Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Paid Incurred and Outstanding Insurance Claims International Comparability

Abstract. International claims statistics join three different observational objects: the monetary transfer made to discharge an insurance obligation, the recognition of a loss during a reporting interval, and the valuation of obligations remaining outstanding at a particular date. Their numerical relationships depend on event cohorts, coverage, expense treatment, valuation objectives, recoveries, and timing. This monograph investigates the comparability of those objects through comparative institutional document analysis, accounting reconciliation, and reproducible constructed numerical experiments. It examines OECD market statistics, European reporting, United States statutory information, Canadian supervisory schedules, Australian prudential valuation and development statistics, United Kingdom aggregate data, and international financial reporting. Its principal finding is that the economically relevant common measure varies with the research question. Payment statistics measure realized disbursement; accident-cohort estimates describe expected loss costs; outstanding liabilities characterize future obligations at a valuation date. A defensible comparison preserves this distinction and reconciles changes in scope and measurement before calculating ratios or establishing rankings. The study shows how unrecognized changes in provisions, development maturity, reinsurance, discounting, risk margins, and product mix can reverse comparative conclusions. Numerical exhibits are constructed rather than observed insurer portfolios. Reported OECD observations are identified separately from these exhibits. The scientific contribution is an integrated interpretation of existing institutional and actuarial evidence, without a claim to a new theory or an independently estimated international ranking.

Keywords: Insurance claims; paid claims; incurred losses; outstanding claims; claims development; insurance reserves; IFRS 17; statutory reporting; reinsurance recoverables; international insurance statistics; stock-flow reconciliation; reporting comparability.

Introduction

An insurance claim is a contractual mechanism through which an insured event or other covered entitlement produces an obligation that may ultimately be discharged by payment, services, or another form of settlement. The monetary record of this obligation develops as information becomes available. An accident may be recognized only after a reporting delay; the first estimate may change when damage is assessed; a court decision or medical development may revise the estimate; and settlement may occur in installments. These processes operate on different clocks. An international statistic becomes meaningful when its underlying clock, institutional perimeter, and valuation convention are specified.

The scientific problem matters because the same numerical expression, such as claims incurred, can indicate an estimated cumulative outcome for a cohort, a calendar-period expense, a regulatory statistic compiled under specific accounting instructions, or a national accounts transaction. The term outstanding can designate the case estimate attached to known claims, all unpaid claims including estimated unreported losses, or a carrying amount containing time-value and uncertainty adjustments. Even paid claims require specification of payments to beneficiaries, expenses, salvage, subrogation, and ceded recoveries. International rankings that ignore these distinctions may turn differences of recognition into apparent differences of economic performance.

The object of this dossier belongs jointly to Global Insurance Markets and Institutions and Claims Science. The former locates reporting within market composition and national supervisory architecture; the latter explains the production of claim records through notification, adjustment, estimation and settlement. The unit of inquiry is the comparability of reported measurements across these institutions. The monograph uses existing actuarial and accounting concepts and develops explanatory examples to demonstrate implications rather than introduce a new named conceptual system.

The research questions concern the meaning of payment and incurred-loss observations, the effect of reserve recognition on comparative inference, and the conditions under which observations can be mapped across markets. The evidence consists of official documentary sources and identified scholarly literature. The reference date is November 19, 2026. Documentary evidence is strongest for the selected reporting schedules and accounting provisions. The analysis does not treat the availability of a downloadable statistical workbook as proof that its underlying records have been examined. Where web retrieval supplied an official indexed excerpt rather than the complete page, claims are restricted to that excerpt. The comparative numerical cases are reproducible counterfactuals. They isolate mechanisms; they neither measure the relative performance of named insurers nor establish causal effects in the observed world.

Concepts, Institutions and Historical Development

The three clocks of insurance claims

The event clock identifies the occurrence, exposure or triggering condition to which the claim belongs. Accident-year allocation generally follows the insured event; underwriting-year allocation attaches experience to the period of underwriting a contract. Claims-made arrangements require attention to the contractual trigger and reporting terms. Calendar-year allocation follows the period of transactions or recognized amounts. Development age measures elapsed time from an origin period to the valuation. All four designations can appear in one statistical table, and each answers a different question.

The settlement clock records transfers. A partial indemnity, an advance, a direct repair payment, and a final payment may occur in several reporting periods. A claim can be closed without payment and may later reopen. Payment counts are therefore distinct from claim counts. An amount paid in year t can relate to a much earlier accident year. Summing payment flows over years gives cumulative cash disbursement for a consistently defined cohort; it does not reproduce calendar-year recognition.

The valuation clock records the estimate of obligations at a given date. An unpaid amount can arise from an established claim file, from an event not yet reported, from subsequent upward revision of a known loss, or from payment-related expenses. The NAIC insurance glossary illustrates the ordinary institutional usage of reserve and incurred loss. It should be distinguished from the precise instructions of a particular financial return. A reserve is an estimate of liability rather than an identified pool of cash dedicated to a single claimant. Asset holdings, funding and liquidity address the capacity to settle that liability.

Let P(a,t) be cumulative paid amounts for origin cohort a at valuation t. Let C(a,t) be estimated unpaid amounts assigned to reported files. Let J(a,t) be the aggregate estimate of additional unpaid amounts outside those file estimates, including unreported claims and, where appropriate, expected development of known claims. Define these quantities consistently on an undiscounted indemnity-only basis. Then the estimated ultimate amount U(a,t) is P(a,t) + C(a,t) + J(a,t). The reported incurred amount, in a common actuarial convention, is P(a,t) + C(a,t). If a source instead uses incurred for the estimate including J, its label must be mapped explicitly. The expressions are definitions within this analytical convention; the symbols impose no universal reporting law.

