confirmed loss can generate a contractually agreed recoverable. An advised loss may be reserved but disputed in coverage or amount. A modeled future event may produce a potential recoverable that has no presently recognized claim. Their expected or accounting weights may differ, and the dashboard should explicitly identify the valuation purpose. Adding all three without labels produces double counting and a false inference of exhaustion.
Potential exhaustion is forward-looking. At each reporting date, simulate scenario-year losses under an agreed reference exposure set, pass every simulated event through the occurrence and aggregate treaty terms, and record the fraction of simulated years in which the cap is reached. Also report distributions of first exhaustion date, uninsured layer-eligible loss after cap depletion, and the conditional tail of final net loss. These measures are conditional on model assumptions and should be accompanied by model limitations. They complement, rather than replace, the documented contractual ledger.
Timing also changes decisions. An exhausted treaty late in December may have relatively little residual unexpired exposure; a similarly exhausted program in February faces a much longer period of potential events. A single headroom percentage is consequently inadequate for comparing reporting dates. One must pair headroom with remaining exposure duration, seasonal hazard mix, recently written risks, and already known catastrophe clusters. When seasons differ by hemisphere and peril, unexpired exposure should be modeled using the correct spatial and temporal portfolio.
Reinstatements and Their Economic Cost
A reinstatement is a mechanism for restoring some specified exhausted or eroded reinsurance capacity. It may be automatic, optional, conditional, free, subject to an additional premium, limited in number, partial, time-prorated, amount-prorated, or written with special terms. It can restore an occurrence limit while leaving an explicit annual aggregate cap unchanged, or provide additional aggregate recovery capacity if the contract so states. Therefore the word “reinstatement” alone does not identify an annual maximum.
William's discussion of reinstated cover notes that excess-of-loss reinsurance often has both per-loss and annual aggregate limits and that reinstatement can provide renewed protection. The Solvency II catastrophe scenario provision, Article 126(2) of Delegated Regulation 2015/35, specifically requires realistic, objective, and verifiable assumptions about management actions relating to reinstatements between two standard-formula catastrophe events where the current contracts permit them. This requirement concerns a defined regulatory calculation. It does not overwrite commercial reinstatement wording or prove that all treaties restore automatically.
A rigorous ledger tracks original available occurrence cover, amounts eroded by each qualifying event, legally reinstatable capacity, reinstatement actually activated, future aggregate recovery ceiling, and reinstatement premium obligation. The premium can be based on a specified percentage of original treaty premium multiplied by the fraction of cover reinstated, sometimes with time adjustment. For a treaty premium of 8, one full amount-proportional reinstatement at 100% of premium would cost 8 if the full 80 layer is reinstated and the contractual charge basis is that simple. The result is a stipulated illustration, not a market quotation or universal clause.
Reinstatement changes the cedent's economics as well as protection. A model that counts the incremental recovery but excludes the reinstatement charge overstates net financial relief. A model that assumes immediate restoration despite a contractual notice or counterparty condition can misstate remaining risk. The reinsurer, meanwhile, must measure the contingent restoration of its own capacity obligations and the potential need for retrocession, collateral, or capital support. Reinstatement should be modeled as a contract-defined state transition, with the cost and timing shown alongside recoveries.
A Reproducible Four-Event Treaty Ledger
Consider a fully hypothetical single-year property excess-of-loss treaty. It covers the same defined set of risks for all four events; the four events are assumed to qualify independently under the loss-occurrence clause. Currency is an abstract monetary unit. Each event is evaluated on ultimate subject loss with zero adjustment for commissions, expenses, salvage, foreign exchange or proportional inuring covers. Per-occurrence retention R=20, per-occurrence limit L=80, and maximum cumulative annual treaty recovery C=160. The contractual wording is stipulated to permit full two-limit annual capacity, including a maximum of one full amount-proportional reinstatement of the original 80 occurrence capacity, charged at 100% of an original premium of 8 without time pro rata. This illustrative wording is internally complete solely for the example.
| Event | Subject gross G | Preliminary layer P | Remaining aggregate before event | Actual ceded recovery | Cedent net event loss | Remaining aggregate after event |
|---|---|---|---|---|---|---|
| One | 70 | 50 | 160 | 50 | 20 | 110 |
| Two | 130 | 80 | 110 | 80 | 50 | 30 |
| Three | 55 | 35 | 30 | 30 | 25 | 0 |
| Four | 180 | 80 | 0 | 0 | 180 | 0 |
| Total | 435 | 245 | — | 160 | 275 | 0 |
The arithmetic can be recomputed directly. Preliminary event recoveries are 50, 80, 35, and 80. Their sum is 245. Actual recoveries are min(50,160)=50, min(80,110)=80, min(35,30)=30, and min(80,0)=0; total ceded recovery is 160 and cedent net loss is 435−160=275. Thus 85 of the 245 otherwise layer-eligible loss falls outside the finite aggregate. This 85 represents 5 of the third event's layer loss and 80 of the fourth event's layer loss.
