RESEARCH DOSSIER
Insurance Firm Production and Cost Functions
Abstract. The production of insurance combines the provision of contingent financial protection, the administration and settlement of claims, risk selection and monitoring, financial intermediation, and the maintenance of capital sufficient to honor uncertain future obligations. Its cost function therefore cannot be inferred from premium volume or operating expenses alone. This monograph establishes an integrated interpretation of insurance production and cost from microeconomic duality, financial economics, actuarial obligations, organizational economics, and international supervisory and accounting evidence. It identifies the conditions under which loss and benefit measures, exposure units, policy counts, and invested assets are informative proxies for output; distinguishes expected claims transfers from the real resources required to administer protection; specifies a multi-output variable-cost framework conditional on risk characteristics and capital; and evaluates scale, scope, technical, allocative, and cost efficiency through econometric and nonparametric frontiers. The empirical review covers foundational U.S. property–liability and life insurance studies, Finnish non-life insurers, European market structure, international comparative efficiency, distribution channels, merger outcomes, and contemporary reporting standards. Original analysis shows that measured cost advantages must be interpreted jointly with output quality, financial resilience, risk heterogeneity, and the timing of contingent obligations. Accounting comparability and frontier identification determine the credibility of cross-firm rankings. The synthesis yields explicit research and supervisory implications without claiming a new economic theory or independently estimated empirical results.
Keywords: Insurance production; insurer cost function; insurance services; contingent claims; risk pooling; risk bearing; financial intermediation; claims settlement; acquisition expense; expense allocation; underwriting; solvency capital; economies of scale; economies of scope; translog cost function; stochastic frontier analysis; data envelopment analysis; cost efficiency; productive efficiency; organizational form; insurance distribution; IFRS 17; NAIC statutory reporting; insurance market comparison.
Introduction
An insurer promises performance in future states of the world and organizes the resources needed to make that performance credible. The promise is produced when coverage is underwritten and continues while the policyholder bears protected risk; its most observable execution may arise years later when a claim is adjusted, defended, settled, or paid. The insurer also invests assets backing obligations, acquires and renews contracts, manages intermediaries, and maintains liquidity and capital. A coherent production analysis must represent these interdependent services and their distinct time horizons.
The unit of analysis is the insurance undertaking as an economic producer, observed at the legal-entity or consolidated-group level with the unit made explicit. This study focuses on production technology, input requirements, economically relevant costs, efficiency measurement, and the interpretation of comparative evidence. Questions of the general level of market competition and the welfare properties of insurance demand are connected to production but are treated in their own scientific dossiers. Here, the central questions concern what the insurer produces, what resources that production consumes, and under which assumptions an empirical cost estimate measures economically meaningful efficiency.
The field's central methodological work is the frontier efficiency synthesis of Cummins and Weiss (2013), which reviews insurance inputs, outputs, costs, stochastic frontiers, and mathematical-programming approaches. Its earlier insurance efficiency working paper explains why pure financial-intermediation measures omit important claims and risk-bearing services. These contributions provide a disciplined starting point for the analysis undertaken here. The 2026 OECD insurance market report supplies a recent international context, including 2025 changes in claims costs, premium growth, and underwriting performance. Aggregate market trends are contextual evidence, not estimates of a firm's production frontier.
The monograph draws on publisher records and accessible original abstracts of peer-reviewed empirical research, the original conceptual and methodological literature, and primary accounting, statistical, and supervisory sources. Where a research article is accessible only in abstract or bibliographic form, its reported findings are described at the level supported by that record. Algebraic illustrations are stated with assumptions and do not purport to estimate a real insurance market. No new named theory, doctrine, statistical data set, or causal estimate is introduced.
Research Object, Evidence, and Analytical Boundaries
Three related economic objects
The insurance production question contains three objects. The first is a technology: the feasible transformation of labor, information, distribution services, assets, and risk-bearing capacity into coverage, claim services, and intermediation. The second is a cost function: the minimum cost of producing a specified service vector at specified input prices, quality standards, risk composition, and institutional constraints. The third is realized financial performance: the accounting outcome generated by actual risks, prices, investment returns, and management decisions. A production frontier is inferred from the first two objects. A financial ratio is principally an observation of the third.
The distinction matters because a high claim payout can mean that an insurer sold more protection, insured a worse risk pool, experienced a catastrophe, paid claims more generously, or settled claims more effectively. These events have different implications for technology and managerial efficiency. Likewise a low expense ratio can reflect automated processes, high premiums, deferred acquisition costs, insufficient claim handling, or underinvestment in customer service. Each comparison must establish its outcome definition, accounting perimeter, risk adjustment, and observational period before it can support a productive-efficiency claim.
