Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Commercial Insurance Pricing Across Multiple Legal Entities

Abstract. Pricing a commercial insurance program spanning several legal entities requires the insurer to estimate the prospective cost of an enforceable portfolio of coverages, while the corporate group must determine how to attribute the resulting economic costs, insurance benefits, retentions, and risk-control incentives among distinct legal persons. The two calculations intersect, but they have different objectives and constraints. A parent corporation, its subsidiaries, separately incorporated operations, and affiliated captive insurers may share economic control without sharing identical insured status, insurable interests, locations, tax obligations, or claims rights. Their losses may also depend on common catastrophes, supply-chain failures, cyber events, and litigation. This monograph develops an integrated documentary and actuarial analysis of these relationships from original supervisory instruments, actuarial standards, accounting and tax guidance, and foundational insurance-economics scholarship. It distinguishes exposure identification, the insurance response function, marginal and allocated risk costs, expense loading, capital and reinsurance, the purchased premium, and internal chargeback. Formal derivations show why the expected loss of a portfolio can be additive before contract limits yet nonadditive after shared retentions and limits; why dependence changes aggregate volatility without necessarily changing marginal expected gross loss; and why a corporate allocation of the actual premium does not determine the insurer's technical price. Three reproducible hypothetical calculations examine premium composition, common-event insurance recovery, and correlated binary losses. The comparative analysis treats European cross-border authorization, worldwide supervisory principles, IFRS insurance accounting, and OECD captive-insurance transfer-pricing guidance as instruments with different scopes. Findings concern defensible inference and transparent design, rather than empirical estimates for an unidentified corporation. The study concludes with concrete documentation requirements, sensitivity tests, and an agenda for comparative evidence.

Keywords: Commercial insurance; multinational insurance programs; multiple legal entities; insurance ratemaking; exposure measurement; aggregate limits; dependence; premium allocation; cross-subsidy; marginal risk; insurance captives; transfer pricing; regulatory jurisdiction; actuarial credibility; economic incentives.

Introduction

A large corporation may purchase property, general liability, cyber, management liability, and specialty insurance through a centralized department while arranging coverage for subsidiaries operating in different territories. The apparent single purchaser can therefore contain many separately incorporated policyholders, insureds, insured interests, and loss beneficiaries. An insurer can quote one negotiated program price, issue local policies, or attach an international excess or difference-in-conditions contract. The economic group may then distribute premium expenses to affiliates through a separate internal arrangement. Each decision uses a different unit of account.

The central research question is how a defensible price for transferred commercial insurance risk can be related to identifiable entities without confusing consolidated ownership, covered legal entitlement, modeled risk, commercial negotiation, or accounting allocation. A technically credible answer must describe which liabilities the insurer actually accepts, the exposure to the relevant insured perils, how multiple entities can lose from the same event, and which cost and capital terms influence the quotation. It must also distinguish an observed premium from an estimate of expected claims and from a later managerial charge.

The CAS ratemaking principles, historically established in 1988 and retained for reference in U.S.-regulated ratemaking after reinstatement in 2021, organize rates around prospective costs associated with risk transfer. The Actuarial Standards Board's ASOP No. 53 separately addresses future-cost indications for property/casualty insurance, reinsurance, captives, and other risk-financing mechanisms; a negotiated transaction price can differ from the actuarial indication. The distinction is decisive in multi-entity portfolios: the mathematical cost of a layer, the market price for that layer, and each affiliate's internal burden form three analytically different quantities.

This is an international comparative research dossier with a source cutoff of November 14, 2026. It uses published legislation, standards, official institutional material, and identifiable scholarship; develops explicit accounting identities and risk-model examples; and evaluates rival explanations for entity-level price differences. Hypothetical numbers are invented solely for demonstration and make no claim about an actual insurer, insured enterprise, or jurisdiction. National tax, contract, licensing, and regulatory applications require their own applicable-date and governing-law analysis.

Primary Foundations and Scope

The insured unit, the economic group, and the policy

An economic group is ordinarily identified through control and consolidation criteria. Insurance coverage, by contrast, operates through contractually and legally specified parties, property, activities, and liability interests. The legal entity named on a declarations page, another corporation included through the insured definition, a newly acquired affiliate automatically covered subject to a condition, and a third party added by endorsement can occupy different legal positions. A broker's group exposure schedule may be economically comprehensive while remaining contractually incomplete.

The first object in pricing is therefore the insured exposure set, with attributes that identify a legal person and the peril, interest, territory, coverage period, and version of the wording. In an idealized representation, entity i has exposure data E_i, hazard and operational features H_i, applicable coverage C_i, and possible random economic loss X_i. An insurance payment is a contract-dependent transformation of the underlying losses, not their synonym. Writing a payout as Y = g(X_1,...,X_n; C) makes the contract's coverage conditions part of the pricing object.

The policyholder who pays the premium can differ from the entity that suffers an event or receives an indemnity. Moreover, some master-program extensions respond only under specified circumstances, and a purported global program may require local issuance, limits, and documentation. One corporate invoice by itself establishes neither a universally identical insured interest nor a universal right to reimbursement. Specific master and local wording, endorsement schedules, controlling law, and claims-payment instructions determine what is contracted.