At maturity, all payments and recoveries included in the selected basis have been recorded and no obligations remain. Ultimate equals cumulative paid in this limiting situation. Before maturity, the estimated ultimate can change as experience becomes known. A comparison of early development ages therefore combines the quality of loss measurement with the maturity of the information.

Institutional evolution of reported loss

The development of claims accounting reflects the mismatch between the inception of coverage and the eventual completion of settlement. Recognizing an expense when the relevant obligation arises permits measurement of a period's underwriting activity even while cash remains unpaid. Prudential institutions evaluate additional questions: whether future settlements are adequately valued, what uncertainty remains, and whether the undertaking can meet obligations under adverse conditions. National accounts aggregate transactions for the analysis of economic production and distribution rather than the solvency of an insurer.

The IFRS Foundation's IFRS 17 overview records the replacement of IFRS 4 and the general 2023 effective date for annual reporting under IFRS 17. The change represents a significant break in the presentation of insurance contracts. Historical series can include premium-based financial statements and subsequent service-based presentation. A researcher should establish the effective period for the actual entity or jurisdiction and distinguish original historical accounts from later comparative restatements. The existence of an international standard does not establish its mandatory use by every entity in every country.

The European System of Accounts (2013), chapter 16, describes insurance transactions and the addition of unpaid incurred claims to outstanding reserves. Its treatment of adjusted claims in measuring insurance output addresses a macroeconomic measurement problem. An exceptional catastrophe can dominate an annual insurer expense, while national accountants need a service output measure that can accommodate unusual volatility. The analytical consequences follow from the difference in research objective: a claim resource transfer, a profitability expense, and the measurement of economic output cannot be substituted mechanically.

Claims populations and product differences

The primary focus of the comparative analysis is non-life insurance. Property losses following a storm can give rise to a surge in payment volumes after the event. Liability losses may require prolonged investigation and adjudication. Health coverage can generate claims in large numbers with heterogeneous delays and contractual allocation between primary and secondary payers. On the life side, benefits and maturities may have partly predictable timing; surrenders can reflect policyholder decisions and financial conditions. A combined life and non-life payout statistic thus mixes economically different payment processes.

Geographic population is equally consequential. A domestic risk written by a foreign branch can enter one supervisory population according to authorization while another source classifies the transaction by location of risk. Consolidated groups may include foreign subsidiaries; solo regulatory returns can include branches and internal reinsurance. A market aggregate must define whether it represents risks resident in the jurisdiction, business written by insurers resident there, or a regulated institution population. Its monetary size is not automatically the size of the resident household insurance protection system.

Measurement perspective and institutional incentives

Operational claim records serve settlement authority and workflow. Financial accounts measure assets, liabilities and performance under the applicable basis. Supervisory reports serve monitoring and prudential assessment. A single administrative system may feed each report, yet transformations occur between transaction records and published statistical fields. Case estimates respond to available information, administrative guidelines and settlement expectations. Aggregate actuarial adjustments can recognize development patterns not apparent in individual files.

This creates an identification problem. Observed reserve growth can reflect a worsening risk environment, improved recognition of existing liabilities, change in discount assumptions, shifts in product portfolios, or changed reinsurance structure. None of these interpretations follows from reserve growth alone. The evidence required for interpretation includes the roll-forward, method changes and exposure composition. Institutional incentives matter, but a rise or fall in outstanding reserves by itself cannot establish deliberate reporting behavior.

Primary Market Sources and Comparative Research Design

Documentary evidence and units of comparability

The OECD Global Insurance Market Trends 2026 identifies 2025 as its reference year; its foreword describes coverage of 67 jurisdictions through national authority inputs. Individual statistical series may cover fewer jurisdictions. International coverage offers a broad starting point and creates a need to preserve country-specific definitions. National returns provide greater definitional detail but apply particular supervisory rules and may describe populations different from the OECD aggregates. The research therefore uses documents at the level at which their definitions can be interpreted, without treating their coverage as interchangeable.

The comparative observation is a field in a dated reporting instrument or an explicitly defined published statistic. A minimal record identifies reporting authority, jurisdiction, population, line of business, original label, accounting basis, period, origin-year convention, development age, gross or net status, currency, expense inclusion, valuation basis, and data revision status. Each field retains its original description and its mapped analytical variable; a transformation is documented before its result becomes an observation.

The case selection is purposive: the EU/EEA offers harmonized supervisory collections, the United States a statutory insurance reporting tradition, Canada a detailed IFRS-related return, Australia a prudential insurance liability standard and historical development publication, and the UK a separately published development series. The OECD is a cross-market reporting layer, while IFRS and national accounts provide functionally different measurement perspectives. These cases illuminate distinctions rather than provide a probability sample of world insurance markets. The documents were inspected through accessible official HTML, or identified indexed passages where direct access failed. Full dataset extraction and construction of an insurer panel are outside the completed evidence scope.