Evidence hierarchy and scope
The analytical synthesis relies on economic identities and conditional mathematical implications; empirical conclusions are attributed to their original samples and methods; regulatory descriptions are anchored to the relevant source and edition. The 2010 international comparison by Eling and Luhnen reports frontier results for 6,462 insurers in 36 countries during 2002–2006. It provides evidence of cross-country variation, while its observation years and methodology bound generalization. The 2022 review by Rubio-Misas documents the influence of the Cummins–Weiss research program and its choices of outputs, inputs, and efficiency estimators. It is interpreted as a secondary critical synthesis rather than a substitute for the underlying studies.
The analysis treats life and non-life insurance as potentially distinct technologies, and it further distinguishes personal from commercial insurance, short-tail from long-tail liabilities, and insurers with different distribution and reinsurance structures. The same legal and accounting labels may cover different service bundles. International comparisons require a consistent currency basis, risk adjustment, time period, reporting perimeter, and price normalization. A national sample containing only large licensed insurers is not an observation of every insurance provider in that economy.
Insurance Production as a Contingent Service
A multidimensional output
Insurance production is best represented as a bundle of services performed over time. Risk pooling and bearing transform many uncertain individual claims into a financed portfolio of contingent liabilities. The insurer contributes contract design, underwriting, diversification, monitoring, financing, and an enforceable promise. Claims operations produce verification, investigation, defense, repair coordination, settlement, and payment services. Financial intermediation transforms premium and capital inflows into assets held to support current and future obligations. These functions may be complementary, yet their costs are generated by different activities and prices.
In the value-added tradition, an economically useful output vector can be written as y = (yᴿ, yᶜ, yᶠ), where yᴿ denotes risk pooling and risk-bearing services, yᶜ claim-related real services, and yᶠ financial intermediation. The components are service categories, not universally measured physical units. This formulation follows the insurer-output distinctions surveyed by Cummins and Weiss (2013). A practical study must select operational proxies and disclose whether the variables measure volume, value, quality, or realized losses.
Insurance resembles other service industries in its use of staff, software, facilities, procurement, and marketing. Its distinctive characteristics are the stochastic timing and severity of future performance, correlated portfolio losses, reverse sequencing of cash collection and claims payment, reliance on solvency resources, and opportunities for moral hazard and adverse selection. The production set must include these constraints if the comparison is meant to capture the cost of delivering equivalent protection. A policy count alone cannot describe a portfolio's loss severity or duration of obligations.
Timing and quality
A multi-period representation clarifies the service commitment. Let contract cohort c issue at t₀, receive premiums P(c,t), incur event-contingent indemnities I(c,t,ω), and consume underwriting, administration, and claim-handling resources h(c,t,ω). The economic production decision commits the insurer to a distribution of future services and payments, while actual cash flows appear across years and states ω. A one-year accounting observation therefore combines contracts issued in the year, service on existing contracts, and claims from earlier cohorts. Cohort aging, reserving development, and discounting are indispensable when comparing long-tail insurers.
Quality is part of the output specification. Claim accuracy, timeliness, explanation of coverage, contractual certainty, and solvency affect the value of the delivered service. Omitting quality from y can mechanically classify a poorly serviced insurer as efficient if it spends less on claim examination. Conversely, detailed claim investigation can be cost-increasing but socially valuable when it reduces false payments and discourages opportunism. The 1997 distribution-system study by Berger, Cummins, and Weiss shows why costs must be interpreted jointly with service quality and revenue rather than as an independent verdict on distribution technology.
Feasible production and unobserved effort
A feasible technology T describes input and output combinations that satisfy operating, legal, and prudential constraints. Let x = (labor hours, purchased services, technology capacity, capital services), and let z capture the composition of risks, underwriting restrictions, service standards, market conditions, and exposure to tail events. Then T(z) contains combinations (x,y) that are technologically and institutionally feasible for this z. Selection of z is a scientific decision: a catastrophe-heavy book and a diversified personal automobile book can share a premium amount while requiring different capital, reinsurance, and claim expertise.
Effort and quality are only partly observable. Underwriting discipline may prevent future claims without producing a visible current-year output. A claims team's diligence can lower leakage while increasing immediate adjustment expense. An empirical study that observes only the realized indemnity and expense totals will fail to detect part of these effects. This is an identification problem arising from the economic object itself; it must be addressed by appropriate outcome and risk controls rather than by a cosmetic adjustment to a performance ratio.