Risk price, transaction price, and internal allocation

Let L denote the prospective present-value cost of insured losses and loss adjustment expenses for the covered policy period. Let A include acquisition and operating expenses, R the net cost of reinsurance as relevant to the selected gross/net perspective, and K the compensation or allowance for deployed capital and contingencies. A simple technical-cost indication is T = E[L] + A + R + K, with all terms measured consistently and no duplication of recoveries or funding. Where a percentage v of premium covers separately identified proportional costs, an elementary solution is P = T/(1-v), if the component structure warrants this relationship. Actual market premium P may then also reflect competition, coverage negotiation, capacity, strategic underwriting, regulation, and a broker's placement design.

The ASOP No. 29 expense standard distinguishes costs that vary with policies, premium, claims, or exposure, while ASOP No. 30 addresses profit, contingency, and the cost of capital. These standards support the separate identification of costs; they do not mechanically prescribe one universal loading or a single market price.

The insurer's quoted P can be paid centrally, apportioned to several insureds on invoices, or recovered by the parent through internal chargebacks. Define a_i as the assigned amount for entity i and a_0 as any undistributed central portion. A reconciliation requires the accounting identity sum_i a_i + a_0 = P, after defining whether local taxes, fees, and captive charges are included. This identity is necessary for an allocation but says nothing about whether each a_i equals its risk contribution.

A price is prospective. Paid claims are retrospective. For an entity with observed claims Z_i, the historical ratio Z_i/a_i can be informative about experience, but a single-year high ratio does not prove historical underpricing. Changes in coverage, claims development, hazard conditions, and sampling error can make ex-post results differ from ex-ante expectation. The ASOP No. 23 data guidance treats fitness, completeness, consistency, and disclosure of limitations as substantive actuarial issues.

International institutional boundaries

The IAIS Insurance Core Principles establish a framework for insurance supervision; their content addresses insurer governance, risk management, valuation, capital, and supervisory powers. They do not constitute a single worldwide contract law, insurance tariff, or licensing permission. A multinational corporate group must therefore establish coverage locally even when an insurer assesses aggregate risk globally. Group-wide insurer supervision and legal-entity-level policyholder coverage remain different questions.

For the European Economic Area, the Solvency II Directive provides an insurance authorization and passporting framework for appropriately authorized undertakings and classes of business, subject to the Directive's conditions. It does not itself remove all issues about the location of risks, applicable national provisions, compulsory cover, indirect taxes, or contracts outside that legal framework. The amended framework under EIOPA's Solvency II overview identifies Directive (EU) 2025/2 reforms taking effect on January 30, 2027; those future rules cannot be silently treated as applicable in a November 2026 historical comparison.

The EU Insurance Distribution Directive regulates defined distribution activity and contains provisions concerning large risks and product oversight. Its obligations vary with the applicable article, product, and circumstances. A corporate insurance premium analysis must identify the exact legal provision before asserting the disclosure or conduct treatment of a particular large-risk contract.

Accounting and tax boundaries

IFRS 17 Insurance Contracts governs the recognition, measurement, presentation, and disclosure of contracts within its scope by reporting insurers. Its groups of insurance contracts and fulfillment cash flows are accounting concepts, not a prescriptive rule that every affiliate in a multinational insured group should bear an equal share of purchase premium. The accounting policies of the insured parent and subsidiaries, and any consolidation eliminations, require a distinct assessment.

Intra-group captive insurance adds a further question: does the affiliate insurer actually assume insurance risk, possess capacity to pay claims, and achieve diversification? The OECD Transfer Pricing Guidelines (2022), chapter X, section E, explicitly analyze the accurate delineation of captive-insurance arrangements, risk assumption, and diversification. The tax treatment of such arrangements concerns related-party transactions; it cannot be inferred solely from a mathematical allocation of an external insurance invoice. Nor does a corporate internal chargeback automatically constitute a captive insurance premium.

Historical and theoretical lineage

Modern prospective ratemaking connects insurance pricing with the valuation of contingent loss, risk pooling, classification, contract design, and uncertainty. The economic difficulties of asymmetric information identified by Rothschild and Stiglitz (1976) concern insurance-market selection when risk characteristics are imperfectly observed. Holmström's 1979 moral-hazard analysis explains how action observability affects the incentive properties of contingent contracts. Both mechanisms can arise within a corporate insured group when the parent controls procurement but local entities influence loss prevention. They identify mechanisms that require observation or plausible causal evidence in an individual program; they are not proof that a particular subsidiary is selecting against the insurer or underinvesting in safety.

Entity Exposure and Taxonomies

A reproducible entity-peril-policy matrix

A useful research dataset employs a relational rather than a consolidated one-row-per-corporation structure. Its entity table records legal registration identifiers, domicile, relevant controlling organization, business activity, operating currencies, acquisition and divestiture dates, and a documented entity hierarchy with temporal validity. Its exposure table records risk location, peril, value basis, units, building or process characteristics, payroll, turnover, vehicle count, workforce, or other relevant denominator. Its contract table records insurer, policy identifier, named insured and insured-definition text, period, jurisdiction, form edition, endorsements, conditions, exclusions, each-occurrence and aggregate limits, deductibles, retentions, and excess attachments. Its claims table distinguishes event, occurrence, claimant, allegedly liable entity, insured claimant, allocated adjustment expense, coverage position, incurred estimate, payment, subrogation, recovery, and development date.

The keys must survive mergers, reorganizations, changes in business names, and the transfer of an asset between legal persons. A loss incurred in one location, a liability booked by another affiliate, a centralized cash disbursement, and a reinsurance reimbursement may represent one economic event. Treating them as four independent losses produces fictitious frequency and inflated severity. Temporal keys must also distinguish the date of occurrence, claims-made notification, coverage attachment, the date a policy endorsement becomes effective, and the valuation date.