Comparative institutional inventory

Source and coverage Documented observation Required mapping before comparison
OECD, reporting jurisdictions Market premiums and claims-payment growth Reporting population, product boundaries, country notes, nominal/real measure
EIOPA, EU/EEA reporting populations Premiums, claims and expense statistics drawn from S.05.01 Accounting basis, direct/accepted/ceded business, expenses, solo versus group
NAIC, US statutory reporting Schedule P loss, expense and unpaid claim information Statement year, line, exact instructions, gross/net and expenses
OSFI, Canadian P&C return Pages 60.35 and 60.45, net claims and adjustment expenses, prior-year development Undiscounted basis, reinsurance, transition blanks and portfolio transactions
APRA, Australian regulated general insurance historical release Accident-year and development-stage information Historical release coverage, class and tail length, exclusion of Lloyd's operations
Bank of England, UK aggregate reporting IR.19.01 net paid claim development Accident or underwriting origin, development age, net basis
IFRS 17 and AASB 17 Liability for incurred claims, service expense and claims development Cash-flow valuation, risk adjustment, investment components and disclosure limits

This inventory asserts functions supported by the source documents, not a complete legal crosswalk of each underlying return. The NAIC's Schedule P description expressly identifies historical loss, loss-expense and unpaid-claim information. The access obtained here does not establish every cell convention of specific annual-statement editions. A US empirical extraction must first identify those instructions.

EIOPA's insurance statistics identifies the source template for premiums, claims and expenses and warns that provisional observations may change. It also records the exclusion of UK data from datasets with reference dates after 2020. A panel that uses an EEA total before and after that change needs a constant membership variant or an explicit break. A change in sample composition differs from changed claim costs within the same sample.

For Canada, OSFI return instructions give specific checks for prior-year claim development. This improves the possibility of a reconciliation study without establishing that every published Canadian market aggregate uses the same fields. For Australia, APRA's claims development statistics release dated July 19, 2023 provides a historical accident-year source for December 2022; the publication describes its exclusion of Lloyd's Australian operations. The historical release is evidence of its stated population, not a current complete market census.

The Bank of England aggregate report provides cumulative and incremental net claim development based on IR.19.01. The chart population and origin convention must be matched before comparing it with other development triangles. An equivalent gross claim triangle is a different series. A downloaded version of the charts and the applicable instructions would be needed to compare exact market development factors empirically.

From nominal data to shared analytical quantities

The harmonization process first separates flow amounts from valuation-date balances. It next aligns the insurance activity population and line definitions. Contract coverage definitions and claim-count rules are retained, including whether open, closed and reopened cases are separately identified. The researcher then selects the expense boundary, gross/net status, and recoveries convention, and determines whether the monetary quantity is nominal, price-adjusted, discounted or risk-adjusted.

A local currency growth rate uses a consistently defined quantity in consecutive periods. For nominal growth g and price growth pi, the exact real-growth transformation under a common price index is (1 + g)/(1 + pi) - 1. A ratio of claims to exposures may be more informative for risk measurement than a growth rate of the entire market, which also reflects changes in volume. Consumer price indices are useful broad adjustments, but repair costs, litigation awards, medical prices and wage inflation can differ from general consumer prices. A sensitivity test with alternative relevant indices is therefore appropriate if they are available.

Cross-sectional money amounts can be translated at transaction-period rates for flows and an appropriate end-period rate for stocks. This produces a currency conversion effect even when economic obligations are stable. A currency-neutral movement can be built in the original currencies or with fixed rates. Purchasing-power conversion addresses the domestic purchasing power of monetary benefits, a different purpose from valuing foreign currency liabilities. The empirical researcher must decide which question is being answered rather than regard these transformations as interchangeable.

Admissibility, evidence grades and missing observations

A comparison is directly aligned when target definition and coverage match. It is transformable when differences can be quantified with actual component data. It remains a partial comparison when material components are unknown but bounded. It is inadmissible for a given inference when the unknown difference can change the result and no credible bound exists. These are ordinary research judgments about admissibility, not new theoretical categories.

An undisclosed value is missing, not zero. A suppressed cell may be bounded by confidentiality rules if the threshold is known, but the researcher cannot impute zero from its absence. A transition blank can mean that comparable historical data do not exist. An undefined denominator, including zero earned premiums, gives an undefined ratio. The dataset should distinguish structural absence, missing reporting, suppression and true zero, then preserve the consequences in an uncertainty ledger.

Revisions should be captured by immutable vintage. Replacing past estimates with the newest value destroys information needed to identify development, which is specifically the difference between estimates at successive dates. Comparing as originally reported with restated values is legitimate only if the two vintages are identified. The analytical calculations in this monograph use explicitly constructed data to make all transformations visible.

Claim Payments versus Incurred Losses

The calendar-period reconciliation

For a stable population and undiscounted, expense-consistent claim basis, let P(t) be payments during period t and O(t) the unpaid claim estimate at its end. Define I(t) as the recognized calendar-period nominal claim cost, including re-estimation of prior periods, and A(t) as signed movements in the unpaid estimate outside that cost and settlement, such as acquisitions, currency translation, or transfers. Then

Ot=Ot−1+It−Pt+AtO_t=O_{t-1}+I_t-P_t+A_t

and

It=Pt+Ot−Ot−1−At.I_t=P_t+O_t-O_{t-1}-A_t.

These are accounting identities conditional on the chosen sign convention and exactly corresponding scope. They are not estimators of future claims. If only the three reported numbers P, opening O and closing O are available without knowledge of A, the raw sum is not identified as incurred loss. The unexplained difference belongs to the reconciliation, not automatically to underwriting.

In a 100-unit earned-premium portfolio, assume opening outstanding 80, payments 65, closing outstanding 95, and no outside movements. Calendar incurred equals 80, while the paid-to-earned-premium ratio is 65% and the incurred loss ratio 80%. The 15-unit gap reflects the increase in estimated obligations. If a portfolio transfer instead accounts for 12 units of the reserve increase, calendar incurred is 68, and the economically comparable loss ratio is 68%. The raw reconciliation would overstate that cost by 12 percentage points in this example.

A payment from an old claim uses existing resources and reduces the unpaid estimate. It does not imply a new loss of equal size in the payment period. Conversely, a newly incurred large claim can create a substantial expense even before any indemnity is disbursed. Whether the relevant statistic is a cash flow or a period cost must be decided according to its intended use.

