Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Reinsurance Treaty Experience Data and Bordereaux

Abstract. Treaty reinsurance is administered through a sequence of contractual, transactional, actuarial, and financial representations of the same underlying insurance business. Ceding insurers record insured exposures and losses, classify the business into treaty populations, calculate ceded shares or recoveries, send bordereaux or summary accounts, reconcile balances with assuming reinsurers and intermediaries, and use successive versions of that information for pricing, reserving, capital measurement, supervision, and settlement. This monograph examines the data structure that makes these operations demonstrable. Its central analytical conclusion is that the reliability of a treaty experience record depends jointly on contractual eligibility, the unit of observation, temporal assignment, versioned transaction history, and reproducible reconciliation. Improving column-level accuracy while leaving any of these relationships undefined can produce precisely formatted but economically false ceded results. The investigation distinguishes risk, premium, claims, cash, and reserve bordereaux; technical accounts; supervisory templates; and actuarial development datasets. It develops contract-level transformations for proportional and nonproportional treaties, presents transparently hypothetical numerical reconciliations and sensitivity checks, compares international reporting frameworks, and identifies data-governance controls appropriate to distinct treaty structures. Primary regulatory and standards evidence is drawn from the IAIS Insurance Core Principles, European Union Solvency II reporting rules, U.S. statutory reporting instructions, United Nations EDI message specifications, Lloyd's reporting standards, ACORD documentation, and IFRS materials. The study makes an original source-grounded synthesis and explicit computational demonstrations; it claims neither a newly established theory nor empirical estimates from undisclosed industry data.

Keywords: Reinsurance; treaty reinsurance; bordereaux; ceded premium; ceded loss; proportional reinsurance; excess-of-loss reinsurance; experience data; actuarial reserving; data lineage; reporting reconciliation; technical accounts; development triangles; regulatory reporting; counterparty exposure; data governance; retrocession.

Introduction

A reinsurance treaty transforms a population of underlying insurance obligations into an allocation of risk, premium, loss, expense, and settlement between contractual counterparties. The transformation requires evidence of what business was insured, which business was ceded, which losses arose, and which contractual provisions were applied. A bordereau is one established mechanism for transmitting that evidence. The Reinsurance Association of America's glossary definition of bordereau explicitly connects periodic reporting to identifiable risks, reinsurance premiums, and losses. The United Nations REBORD message specification similarly describes individual-policy data and their effect on a reinsurance contract. Together, these sources establish an essential relationship: a reporting record is meaningful because it is attributable to a particular contract, risk, and financial consequence.

The contractual character of the obligation is fundamental. Under IAIS ICP 13, reinsurance transfers insurance risk economically while creating credit, operational, and basis risk. The obligation is between cedent and reinsurer. The mere existence of an underlying policy or recorded loss does not establish an immediately due receivable from a reinsurer; the relevant contract, attachment, exclusions, reporting obligations, settlement terms, and applicable valuation rules must also be considered. This distinction informs every data definition in the monograph.

The study investigates three related questions. First, which units, definitions, and documented assumptions allow an individual ceded transaction to be classified and reproduced? Second, how do the data definitions affect calculated premiums, recoveries, outstanding balances, actuarial estimates, and counterparty exposures? Third, how should organizations control correction histories and reconcile different legitimate views of the same business across treaty forms and jurisdictions? The object is treaty experience data and reporting, including its interface with actuarial and financial measurement. Broader questions of reinsurance market equilibrium, treaty negotiation, and insurer-wide capital optimization belong to adjacent research areas.

The methodology combines documentary comparison, contract-based algebra, transaction-level thought experiments, sensitivity analysis, and evidence-to-claim verification. The mathematical examples are wholly hypothetical, use explicitly stated simplifying assumptions, and represent neither observed portfolio estimates nor model calibration. The document distinguishes authoritative legal requirements from sector standards, educational guidance, voluntary data specifications, and suggested internal controls. Its comparisons are international in scope; no one legal regime is presumed universal.

Concepts, History, and Evidence Sources

The reporting object and its historical continuity

Experience data denotes recorded information about covered exposures, premiums, insured events, claims, incurred and paid amounts, reserves, and corresponding reinsurance transformations. A treaty experience extract is a selected, dated view of that information, constructed according to a particular agreement and analytical purpose. A bordereau typically supplies detail by policy, risk, cession, or claim; an account statement summarizes contractual debits and credits; a regulatory report restates selected measurements according to a prescribed reporting basis. Each is a legitimate representation provided its granularity and accounting convention are declared.

The UN/EDIFACT standard makes this distinction especially clear. The REBORD message identifies policies and individual cessions contributing to premium or claims figures. The RETACC technical account, in contrast, reports one recipient's position in the sender's books with respect to a single reinsurance contract and may support payment. An amount can therefore be traceable from a bordereau row into a technical account without either document being a complete substitute for the other. The two standards also show that electronic structuring of reinsurance messages predates contemporary cloud and API reporting.

Industry practice includes both detailed and aggregate reporting. In the Jacqueline Friedland’s (2023) reinsurance reserving study, premium and loss bordereaux are described alongside summary reporting and reporting-lag difficulties. The distinction matters when reconstructing historical years: a missing individual risk record may reflect a contractual reporting convention rather than an omitted policy. The absence of claim counts and exposure units can nevertheless prevent useful severity, frequency, and closure diagnostics even when aggregate incurred losses are available.