For a one-reinstatement convention, restoration of total cumulative protection from 80 to 160 consumes the single permitted extra 80 of capacity during the events. Under the stipulated full amount-proportional premium basis, a fully used reinstatement produces an additional premium of 8. The cedent's modeled event loss plus that incremental premium is 283 before considering the original treaty premium, operating costs, tax, discounting, or counterparty default. Because this example assumes one reinstatement is included in the maximum C of 160, it must never be added a second time to derive an imaginary C of 240.
Changing the order of identical independent events can change which particular event is reimbursed after the aggregate cap is consumed; with these simplified same-year terms and no other state restrictions it does not change the final sum min(245,160). This distinction matters when event-specific ceded reserve recognition, underwriting-unit attribution, or interim solvency estimates depend on event chronology. Reinstatement timing and the actual clause may also introduce path dependence not captured by the final-year min function.
An audit reproducing the table needs only four input gross losses, the parameters R, L and C, and the sequential headroom rule. It does not need a catastrophe model or a claim database. That reproducibility is deliberate: the example tests financial and contractual arithmetic, while making no claim about the empirical frequency of the four events.
Alternative Aggregate Stop-Loss Calculation
Reevaluate the same gross sequence under a wholly separate hypothetical annual aggregate excess-of-loss contract. Assume annual gross subject loss S=435, annual aggregate deductible A=200, annual cover limit M=160, no exclusion, and no other adjustment. The aggregate stop-loss recovery is min(max(435−200,0),160)=160. The cedent's final retained annual loss is 275 before premium. The final totals happen to equal those of the capped occurrence treaty. Contractual equivalence does not follow.
Under the stop-loss contract, the first 200 of annual subject loss remain below aggregate attachment. The occurrence cover, by contrast, attaches whenever a qualified single event exceeds 20, even when cumulative year-to-date gross losses are low. A future year dominated by one 180 event produces occurrence-layer recovery of 80, but annual aggregate stop-loss with attachment 200 yields zero. A year with many small losses, each below the occurrence retention, can trigger the annual aggregate stop-loss once their combined total exceeds 200, even though the occurrence treaty yields zero. The underlying economic distribution of recovery is therefore different.
Treaty comparators should test more than total annual loss at one selected sample point. They should compare expected ceded recovery, expected cedent loss, exhaustion probability, stress loss, cost of reinstatements, conditional tail outcomes, premium, claims handling cash timing, and counterparty exposures. Reijnen, Albers, and Kallenberg (2005) show in their stop-loss premium comparison that even limited loss dependence can influence stop-loss pricing, with the magnitude sensitive to assumptions. This paper is useful as a warning about model inputs rather than a calibration for the hypothetical contract.
The word “aggregate” is thus insufficient to translate a risk report from one program to another. The monitor must identify whether the aggregate applies before per-occurrence attachment, after per-occurrence payouts, across classes or years, and before or after an inuring layer.
Portfolio Dependence and Stochastic Modeling
An aggregate cover's utilization probability is driven by the joint law of occurrences, severities, event footprints, and insurance interests. In a collective risk specification, annual total subject loss may be represented as S=sum_{j=1}^{N} X_j, where N is an event count and X_j denotes event severity. A compound-Poisson approximation is convenient when its frequency and conditional independence assumptions are defensible; other processes account for seasonality, clustering, changing exposure and common shock. Aggregate Loss Modeling by Panjer explains foundations of the aggregate loss-distribution problem. The actual treaty transformation, including per-event and annual caps, must then be applied at the appropriate level of the simulated event sequence.