Output Measurement and the Value-Added Approach
Competing proxies
Five commonly encountered measures require explicit differentiation. Gross written premiums measure contract consideration and partly reflect rates, coverage, and demand. Earned premiums allocate consideration to accounting periods and remain affected by prices. Policy and exposure counts approximate quantity of protection but ignore insured limits, deductibles, hazards, and service complexity. Incurred losses or benefits are related to the scale of obligations but reflect realized risk, claim management, and estimation. Invested assets proxy the intermediation function but also vary with asset prices, leverage, and business duration. A sixth class, present-value expected claims or expected indemnities, uses modeled prospective obligations as a risk-bearing scale proxy, subject to model risk and appropriate deflation.
The 1998 Cummins–Weiss discussion contrasts the asset/intermediation approach, user-cost approach, and value-added approach, favoring a modified value-added representation for insurers because they furnish real risk and claim services in addition to holding financial assets. The 1993 U.S. property–liability cost study operationalized this service approach when estimating translog cost frontiers and the production of loss-settlement services. A modeling choice should be justified by the research question and data, not merely by its frequency of use in prior articles.
Interpreting expected and realized claims
Let q denote standardized exposures and m(x) the expected indemnity per exposure after controlling for insured risk. An idealized volume proxy for contingent benefits is yᴿ = q m(x). Realized payments are Y = ∑Iⱼ and generally satisfy E[Y | q,x] = q m(x) under a suitable specification. The difference Y − E[Y | q,x] includes process variance, catastrophes, claim development, and potentially estimation error. Treating Y directly as output allows the stochastic loss draw to shift measured productivity without any change in managerial practice. Replacing Y with a model-based expectation reduces that problem while creating dependence on model specification, underwriting data, and forecast validation.
Premium income offers another tempting proxy, but if two otherwise identical insurers sell the same protection at different prices, premium-defined output rises for the higher-priced insurer. An apparent movement toward the frontier may then represent price power rather than service productivity. In panel data, separating rate changes from coverage and exposure changes requires reliable price or risk-adjusted volume indices. Inflation adjustments based on an economy-wide price index may fail to capture changes in medical, legal, construction, or repair costs that determine insured claims.
Financial intermediation
Financial intermediation should enter the analysis with a defined role. Investment assets may be modeled as an output of portfolio-management services, an intermediate state supporting claims liabilities, or a quasi-fixed input depending on the question. If assets are counted as output and the returns they generate are also treated as output without an explicit production accounting, the analysis risks double counting. Where the objective is cost efficiency in underwriting and claim administration, the intermediation function may be represented separately from insurance service production with appropriate segment expenses and shared-cost allocation.
The 2009 analysis of endogenous risk management and intermediation explicitly treats these activities as intermediate outputs and estimates their shadow prices in U.S. property–liability insurance. Its conceptual implication is important: risk management can increase service value or economize on other resources even where recorded expenditure on the risk-management function increases. This warrants attention to production complementarities rather than interpreting each expense category as an isolated efficiency target.
Claims, Expected Benefits, and Real Resource Costs
Indemnities and administrative production
The total economic burden of writing insurance is commonly represented as expected promised payments, underwriting and servicing expenses, acquisition and distribution charges, risk-financing requirements, reinsurance expense, taxes, and required returns to capital providers. However, an indemnity paid to a policyholder is a transfer that fulfills the contingent promise; it is conceptually different from the labor, technology, and materials used to determine and execute that transfer. In a social-resource-cost analysis, an indemnity's transfer component is separated from the resources consumed in producing and financing it. In a private-insurer cash-flow or pricing analysis, promised benefits plainly remain a major economic cost to the insurer. These are complementary cost concepts rather than contradictory accounts.
Write an illustrative private economic commitment cost as Cᴾ = E[I] + Cᴴ + Cᴬ + Cᴳ + Cᴷ + Cᴿ + T, where E[I] is discounted expected indemnity, Cᴴ claims handling, Cᴬ acquisition, Cᴳ other administration, Cᴷ financing and capital services, Cᴿ net reinsurance cost, and T taxes and legally required charges. This is an analytical classification, not an IFRS income-statement identity. Accounting recognition, expense allocations, the financing effect of premium timing, ceded recoveries, and intercompany transactions can make booked totals differ from the terms. Some empirical frontier models deliberately exclude E[I] from the operating-cost dependent variable and use expected or realized loss-related quantities as output measures. That specification measures operating resource costs, not the full economic cost of promised insurance protection.