Accident-cohort versus calendar-period incidence

Consider the following synthetic recognized ultimate and paid amounts. Both cohorts are measured on the same nominal indemnity basis. All numbers are monetary units; no discount, expenses, transfers or recoveries are included.

Accident cohort Ultimate estimate at end of 2024 Paid cumulative end of 2024 Ultimate estimate at end of 2025 Paid cumulative end of 2025
2024 and earlier 160 90 175 130
2025 0 0 120 50
Total 160 90 295 180

Opening outstanding is 70. Closing outstanding is 115: 45 for the old cohort and 70 for 2025. Payments in 2025 equal 90, with 40 arising from prior events and 50 from 2025. Calendar incurred is 135, comprising the 120 estimated new-event cost and 15 adverse development of previous cohorts. The stock-flow formula yields 90 + 115 - 70 = 135. Three different measurements are therefore simultaneously correct: 90 paid, 120 current-cohort ultimate and 135 calendar incurred. None can replace the other two without modifying the research question.

An alternative experience with faster payments can produce a higher cash ratio and unchanged current-cohort ultimate. If 30 additional units of the 2025 cohort are paid during 2025, rather than after the year end, paid becomes 120 and closing outstanding becomes 85. Incurred remains 135. This counterfactual holds estimates fixed and isolates payment timing. It neither assumes that accelerated settlement is costless in practice nor asserts that fast payment proves superior service.

Reporting and payment development

Claims development is represented by a matrix whose rows are origins and columns development ages. A cumulative paid triangle represents the history of settlement progress; reported incurred triangles incorporate file estimates; ultimate estimates additionally depend on the selected actuarial method. The actuary's predictive problem involves future payments and may draw on both types of information.

A basic loss development estimate can multiply the cumulative value at a given age by selected age-to-age factors. The factors reflect historical development among observed cohorts. A simplified two-age illustration uses mature comparable cohorts whose paid amounts at age one are 40 and 60 and at age two 80 and 90. The volume-weighted factor is (80 + 90)/(40 + 60) = 1.7. A new cohort with paid 50 at age one has a projected age-two paid amount of 85 under that factor. If age two represents ultimate settlement under the illustration, unpaid is 35. A factor of 1.5 instead yields ultimate 75 and unpaid 25; a factor of 2.0 yields 100 and 50. The calculation exposes sensitivity to historical development selection. A real application would also accommodate later ages, tail estimation, changes in portfolio mix, and the dependence of estimates on calendar conditions.

The actuarial literature supplies tools for evaluating reserve prediction. Mack's (1993) distribution-free standard error paper develops a formula for uncertainty in chain-ladder reserve estimates; this statement is based on the publisher abstract. No full derivation from that paper is claimed here. Friedland's (2010) unpaid-claims estimation text is a practitioner literature reference for the broader family of reserving methods. The worked examples here are independently constructed accounting and development calculations rather than replicas of figures from the literature.

The model assumption matters more than the apparent numerical consistency of a triangle. Historical age-one payment proportions may change when new digital settlement tools accelerate payments, when inflation makes remaining repairs more expensive, or when courts accumulate case backlogs. A factor derived from a different class of risk or legal system can be statistically precise but substantively inappropriate. When lag patterns are changing, alternative selected development patterns, accident-period exposure information, and subsequent realized values support a more useful comparison than an unqualified point estimate.

Claims counts, severity and settlement quality

A paid amount can be divided by claim payments, closed claims, reported claim files, or exposure. Each denominator defines a different frequency or severity quantity. Partial payments for a large liability claim can generate several transactions; dividing by transaction count produces an average disbursement rather than average ultimate severity. A comparison of payment amount per closed claim also depends on whether the numerator includes payments for claims still open. Matching numerator and denominator populations at the same development age is necessary.

Payment speed can be measured using the share of ultimate expected payments already disbursed, or using distributions of time to first or final payment. The first measure depends on estimated ultimate; the second depends on censoring, closure rules, and treatment of reopened claims. A mature cohort with complete settlement is preferable for inference about final duration. Immature cohorts can be analyzed with censoring-aware methods if suitable claim-level data exist. A higher aggregate paid share might reflect faster settlement, a short-tail product mix, or underestimated ultimate losses. A service-quality interpretation additionally needs claim disputes, partial payments and customer outcome evidence.

Reserve Recognition

Outstanding claims and liability valuation

For an existing set of claims, the undiscounted expected future cash flow is the sum of expected settlement and included expense cash flows across future periods. Denote positive expected outward amounts at future horizons k by X(k); for a selected discount term structure r(k), a simple annually compounded present value is the sum of X(k)/(1+r(k))^k. The difference between the nominal sum and the present value depends on settlement timing and the specified curve. A risk adjustment or prudential margin may further change the carrying amount. These are separate dimensions of valuation.

The financial accounting concept of a liability for incurred claims concerns past service. The IASB's IFRS 17 presentation distinguishes service results and finance effects. The accessible Australian implementation, AASB 17, paragraphs 40-42, 59(b) and 130, provides the relevant detailed text. It uses fulfillment cash flow measurement for the incurred-claim liability and distinguishes changes relating to claims and expenses from finance effects. Under the premium allocation approach, paragraph 59(b) includes relief from time-value and financial-risk adjustment where the relevant future cash flows are expected to be paid or received within a year of the claim being incurred. A portfolio with a short contract coverage period can still have long claim settlement duration; the latter condition must be assessed separately.