For property exposures, a sum-insured figure can have a different relationship to loss than replacement value, maximum foreseeable loss, or modeled probable maximum loss. A manufacturing location's payroll is a weak direct denominator for building fire loss, while revenue may be relevant to a specified business-interruption measure only when the coverage wording and valuation assumptions align. For liability, operations, jurisdiction, policy trigger, and potential defense costs may dominate simple asset value. A single group-wide revenue metric therefore creates classification error when used indiscriminately across lines.

Classification as a hypothesis about cost

ASOP No. 12 on risk classification supplies a basis for separating meaningful risk characteristics from arbitrary administrative categories. Entity identity may proxy for industry, hazards, territory, limit, governance, or historical exposure; a legal-entity indicator alone does not explain which mechanism caused a higher expected loss.

For a frequency-severity representation, conditional expected gross loss may be written E[X_i | z_i] = E[N_i | z_i] × E[S_i | z_i] only when N_i is a claim count and the selected mean severity is defined consistently for the compound process, with the needed conditional assumptions. A practical specification may be E[N_i|z_i] = e_i exp(β₀ + β'z_i + u_i), where e_i is exposure and u_i an entity effect. Severity, coverage attachment, and aggregate limit must then be modeled separately. The Ohlsson and Johansson (2010) GLM treatment supplies a well-established statistical foundation for non-life tariff analysis. Its methods do not establish the adequacy of a particular multinational entity dataset.

In mixed enterprise portfolios, the insurer should investigate entity effects after controlling for industry, location, occupancy, insured values, risk management, and policy design. Entity coefficients are often unstable for small subsidiaries with scarce claims. Hierarchical or credibility approaches borrow strength across comparable observations; the interpretation depends on whether comparable exposures and coverage bases exist. The guidance in ASOP No. 25 on credibility is directly relevant when subject experience is blended with wider experience.

The denominator problem

An entity-level loss cost per $1,000 of insured asset value can be calculated only when the asset-value denominator is tied to the insured interest and time. If subsidiary A operates a plant for half the policy year following an acquisition, an annual exposure count must reflect entry date or another justified exposure convention. If the parent holds title while a subsidiary operates the site, both entities may appear in exposure schedules; recording the full property value twice exaggerates the denominator. If they have separate legally insured interests, the dataset must represent those interests without assuming that two interests equal two physical buildings.

Losses and exposures must use a common price basis: original currency, translation date, claims inflation, coverage edition, and value index. For international portfolios, analysts should present nominal local currency, translation assumptions, and the reporting currency separately. Exchange-rate variability may affect a local indemnity, group budget, or reinsurer reimbursement in different ways depending on contractual settlement currency.

The measurement problem becomes particularly severe when a central corporate broker uses historical premium shares to estimate entity exposure. Such a circular allocation may be useful for budgeting but cannot independently validate risk-weighted pricing: it recycles the very prices under examination. A scientifically stronger design builds exposure from underlying operational and contract records, then compares observed premium shares with predicted risk contributions.

Missing entities and the coverage boundary

An unlisted subsidiary, newly acquired business, dissolved entity, outsourced activity, or joint venture should receive an explicit coverage status. The statuses are documentary claims: affirmatively included, conditionally included, coverage disputed, expressly excluded, or not determined. These are not interchangeable with actuarial zero exposure. An excluded business may still affect group liquidity and correlated event losses, but its uninsured loss belongs outside the covered-claims expectation until the relevant benefit is established.

The entity-peril-policy matrix makes two important cross-checks possible. First, the sum of selected operating units can reconcile to the group's reported exposure inventories, after eliminations and ownership adjustments. Second, the contract response can be reconciled to the covered persons and insured property interests on the effective date. Material breaks between the two sets identify either administrative deficiencies or actual differences in risk transfer. An insurer who receives the exposure file may price one insured population while the insured group assumes another, creating basis risk.

Empirical measurement and data provenance

An empirical study should pre-register the observation unit and avoid pooling frequency measures across incomparable triggers. Recommended metrics include exposure-adjusted claim frequency, mean severity after inflation adjustment, large-loss exceedance frequency, event clustering, loss adjustment expenses, percentage of locations with verified insured status, fraction of reported values reconciled to assets, missing legal-entity identifiers, currency translation residual, and the number of policy-version mismatches. Each metric needs an auditable denominator and time window.

Data provenance is particularly important for cyber and contingent business interruption, where a single event can create claims at many legal entities and where causation and coverage may remain unsettled. Claims observations should distinguish notifications, coverage acceptance, case reserves, settlements, and collected indemnity. Otherwise an insurer's loss-development pattern or a disputed claim can falsely appear as a shift in underlying hazard.

The ASOP No. 56 modeling standard places particular significance on model purpose, inputs, validation, and limitations. A model is credible for the defined estimand only when its exposure population, input treatment, version, and response variable correspond to the decision being evaluated. Model sophistication cannot repair a missing insured-status definition.

Joint Losses and Pricing Relativities

Gross-loss additivity and dependence

For integrable entity gross losses X_1,...,X_n, the group's gross loss is X = sum_i X_i, and linearity gives E[X] = sum_i E[X_i]. This identity holds regardless of dependence. Dependence changes the distribution of X, its variance, tail probability, and many capital or reinsurance costs. In particular, Var(X) = sum_i Var(X_i) + 2 sum_{i<j} Cov(X_i,X_j). Thus two entities with unchanged standalone expected gross losses can raise the cost of aggregate capital or a tail-attaching reinsurance layer when a common event makes them lose together.