Claim handling technology
Claims costs contain fixed infrastructure and claims-sensitive variable effort. A stylized claim-processing activity may have Cᴴ(N,s,k) = Fᴴ + aN + b∑g(sⱼ,kⱼ), where N is claim count, s severity, and k legal or technical complexity. A severe bodily-injury claim may require legal defense, medical evidence, and years of reserve updates, while a small automated property claim may demand few staff minutes. The derivative of Cᴴ with respect to N is therefore conditional on claim mix. A common cost-per-claim figure can misrepresent productivity if claim mix, complexity, legal rules, and service quality differ.
The NAIC definition of loss adjustment expense distinguishes the costs associated with adjustment and recording of losses from losses themselves and discusses defense-and-cost-containment and adjusting-and-other categories. The NAIC 2025 property–casualty statement form provides separate reporting locations for incurred losses, loss adjustment expenses, underwriting expenses, and investment results. These classifications are useful evidence about reporting architecture. They do not make costs fully comparable across legal and national systems without reconciliation of definitions and expense allocation.
Endogeneity of settlements and prevention
The claim outcome may itself respond to insurer effort. Faster settlement can reduce litigation and customer hardship, while indiscriminate payment may increase leakage or future opportunism. More investigation can reduce fraudulent indemnity at the cost of adjustment labor and possible delay. The relevant performance objective is therefore expected total economic cost conditional on lawful and appropriate settlement quality, not the mechanical minimization of either claim payout or adjustment expense alone. Comparing insurers requires an explicit treatment of service requirements and potentially endogenous fraud exposure.
In long-tail business, claims paid this year may originate from many underwriting years. A fall in current claims expense could reflect reserve strengthening or release, changed inflation assumptions, or a shift in settlement timing. Underwriting-year or accident-year development can improve comparability, although even detailed development triangles do not fully identify management quality or the counterfactual cost of alternative settlements. The study's empirical claims therefore retain the basic restriction that reported accounting quantities are proxies for economic service outcomes.
Acquisition, Administration, and Claims Settlement
Acquisition technology and channel structure
Acquisition involves advertising, prospecting, underwriting applications, distribution commissions, broker relations, product explanation, customer onboarding, compliance, and contract issuance. Costs depend on distribution technology, risk complexity, customer churn, and contract duration. A direct insurer can have high up-front technology and advertising expenses but low marginal transaction costs; an intermediated insurer can have higher recorded commissions while providing comparison, advice, servicing, and customer acquisition capabilities that would otherwise require insurer-side expenditure. The quality of distribution output is therefore essential to an efficiency comparison.
The empirical Berger–Cummins–Weiss distribution comparison finds that independent-agency property–liability insurers incurred higher costs than direct writers in the U.S. sample, but interprets their coexistence through the evidence of higher-quality services and corresponding revenue advantages. This result directly rejects a universal inference from high acquisition expense to inefficient distribution. A cost-only comparison and a profit or revenue-efficiency comparison answer different questions. The chosen performance objective must match the service whose provision is under examination.
Fixed costs and contract life
Many costs of designing an insurance product, building policy administration systems, satisfying regulatory reporting, and maintaining security are incurred before an additional policy is issued. In a simplified technology, acquisition-administration cost can be Cᴬ(q) = F + aq + d(q,z), with q standardized policies, F quasi-fixed infrastructure, a per-policy servicing cost, and d varying with complexity and regulation. Average expense F/q + a + d/q falls with q only over the range in which capacity is not binding and the mix does not deteriorate. Once customer support, staff supervision, cyber controls, and technology resilience require further investment, economies can flatten or reverse.
Long-duration life products complicate the comparison. Acquisition expenditure may be concentrated at inception, while servicing and benefit liabilities continue for decades. Comparing one-year acquisition ratios across a new-business-heavy firm and a mature closed book without cohort adjustment confounds lifecycle composition with technical efficiency. A defensible study uses new-business indicators, policy duration, surrender and lapse experience, and appropriate amortization or cash-flow representations. Shared distribution and administration services further raise questions of cost allocation across lines.
Allocation is a modeling decision
A single claims system, investment department, legal function, or group brand can support several lines. Allocating its expenses by earned premiums is convenient, but allocation by headcount, claim count, complexity, transactions, or time-driven activity measures can produce materially different line-level cost curves. Such allocations are accounting conventions or activity-based approximations rather than direct observations of the underlying production technology. The NAIC statement's expense groups make these categories visible in U.S. statutory reporting, yet a cross-country study must disclose local definitions and practices.