Modern data standards seek to reduce semantic friction across multiple counterparties. ACORD GRLC specifications include data dictionaries, codes, schema files, and process-oriented guides for reinsurance and large commercial transactions. ACORD's 2025 GRLC Generation 2.0 announcement distinguishes placing, accounting and settlement, and claims messaging and describes later updates as forthcoming. The statement is evidence of an announced standards program, not evidence that every proposed message had already been implemented across markets. Implementations must be versioned and assessed against the actual agreed specification.

Contractual authority and data authority

A dataset has several distinct sources of authority. The signed treaty and endorsements determine the contractual scope and formula. A primary system records policies, exposure units, and financial movements. A cedent's accounting close gives a controlled reporting snapshot. A broker may translate or aggregate the data, while the reinsurer may independently classify and validate it. Statutory accounting rules, supervision, and financial reporting each may require another measurement or grouping. A data field can be technically valid yet be contractually inapplicable when used with the wrong treaty period, currency basis, or risk-attachment condition.

The IAIS guidance on reporting requires material reports due between counterparties to be timely and complete and ties reinsurance effectiveness to documented contracts and internal controls. It establishes a prudential expectation rather than a globally identical bordereau template. In the delegated-authority context, Lloyd's reporting standards establish core risk, premium, tax, regulatory, and claims data for relevant coverholder and claims-administration arrangements. Those are primary-to-market delegated reporting requirements and should not automatically be described as requirements governing every cedent-to-reinsurer treaty. Their detailed field-level controls provide a useful comparator for standardized downstream reporting.

Evidence hierarchy and interpretive method

The analysis assigns the greatest authority to a current operative treaty for case-specific coverage and to binding legal or supervisory instruments for their actual jurisdictional mandates. Institutional standards describe interoperable data structures and supervisory expectations; association glossaries and actuarial texts establish terminology and techniques; selected reporting examples demonstrate possible implementation patterns. The Society of Actuaries' treaty provisions material illustrates why life reinsurance agreements may specify the reporting of new business, policy changes, claims, premium calculations, and statistical information, as well as timing and sample formats. It is a drafting reference, rather than a law binding every treaty.

The source base supports institutional and mathematical analysis but does not supply private, audited pairs of source-system and reinsurer-ledger transactions. Thus, statements about frequency of reporting errors, the cost of automation, or quantitative gains in reinsurance recoveries cannot be inferred for the world market. Numerical results below are author-computed conditional examples.

Data Dictionary and Coverage Units

Identity and granularity

The minimum meaningful conceptual distinction is among treaty, section or layer, underwriting period, insured risk, policy, cession, occurrence, claim, transaction, and accounting snapshot. These may have one-to-many or many-to-many relationships. A policy may cover multiple locations, a location may contribute to an aggregate, a claim may involve several coverage sections, and a single occurrence may combine many claims for a catastrophe layer. A reinsurance program can involve multiple subscribing counterparties. The primary business key is therefore rarely a single policy number.

A reproducible reporting schema should preserve at least the following logically separate entities:

Entity Stable identity and defining attributes Critical analytical use
Treaty and section Treaty ID, legal counterparties, type, effective dates, signed share, wording and endorsement versions Determine applicability and payment obligations
Underlying coverage Policy ID, insured object, location, perils, limit, deductible, currency, written and effective dates Determine insured exposure and coverage
Exposure and cession Exposure ID, ceded share or risk amount, treaty year, cession date, policy endorsement ID Reproduce allocated premium and risk
Occurrence and claim Occurrence ID, claim ID, date of loss, peril, coverage section, paid and case-reserve states Avoid misaggregation and identify recoveries
Financial movement Immutable movement ID, event type, amount, currency, posting and effective dates, reversal link Reconcile cash and accounting periods
Snapshot and transmission Report ID, as-of date, delivery time, source extract revision, approval and recipient Reproduce what each party knew when

Each field needs a type, allowable values, currency convention, units, null semantics, whether it is mandatory for a particular treaty and report, and the system of origin. Reference data should identify the version of peril codes, geographic codes, company identifiers, and line-of-business classifications. The RAPA data management initiative explicitly emphasizes consistent treaty codes, product and plan codes, claims reasons, underwriting categories, and business-partner communication of changed systems. A common vocabulary reduces a class of avoidable matching failures, although semantic agreement still requires contractual interpretation.

Dates as distinct dimensions

An experienced analyst distinguishes policy inception, exposure start, accounting booking, cession attachment, loss occurrence, notification, case-reserve change, reinsurance presentation, remittance, settlement, and reporting snapshot dates. The choice of an underwriting-year, accident-year, occurrence-year, or reporting-year cohort must be specified. A risk-attaching treaty can classify policies by attachment or inception and cover losses occurring later; a losses-occurring treaty generally applies by qualifying loss occurrence within its period. Claims-made or discovery triggers and special wording require their own rule. Contract labels alone are insufficient; the wording determines the actual mapping.