For a stylized pair of operations, let each lose $10 million with marginal probability 0.10 during a year and zero otherwise. Each has expected loss $1 million and variance 9 in squared-million-dollar units. If their joint loss probability is 0.01, they are independent: E[X]=2, Var(X)=18, and the probability of $20 million is 1%. If their joint probability is 0.05, the marginal frequencies remain 10%, but covariance becomes 100(0.05)−1=4, group variance rises to 26, and the probability of $20 million becomes 5%. Both illustrative cases are mathematically feasible. In the independent case the 99th percentile under the lower-quantile definition is $10 million because the cumulative probability at $10 million equals 99%; in the second case it is $20 million. Endpoint conventions must be declared for discrete quantiles.

The illustration establishes a narrower and more useful result than a blanket statement that correlation increases premiums: correlation can increase the tail and required risk capital, yet the premium effect depends on the coverage response, insurer funding and pricing approach, contract layers, reinsurance cost, and capital allocation. For some capped and deductible contracts, a change in dependence can even move expected indemnity in a direction that is not determined by correlation alone.

Contract transformations destroy simple allocation

Consider a shared-limit group cover where three subsidiaries suffer covered gross losses of $8 million, $7 million, and $3 million from one event. For a deliberately simplified event-level deductible of $2 million and a total response limit of $10 million, aggregate indemnity is min(max(18−2,0),10) = $10 million. If instead every entity had an independently attaching deductible and limit, total indemnity would depend on the three separate contracts and could differ substantially. The same total gross loss therefore does not determine payout without the contract's attachment architecture.

Where the contract pays $10 million, a hypothetical managerial allocation proportional to the three gross losses produces $4.444 million, $3.889 million, and $1.667 million, rounded to three decimals. Those numbers are an administrative illustration, not a legal determination of each insured's entitlement. Actual allocation could depend on priority rules, chronological payments, sublimits, claims handling, the identity of the claimant, legal interests, or a negotiated settlement. The observed distribution of paid claims is especially different from a cost allocation when one entity bears uninsured loss or the underlying claim is rejected.

Write the covered payout Y = g(X_1,...,X_n). In general E[g(X)] is not equal to g(E[X]) and is not equal to sum_i E[g_i(X_i)] when g includes shared aggregate limits, attachment thresholds, reinstatements, or interacting exclusions. The difference is a mechanical consequence of nonlinear contract response, not by itself evidence of pricing failure.

Common-event structure and catastrophe exposure

A dependence model should separate shared event indicators from entity-level vulnerability. A flood, storm, earthquake, regional power failure, ransomware campaign, supplier collapse, or civil disturbance can affect legally separate affiliates through common physical or operational channels. The probability and severity of those shared losses require hazard and network evidence, not merely observed positive correlation in a short accounting panel.

For each event k, one can identify an event indicator I_k, entity vulnerability V_ik, gross loss Z_ik, and a coverage response dependent on limit and attachment. The relevant net risk is constructed from simulated event losses and the actual insurance response. Aggregate dependence is also influenced by shared reinsurance treaties and reinstatement provisions. A model that treats each entity loss as independently distributed can be materially misleading when common-event exposures are strong.

ASOP No. 39's catastrophe guidance addresses assessment of catastrophe losses in property/casualty ratemaking. Its presence does not substitute for a documented catastrophe model, hazard assumptions, or location-level vulnerability. Analysts should distinguish an adopted actuarial standard from an exposure draft of a prospective revision.

Pricing relativity and the choice of benchmark

A relativity r_i can express estimated cost per normalized exposure unit for entity i compared with a chosen benchmark. It becomes interpretable only when the numerator and denominator have consistent coverage, claims-development, and inflation bases. An adjusted relativity may compare loss costs at equal attachment points and limits; an observed invoice relativity may instead compare premiums affected by negotiated conditions, taxes, or fronting fees.

There are at least four legitimate but distinct benchmarks. The standalone marginal loss cost of an entity estimates its covered loss without others. The incremental portfolio cost T(G)−T(G{i}) estimates the effect of adding entity i to the program under a fully specified technical cost function. The proportional portfolio allocation assigns a budget pool using a predetermined exposure or cost weight. The competitive market quote reflects what an insurer is willing to contract for the particular combined submission. These benchmarks need not coincide and answer different questions.

A proposed insurer price can be tested by recalculating the same program under two portfolios, using the same economic assumptions. But removing a subsidiary can change the aggregate deductible, underwriting appetite, minimum premium, reinsurance purchasing, and legal coverage design. A valid incremental comparison must identify which contract terms are held fixed and which are reoptimized. Otherwise observed price difference confounds marginal risk with program redesign.

Capital and allocation

Diversification can lower portfolio risk capital compared with the sum of standalone requirements, although concentration can remove much of the benefit. Allocation of a shared capital amount is a governance and measurement problem with several established approaches. In his coherent capital-allocation analysis, Denault (2001) investigates fair allocation using game-theoretic and Aumann–Shapley ideas. Such literature demonstrates that several rational allocation criteria are possible; it does not establish a unique universal insurance premium split.