The essential analytical question is the incremental cost of providing an additional unit of the relevant service under the feasible alternative allocation of shared resources. This requires avoiding the assumption that every dollar of recorded general expense is attributable to a particular product. Where shared cost dominates and line-level allocation cannot be credibly identified, aggregate firm-level cost measurement may be more reliable than purportedly precise line-specific frontiers.
Underwriting, Investment Intermediation, and Capital Inputs
Underwriting information as a productive input
Underwriting consumes specialized labor, data, statistical models, monitoring infrastructure, audits, and legal expertise. It creates risk classification, contract terms, acceptance decisions, and a portfolio with desired aggregate loss properties. Better underwriting can reduce expected indemnity and capital requirements; it can also alter the composition and scale of coverage sold. If an efficiency study conditions only on premiums and staff expense, a firm that takes greater risk may appear to achieve more output with fewer inputs even though it exposes capital providers to larger tail losses. A credible cost comparison must control for the volume and risk of promises jointly.
A conceptual underwriting service output is a vector of contracts indexed by exposures, limits, deductibles, hazard distributions, and duration. Aggregate standardized risk units can be constructed from this vector, but the weights must be justified for the relevant business. Catastrophe dependence, uncertain legal inflation, and policyholder behavior can invalidate a scalar reduction. The most useful empirical simplification is the one that preserves the aspect of risk most material to the inference being drawn.
Financial capital and quasi-fixed capacity
Capital can be treated as an input whose services support insurance promises, as a quasi-fixed constraint on underwriting capacity, or as an endogenous choice jointly determined with risk exposure. The economic price of equity capital is not the nominal balance-sheet value of equity and is not simply interest expense; it reflects investors' required risk-adjusted return and the opportunity cost of committed financial resources. Debt-like liabilities, policyholder funds, reinsurance, and invested assets should likewise be defined consistently to prevent the same financing activity being counted multiple times.
A conditional operating-cost function Cᵛ(y,w;K,z) takes required capital K as given while minimizing variable inputs. A longer-run cost function minimizes over K subject to solvency and funding constraints. The distinction matters for estimating scale economies: an insurer can look inefficient because it maintains temporary excess capital or because its supervisory jurisdiction requires stronger buffers; alternatively an artificially low observed capital stock may signal fragility rather than technological advantage. Capital adequacy belongs to the feasible production set.
The IAIS Insurance Core Principles establish an international supervisory architecture, while the December 2024 ICP 17 requirements address capital adequacy using a total-balance-sheet approach. These supervisory principles are standards for national implementation, not identical legal requirements in every jurisdiction. Economic measurement should compare observed capital to the applicable requirements and the insurer's actual risk rather than assume one world-wide capital price.
Reinsurance and intermediate services
Reinsurance can transfer part of claim volatility and reduce retained capital needs, while introducing premiums, commissions, counterparty exposure, collateral requirements, and transaction costs. A gross-basis production study may count all originally underwritten exposure as output; a net risk-bearing study may emphasize retained losses and capital. Both have research uses, but cross-firm rankings can change if some companies systematically reinsure more risk. A consistent analysis records direct, assumed, and ceded business and identifies the economic unit that ultimately retains the contingent liability.
The Cummins–Dionne–Gagné–Nouira analysis makes risk management and financial intermediation endogenous components of insurance performance. Its implication for cost modeling is that inputs and intermediate outputs may be jointly chosen. An empirical model that treats them as strictly exogenous can attribute cost savings to size when they partly reflect deliberate changes in risk management, reinsurance strategy, or asset-liability structure.
Cost Duality and the Multi-Output Cost Function
Technology, conditional minimization, and economic prices
A production analysis begins with a feasible technology T consisting of pairs (x,y), where x is a vector of inputs and y a vector of specified insurance service outputs. The output vector may contain separately measured underwriting-risk pooling, claim settlement, contract administration, and intermediation services; output definitions must be mutually interpretable before they enter a regression. For strictly positive input prices w, the conditional cost function is C(y,w,z) = min{x: (x,y) belongs to T(z)} w·x, where z denotes environmental, contractual, risk, and quality variables that shift the feasible set. This is the standard duality of production economics applied to a carefully specified insurance technology. The expression assumes that the measured outputs are feasible and that the firm minimizes costs for those outputs at prevailing input prices. It does not itself establish that actual insurers behave as frictionless minimizers.
The requirement that the input price vector represent opportunity costs deserves emphasis. Salaries, purchased services, rental or imputed property costs, data services, and financing resources may be included. A financial insurer also operates with capital, reserves, reinsurance and liquidity constraints, so the cost minimum is conditional on a risk-bearing feasible set. If capital K is quasi-fixed in the observation period, the short-run function C(y,w;K,z) and a long-run function that also optimizes K differ. In the long run, a separate user cost of capital can enter the minimand if the data and theory support a credible market price for its services. There is no universal accounting variable that automatically equals that price.