Suppose a policy attaches on December 20 of Year 1, a loss occurs on January 5 of Year 2, the claim is reported February 1, and the ceded share is booked March 1. The same item may appear in Year 1 underwriting exposure, Year 2 accident loss, and Year 2 reporting and financial movements. Combining these bases in an apparently consistent development triangle creates a spurious emergence pattern. The correction is to retain each date and define a separate cohort transformation, rather than overwrite the raw transaction date to force agreement.

Currency and value basis

A bordereau should distinguish original currency, transaction currency, settlement currency, reporting currency, source amount, exchange-rate date, and rate source. Gross premium written, earned premium, premium due, premium collected, return premium, reinstatement premium, and ceded premium each describe different economic quantities. Similarly, gross paid, net paid, salvage and subrogation, case outstanding, incurred-but-not-reported estimates, allocated expenses, and reinsurance paid or recoverable are separate measurements. The IFRS 17 standard overview concerns recognition and measurement of insurance contracts, including reinsurance contracts held; its presentation and fulfillment-cash-flow measures cannot be equated mechanically with a treaty's quarterly cash account.

For a stated local convention, an analyst may report a movement as A(r)=A(o)×FX(o→r,t,b)A^{(r)}=A^{(o)}\times FX(o\to r,t,b), where tt denotes the rate date and bb the selected valuation basis. The result is reproducible only if all arguments are recorded. Rounding should occur at a declared stage: transaction, currency aggregation, or final statement. Different rounding policies can generate small reconciling differences that should be separately classified from missing business.

Exposure completeness and denominator discipline

An expected loss ratio generally compares a defined loss measure with a defined premium or exposure denominator. The classic reinsurance pricing discussion by Clark stresses the individually tailored nature of treaties and the limits of simple experience rating. If a bordereau contains only claims above a reporting threshold while the premium denominator covers all risks, the apparent loss experience has been left-censored. If only cessions accepted by a reinsurer are supplied, that subset is not necessarily representative of every policy in the cedent's book. Exposure counts and limits also need consistent treatment of renewed, canceled, reinstated, and endorsed policies.

The data dictionary must therefore record both the population definition and the observation mechanism: what qualifies for inclusion, how missingness arises, and which financial or claim thresholds trigger detail. A completeness percentage measured over records is a poor substitute for completeness measured over insured exposure, ceded premium, or ultimate loss when missing rows are systematically large.

Cession and Claim Bordereaux

Proportional treaties: transaction algebra

Let ii index a qualifying underlying risk, GiG_i its eligible gross premium, qiq_i the contractual ceded percentage, and Li,tL_{i,t} the qualifying loss movement at time tt. Under a simple quota-share agreement with a constant share, no exclusions, no limits, and no unusual expense provisions, ceded premium is Pic=qiGiP_i^{c}=q_iG_i and ceded loss is Li,tc=qiLi,tL_{i,t}^{c}=q_iL_{i,t}. For a variable percentage or surplus-share treaty, qiq_i must be calculated from the insured amount, retention and accepted treaty line for the specific risk. Commission calculations and cash accounts are additional operations and should be recorded separately.

A defensible premium bordereau contains policy/risk identity, treaty and treaty-year identity, type of premium movement, period, gross basis, allowed adjustments, cession percentage, ceded premium, commission basis, currency, and an immutable movement identifier. A loss bordereau contains claim/occurrence identity, relevant coverage, loss and reporting dates, gross payment and reserve changes, recoverable share, paid-to-date, salvage, expense allocation, and event and layer references when relevant. It must distinguish cumulative balances from incremental changes.

The RAA reinsurance glossary and UN REBORD definition both support reporting detailed premium and claim contributions. The exact number and nature of fields are set by contractual and operational arrangements. A statement of account can be sufficient under some treaty wordings, as explained in the IRMI discussion of bordereau provisions; detailed records are therefore a scientifically useful audit trail, not automatically a legal delivery obligation in every treaty.

Reproducible proportional example

Consider two fully eligible policies in one currency, one reinsurer, no tax, funds withheld, premium adjustments, interest, salvage, or minimum premium. The treaty provides a 25% ceding commission on ceded written premium; all stated paid claims have been accepted under its terms. For simplicity, amounts are dimensionless monetary units, and loss movements are gross of reinsurance. This example does not model the complete contractual profit commission or settlement timetable.

Policy Gross written premium Ceded share Ceded premium Gross paid loss Ceded paid loss Gross case outstanding Ceded case outstanding
A 1,000 40% 400 300 120 200 80
B 800 25% 200 0 0 600 150
Total 1,800 — 600 300 120 800 230

Ceding commission is 0.25×600=1500.25\times600=150. If the hypothetical treaty account nets ceded premium against commission and amounts already paid by the cedent and reimbursable by the reinsurer, the simple balance payable to the reinsurer is 600−150−120=330600-150-120=330. Case outstanding of 230 is an informational measure of expected future claim payments, not an immediately payable cash entry under these assumptions. The precise sign convention and settlement responsibility must be fixed by the actual technical-account clauses. The RETACC message distinction between financial and informational account entries corroborates the need to separate these amounts.