Euler-style allocation can allocate a differentiable, positively homogeneous risk measure K(X) using contributions k_i = x_i ∂K/∂x_i in a portfolio represented by weighted positions, subject to its mathematical assumptions. A covariance allocation may use Cov(X_i,X)/sqrt(Var(X)) for portfolio standard-deviation capital, provided variance is positive and the selected premium component is correctly linked to this risk measure. Under these definitions, entity allocations can be negative when diversification effects are sufficiently favorable; a negative contribution to incremental capital is conceptually different from a negative total commercial premium.

A credible cost analysis therefore reports the selected risk measure, confidence level or cost-of-capital approach, chosen allocation rule, treatment of catastrophe concentration, reinsurance and limits, and whether capital contributions are gross or net. The capital rule must be interpreted as part of the pricing framework, not an automatic legal requirement to impose that exact internal entity charge.

Premium composition: a reproducible illustration

Suppose a hypothetical insurer prices a group program in millions of dollars with expected covered losses of 1.20, 0.80, and 0.50 across three entities; loss-adjustment and other non-proportional expense allowance of 0.25; centralized fixed cost of 0.30; net reinsurance cost provision of 0.18; and capital/contingency provision of 0.40. Assume each term is defined on a consistent net-of-recovery basis. Their sum is 3.63. If a separate commission and premium-related cost provision equals 8% of the final price, solving P = 3.63 + 0.08P yields P = 3.945652... million, approximately $3.946 million.

This is a pricing identity with assumed quantities, not an empirical price indication. It demonstrates how three standalone loss expectations can form only part of the price. It also demonstrates why dividing every expense by expected losses or allocating all costs by asset value can misstate entity contributions. A minimum fronting charge may attach to each local policy, the cost of catastrophe reinsurance may relate to concentrated locations, and premium taxes may follow an independently established tax base. If these terms vary, the 8% simplification must be replaced by a schedule of relevant proportional and fixed costs.

Cross-Subsidy and Incentive Design

Distinguishing group pooling from cross-subsidy

Risk pooling deliberately shares uncertainty. Cross-subsidy is a comparison-dependent claim about one group's contribution supporting another group's costs. An entity can pay less than its estimated standalone technical premium yet more than its incremental portfolio cost because pooling generates diversification benefits. Neither observation alone proves an economically improper transfer. Before assessing cross-subsidy, analysts must identify the reference counterfactual, the covered risks, shared fixed costs, cost-of-capital rule, reinsurance terms, and objective of the allocation.

For entity i, define p_i^stand as the technical cost of separate insurance under comparable coverage, p_i^incr as the additional technical cost of including it in the existing group program, and a_i as the charged amount under an internal allocation. A negative a_i−p_i^stand indicates a lower charge than standalone cost, but this may arise from genuine group purchasing efficiencies or diversification. A negative a_i−p_i^incr may indicate that other affiliates absorb an incremental cost under the specified counterfactual, but the difference may be chosen transparently to support investment or smooth volatile budgets.

A rigorous claim about subsidies should report at least one defensible counterfactual and show how allocation results change when fixed and common costs are treated differently. If there is only one negotiated quote and no comparable standalone quotes or independent technical model, the apparent subsidy is only an untested hypothesis.

Centralization, prevention, and moral hazard

Centralized purchasing can reduce duplication, improve data quality, expand placement options, or pool capacity. Local management, however, often controls maintenance, safety, security, and operational continuity. If every affiliate receives its premium charge in proportion to revenue regardless of its loss-prevention investments, local managers may perceive only a small fraction of the cost of their decisions. A chargeback using recent claims creates the opposite risk: a severe random catastrophe can trigger a large punitive budget shift that overwhelms more reliable exposure-based evidence.

Holmström (1979) identifies information conditions under which additional performance signals can improve incentives. In the corporate insurance setting, inspection outcomes, near-miss records, site control compliance, and independently measured vulnerability may be more decision-relevant than a short claims period. Yet each signal can be manipulated or affected by local reporting intensity. Incentive design therefore requires a pre-specified link between observable action, loss mechanisms, audit reliability, and managerial control.

A reasonable hybrid chargeback can combine a relatively stable expected-loss component with a capped experience adjustment, separately allocated common costs, and a transparent reward for verified risk improvements. That is an ordinary managerial design option, not a newly asserted universal theory. The actuarial weights require estimation, and the behavioral consequences require testing rather than assumption.

Selection and adverse selection inside groups

The Rothschild–Stiglitz insurance model shows why private information about type can complicate insurance-market equilibria. A multinational insurance buyer may face an analogous information problem if the central risk function lacks information about one subsidiary's condition, new hazardous operations, or incident history. Conversely, a subsidiary may resent an allocation rule that rewards underreporting exposure. Robust central procurement should make accurate information provision an explicit contractual or organizational requirement and verify risk-classification variables independently where material.

Acquisitions and divestitures can trigger strategic timing effects. A business entering a group just before renewal might benefit from a pre-existing limit without bearing an adequate estimate of incremental risk; a departing subsidiary may leave unpaid long-tail liabilities or historical loss development. Pricing and allocation should therefore specify temporal eligibility, retrospective adjustments if contractually permitted, experience-period rules, and responsibility for run-off exposures.