Assume a synthetic undiscounted expected payment of 100 at the end of year four, with no other cash flows. At 0%, 3% and 6%, the present values are 100.00, 88.85 and 79.21 respectively (rounded to cents). Adding a purely illustrative 8-unit margin gives carrying values of 108.00, 96.85 and 87.21. The component amounts are assumptions of the experiment rather than prescribed margins. Identical nominal obligations can generate different valuation-date stocks solely through the rate and measurement basis. Subtracting two carrying values and attributing the difference to claim incidence would be erroneous.

Reinsurance and other recoveries

Let gross paid indemnity for a period equal 100 and ceded cash recoveries equal 25. The net paid amount under a simplified cash definition is 75. Let gross outstanding expected indemnity rise from 60 to 90 and outstanding expected reinsurance recoveries rise from 20 to 35. Net outstanding rises from 40 to 55. With no other movements, gross incurred equals 130; net incurred equals 90, since gross incurred less ceded incurred of 40 gives the same result as 75 + 15. The calculation assumes matched claim populations and identical treatment of expenses, discount, credit risk and timing.

This distinction is vital for international market aggregates. A domestic primary insurer may cede business to a foreign reinsurer. Gross data describe the claim obligations before reinsurance recovery, while net data describe the retained financial result under a specified basis. Reinsurance assumed and direct business must be separated to avoid double-counting the same underlying insured event when primary insurer and reinsurer populations are combined. The primary insurer's obligations to the policyholder remain relevant even when financing is supplied by its reinsurer.

Salvage and subrogation recoveries can be deducted from incurred amounts or reported separately depending on the dataset. Coverage limits, deductibles and reinstatement premiums alter which economic costs the insurer actually bears. These transformations should retain the original reporting line and signs rather than infer gross from a net number using a broad retention percentage. A catastrophe treaty with an attachment point and limit generates a nonlinear relation between gross and retained loss; the same gross average across markets can therefore imply very different net outcomes.

Claim handling expenses and differences in purpose

The boundary between indemnity and claims management expense is a major measurement variable. Legal fees, loss adjuster fees, internal staff expenditures and other allocated or unallocated costs may be recognized within or outside the reported claims measure. Their inclusion can materially affect liability business in which expenses develop over long horizons. Direct comparison requires equivalent cost boundaries or an adjustment supported by reported components.

Historical EIOPA Q&A 881 shows why a single supervisor's templates may use different measurement perspectives. Its answer identifies the accounting basis of S.05.01 and the distinct Solvency II perspective of S.14.01, and explains the exclusion of specified claim management expenses from the cited gross incurred claim field. Later EIOPA Q&A 2911 addresses IFRS 17-related reporting of liability movements; the indexed answer indicates that, where no incurred-claim liability or movement is recognized, reported incurred claims equal payments in the circumstance described. The later answer addresses reporting under Commission Implementing Regulation (EU) 2023/894; it does not express an opinion on the correct application of IFRS 17. The historical answer should not be read as an assertion that the 2015 annex remains the current instruction for every reporter. A precise application requires the template version applicable at the reporting date.

The Australian prudential GPS 340 standard illustrates a different objective: central estimates and risk margins are identified separately for outstanding claims and premium liabilities under its requirements, with class-of-business and materiality qualifications. Its Australian-dollar discount-rate provisions relate to government securities yields. This prudential valuation is a distinct purpose from a cash-payment statistic, and its margins cannot be assumed numerically equivalent to the risk adjustment in financial reporting.

Prior-year development and the audit of estimates

The Canadian instructions on page 60.35 define the excess or deficiency of prior-year liabilities on an undiscounted basis after deducting prior-year claim and adjustment expense payments during the current year and the remaining unpaid amount for earlier claims. A positive excess under that sign convention indicates that the opening estimate exceeded the updated amount of paid plus remaining obligations; a negative result indicates a deficiency. The source is the Canadian prior-year development schedule. Page 60.45 separately identifies discount and nonfinancial risk adjustment and allows transition-related blank cells within its terms. It also includes portfolio acquisitions and dispositions rather than providing them as a separately reported amount in that schedule.

For example, opening prior-cohort unpaid liabilities equal 120. During the year 45 is paid and 90 remains outstanding on the same basis. Excess equals 120 - 45 - 90 = -15, a deficiency of 15. If the remaining amount is 60 instead, the result is +15. Both cases have the same payments; the direction of reserve development differs. Outside population movements would invalidate a simple causal interpretation of the difference unless controlled. The Canadian note on acquisitions/dispositions demonstrates why the researcher must record such transactions as part of scope rather than interpret every movement as uncertainty about prior losses.

The ASOP 43 standard (2010) explicitly addresses the intended use, assumptions, sensitivity, recoverables, uncertainty and distinctions among accounting, valuation and review dates in unpaid-claim estimates. It is a US actuarial practice standard with defined scope, not a globally binding rule. Its methodological relevance extends to comparative research: knowledge available after a balance sheet date may inform a review without constituting a cash transaction at that earlier date. An international dataset of estimates requires a consistent information cutoff when evaluating forecast performance.

Cross-Market Reporting

The empirical documentary findings

The OECD's 2026 report describes a broad international collection, but the statistics remain based on reported jurisdictional inputs. Its non-life underwriting chapter reports average growth in non-life claims payments of 7.9% nominal and 3.1% real in 2025 and gives 3.5% as the corresponding real growth figure for 2024. These are reported summary observations from the original report, not an independent calculation from country-level data. They describe changes in payments, which can be influenced by exposure growth, settlement timing, inflation and prior events. Their interpretation as changes in expected ultimate claims would require additional data.