Now policy B receives an endorsement increasing gross eligible written premium by 100 while its ceded share remains 25%. The incremental ceded premium is 25, incremental commission 6.25, and the new net account balance—holding every other variable fixed—is 625−156.25−120=348.75625-156.25-120=348.75. The difference is 18.75. A current cumulative export showing only 625 and 156.25 cannot explain the original 330 unless the endorsement is separately retained. A versioned movement ledger with an original position and a delta can reconstruct both reports.

Nonproportional treaties: aggregation changes the unit

For a single occurrence with qualifying underlying loss XX, attachment aa, limit uu, and no other special provision, the elementary occurrence excess-of-loss recovery is

R(X;a,u)=min⁡{max⁡(X−a,0),u}.R(X;a,u)=\min\{\max(X-a,0),u\}.

For a hypothetical occurrence loss of 8 million, attachment of 3 million and layer limit of 4 million, the layer recovery is 4 million. Net loss is 4 million before considering other reinsurance. At 2.5 million gross loss, recovery is zero. An additional occurrence may consume a fresh per-occurrence limit or interact with an aggregate limit and reinstatement clause. That effect depends on the treaty. A correct calculation therefore needs occurrence grouping, contract layer and basis, aggregate exhaustion, any reinstatement, inuring protection, and post-loss adjustments. A claim-by-claim calculation that applies the occurrence attachment to each claim can materially understate coverage for a multi-claim event.

A nonproportional premium may be fixed or adjustable using a declared exposure measure, deposit premium, minimum-and-deposit provision, rate-on-line, subject premium, or other agreed basis. Consequently, a bordereau of individual ceded percentages may have no direct analogue. The appropriate experience record may be an underlying-loss listing, event summary, subject-premium statement, adjustment-premium calculation, or cash-call presentation. Clark's excess-layer pricing treatment is a useful actuarial point of reference, but it does not prescribe each contract's legal loss definition.

Claims movements and reserve development

Claim experience should have a clear state model. A paid movement changes cumulative paid; an opening or revised case estimate changes case outstanding; a closure may remove outstanding reserves and may involve final payments or recoveries. Gross incurred at a valuation date is often defined for a chosen reporting dataset as paid-to-date plus case outstanding. Estimated ultimate and incurred-but-not-reported amounts are actuarial estimates built on further assumptions and datasets. Their definitions should be explicit, especially when legal claims expenses or salvage are tracked separately.

Actuarial development analysis depends on stable valuation dates and accident or underwriting cohorts. Friedland's basic reserving techniques explain transaction and triangle structures and illustrate how definitions affect methods. A triangle prepared from each monthly received file without historic snapshots may fail to reproduce prior evaluation points after corrections. Retaining transaction-effective dates, ingestion dates, and frozen valuation snapshots makes the triangle auditable.

For proportional business, a claim may have both gross and ceded cumulative views. For excess business, the ceded view depends on the evolving event estimate, layer position, exhaustion, reinstatements, and possibly other treaties. The formula for a ceded incurred triangle therefore should be documented as a transformation of current loss information and contract terms, rather than assumed to be a stable percentage of the primary triangle. An adjustment in occurrence aggregation can change ceded results without changing any underlying gross claim payments.

Life, health, agricultural, and cyber-specific extensions

Treaty experience has line-specific observation units. Life treaties may report policies, lives, sums at risk, attained ages, deaths, lapses, reinstatements, recaptures, and premium modes. The SOA treaty reporting provisions provide explicit examples of these elements. Health arrangements may require member-months, claim service and adjudication dates, provider classifications, benefits, and recoveries; personal-data restrictions can influence permissible transfers. Agricultural programs may report insured acreage, crop season, yield, geographic exposure, event definition, and government-program participation. Cyber treaty datasets may require event clustering across insured entities, dates of discovery, systemic service-provider dependencies, and coverage sublimits. These are analytical requirements arising from the risk structure, not claims that a globally identical compulsory field list exists for these lines.

The economically correct reporting unit follows the treaty's insured event and ceded obligation. A single universal bordereau can standardize data typing while still preserving multiple contract-specific schemas and aggregations.

Data Lineage, Errors, and Corrections

A ceded result is a derived record

A ceded premium or recovery should be reconstructible as the output of a transformation over identified underlying events, a treaty version, a valuation rule, and a defined observation period. In general notation, write

Ct,v=F(Dt,v,Tv,Av),C_{t,v}=F(D_{t,v},T_v,A_v),

where DD contains the identified source records, TT the applicable legal treaty terms, AA the valuation and aggregation assumptions, tt the observation or valuation cutoff, and vv the version of the inputs and computational rules. This is a representation of accounting and computation, not an invented insurance doctrine. It makes explicit that a calculation can change because of a corrected source record, an endorsed contract, a changed valuation basis, or a revised grouping rule. Those causes must be distinguishable in the audit trail.

The W3C PROV-O provenance standard distinguishes entities, activities, and responsible agents, and defines relationships such as derivation, use, and generation. In treaty administration, a source policy record, an extract, a validated cession record, a technical account, and an actuarial dataset can each be represented as separate versioned entities. Extraction, transformation, approval, transmission, and reconciliation are activities. This provenance vocabulary offers a transferable method for recording data origins while leaving insurance-specific contract interpretation with the insurer and reinsurer.