When differentiability and an interior optimum hold, Shephard's lemma gives ∂C/∂wᵢ = xᵢ(y,w,z). Dividing by cost yields input expenditure shares. This provides the economic basis for estimating a cost function jointly with cost-share equations, a strategy used in the original Cummins and Weiss cost-frontier study. The implication must be read conditionally: off-frontier firms, measurement errors and changing output quality may make observed shares diverge from optimal shares.
Flexible functional forms
A common empirical approximation is the translog multi-output function:
ln C = α₀ + Σᵢ αᵢ ln yᵢ + Σₖ βₖ ln wₖ + ½ΣᵢΣⱼ γᵢⱼ ln yᵢ ln yⱼ + ½ΣₖΣₗ δₖₗ ln wₖ ln wₗ + ΣᵢΣₖ θᵢₖ ln yᵢ ln wₖ + λ'z + ε.
At a normalization point, αᵢ represents the partial elasticity of estimated cost with respect to output i, subject to the estimated interactions and environmental covariates. A cost function homogeneous of degree one in input prices imposes Σₖβₖ = 1, Σₖδₖₗ = 0 for each l, and Σₖθᵢₖ = 0 for each i, with corresponding symmetry restrictions on interaction coefficients. These constraints matter because rescaling all input prices while holding output fixed should proportionately rescale the minimum cost. Output and price quantities should be normalized, with care about zero output values because a logarithmic translog requires strictly positive arguments or an explicitly justified treatment of zeros.
One flexible alternative uses a Fourier extension to accommodate nonlinearities beyond a standard translog. The Fenn and colleagues' European insurance study fits Flexible Fourier cost frontiers for life, non-life, and composite insurers, and simultaneously models heteroskedasticity in frontier errors. Its methods illustrate why the specification of stochastic variation is part of the economic model rather than a cosmetic technical adjustment.
Scale economies and marginal cost
For a proportional expansion of all outputs, define the aggregate output elasticity εCy = Σᵢ ∂ln C/∂ln yᵢ, with environment and quality fixed. A common measure of scale economies is SE = 1/εCy. When εCy is below one, proportional output expansion raises total cost less than proportionally within the relevant local neighborhood; SE exceeds one. When εCy equals one, local constant returns to scale hold. If capital and risk must rise disproportionately with portfolio size, a calculation based on operating expenses alone may overstate long-run economies.
The marginal cost of output i is MCᵢ = ∂C/∂yᵢ. The elasticity of C with respect to yᵢ equals MCᵢ·yᵢ/C. A premium-based proxy for yᵢ therefore changes the apparent elasticity if prices, exposure risk, or the policy mix change at the same time. A firm may show an increasing premium scale with constant number of policies because insured values or market rates rose. Without suitable exposure and risk adjustments, an estimated cost-output slope then combines technology with market pricing and portfolio composition.
Scope economies and joint production
For two outputs a and b, a conventional two-product scope measure is SC = [C(a,0)+C(0,b)−C(a,b)]/C(a,b), evaluated at comparable output quantities and input prices. Positive values indicate that joint production costs less than separate production under the specified counterfactual technologies. This comparison is conceptually distinct from scale economies. It can be driven by shared underwriting information, claims systems, investments, distribution, brand, or capital resources, while specialization can economize on complexity and governance.
Estimating C(a,0) and C(0,b) from a translog model with ln y requires a defensible zero-output treatment; mechanically extrapolating the logarithmic function to zero is invalid. Scope studies often use frontier techniques, alternative functional forms or careful comparison among specialist and diversified firms. The Cummins–Weiss–Xie–Zi scope study explicitly tests financial-service diversification versus specialization in U.S. insurance over 1993–2006. Its empirical design is relevant because a favorable joint-production cost difference cannot be inferred from the mere existence of conglomerate insurers.
Heterogeneous output and quality
Quality is part of the production specification whenever additional service requires resources. Important dimensions include settlement timeliness, accuracy, complaint handling, long-term solvency, availability during stress, effective coverage, and distribution advice. A cost regression comparing two insurers with identical premiums but different contract quality is a mixture of technology and output design unless controls capture those differences. Omitting quality may create the statistical appearance that expensive but higher-quality insurers operate farther from the frontier. Conversely, cheap but under-reserved coverage may be mistaken for efficient production. The frontier is meaningful only after identifying the service delivered and the constraints under which it was produced.