A further qualification is established by the report's Figure 2.3 notes and Annex B. The Romanian observation is based on gross incurred claims rather than gross paid claims. The Croatian series includes claims payments and changes in technical provisions. The statistical label consequently does not establish a uniform payment concept across the reporting population. Removing the Romanian and Croatian observations would change the estimand and potentially the composition of the simple average; converting them to a payment basis would instead require the relevant reserve movements and reconciliation adjustments. Neither alternative is computed here because the required underlying observations have not been independently collected. The documented exceptions establish the measurement problem directly without establishing its quantitative impact on the reported average.

The report's methodological notes identify an explicit comparability qualification for Korea, concerning excluded refunds for long-term products in the gross claims-paid measure while long-term premiums are included. This matters because even a ratio with a seemingly consistent gross label can combine different populations of insurance transactions. An analyst should treat the reported ratio under its actual definition or use a documented alternative with aligned inclusion. A claim that the resulting difference measures efficiency is unsupported without that reconstruction.

EIOPA's notice about the UK coverage break and provisional vintages provides a concrete example of reporting heterogeneity despite standardized templates. The UK and Australian development publications show that public reporting can support analysis by origin and development age. The Canadian schedules disclose a more direct quantitative route to separating development from valuation components. The inspected documents establish differences in institutional measurement and available fields. They do not provide a completed cross-country dataset of comparable paid, incurred and outstanding amounts.

Statistical rankings and settlement timing

Construct two markets A and B with earned premiums of 200 each and stable portfolios on a common nominal, gross indemnity basis. Opening outstanding is 60 for both. A pays 100 and closes with outstanding 95. B pays 125 and closes with outstanding 60. Calendar incurred for A is 135 and for B 125.

Measure A B
Paid claims 100 125
Opening outstanding 60 60
Closing outstanding 95 60
Incurred claims 135 125
Paid / earned premiums 50.0% 62.5%
Incurred / earned premiums 67.5% 62.5%

The ranking reverses: A has lower payments relative to earned premium but a higher incurred cost. Its closing outstanding is also greater. All three results follow from the same data. Faster settlement in B could explain part of its higher payments, but the example does not identify the cause of the different reserve movements. The reversal proves the logical possibility of misranking from mismatched definitions, not that any specific observed markets have been misranked.

Suppose an unknown positive acquisition movement in A's reserve can range from zero to 20. A's adjusted incurred cost then ranges from 115 to 135, and its incurred ratio ranges from 57.5% to 67.5%. B remains at 62.5%. The ranking is indeterminate across the admissible bounds because the acquisition amount required for equality is 10. If documentary evidence limited the movement to less than 10, A's incurred ratio would remain greater than B's within these assumptions. This provides a simple sensitivity calculation that is more informative than silently treating missing transfer data as zero.

Product composition and standardized comparisons

A market's aggregate claims ratio is a weighted mean of line-level ratios when the weights are positive shares of the relevant denominator and the line amounts add to the market total. Consider two lines: short-tail and long-tail. Market A has ratios 40% and 80%; B has 45% and 85%. Suppose A's long-tail earned premium weight is 75%, B's is 25%. A's actual mixed ratio is 70% and B's is 55%. Under equal common weights the ratios become 60% and 65%. A has lower ratios within both lines but a higher overall ratio because of portfolio composition.

The substantive research question determines whether the original aggregate or the standardized comparison is useful. The actual aggregate describes the insurer population's realized composition. A standardized ratio measures a counterfactual composition. Common weights may be chosen from a pooled portfolio or an external target population, but the weight source and universe must be stated. Where line definitions differ, weights cannot repair the absence of a common line definition. Where a line has no comparable data, an explicitly limited subset is preferable to invented component ratios.

The same logic applies to development ages. A weighted payment-completion rate across motor, property and liability portfolios reflects the relative size of those portfolios. Applying a single market-wide development factor to a substantially different product mix can produce misleading estimates even if each line's own development pattern is stable.

Denominator discipline and IFRS transition

Claims paid divided by written premiums is a payment intensity measure relative to written business, subject to the payment and premium population being aligned. Incurred indemnity divided by earned premiums is a conventional underwriting loss ratio when numerator and denominator use compatible recognition and expense boundaries. IFRS insurance revenue measures service activity and excludes receipts of investment components; it is not generally identical to written premium. The IFRS Foundation's insurance contract overview explains the separate service and financial presentation and investment component treatment. A researcher must specify the denominator instead of replacing premium figures with insurance revenue without a bridge.

Let written premium equal 150, earned premium 120, insurance revenue 110 and incurred indemnity 75 in a synthetic portfolio. Ratios are 50%, 62.5% and approximately 68.18% respectively. The differences do not establish three different underlying levels of loss. They are caused by the selected measures of volume and recognition. The last ratio may be analytically relevant for a service-result comparison if both numerator and denominator match insurance service scope; it should be labeled accordingly.

AASB 17 paragraph 130 sets out disclosure of actual claims against earlier estimates on an undiscounted basis, with up to a ten-year horizon and an exception for claims where uncertainty is ordinarily resolved within one year. This improves access to development evidence while permitting variation in observable history and coverage. A sample based solely on published long-horizon development tables can overrepresent slower-settling risks or larger firms, depending on disclosure and materiality. The absence of a table does not establish the absence of uncertainty in all claims.

Currency, inflation and stocks

Suppose a market has outstanding obligations of 100 in local currency at both calendar dates, and no nominal change in expected claim cash flows. If the translation rate changes from 0.80 reporting units per local unit to 0.90, the translated amount rises from 80 to 90. The change of ten reflects currency translation alone in the illustration. A reconciliation performed directly in the reporting currency without recognizing the translation movement would misclassify the increase as a claim-cost movement. A corresponding payment flow translation must use a rate consistent with the defined reporting convention.