At minimum, a material derivative amount should retain a source-system reference, extraction timestamp, snapshot or change-sequence ID, contract version, processing version, transformation log, approval record, currency-conversion basis, and reconciliation result. A hash proves that bytes did not change between two measured states, not that the source records were complete or the contractual calculation was economically correct. A chain of signed files can faithfully preserve an error. Substantive verification must still compare the data with authoritative contractual and transaction records.

Structural defects and their financial effects

Errors can arise from at least eight distinct mechanisms: missing rows, duplicate rows, wrong treaty attribution, incorrect ceded percentage or layer parameters, date-basis mismatch, currency transformation, inconsistent claim-status definitions, and an unrecorded correction. These mechanisms have different effects. Duplicated gross premium increases the apparent exposure denominator and may inflate ceded premium payable; an omitted large claim decreases paid or incurred loss and expected recovery; an incorrectly grouped catastrophe can overstate or understate excess-layer exhaustion; an old signed share can misallocate balances among reinsurers even if the total ceded amount appears correct.

For a simple quota-share dataset with true eligible gross premium GG, ceded proportion qq, and source premium error δG\delta G, the cession error is δPc=qδG\delta P^c=q\delta G, assuming qq remains correct. If the source premium is understated by 5% on a hypothetical portfolio with 10 million eligible gross premium and a 40% quota share, the ceded premium understatement is 0.40×0.05×10 million=0.20 million0.40\times 0.05\times10\text{ million}=0.20\text{ million}. If the original ceded premium is 4 million, this is a 5% relative error. The result is mechanical and conditional; it does not estimate the actual frequency of underwriting omissions.

For an occurrence excess layer, local sensitivity behaves differently. In the interior interval a<X<a+ua<X<a+u, a marginal increase in accepted loss produces an equal increase in layer recovery, while below attachment or beyond full exhaustion the marginal response of the simple capped formula is zero. At attachment and exhaustion boundaries the derivative changes discontinuously. This means that reporting an inaccurate event total can have negligible or substantial effects depending on proximity to the threshold. The sensitivity is contract- and state-dependent, not proportional to the underlying-data error.

Transaction history, restatements, and version discipline

A critical distinction exists between corrected knowledge of the past and a new financial event. An endorsement may create a legally effective premium adjustment. A late claim report adds previously unreported experience. A corrected currency code repairs a prior processing mistake. A reserve strengthening changes a current estimate of future payments. A data-quality system should classify these changes separately, although a report may aggregate their financial impact.

A practical immutable-movement approach records each approved change as an additional event with its own ID, effective date, posting date, reason, source, and explicit reversal or supersession relation. A source error may be corrected by a negative reversal plus a replacement amount, while a genuine subsequent endorsement is recorded as a delta attributable to that endorsement. The current position is obtained by summing eligible movements as of the specified cutoff. The prior delivered report remains reproducible from the prior cutoff and version. This technique is consistent with the provenance logic of the W3C model for generated records; it is a recommended audit implementation, not a prescribed universal ledger format.

Three state views should coexist. The business-effective view records when a premium or coverage change legally applies. The accounting view records when a movement was posted into a controlled close. The information view records what was known and transmitted at a particular time. These views explain why a reinsurer may have received a later but correct bordereau that disagrees with the cedent's earlier closed account. A reconciliation must establish whether the difference is timing, permissible estimate change, classification, or true discrepancy.

Missingness, selection, and estimation

Data incompleteness is not random by default. Large-loss notifications may be accelerated, small losses may arrive only in periodic batches, closed files may omit withdrawn claims, and catastrophe exposures may be corrected after geocoding. A bordereau based only on notified losses can therefore misstate development patterns. If the probability that a record appears depends on its magnitude, transaction date, broker route, or claims status, simple imputation under an independent-missingness assumption may be biased.

The actuarial remedy begins with an explicit selection mechanism, cohort definition, and assessment of what is not reported. Clark's experience-rating limitations and Friedland's loss-development methods motivate the use of multiple diagnostics and assumptions. A complete audit record should separate observed exposures and claims from modeled ultimate losses and show the sensitivity to reporting delays and exposure adjustments. The monograph does not propose a universal imputation factor; such a factor requires portfolio-specific testing.

Personal data, commercial confidentiality, and security

Individual bordereaux may contain insured names, claimants, health data, location identifiers, and confidential underwriting prices. Where European personal-data law applies, the GDPR principles in Article 5 require purpose limitation, data minimization, accuracy, and storage controls; its Article 32 addresses appropriate technical and organizational security measures. Treaty reporting must therefore specify permissible purposes, roles, transfer mechanisms, and controls tailored to the data and relevant legal regimes. A report transmitted to a reinsurer need not contain every raw personal identifier required internally to trace a claim. Controlled pseudonymous keys can preserve reconciliation where counterparties also have an authorized mapping process.

Data minimization and auditability can be jointly achieved through stable pseudonymous identifiers, restricted mapping tables, retention schedules, access logging, encryption, and selective disclosure. Encryption of a complete data dump does not settle whether each field was necessary or lawfully transferred. The legal analysis depends on jurisdiction, lawful basis, contractual roles, and any cross-border transfer rules; this work provides the data-governance implications rather than individual legal advice.