Captives, related-party pricing, and true risk transfer

A captive may retain part of group risk, purchase reinsurance, or front policies through locally licensed commercial insurers. However, internal reserve funding, an accounting transfer, an indemnity agreement, and an insurance contract have different legal and financial properties. The OECD captive-insurance discussion in chapter X asks whether insurance risk is truly assumed and diversified and whether the captive has a realistic capacity to meet claims. The document addresses arm's-length pricing under its tax framework; a group cannot establish adequate risk transfer merely by naming an affiliate an insurer.

Where a captive assumes a defined first layer, the group may pay a commercial fronting insurer for policy issuance, administrative services, collateral or credit exposure, and external reinsurance. Each cost has a different economic basis. A defensible analysis maps premium flows, ceded and retained limits, claims obligations, capital, collateral, recoveries, and the tax or legal relationship between entities. An external commercial quote should not be mechanically equated to an arm's-length related-party premium when the risks, service functions, coverage and capital differ.

International comparability is especially sensitive to governing law. The Solvency II authorization framework applies within its stated legal jurisdiction and should not be read as permission for unrestricted non-admitted insurance globally. Entities outside its coverage require jurisdiction-specific research on direct insurance placement, premium tax, policy issuance, and payment restrictions. This dossier does not supply legal clearance for any unidentified multinational program.

Organizational accounting and the premium ledger

A group allocation schedule should separate the insurer's billing unit from the internal cost center. A central parent may pay a gross external invoice, obtain a consolidated discount, fund deductibles, and distribute costs to subsidiaries. But reporting currency, intercompany agreements, value-added or premium taxes, local deductions, and accounting eliminations may change the amounts recognized by particular entities. The policyholder's cash settlement and each legal insured's insurance entitlement can be different again.

IFRS 17's group-measurement requirements help describe the reporting insurer's insurance contracts, but they do not set the internal transfer-pricing standard for insured subsidiaries. The OECD arm's-length framework is likewise purpose-specific. The strongest administrative practice maintains three reconcilable ledgers: the insurer's technical/pricing components; enforceable coverage and claims transfers; and the corporate financial allocation. A fourth ledger may be necessary for captive and intercompany transactions under relevant law.

Comparative institutional evidence and its limits

The FERMA/PwC 2026 risk-manager survey reports more than 1,300 practitioner responses across 79 countries and documents perceived premium increases, changing insurance-buying behavior, and attention to interconnected threats. This is useful evidence of respondent perceptions and management choices. It is not an observed entity-level insurer rate database, a representative audit of the legal completeness of multinational programs, or a causal test of a particular cross-subsidy mechanism. The published survey methodology identifies 1,329 participants; geographical and self-selection limitations must be considered before making population claims.

A disciplined comparison therefore separates survey responses, insurers' quoted premiums, regulatory filings, audited loss data, and standardized exposure estimates. Cross-country premiums without consistent policy terms, currency valuation, and tax treatment often show differences that arise from measurement rather than underlying risk.

Evidence Robustness and Source-Critical Validation

Evidence hierarchy and claim matching

Evidence should be graded according to the question it actually resolves. An executed insurance contract and its operative endorsements can substantiate who was promised what coverage for a defined period. A dated claims record can substantiate notification, booking, and payments, subject to its reliability; it cannot establish that an unpaid or denied claim will ultimately be covered. Actuarial standard text documents professional guidance within its stated jurisdiction or profession. A statute and consolidated amending text establish legal rules at specified dates, subject to implementation and interpretation. A peer-reviewed economic model supports a general mechanism under its assumptions. A risk-manager survey establishes respondent patterns, rather than causal effects or specific policy rights.

The research questions should be made falsifiable. To test whether a subsidiary's high charge is driven by greater loss exposure, construct otherwise comparable entity-peril observations and adjust for different coverages, limits, locations, claims development, and expense terms. To test whether shared catastrophes explain price loadings, compare insurer-consistent event simulation under observed dependence and a justified alternative dependence assumption while holding marginal losses and policy response fixed. To test whether chargebacks influence prevention, examine a documented change in the chargeback rule and compare behavior with a credible counterfactual, accounting for simultaneous safety initiatives, acquisitions, and market-cycle changes.

This dossier performs analytical derivations and source-critical synthesis; it does not claim to execute these empirical studies because the underlying proprietary policies, claims and quotes have not been provided. A theoretical or regulatory mechanism remains a supported explanation to be tested, rather than an observed causal finding about actual corporate portfolios.

Contract, insurer and accounting version control

For each quoted period, record proposal date, binding time, inception, expiry, endorsement dates, the wording actually issued, and subsequent endorsements. Claims-made and occurrence contracts require different time logic. A later policy edition should not be substituted for the version governing an earlier loss. Historical premium changes should separate exposure growth, attachment changes, limit changes, tax changes, claims development, and pure price movement.

An insurer's reported accounting result should not be treated as a direct observation of its initial expected cost. IFRS 17 requires insurance-contract recognition, measurement and release of insurance service results according to accounting rules; claim payments and premium cash receipts can occur on a different timeline. Similarly, a group may allocate future renewal premium using a forecast while later transfer-pricing or tax adjustments refer to a different accounting period.

For European legal comparisons, distinguish the original Solvency II Directive, its then-effective consolidated text, and the 2027 application of reforms. For international benchmarks, distinguish IAIS voluntary supervisory standards from the binding law as implemented in a particular jurisdiction. Country labels alone are an inadequate legal version citation.