For payments of 100 followed by 110 in local currency, nominal growth is 10%. With general inflation of 5%, real growth is approximately 4.76%; with an alternative claim-cost index growing at 12%, the adjusted change is approximately -1.79%. The choice of index reflects the inquiry. A general price adjustment approximates broad purchasing power. A line-specific cost index addresses the cost of insured repairs or services. Exposure-adjusted frequency and severity require additional denominators. Even after price adjustment, growth can be caused by changes in the number or quality of covered claims.

Stocks should be interpreted with maturity and future financing obligations. A high ratio of outstanding claims to annual payments can result from a longer settlement tail, stronger new business growth, a change of estimates, or a fall in annual payment activity. It is not a direct estimate of the time required to settle claims unless steady state assumptions and consistent populations are established. Under stationary annual claim generation with stable settlement timing, a stock-flow relation can give information about average delay, but actual markets often depart from stationarity and valuation-date risk adjustments complicate it.

Integrated Scholarly Synthesis and Research Frontiers

Appropriate inferences from each measure

International comparisons become reliable by connecting the measurement to the economic question. Cash payments bear directly on liquidity demands, the timing of compensation and realized transfers to claimants. They are also relevant to catastrophe recovery, although the household or enterprise recovery outcome depends on timing, sufficiency of coverage, repair capacity and other resources. A study of payment timeliness requires claim-level or consistently defined cohort duration observations.

Incurred amounts characterize recognition of insurance costs in a period or event cohort, depending on the source. A period loss ratio includes development of earlier estimates and can reveal pressure on underwriting performance, provided the denominator and accounting treatment are aligned. An accident-year ultimate loss ratio supports risk pricing and exposure comparisons but remains an estimate until settlement completes. Distinguishing these two perspectives improves explanations of when losses arose and when institutions recognized their economic magnitude.

Outstanding claim amounts are forward obligations estimated on a specified basis. Their significance for solvency requires consideration of capital, recoveries, liability cash-flow structure, and uncertainty. A short-tail insurer may have smaller outstanding obligations relative to annual cost than a long-tail specialist; a comparison of reserve amounts alone cannot establish financial soundness. Reserve adequacy assessment needs subsequent development, model diagnostics, and information on the balance-sheet valuation purpose.

The integrated interpretation follows from the stock-flow reconciliation and institutional crosswalk: payment differences, recognition differences and valuation differences are empirically separable when reporting supplies the required components. Where those components are missing, observed rankings are conditional. It is methodologically preferable to retain multiple well-defined measures than force them into a total-loss measure that cannot be validated.

Research protocol for a future matched-market dataset

A complete empirical study would start from a dated extraction of the statistical release and the applicable reporting instructions. It would construct a record of reporting institutions and reconcile subsidiary, branch, direct and assumed reinsurance populations. Each document and dataset would have an immutable version identifier or checksum. The dataset would retain original labels, raw values, units and transformations. Mapping to common classes would be reversible, with unresolved non-equivalences explicitly recorded.

For each country and line, collect cumulative and incremental paid amounts by origin/development, current-year ultimate and prior estimates, opening and closing outstanding liabilities, recoverables, adjustment expenses, and measurable non-claim movements. Obtain a common denominator, preferably exposure or earned premiums according to research aim, with separate records of product and coverage boundaries. Where possible, retain both gross and net measures and both undiscounted and carrying amounts rather than reconstructing the former from a single stock.

Perform origin-year and calendar-year reconciliation; verify that aggregate incremental payments reproduce cohort diagonals under consistent inclusion. Compare paid and reported patterns at matched ages. Investigate unusually large residuals in the stock-flow identity as potential data issues or genuine portfolio changes. Classify them as explained, bounded or unresolved rather than assigning them automatically to new claim recognition.

Use sample-specific sensitivity analyses for inflation, fixed exchange rates, reserve adjustment, product weights and development patterns. Assess estimate performance using original valuation vintages and future outcomes, with holdout cohorts where sufficient comparable data exist. Account for overlapping origin years, correlated catastrophe events and repeated development estimates when selecting uncertainty methods. Report the identification limitations of observational comparisons and separate correlation from causal claims.

The present monograph provides the documentary basis and reproducible arithmetic for that design. It has not collected the full insurer-level or market-level panel needed to implement the empirical protocol. Therefore its examples demonstrate the mechanisms that can distort comparisons and its documentary observations establish that relevant reporting differences occur; its conclusions do not quantify the frequency or magnitude of those distortions across the global insurance industry.

Implications for international statistical publication

Market statistical agencies can increase comparability by attaching field dictionaries and population metadata to each vintage. Where paid, incurred and outstanding amounts are all available, a reconciliation of their movement, including changes of scope and exchange rates, permits users to identify meaningful discrepancies. Reporting gross and net data separately makes the distribution of risk through insurance and reinsurance more interpretable. Separate adjustment-expense and valuation components help link operational claims processes with financial reporting and supervision.

Cross-market summaries should retain explicit definitions for numerators and denominators, distinguish preliminary from final data, and identify whether an average is weighted by market size. Changes in legal reporting perimeter and accounting standards can be marked as statistical breaks. National differences in coverage should remain visible, including explicitly documented exceptions, rather than being absorbed into a uniform label. Standardization is most useful when it preserves the original mapping.

For researchers, a rank should be published only for a clearly defined common measure. If a reasonable sensitivity range changes the ordering, the conclusion must acknowledge that dependence. For insurers and supervisors, the same reasoning favors auditable reconciliations over a single preferred indicator: no cash statistic establishes the amount of future liability, and no prudential carrying value establishes when claimants actually received compensation.