Contract-Level Reproducibility Controls

Reconstructing the treaty calculation

Reproducibility begins with the treaty instrument. For each period, the system should retain an identifiable signed treaty, section, schedules, endorsements, authorized participants, signed shares, and the exact coverage and settlement provisions used in computation. Business rules should map contract concepts to data fields: eligible premium, covered peril, risk attachment, occurrence definition, retention, limit, aggregate, annual period, reinstatement, allowable expenses, salvage, commission, currency, and payment terms. Where contractual language requires judgment, the approved interpretive decision belongs in the record with the relevant clause and decision date.

The IAIS control guidance in ICP 13 calls for clear, prompt documentation of economic and coverage terms and complete, timely reporting. It also highlights the effect of mismatched underlying and reinsurance terms. A control that merely checks whether the ceded percentage is numeric misses those contractual issues. For example, an otherwise correct 40% share can still be applied to a liability risk that falls outside the treaty's territorial scope, or to a policy incepting after the relevant risk-attaching period.

A defensible audit chain has four computational stages. First, select underlying policies and claims by explicit eligibility rules. Second, transform eligible gross movements into ceded financial and exposure movements using the applicable treaty version. Third, allocate assumed interests among signed participants and any broker or funds-withheld accounts without treating intermediaries as the risk-bearing party. Fourth, aggregate the resulting items into snapshots for cedent accounts, reinsurer accounts, actuarial analyses, financial statements, and supervisory returns, retaining their distinct bases. Each stage must be independently inspectable.

Required reconciliations

There are at least six economically distinct reconciliations:

  1. Underlying policy ledger to selected exposure and premium bordereau, with omissions and exclusions disclosed.
  2. Gross claim ledger to selected claim, occurrence, and loss bordereau, respecting cumulative and movement conventions.
  3. Selected bordereaux to contractual ceded allocations, including shares, caps, and signed lines.
  4. Ceded movements to treaty technical account, documenting commission, expense, settlement, interest, tax, and informational reserve balances.
  5. Technical-account balances to bank payments, broker settlements, cash calls, outstanding receivables/payables, and aging.
  6. Contract-level balances to accounting and supervisory outputs after declared valuation and grouping adjustments.

These reconciliations answer different questions. Equality between premium bordereau total and technical-account cash balance is generally neither expected nor sufficient: commissions and claims settlements may intervene. Likewise, case outstanding in a claims bordereau may be relevant to actuarial incurred loss while a technical account carries only current-period cash settlements. The UN RETACC specification explicitly provides for financial and informational entries; their conflation creates false reconciliation failures.

A useful reconciliation record has opening state, period additions, period reversals, foreign exchange effects, other contract-specific adjustments, closing state, unmatched differences, tolerance rules, responsible approver, and supporting document IDs. Let B0B_0 be the previously approved balance, Δj\Delta_j the signed movements, and B1B_1 the new closing balance. Then B1=B0+∑jΔjB_1=B_0+\sum_j\Delta_j must hold on the declared reporting basis. A formally balanced ledger can nevertheless be incomplete if both a source transaction and its expected accounting entry are missing. Accordingly, an independent exposure and claim completeness test is required alongside the arithmetic control.

Two-party comparison and dispute taxonomy

The cedent and assuming reinsurer maintain different books and may process the same transaction on different dates. A two-way comparison should use persistent treaty and reinsurer identifiers and an explicit reporting cutoff. Differences should be classified as source-record omission, timing difference, unapproved amount, contractual eligibility dispute, currency or rounding difference, participant allocation mismatch, unsettled cash, and prior-period restatement. The classification directs the remedy. A timing difference may reverse naturally, while a coverage dispute requires documented interpretation. Reconciliation should preserve unresolved material amounts rather than force them into a balancing journal with no provenance.

Control thresholds should be risk-based. An absolute tolerance useful for high-volume low-premium accounts can be dangerously permissive for a single large catastrophe claim. Dual thresholds—absolute monetary amount and percentage of an appropriate denominator—can prioritize exception review, subject to explicit coverage and materiality rules. Near an excess-layer attachment, a small source discrepancy can have a large relative effect on the ceded recovery. In such cases, a boundary-sensitive control is preferable to a uniform percentage threshold.

Reproducible hypothetical end-to-end test

The two-policy quota-share example above allows an executable set of assertions. The source contains two distinct risk records, a single underlying claim on policy A, a case reserve on each policy, and a later premium endorsement on B. The agreed ceded percentages and commission are version-controlled. The initial pass produces ceded premium 600, commission 150, paid recovery 120, case outstanding 230, and hypothetical net amount payable to reinsurer 330. A later version adds gross premium 100 to B and produces ceded premium 625, commission 156.25, paid recovery 120, and net amount payable 348.75. The change to the net payable is 18.75. Replaying the former data version must produce the initial result exactly.

A further test should duplicate the endorsement movement ID. An idempotent ingestion process rejects or quarantines the duplicate and preserves 348.75. An incorrect ingestion that accepts it doubles the premium adjustment to 50 and commission adjustment to 12.50, resulting in a spurious payable of 367.50. A rejection test can therefore detect a 18.75 overstatement. These numbers are not empirical observations: they follow from the stated algebra. Independent computation, which this research package records, is necessary to guard against arithmetic transcription errors.