Correlation in underlying losses is not a cosmetic parameter. Two cedents insured in the same storm footprint may trigger one reinsurer simultaneously; several business lines can be affected by one earthquake; pandemic or cyber interruptions may cross geographic boundaries. Mata (2003), in her study of dependent reinsurance aggregates, examines joint distributions under specified dependence assumptions and demonstrates the importance of the dependence structure to the reinsurance aggregation problem. Its asymptotic setting is narrower than the entire practical market and does not supply a universal catastrophe correlation.
Where stochastic event catalogs are used, each simulated year should be internally consistent: common event identifiers must attach to all insured risks affected by the same simulated event, and contract definitions must determine how those losses are grouped. Application of a reinsurance formula independently to every policy and subsequent summation can incorrectly multiply per-occurrence limits. Likewise, applying a single arbitrary annual loss severity to every contract can erase legal event allocation and reinstatement timing. Model architecture should reflect peril, geography, policy conditions, and treaty period.
Define p_exh=P(T=C) for a simple cap with deterministic C, with a caveat that accounting estimates and contractual disputes may make actual utilization uncertain. Conditional distributions of the first exhaustion time and of residual portfolio net loss can illuminate decisions more directly than an unconditional expected annual loss. In a Monte Carlo setting, report sampling error, assumed horizon, exposure valuation date, base currency, inflation basis, and the results of parameter stress. Scenarios with zero event-model probability because they are outside the catalog should be discussed as model coverage gaps, not treated as physically impossible.
Interdependence extends beyond insured claims to recoverability. A catastrophic event can produce large reinsurance claims and simultaneously impair the same counterparties, collateral values, liquidity or payment infrastructure. A model that calculates treaty recovery without credit or operational channels is a useful gross-to-net engineering approximation, not a full assessment of financial protection.
Catastrophe Scenarios and Multiperil Accumulation
A deterministic catastrophe scenario specifies an event, physical footprint, insured exposure translation and financial loss mechanism. Lloyd's RDS 2025 scenario definitions include compulsory scenarios and event descriptions for both elemental and other catastrophic risks; the document describes common assumptions and allows managing agents to use their own loss calculations within reporting conditions. This structure is valuable for comparing gross and net results at a standardized reference point. It cannot by itself establish an individual insurer's full portfolio probability distribution, its treaty coverage response, or the number of possible successive catastrophes in the year.
Lloyd's exposure management model tests connect aggregate exceedance probability final net loss, catastrophe appetite, model completeness, sensitivity and realistic disaster scenarios to its capital review. That evidence illustrates how financial capital measurement depends on exposure reconciliation. It also underlines a methodological distinction: an aggregate exceedance probability curve describes modeled annual loss behavior, while a treaty annual aggregate cap is a legally enforceable ceiling on certain recoveries. The same word “aggregate” names a statistical distribution in one context and a contractual capacity in the other.
Syndicate accounts provide bounded observations rather than a census. Blenheim Syndicate 5886's 2025 syndicate annual accounts report scenario exposure monitoring and an illustrative Florida windstorm RDS with estimated 260.5 million US dollars gross versus 91.6 million net of reinsurance recoveries and reinstatement costs, based on a July 2025 RDS submission. Those figures are the syndicate's disclosed estimates for a specified scenario and valuation context, not actual paid losses or universal market data. They demonstrate why net reporting can combine catastrophe recoveries and reinstatement economics.
Similarly, the Syndicate 2988 accounts describe territorial and aggregate underwriting limits, catastrophe modeling, scenarios and oversight against risk tolerance. The accounts show that exposure aggregation is a board-level control topic. Neither disclosure supplies the proprietary event-loss catalog, all treaty wordings, or the full counterparty recovery schedule necessary to independently replicate every reported net scenario result.
A robust stress catalogue includes natural perils, terrorism, large industrial and marine accumulations, cyber outages, multi-line liability shocks, and interacting hazards as appropriate to the portfolio. Each relevant scenario should show the gross event loss, primary-policy limitations, treaty eligible loss, recovery under every program layer, reinstatement charge, annual cap utilization, counterparty distribution, and resulting net loss. Comparators should present a base-case treaty and explicitly documented alternative terms, instead of using unnamed adjustments.
Counterparty, Collateral, and Recoverability
An exhausted contractual cap and a failed reinsurer create different shortfalls. In the first, the agreement no longer transfers additional eligible layer loss after the cap. In the second, the treaty may still promise recovery, but the cedent faces risk that the reinsurer cannot pay promptly or in full. A monitoring system must therefore separate contractual recovery entitlement from expected collectibility, collateral coverage, cash receipt, and liquidity timing.