Economies of Scale, Scope, and Organizational Boundaries
Local technology rather than a universal firm-size law
Fixed costs support increasing returns for small or expanding insurers in some ranges. At larger size, coordination, organizational complexity, regulatory capital, catastrophe concentration, distribution conflict, and accumulated legacy systems can offset the savings. These mechanisms imply that economies of scale are conditional on product, country, time period, firm size, and the definition of output. They do not imply a single optimum premium volume applicable to life insurers, property–casualty carriers, mutuals, composites, and global groups.
The Finnish non-life cost analysis by Toivanen provides an early country-specific test of scale and scope. The Cummins and Weiss U.S. frontier study separately estimated size-stratified cost frontiers and reported economies of scale for small and medium property–liability insurers in its 1980–1988 observations. Their findings apply to those samples and technologies; they should not be transformed into a global statement about unlimited consolidation efficiency. The proper empirical question is whether the estimated economies persist after updating output measures, regulatory conditions, claims technology and risk mix.
Life insurance and organizational development
Long-tail life insurance involves long-duration liabilities, costly acquisition, mortality and lapse modeling, investments, and policy administration. Consolidation can create shared technology and distribution while imposing integration costs and changing customer relationships. The Cummins–Tennyson–Weiss study of U.S. life insurance consolidation evaluates merger and acquisition activity using data-envelopment and productivity methods. Its empirical relevance is the need to distinguish a merger's realized productivity effect from the engineering possibility of shared systems. Selection into mergers and changes in the business environment complicate causal attribution.
The Cummins–Xie study of U.S. property–liability mergers likewise links merger outcomes to productivity and efficiency. The fact of merger completion alone does not prove that promised economies materialized, since purchase selection, subsequent business mix and time-varying industry conditions jointly determine observed results. A proper ex post study follows consistent entities through a transition period and separates integration expenditures from ongoing technology.
Mutual and stock firms
Mutual insurers and investor-owned insurers can operate with different governance, surplus distribution and business objectives. In a mutual, policyholder and residual-claimant roles can partly coincide; a stock insurer has external shareholders and potentially stronger market discipline. Their organizational costs depend on agency relationships, portfolio scope, regulation and market segment. It is therefore an empirical error to assume that legal form mechanically determines productive efficiency. The Cummins–Weiss–Zi study of organizational form examines the coexistence of stock and mutual U.S. property–liability insurers using efficiency methods. Cross-form comparisons must account for differences in lines, ownership rights, business objectives and risk retention.
The international evidence of Eling and Luhnen (2010) specifically reports that its findings do not consistently support the expense-preference hypothesis under which mutuals are presumed systematically less efficient. That observation further supports treating governance as a potential explanatory variable, not a shortcut for ranking individual firms.
Multi-line specialization and conglomeration
Diversification can reuse a brand, analytic infrastructure, management systems, and financial resources. It can also enlarge product complexity, create internal transfer pricing disputes, introduce cross-selling frictions, and complicate governance of underwriting risk. Technical scope economies therefore need separate examination from revenue opportunities, capitalization, and strategic focus. The 2010 U.S. insurance diversification study evaluates whether insurers offering multiple major insurance segments outperform focused alternatives. The study's sample and method should accompany any interpretation; related empirical literature has found context-dependent coexistence of both conglomerate and focused strategies.
A comprehensive economic comparison evaluates at least three effects: the reduction in common resource expenditure; the change in revenue or service quality through joint distribution; and the change in risk financing and required capital. The first is a productive scope claim. The second may be a demand or distribution advantage. The third is a financing and risk-bearing effect. Collapsing all three into one asserted economy of scope obscures the mechanism and can overstate the relevance of a cost-only frontier for corporate strategy.
Cost Frontiers, Efficiency, and Productivity
Technical, allocative, and cost efficiency
Technical efficiency concerns whether the observed input bundle uses the minimum feasible resources for the specified outputs, given technology. Allocative efficiency concerns whether the mix of inputs is cost-minimizing at their economic prices. Cost efficiency combines technical and allocative components in the conventional framework; actual results depend on normalization and methodology. These are relative concepts requiring a defined comparison set and observation window. An efficiency score of 0.85 is not an accounting assertion that exactly 15 percent of total reported costs could be eliminated next year, because environment, sample, quality and adjustment costs may constrain the feasible change.
The field's methodological development is reviewed in Cummins and Weiss's frontier-methods chapter. Insurance studies differ in the inclusion of claims and financial intermediation outputs, input prices, regulatory environments, capitalization, treatment of risk, and choice of statistical frontier. Each choice alters the economic meaning of the estimated gap.