Statistical uncertainty and loss-development bias

A low-frequency, high-severity insured portfolio contains relatively little annual data about its tail. Parameters for regional correlation, fire vulnerability, cyber aggregation, claims inflation, and litigation development can dominate the apparent entity-level risk ranking. The ASOP No. 53 future-cost guidance requires considered selection of historical data, trend, new exposures, models, and communication of limitations. The ASOP No. 23 data-quality standard emphasizes whether available observations fit the actuarial purpose and what defects remain.

For a hypothetical dataset, uncertainty would be reported through parameter intervals, alternative admissible dependence structures, coverage uncertainty, exposure revaluation, and scenario comparison. A validation split must respect the entity, event, and calendar grouping: random division of claim rows generated by one catastrophe between training and test sets leaks event information, creating inflated apparent accuracy. For portfolio acquisition studies, the test period should occur after the training period, with changes in legal structure captured explicitly.

A frequency-severity model's calibration must be checked separately by line, peril, geography, entity size, and claims maturity, where data support these comparisons. Reporting only overall loss-ratio fit can conceal opposite errors in different subsidiaries. The loss distribution's extreme quantiles require special checks beyond aggregate average error. A pricing statement about capital or aggregate limits should expose sensitivity to the tail model rather than hiding it within a single premium figure.

Reproducing the illustrative calculations

The premium example is an arithmetic identity in million-dollar units: losses 1.20+0.80+0.50=2.50; other fixed components 0.25+0.30+0.18+0.40=1.13; base cost 3.63; commission loading 0.08 times premium; premium 3.63/0.92 = 3.945652173913. No discount-rate or investment-income assumption is implicit. The commission is not counted in the fixed component, and the reinsurance term is explicitly a net cost. If changing to a different gross/net claim basis, each component must be recalibrated to avoid counting recoveries twice.

For the common-event claim, the assumed one-event group-wide $2 million deductible and $10 million payment limit apply to total gross loss of 18, giving 10 in payment. A hypothetical proportional administrative recovery allocation is 10×(8,7,3)/18 = (4.444444...,3.888888...,1.666666...). The rounded sum 10.000 is preserved. No assertion is made that insurers or courts would allocate actual legal entitlements in this manner.

For the Bernoulli example, E[X_i]=0.1×10=1; E[X_i²]=0.1×100=10; variance 10−1=9. With joint probability 0.05, E[X_1 X_2]=0.05×100=5, hence covariance 5−1=4 and total variance 9+9+2×4=26. The independent joint probability 0.01 yields covariance 0 and total variance 18. The joint probability is within the Fréchet interval [0,0.1] given equal marginal probabilities of 0.1. Discrete VaR requires a defined quantile convention; for both distributions the specified lower quantile is the smallest loss whose cumulative probability is at least 0.99.

Counterfactual and sensitivity register

A replicable insurer-cost investigation needs a frozen comparison table. At minimum, the table specifies a baseline portfolio with identified insured legal persons and coverages; an entity-removal scenario; a changed aggregate limit; a changed deductible; one independently validated catastrophe correlation stress; an inflation and currency stress; a revised claim-development estimate; and an alternative allocation of fixed and capital costs. Each scenario reports technical cost, expected paid loss, loss limit response, capital, reinsurance provision, taxes where applicable, premium, and entity charges as distinct fields.

A model comparison should change one factor at a time for sensitivity, while scenario combinations may explore interaction. For example, increasing a shared limit with a simultaneously changed insured population does not identify the price effect of the limit alone. The scientific design should record which constraints are fixed and which terms would be renegotiated in market practice. That distinction prevents a hypothetical model output from being misrepresented as an insurer's actual renewal offer.

Scholarly and editorial limitations

The primary sources establish standards, institutional rules, bibliographic evidence, and economic mechanisms. The dossier's mathematical results are derived internally under explicitly stated assumptions. The official ASOP HTML pages have been used for section-level reading, and their official PDF forms govern where textual precision is required. OECD section E of chapter X is read against the published 2022 edition; the publisher landing page supplies the stable bibliographic link. Material outside the available text of certain scholarship is not treated as independently reanalyzed. The comparative legal reading is limited to cited EU-level provisions and international standards, without an exhaustive country-by-country statutory survey.

The evidence ledger records the specific claim each source supports, its institutional type, and the verification boundary. Source completeness is a finite property of this dossier's reviewed citations; it is not a claim that every insurer's contractual form, market quote, or national law has been examined. This is an original synthesis of published material and worked examples, with no claimed newly discovered empirical regularity, new proprietary theoretical model, or external peer-review outcome.

Integrated Research Synthesis and Unresolved Questions

A coherent causal reading of multi-entity price variation

The dossier's central synthesis is a sequence of testable distinctions. The legal definition of covered persons and interests identifies what the insurer must potentially pay. Exposure measurement identifies who is at risk and how the relevant peril could produce an insured loss. Dependence and contract response map entity events into aggregate claims. Actuarial cost estimation adds claim handling, funding, reinsurance and capital under documented methods. Underwriting negotiation converts an indication into a transaction price. Finally, the corporate organization allocates the cash and economic burden among entities. Errors arise when data or interpretations move between these stages without a reconciliation.

An observed high affiliate charge can result from a genuinely hazardous location, wider contract conditions, low deductible, high insured value, higher cost of local issuance, catastrophe concentration, tax treatment, a parent's deliberately redistributive allocation, or an arbitrary historical billing percentage. Competing explanations are separable only with exposure, contract, price component, and chargeback data. The same invoice is consistent with several causal histories.