The Solvency II counterparty default adjustment, Article 42 of Delegated Regulation 2015/35, addresses expected changes in cash flows due to default, with separate calculation by counterparty and line of business under the specified framework. The IAIS reinsurance guidance identifies reinsurer credit quality, collateral or security, settlement reporting, and the possible consequences of a reinsurer failure for the cedent's solvency and liquidity. A highly concentrated treaty placement can generate a second-order concentration: a single catastrophe exhausts cover at the same time that it maximizes demand for payment from a small set of reinsurers.
In the United States, the NAIC 2025 property and casualty statement blank includes Schedule F information about ceded reinsurers and reinsurance recoverables, distinguishing paid, unpaid, incurred-but-not-reported, other items and related credits. This is a reporting template for the stated reporting year and line of business, not a global contract rule or a complete real-time utilization register. State law, accounting recognition and security regimes require their own analysis before a conclusion about admissible balance sheet credit is drawn.
Operational controls should group related counterparties and track contract shares, amounts owed, disputed balances, funds withheld, collateral instruments, collateral legal status and the currency and maturity of amounts due. Concentration by reinsurer legal entity should be distinguished from concentration by group and from exposure to a common collateral bank. For a cash-shortfall stress, the analyst may haircut a recoverable as a defined assumption, but must document that assumption rather than silently rewriting the contractual cap.
Reporting and Accounting Boundaries
A contract-cap monitoring ledger is neither identical to the financial reporting balance sheet nor subordinate to it. The contract ledger answers what losses are covered and how capacity is consumed. A valuation ledger estimates recoverables at a reporting date. An accounting presentation applies recognition, measurement and disclosure rules. A liquidity forecast estimates the date and likelihood of cash receipts. Reconciliation across these representations is necessary, but equality of every line is not expected.
The IFRS Foundation's guide to reinsurance contracts held explains that IFRS 17 treats reinsurance held as a separate contract from underlying insurance contracts, and that the primary insurer remains responsible under its own contracts even if the reinsurer fails to perform. The same guide also describes separate presentation of reinsurance-held assets and liabilities rather than netting them directly against underlying issued-insurance positions. Consequently, a numerical “net catastrophe loss” dashboard should explicitly state whether it means gross claims less contractually modeled recoveries, gross claims less accounting-recognized reinsurance amounts, or a cash-adjusted net financial loss.
The amount provisionally reported as a reinsurance recoverable may change when case estimates develop, when the cedent and reinsurer disagree on covered events, or when loss information arrives after a financial close. For a contract with an annual cap, recoverables reserved on earlier events may change the residual cover available for later events. Reconciliations must record valuation-date source versions and the reasons for changes rather than overwrite historical data in place.
The monitoring system should also avoid confusing legal claim settlement, claims accounting, capital model gross-to-net transforms and the treasury's eventual receipt. A cash payment of 30 against a recognized recoverable of 50 affects cash and receivables; it does not by itself prove that 20 of annual aggregate treaty capacity has returned. Equally, a reduction of an erroneous earlier case estimate can modify modeled aggregate utilization without implying that the treaty renewed. These distinctions belong in audit records and validation test cases.
Supervisory Standards and Cross-Jurisdiction Comparison
International IAIS standards provide a common vocabulary for control quality. ICP 13 and ICP 16 require effective program management and an enterprise-level understanding of risk interdependencies, while leaving contract rights and legally binding supervisory powers to domestic frameworks. Within the European Union, Solvency II Directive Article 44 addresses continuous risk monitoring individually and in aggregate. Delegated Regulation Article 260 sets out specific concentration and reinsurance controls, including concentration risk identification and monitoring and reinsurance-related credit assessment.
Solvency II has more detailed tests for recognizing risk mitigation in capital calculations. Article 208's no-double-counting requirement concerns combined scenario-based effects of coverage. Article 209's qualitative recognition conditions include legal enforceability, risk monitoring and counterparty rights. Article 210's effective risk transfer criteria address clarity of coverage and material basis risk. Article 126's reinstatement scenario treatment specifies how certain future management actions are considered between standard-formula catastrophe scenarios. These are distinct legal and technical provisions. A favorable model output does not override an unenforceable treaty term or a regulatory condition on recognition.