For the excess-layer example, one should test a gross occurrence below attachment (2.5 million), within the layer (5 million), and above full exhaustion (8 million): the simple formula yields recoveries of 0, 2, and 4 million respectively. An event represented by multiple claims should yield the same recovery after aggregation as an equivalent single event amount, provided all claims share a valid occurrence grouping. Separate occurrences must remain separate where the treaty treats them separately. These tests establish software arithmetic and interpretation under stipulated assumptions; actual claims recoverability still depends on enforceable wording and claims evidence.

Financial reporting and supervisory bridges

Financial reporting imposes its own scope. The IFRS 17 requirements measure groups of insurance contracts using specified estimates of future cash flows and presentation principles. In particular, reinsurance contracts held are separately recognized and presented within the standard's architecture. A ceded claim cash account is therefore input evidence for one part of financial reporting, not the complete measurement of a reinsurance contract held. Differences may arise from projected future cash flows, discounting, risk adjustment, contract grouping, and reporting-period recognition; bridging schedules must explain those differences.

The European Commission's Implementing Regulation 2023/894 sets Solvency II supervisory reporting templates. Its Article 23 identifies prospective reinsurance program and counterparty templates, with certain S.30 reporting conditioned on the ratio of reinsurance recoverables to total best estimate, and specifies S.31.01 information on reinsurers' shares. This is supervisory classification of contracts and recoverables, not a prescription for policy-level bordereau columns. The EIOPA supervisory reporting overview documents the operative ITS framework. EIOPA's recoverables-versus-receivables clarification illustrates how the accounting state of policyholder payments affects regulatory reinsurance reporting categories. The proper mapping depends on the actual valuation category, not a raw bordereau field called 'reinsurance balance.'

In the United States, the NAIC's 2026 Property/Casualty Annual Statement Instructions include Schedule F reinsurance reporting and specific instructions for ceded reinsurance premiums payable and funds held. Schedule F records regulated financial relationships and classifications, potentially with several categories for the same counterparty arising from historical contract and status differences. A source system that aggregates solely by reinsurer name without the contract, entity identifier, and applicable legal status may be unable to reconstruct the required classifications. NAIC statutory reporting and EU Solvency II reporting remain legally distinct; a shared transactional foundation can feed both only through documented jurisdiction-specific transformations.

Cross-Framework Synthesis, Research Gaps, and Conclusions

Standards converge on traceability while serving different goals

The cross-framework comparison establishes a layered picture. Contract documentation establishes covered risk and the payment relationship. Reinsurance transaction standards establish structured exchange. Actuarial literature establishes evaluation methods and assumptions. Financial accounting establishes recognition and measurement. Prudential reporting establishes specified aggregate disclosures and supervisory classifications. Data-protection law constrains the permissible contents and handling of transmitted information. These layers share source facts while retaining their independent definitions and authorities.

Framework or institution Principal question answered Main data granularity Limitation for treaty reproduction
Signed treaty and endorsements What has been transferred and on what terms? Contract, section, claim/occurrence and covered business Requires correct transaction population
UN/EDIFACT REBORD Which individual policies or cessions contribute to the figures? Policy, cession, premium/claim contribution Message structure alone cannot determine legal coverage
UN/EDIFACT RETACC What is one contract's technical-account position? Treaty and account entries Aggregates may conceal missing or duplicated risks
ACORD GRLC How can trading partners exchange placing, claims, and account messages consistently? Defined business messages and referenced data Version, implementation scope, and bilateral mapping still matter
IAIS ICP 13 Is reinsurance effectively managed and supervised? Risk program, counterparties, contracts and controls Principle-based supervision does not supply a single transaction schema
Lloyd's delegated reporting What delegated risk, premium, and claims information must be collected? Coverholder/delegated data Not universally a treaty reporting mandate
EU Solvency II ITS What information is required for specified EU supervision? Standardized supervisory fields and treaty/counterparty data Regulatory ratios and classifications differ from treaty cash movements
NAIC Schedule F What statutory reinsurance balances and counterparties are reported? Insurer, reinsurer and specified statutory categories State-specific statutory basis and statement conventions apply
IFRS 17 How are insurance and reinsurance contract groups recognized and measured? Contract groups and fulfillment cash flows Values are not identical to a periodic treaty bordereau

The primary IAIS reinsurance standard, UN bordereau message, ACORD GRLC documentation, and Lloyd's reporting guidance collectively support structured, attributable communication. Their different purposes explain why harmonization must preserve contract rules, message versions, legal basis, and aggregation units. The comparable feature is the provenance of facts, not an assumption of identical definitions.

Original analytical findings

First, treaty data quality is relational. An accurate claims amount with the wrong occurrence ID is an inaccurate ceded recovery input. A correct cession percentage attached to the wrong signed treaty is an incorrect counterparty exposure. Field accuracy, identity, eligibility, aggregation, and timing jointly determine reliability. The case calculations show that the effect of a source error depends on contract structure: linear for a simple quota-share percentage and piecewise for an occurrence excess layer.

Second, reporting completeness should be defined with respect to the decision. For a premium account, the relevant question may be whether all period-eligible premium movements have been delivered. For experience pricing, the question is whether the observed exposure-loss population and its reporting delay are sufficiently representative. For solvency or counterparty credit assessment, the question includes prospective contractual rights, valuation adjustments, and reinsurer identity. A single generic 'completeness score' cannot prove all three.