Portfolio value and its distribution

Common purchase may create savings from lower administrative duplication or favorable negotiated terms and may permit risk pooling. It can also create shared aggregate-limit congestion, correlated catastrophe exposure, minimum local issuance costs, or disputes about which insured receives protection when limits are exhausted. A claim that the combined program is cheaper must therefore specify the metric: group total premium for equal coverage, actuarial technical cost, expected net cash cost after retentions, tail risk of uninsured loss, or total cost of risk including corporate capital.

The most useful comparison pairs a joint program with a carefully specified alternative collection of separate programs. The alternatives must keep risk ownership, coverage scope, event periods, and valuation currency commensurable. Even then, assigning the total group advantage to subsidiaries requires a convention: standalone saving, incremental saving, proportional exposure, negotiated bargaining weight, or some other specified administrative basis. That assignment is a decision about benefits and incentives rather than a theorem that follows uniquely from the aggregate quote.

Robust incentives and evidence sufficiency

A stable allocation rule that tracks causally relevant exposure can improve the local visibility of risk costs. A contemporaneous, independently measured risk-control component may create a stronger incentive than raw short-period claims experience. However, where an affiliate cannot influence climate hazard, the group can justifiably centralize catastrophe-capital charges rather than penalize local management for irreducible geographic exposure. Conversely, entity safety actions that materially affect claim severity should be made visible when the underlying evidence supports the connection.

Information governance is part of price governance. A parent that cannot reconcile legal identifiers, covered asset interests, historical claim versions, and invoices cannot claim a high-confidence estimate of entity subsidy. The research answer is conditional: first construct the evidence base, then estimate a cost, then choose and disclose an allocation objective. The decision rule should remain revisable as new data arrive.

International comparisons as conditional institutional analysis

International comparison works best through functions rather than country slogans. Ask who may lawfully issue a policy; what entity may hold the insured interest; how payment obligations and territorial provisions operate; what reinsurance or collateral is used; what reporting or tax obligations attach to each flow; and when each legal text was effective. The IAIS supervisory framework, European authorization provisions, IFRS insurance-contract standard, and OECD transfer-pricing treatment each govern a different question. Their simultaneous relevance does not turn them into one worldwide pricing code.

An entity-level historical study should accordingly record an applicable-jurisdiction matrix and avoid projecting a future directive's effective date backward. The European reform identified as applying from January 2027 is an explicit case where otherwise plausible statements about "current" insurance law could become temporally incorrect.

The future empirical agenda

The strongest next study would collect anonymized multi-year exposure and policy-version panels, insurer technical indications, program quotes and renewals, complete claim-development triangles, entity acquisition histories, risk-control measures, catastrophic-event keys, and internal allocation agreements. It would report data completeness before results. It could test whether entity-level marginal-cost estimates predict renewal changes better than simple premium-share allocation; whether alternative capital rules materially change internal cross-subsidy classifications; and whether improved exposure governance precedes better loss experience.

A second investigation should compare an actual master/local portfolio with otherwise equivalent independent local policies under local legal constraints, rather than assume every international placement is interchangeable. A third could examine behavioral responses when risk-control-sensitive chargebacks replace pure revenue allocation, using an explicit difference-in-differences or other defensible counterfactual where the identifying assumptions can be supported. Such research would be empirical extensions of existing insurance economics, not proprietary theories asserted by this monograph.

Resolved research propositions

The literature and derivations support five bounded results. First, legal-person distinctions matter because the identity of an economic controller does not determine a policy's insured population. Second, an additive gross expected-loss total does not imply additive insured payments under shared deductibles and aggregate limits. Third, dependence affects aggregate tail and capital even when marginal gross expected losses are constant. Fourth, an affiliate's share of corporate insurance spending is not an observed measure of its standalone insurer risk cost. Fifth, a well-documented internal allocation can be economically justified by different objectives, but the selected objective and the counterfactual must be visible before a subsidy claim has a defined meaning.

These propositions are analytical conclusions derived from established contract reasoning, probability, actuarial cost estimation, and economic scholarship. Their application to a particular insurance contract remains dependent on its facts and governing provisions. No universal numeric premium relativities or empirically validated behavioral coefficients are claimed.

Conclusion

Commercial insurance pricing across multiple legal entities is a problem of identifying an enforceable transfer of risk, measuring exposures at the correct legal and operational level, aggregating dependent losses, and attributing the prospective cost of insurance and retained risk. A consolidated insurance invoice cannot answer each of these questions, and a legal-entity chargeback is neither a direct observation of actuarial risk nor a legal adjudication of claim entitlement.

The documentary foundation begins with a dated entity-peril-policy matrix and a reconciled claims and exposure record. Mathematical analysis separates additive gross expectation from nonlinear insurance payouts, and marginal loss expectation from dependence-sensitive capital and reinsurance costs. Economic analysis distinguishes genuine pooling benefits from transfers that depend on a chosen allocation benchmark. The governance analysis distinguishes external insurance contracts, group financial reporting, local licensing, and related-party captive transactions.

The practical scientific result is a transparent method of explanation: specify the insured population; reconstruct the contract's payment function; estimate and validate prospective insured losses and expenses; identify the dependence and capital method; reconcile the commercial quote; and disclose the objective, data, and counterfactual used to allocate costs among affiliates. Each stage can be independently checked and revised as evidence changes. The boundary between supported results and future empirical investigation is explicit. The dossier contributes original integrated interpretation and reproducible demonstrations within established actuarial and economic scholarship.

References