Lloyd's capital submission instructions describe syndicate-specific modeling and exposure reporting, particularly use of aggregate exceedance probability, model completeness and sensitivity. The NAIC Schedule F reporting structure documents recoverables and credit exposures in an American statutory setting. Their purposes and populations differ. A defensible comparison therefore asks whether each framework demands accurate inventories, clear reinsurance contract mapping, sensitivity to exhaustion, and assessment of the counterparty, without pretending that all frameworks use an identical formula or compulsory dashboard.
The institutional synthesis is consistent with contractual analysis: a usable aggregate monitor is a documented control over the translation from underlying exposure to legal recovery and ultimately to residual financial risk. Regulators may demand capital or risk-management evidence, but the practical improvement comes from maintaining one traceable contract-to-event chain capable of supporting different regulatory and accounting calculations.
Monitoring Dashboards and Decision Controls
A mature dashboard should display treaty headroom in monetary units and percentage, with explicit contractual period and basis. Beside that value should appear event-layer eligibility, recognized cumulative utilization, disputed utilization, reinstatement availability and charges, unexpired gross exposure, and annual exhaustion probability under specified models. A single green “capacity remaining” indicator without the source of each number can be materially wrong when a major claim has been advised but not finalized.
At a portfolio level, a concentration dashboard should permit drill-down by peril, geographic footprint, line of business, insured group, cedent, reinsurance layer, reinsurer legal entity and group, and period of attachment. It should show gross and net modeled losses on the same versioned scenario basis. Losses from one event should be grouped according to contract terms, while multi-event annual runs should separately accumulate aggregate cap utilization. The capability to reproduce a particular headline amount from constituent risks and signed contract terms is more informative than additional decorative charts.
Controls should define what triggers mandatory action. Examples are a proposed risk pushing an internal gross accumulation threshold, a new catastrophe reducing treaty headroom below a risk appetite tolerance, a reserve deterioration consuming the last reinstatement, a concentration of recoverables at a reinsurer facing material credit deterioration, or a model validation failure that undermines net-loss estimates. These are examples of organization-specific risk limits; no universal numerical threshold is asserted in this monograph.
The oversight cycle should reconcile daily or event-driven bordereaux and claim estimates, periodic exposure model outputs, treaty amendments and financial reporting extracts. Responsibility for underwriting data, claims estimates, contract wording, actuarial calculation, reinsurance placement, credit risk and independent validation should be named. A dashboard's output should retain the calculation timestamp, original inputs, scenario version, contract version, validation state and approving person. Independent checks must be capable of finding a duplicate loss, a missing facultative acceptance, a currency mismatch, or a recovery that improperly exceeds the treaty's maximum.
Sensitivity Analysis and Reverse Stress
Start with the four-event worked ledger and vary the annual cap without changing occurrence terms. If C=80, preliminary eligible loss totals 245 but ceded recovery is 80 and cedent loss becomes 355. If C=160, ceded recovery is 160 and cedent loss is 275. If C=240, ceded recovery is 240 and cedent loss is 195. If C=245 or more, every 245 unit of preliminary layer loss is recovered and cedent loss is 190. The intermediate numbers follow min(245,C). They deliberately ignore different premiums or reinstatement constraints: a real alternative with greater C would require additional capacity to be legally available and priced.
Attachment sensitivity differs. Holding C=160, increasing per-occurrence retention from 20 to 30 produces preliminary recoveries 40, 80, 25, 80; the total 225 still reaches the same annual cap. The selected sample's ultimate cedent loss remains 275, even though interim eligibility and the scenario conditions change. This example shows why sample-year net loss is insufficient to infer that a deductible has no economic effect: in other event sequences, different retention changes whether and when annual capacity is consumed.
Reverse stress asks what combination of events, contract interpretations and counterparty outcomes would produce a selected unacceptable net loss or complete exhaustion early in the contract year. One scenario could place two major events in the same quarter, another could treat multiple individual losses as one occurrence, and a third could reduce expected collection on the remaining treaty cover after a counterparty shock. Each scenario should use a clear causal narrative and record the assumptions causing the net result.
Where estimated exhaustion probability is based on simulation, examine alternative event frequency, severity inflation, secondary peril loadings, location uncertainty, dependence and previously unmodeled exposures. Outputs should be compared on consistent exposure and financial years. A model with a stable expected annual gross loss may still exhibit markedly changed high quantiles or timing of cap depletion when event clustering increases.