Third, reconciliation is a collection of controlled transformations rather than one equality test. The gross-to-cession, cession-to-account, account-to-cash, and account-to-regulatory bridges have distinct adjustment categories. Explicit bridges make legitimate measurement differences explainable while keeping unresolved contract disputes visible. Mathematical balance checks and semantic checks are complementary.

Fourth, version history is part of scientific reproducibility. The data available at the time of a pricing decision should be reconstructible even after later corrections. The research analyst should preserve both the original reporting state and the restated state, report changes and classification logic, and avoid silently replacing the prior data population. This is necessary for reproducible retrospective testing of rating assumptions and operational controls.

Fifth, modernization yields its strongest evidential contribution when it carries stable contractual identities and transparent change histories across organizations. A validated exchange schema can reduce syntax errors while leaving treaty attachment disputes, semantic confusion, and counterparty allocation mistakes unresolved. Data standards can support accuracy, but agreement on terminology and checks against the underlying contract remain indispensable.

These results are an original comparative synthesis, mathematical elaboration, and editorial inference from documented mechanisms. They are not introduced as an independent new theory, principle, doctrine, or universally quantified empirical finding.

Counterexamples and boundary cases

The analysis accommodates legitimate differences. A treaty may require only quarterly summary accounts; demanding policy-level rows may exceed the contractual obligation. A program of facultative placements may have detailed individual slips and claims records but no periodic treaty bordereau. A parametric arrangement may pay on an externally measured index without an indemnity claim-by-claim calculation. An aggregate excess contract may require policy detail to verify subject premium while using portfolio-level attachment in its loss calculation. Life reinsurance may demand detailed in-force and termination records where a property catastrophe excess treaty relies heavily on event exposure aggregates. These cases confirm that the appropriate data grain follows the contract and decision purpose.

A second boundary involves counterparty identity. Brokers often transmit and reconcile messages, but a broker's role in a cash settlement does not automatically make it the risk-bearing reinsurer. The NAIC statutory instructions distinguish risk-bearing counterparties for relevant reporting purposes. Participant shares, legal entity identifiers, and potential multiple reinsurer-status classifications should be retained independently from settlement-channel and intermediary metadata.

A third boundary concerns legal and personal data. A highly detailed claim report might improve forensic traceability yet expose sensitive health or personal identifiers beyond what is required. The GDPR's data minimization principle demands context-specific assessment. Scientific reproducibility can be supported by controlled identifiers, private mappings, and protected audit records, rather than unrestricted disclosure of raw claimant details.

Research limitations and agenda

The principal limitation is the unavailability in public sources of a large matched panel of raw cedent transactions, delivered bordereaux, reinsurer acceptance records, signed wording, and settled cash. The documentary analysis can establish mechanisms and normative requirements; it cannot calculate cross-market error rates, average settlement delay, or the realized performance of standardization without such matched evidence. Public standards also differ in implementation rights and licensing. The presence of a specification on a website does not establish universal market adoption.

A future empirical investigation should collect consented, anonymized treaty-year panels across proportional and nonproportional programs, with policy/exposure IDs, claims movements, report versions, treaty terms and settlement outcomes. Its primary estimands should be specified before sampling: for example, weighted premium completeness, false-positive ceded claims, error in ceded ultimate estimates, lag-adjusted reconciliation rates, and settlement differences by treaty structure. Sampling must account for insurer size, business mix, reporting frequency, technology, data loss, and claims development. Reinsurer and cedent reconciliation totals should be compared before adjudicating error attribution.

A separate controlled experiment could compare two implementations of the same source extract: one using flat periodic spreadsheet reporting and one using versioned event-based messages. Pre-registered outcomes might include reproducibility, monetary exceptions, time to resolution, and false rejections. Such a study must track the cost of implementation and evaluate systemic error propagation. The present monograph establishes the definitional and computational prerequisites for that research, not its results.

Conclusion

Treaty reinsurance experience data is a chain of accountable transformations from underlying insured business to contractual cession, claim recovery, technical account, settlement, and subsequent financial or supervisory representation. Bordereaux are a central instrument in that chain when the treaty and reporting arrangements call for individual risk or claim detail. They are complemented by aggregate account statements and valuation-specific reporting. The governing test of quality is whether every material ceded figure can be traced to a defined population, the applicable contract version, correctly classified transaction movements, explicit dates, and a reproducible computation.

The documentary sources show a consistent concern with contract clarity, timely reporting, attributable transaction data, supervisory oversight, and controlled accounting. The hypothetical computations establish that errors transmit differently through simple proportional and capped nonproportional contracts; that informational reserves and cash balances must remain distinguishable; and that revisions must preserve earlier reporting states. Effective controls join completeness, contractual eligibility, event-level aggregation, bidirectional counterparty reconciliation, and protection of sensitive information.

The scientific contribution of this dossier is a tested comparative account of those relationships and their consequences for actuarial and operational reliability. Its conclusions are conditional on the stated sources and mathematical assumptions. Empirical validation of frequency, efficiency gains, and error distributions remains a separate research undertaking with its own data-access, sampling, and identification requirements.

References