Limits of the research and directions for extension

The comparative source set is intentionally selected rather than exhaustive. It does not contain representative insurer microdata, independently downloaded OECD or EIOPA country tables, or the complete contemporaneous filing instructions of every jurisdiction. The OECD report was retrieved as publisher PDF text, including the methodological notes; direct HTML and PDF-page image retrieval were unavailable. The newer EIOPA Q&A was inspected through the official indexed question and answer, while direct page access remained rate-limited. Claims are restricted to the examined text. The Mack article was read at abstract and metadata level; the Friedland reference at publisher-description level. No quantitative theorem from inaccessible article text is attributed to either.

Future independent research can investigate reporting in Asia, Latin America and Africa with jurisdiction-specific supervision and financial statements. Longitudinal analysis of IFRS 17 adoption can examine restatements and movements in presentation, conditional on a credible sample and accounting bridges. Claim-specific payment durations and development estimates could be linked with measured recovery outcomes and exposure, subject to confidentiality and censoring. These extensions require additional primary datasets and independently assessed uncertainty. The present findings remain useful as comparative document analysis and controlled reconciliation rather than a completed global empirical survey.

Conclusion

Paid claims, incurred claims and outstanding claims are related through the insurance obligation and its settlement history. They occupy distinct temporal and valuation positions. Calendar-period incurred amounts equal payments plus the movement in consistent outstanding estimates after relevant outside movements are removed; cohort ultimate estimates combine accumulated payments and remaining nominal expected obligations. Financial or prudential carrying amounts incorporate their applicable measurement adjustments and must be reconciled before being inserted into these nominal identities.

The institutional evidence establishes consequential differences in the way international insurance information is assembled. OECD payments describe market disbursement trends with documented qualifications. EIOPA's statistical and reporting guidance distinguishes source templates, accounting perspectives, provisional data and coverage changes. NAIC loss and expense reporting, Canadian reconciliation schedules, Australian liability valuation and development releases, and UK paid development charts offer different points of access to claims experience. IFRS 17's service-based accounting and claims development provisions further clarify why accounting results and market premium statistics require separate interpretation.

The numerical experiments prove that rankings may change when the selected measure, portfolio mix, reserve movements, discount rates or denominators change, even while the relevant underlying claims experience is held fixed in the respective experiment. The conclusions are algebraic and conditional, not empirical rankings. International comparability is achieved through documented scope, consistent cohort and time conventions, explicit gross/net and expense treatment, appropriate valuation bridges, and transparent uncertainty about missing components. This integrated interpretation links market statistics with claims science and supports research in underwriting performance, settlement behavior and outstanding obligations while preserving the integrity of each measurement.

References

  • OECD (2026). Global Insurance Market Trends 2026. OECD Publishing, Paris. 59 pp. doi:10.1787/c8be6585-en. Full report

  • European Insurance and Occupational Pensions Authority (EIOPA) (n.d.). Insurance statistics. Frankfurt am Main. Accessed 2026-11-19. Source record

  • European Insurance and Occupational Pensions Authority (EIOPA) (n.d.). Question and Answer 881, reporting templates S.05.01 and S.14.01. Frankfurt am Main. Accessed 2026-11-19. Source record

  • European Insurance and Occupational Pensions Authority (EIOPA) (n.d.). Question and Answer 2911, claims incurred under IFRS 17. Frankfurt am Main. Accessed 2026-11-19. Source record

  • International Accounting Standards Board (IASB) (2017, amended 2020 and subsequently). IFRS 17 Insurance Contracts. IFRS Foundation, London. Official standard overview. Source record

  • Australian Accounting Standards Board (AASB) (2022). AASB 17 Insurance Contracts, December 2022 compilation, paragraphs 40-42, 59(b), 100-105 and 130. Melbourne. Source record

  • Office of the Superintendent of Financial Institutions (OSFI) (n.d.). Property and casualty insurance return, Section IV: Detailed instructions. Ottawa. Pages 60.35 and 60.45. Accessed 2026-11-19. Source record

  • National Association of Insurance Commissioners (NAIC) (n.d.). Schedule P. Kansas City, MO. Product and service description. Accessed 2026-11-19. Source record

  • National Association of Insurance Commissioners (NAIC) (n.d.). Glossary of Insurance Terms. Kansas City, MO. Accessed 2026-11-19. Source record

  • Australian Prudential Regulation Authority (APRA) (2023). General insurance claims development statistics: December 2022 release. Sydney. Release page, 19 July 2023. Source record

  • Australian Prudential Regulation Authority (APRA) (2023). Prudential Standard GPS 340 Insurance Liability Valuation. Sydney. Source record

  • Bank of England (n.d.). Insurance aggregate data annual report. London. Charts 2.3 and 2.4, template IR.19.01. Accessed 2026-11-19. Source record

  • European Union (2013). European System of Accounts 2010, Chapter 16: Insurance. Annex A to Regulation (EU) No 549/2013. Eurostat. Source record

  • Actuarial Standards Board (ASB) (2010). Actuarial Standard of Practice No. 43: Property/Casualty Unpaid Claim Estimates. Washington, DC. Source record

  • Mack, T. (1993). Distribution-free calculation of the standard error of chain ladder reserve estimates. ASTIN Bulletin, 23(2), 213-225. doi:10.2143/AST.23.2.2005092. Source record

  • Friedland, J. F. (2010). Estimating Unpaid Claims Using Basic Techniques. Casualty Actuarial Society, Arlington, VA. Publisher description. Source record