Contract Wording, Disputes, and Operational Failure
A scientifically coherent contract engine is necessary but insufficient if its rulebook is wrong. Aggregation disputes can concern whether losses constitute one event, one original cause or multiple independent occurrences; allocation can concern which treaty year, layer or insured entity bears a loss. William's aggregation scholarship analyzes the effects of wording and court interpretation. Consequently, a calculated aggregate recovery can be conditional on contested legal classifications. The dashboard should store base and alternative coverage outcomes with the precise source of each uncertainty and its financial effect.
Unconfirmed placements are another failure mode. If an underwriter assumes facultative protection will become effective but the reinsurer has yet to bind the risk, a modeled net line may overstate protection. IAIS facultative placement guidance explicitly connects procurement controls to maximum net capacity and aggregation management. Endorsements, side letters, and changes in participation can similarly change remaining headroom. Contract administration is part of exposure science because the coverage transformation has legal inputs.
A claim-reporting delay is also material. A gross event may be known and the expected ceded amount estimated before a formal bordereau or reinsurer notice is completed. Missing claim data can leave utilization understated. Conversely, duplicate records or an uncorrected superseded estimate can overstate consumption. Controls should detect both directions and should distinguish pending administrative action from legal recovery risk.
A final failure mode occurs when dashboards sum per-treaty nominal limits and report the total as portfolio protection. The same event can touch multiple layers, reinstatements and inuring arrangements; counting every nominal limit without priority, overlap, and annual utilization can double count protection. Article 208's scenario allocation provision is a regulatory example of the need to capture the economic effect of multiple protections without double counting. This principle of calculation can also guide internal reconciliations outside its immediate legal scope.
Underwriting Capacity and Management Responses
Net underwriting appetite should be connected to future residual cover rather than treated as a fixed share of gross sums insured. A cedent with substantial unused occurrence cover but nearly exhausted annual aggregate protection may reasonably restrict new exposure in the same catastrophe zone, purchase additional cover, use facultative protection for selected risks, adjust gross limits, revise retention or accept a documented increased net risk. The appropriate response depends on cost, contractual feasibility, timing, capital, liquidity and risk strategy. A single proposed action cannot be inferred universally from a threshold breach.
Reinsurers face corresponding constraints. A reinsurer accepting multiple cedent programs must identify common perils, geography, business entities, and program interactions. Individual contracts may appear modest but consume a shared catastrophe or cyber aggregate under a common shock. Inwards portfolio monitoring should also incorporate retrocession, collateral and reinstatement obligations, with protection terms measured against the reinsurer's own acceptance limits. A cedent's modeled recovery is a reinsurer's correlated potential payment and may connect both firms through the same market shock.
Good decisions preserve alternatives. Before changing underwriting strategy, estimate the exposure affected, model the expected and tail consequences, review treaty wording and placement feasibility, consider collateral and claims liquidity, record the decision authority, and identify a subsequent observation date. A contingency plan should specify how to respond to cover exhaustion between planned reporting cycles. IAIS enterprise risk guidance treats risk limits and contingency planning as continuing governance elements, not merely year-end documentation.
The control objective is the ability to answer, at any time, the following practical question: how much additional contractually eligible loss can the portfolio generate before the aggregate runs out, which exposures can generate it, under which event groupings, and what will remain with the insurer if a reinsurer fails or pays late? The answer is a conditional statement with traceable data and contractual assumptions, not one untethered aggregate number.
An Integrated Audit Specification
The research supports a concise operational audit specification expressed in existing actuarial and internal-control terminology. Identity control establishes the exact treaty version, coverage period, attachment and aggregate limit. Exposure control establishes completeness of underlying policies, sites, sums, peril mappings, and insurer shares. Loss control establishes event identity, claim grouping, development, duplication, expenses, and timing. Utilization control reconciles preliminary layer recoveries, contractual aggregate headroom, reinstatement and any disputed coverage. Recoverability control maps every recognized or modeled recovery to counterparty, collateral and timing. Scenario control independently recalculates gross and net outcomes, then tests uncertainty and model limitations.
Each control has a documentable failure test. A missing location should appear in a geocoding completeness exception. A prior-year claim should not consume a current-year cap unless the contract provides that result. An unreinstated layer should not reappear automatically in scenario two. A reinsurer claim outside its participation share should fail a numerical constraint. An aggregate recovery greater than C under the stated two-limit contract should be flagged, rather than offset by an unrelated treaty's capacity. An unresolved interpretation of an hours clause should appear with alternative exposure estimates.
The resulting evidence chain can be expressed as: policy and risk snapshot → contractually defined occurrence → event gross loss → preliminary ceded layer → annual utilization and restoration → legal recoverable → counterparty collectible amount → financial and management view. This is a synthesis of already established concepts, proposed here as a transparent audit workflow rather than a distinct theoretical system. It permits reconciliation of model outputs with contract files and financial reporting without erasing the proper meaning of each representation.
A well-designed audit records corrections to old figures as new versioned observations. If an earlier event's reserve moves from 60 to 75, the change may alter preliminary recoveries and residual cap; the system should show the before-and-after calculation, reviewer, reason and downstream changed reports. Such controls are valuable precisely when catastrophe claims develop for years and estimates remain provisional.
Research Limitations and Further Evidence Needs
The analysis relies on selected publicly accessible standards, a subset of legal and mathematical literature, and specifically identified Lloyd's financial disclosures. Many treaties, placement slips, bordereaux, full stochastic models, capital submissions and reinsurer claims files are confidential. Neither the four-event example nor the selected syndicate disclosures can estimate marketwide catastrophe risk, price a treaty, identify the true historical distribution of reinstatement charges, or establish universal contract language.
Some empirical material is disclosed only at an entity- and reporting-date level. The RDS values from Blenheim are scenario estimates and their gross-net differences include assumptions not independently reconstructible from the public accounts. Actuarial papers on dependence and aggregate pricing are theory with specified distributions; they do not provide direct calibration for all geographic or systemic perils. The legal monograph on aggregation emphasizes English doctrine. EIOPA's cited articles apply in their defined EU legal setting; IAIS texts are supervisory standards; NAIC statement forms are reporting instruments. No claim of exhaustive global legal coverage or external peer review is made.
Further research would benefit from anonymized event-level treaty and claims transactions with legally validated occurrence identifiers; contract-version histories; reserve-development trajectories; observed reinstatement invoices; model-event catalogs; collateral and reinsurer payment histories; and independent validation records. Such data could test how often dashboard headroom diverges from ultimate legal recovery, which missing data fields most affect exhaustion timing, and whether contemporary catastrophe models understate joint peril or counterparty effects.
The central methodological boundary is deliberately preserved: agreement wording, stochastic loss assumptions, and cash collectibility are separately uncertain and cannot be validated by one numerical fit statistic. The dossier's formulas and ledger are exact only within their stated hypothetical assumptions; the institutional comparisons are accurate within the verified documents' stated scope.
Conclusion
A reinsurance portfolio aggregate limit is a cumulative contractual boundary whose economic significance evolves through event frequency, occurrence classification, claims development, reinstatement and the concentration of the remaining insured exposure. A nominal annual cap cannot be interpreted as immediately available protection without first identifying the treaty period and utilization state. Per-occurrence limits, aggregate caps, aggregate stop-loss attachments and internal underwriting risk limits remain distinct mechanisms, even when an example produces the same annual final net loss.
The documented mathematical example makes the mechanism visible. Four eligible occurrence recoveries of 50, 80, 35 and 80 total 245, while a contract capped at 160 transfers only 160; the cedent retains 275 of 435 gross losses and an additional assumed reinstatement charge of 8. The residual 85 of otherwise layer-eligible claims shows how exhaustion changes the cost of later events. The stop-loss comparison demonstrates that equal sampled totals can conceal radically different response paths and future-risk consequences.
The primary standards and reporting evidence converge on the practical requirement for transparent, repeatable reconciliation. Contractually eligible recovery, capacity consumption, future restored cover, modeled tail protection, accounting-recognized recoverables and counterparty cash are separate measurements with common source data and distinct decision purposes. Underwriting, claims, actuarial, reinsurance, financial and supervisory functions should reconcile them through versioned event and contract records, explicit sensitivity tests and independent review.
The research contributes an original comparative and analytical synthesis of these requirements within the approved scope of Insurance by Bogdanova. It establishes the scientific and operational significance of aggregate exhaustion monitoring while retaining the contractual, legal and empirical boundaries of each conclusion. The manuscript is an internally researched scientific dossier reserved for later authorized batch publication.
References
International Association of Insurance Supervisors (IAIS). (2024, December; online edition accessed November 14, 2026). Insurance Core Principles and ComFrame, especially ICP 8, ICP 13, and ICP 16. IAIS online standards tool. https://www.iais.org/icp-online-tool/13524-icp-13reinsurance-and-other-forms-of-risk-transfer/
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 126, Interpretation of catastrophe scenarios. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5745_en
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 208, Methods and assumptions. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5861_en
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 209, Qualitative criteria. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5862_en
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 210, Effective transfer of risk. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5863_en
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 42, Counterparty default adjustment. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2455_en
European Parliament and Council. (2009, consolidated text accessed November 14, 2026). Directive 2009/138/EC, Article 44, Risk management. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2127_en
European Commission. (2015, consolidated rulebook accessed November 14, 2026). Commission Delegated Regulation (EU) 2015/35, Article 260, Risk management areas. EIOPA Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5922_en
Lloyd's. (2025, January). RDS 2025: Realistic Disaster Scenarios—Scenario Specification, EM 518, version 1.0. Exposure Management. https://assets.lloyds.com/media/e0c22427-dc75-419d-b9c1-5feb0823bc80/2.%20RDS%20Scenario%20Specification%20-%20January%202025.pdf
Lloyd's. (2024, July). 2025 LCR Instructions: Instructions for Submission of the 2025 Year of Account Lloyd's Capital Return and Supporting Documentation, especially Section 7. https://assets.lloyds.com/media/518ea204-2f36-4ed6-a7b0-fa721173db86/LCR%202025%20YOA%20Instructions.pdf
Blenheim Underwriting Limited / Lloyd's Syndicate 5886. (2026, February; reporting year ended December 31, 2025). Syndicate Annual Accounts 2025. Lloyd's published electronic report. https://assets.lloyds.com/media-651c0e64-c1d0-4f97-90f7-883c69fe2ef2/31ffcc4c-963c-49a2-bcd2-6244fb4693ed/5886_Accounts_2025Q4_20260219_131207.html
Lloyd's Syndicate 2988. (2026, February; reporting year ended December 31, 2025). Syndicate Annual Accounts 2025. Lloyd's published electronic report. https://assets.lloyds.com/media-651c0e64-c1d0-4f97-90f7-883c69fe2ef2/9542f82e-72b1-442d-ac82-3e063e941c34/2988_Accounts_2025Q4_20260220_164810.html
National Association of Insurance Commissioners (NAIC). (2025). Official NAIC Annual Statement Blank: Property/Casualty, especially Schedule F, Parts 1–6. NAIC. https://content.naic.org/sites/default/files/publication-asb-prop.pdf
William, Oliver D. (2021). Reinsurance and the Law of Aggregation: Event, Occurrence, Cause. Routledge, Contemporary Commercial Law series. Open-access edition, ISBN 978-1-003-08048-0. https://www.ius.uzh.ch/dam/jcr%3A54ab6629-a5e9-4636-bf1d-8a0845f4cf93/William_Reinsurance%20and%20the%20law%20of%20aggregation_event_occurrence_cause_ebook%20%281%29.pdf
Mata, Ana J. (2003). Asymptotic dependence of reinsurance aggregate claim amounts. ASTIN Bulletin, 33(2), 239–263. https://www.casact.org/sites/default/files/old/astin_vol33no2_239.pdf
Panjer, Harry H. (2014). Aggregate loss modeling. In Wiley StatsRef: Statistics Reference Online. Originally published in Encyclopedia of Actuarial Science (2006). https://doi.org/10.1002/9781118445112.stat04301. https://onlinelibrary.wiley.com/doi/10.1002/9781118445112.stat04301
IFRS Foundation. (2018, July). IFRS 17 Pocket Guide on Reinsurance Contracts Held, prepared with Darrel Scott. IFRS Foundation. https://www.ifrs.org/content/dam/ifrs/supporting-implementation/ifrs-17/ifrs-17-pocket-guide-on-reinsurance-contracts-held.pdf?hash=2EA790701F3D51AF62772AB48A6F74471B9C0035&la=en
Reijnen, Rajko, Willem Albers, and Wilbert C. M. Kallenberg. (2005). Approximations for stop-loss reinsurance premiums. Insurance: Mathematics and Economics, 36(3), 237–250. https://doi.org/10.1016/j.insmatheco.2005.02.001. https://doi.org/10.1016/j.insmatheco.2005.02.001