Stochastic frontier analysis
For a cost frontier, a simple stochastic model is ln Cᵢ = f(ln yᵢ,ln wᵢ,zᵢ;β) + vᵢ + uᵢ, where vᵢ is a two-sided statistical noise component and uᵢ≥0 is a one-sided nonnegative cost-inefficiency term. With an appropriate stochastic specification, a conditional cost efficiency index can be written CEᵢ = exp(−uᵢ). Estimation requires distributional assumptions and credible separation of noise from inefficiency. The Aigner–Lovell–Schmidt original frontier model establishes the methodological foundation; Kumbhakar and Lovell's systematic treatment explains model estimation and interpretation.
Insurance expenses are exposed to rare catastrophic shocks, reserve revisions, and legal claims. A high observed cost can therefore reflect adverse states of nature rather than inferior management. The symmetric noise term is an attempt to distinguish random shocks from systematic one-sided inefficiency. Identification remains sensitive to distributional shape, heteroskedasticity, persistence, panel structure, and the omission of exposure and risk variables. The Fenn et al. European application explicitly models heteroskedasticity in the stochastic components while examining scale and inefficiency.
Data envelopment analysis
Data envelopment analysis (DEA) constructs a piecewise-linear frontier from observed inputs and outputs under explicitly chosen returns-to-scale assumptions. The foundational Charnes–Cooper–Rhodes formulation develops an input–output efficiency measure using linear programming. Applied to insurance, it can accommodate multiple output indicators without imposing a single translog functional form. Its measured frontier is nevertheless generated by the firms included in the sample: outliers, common shocks, measurement noise and differences in quality can move that frontier.
DEA under constant returns to scale and variable returns to scale yields different efficiency decompositions. Researchers must state whether the optimization is input-oriented or output-oriented and whether observed input prices make the objective cost-minimization or physical-input reduction. A firm with high claims payments may appear highly productive if realized indemnity is treated as desirable output, even if much of that payment reflects adverse selection or inefficient settlements. Correctly specifying the claim-servicing output is therefore a substantive prerequisite to choosing DEA.
The Cummins–Zi comparison of frontier methods demonstrates the methodological relevance of comparing techniques within U.S. life insurance. Differences among methods should be treated as evidence about estimator dependence and research design, not automatically averaged into a definitive managerial ranking.
Productivity over time
Productivity measures describe shifts in output relative to input through time, while technical change refers to movement of the feasible frontier under a maintained model. Malmquist indices constructed from distance functions are one way to decompose change, with restrictions related to input-output measurement and returns to scale. Advances in digital policy issuance, automated settlement, and risk classification may shift the attainable service bundle. Simultaneously, inflation, catastrophe risk, privacy rules, and changes in required capital can alter conditions of production. A decline in expense ratios during a price increase is not equivalent to technological progress.
The studies of U.S. life-insurance consolidation and property–liability mergers apply productivity analysis to structural changes in the industry. Their results motivate longitudinal designs that track the same output definition and adapt appropriately to mergers, exits, inflation and changing underwriting exposures.
Identification, Data Quality, and Model Specification
Simultaneity and endogenous output
The quantity and composition of insured exposures are management choices jointly influenced by premiums, underwriting appetite, capacity and risk selection. Firms may expand into higher-risk lines because they possess unusual technical expertise, or contract when rates fail to cover prospective costs. A cross-sectional correlation between cost and output consequently need not identify a technological cost curve. Premium income is simultaneously a proxy for activity, a price times quantity, and an outcome of competitive conditions. Econometric specification must consider endogenous output, omitted managerial quality, and the firm's exposure to claims distributions.
A stronger empirical design uses policy counts, sum insured, contract limits, earned exposures, claims-service workload, or risk-adjusted present values in explicitly declared combinations. Instrumental variables, panel fixed effects, structural restrictions or quasi-experimental institutional changes may help address endogeneity when their identifying assumptions are credible. No single method can rescue a fundamentally invalid output measure. A source-based review should describe what the published paper actually identifies rather than upgrade associations to causal effects.
Noise, tail events, and reserve development
Observed claims, expense ratios and reserves react to catastrophe seasons, inflation, litigation, changes in social norms, and changes in accounting estimates. An insurer writing coastal property is exposed to a different conditional distribution from a similar-sized firm writing inland household policies. Cost models should control for geographic concentration and catastrophe years, test sensitivity to removing outlier years, and consider tail-dependent risk. Reserve development needs accident-year and calendar-year context. Excluding a catastrophe year solely because it raises cost can remove the very risk production the insurer was organized to bear.
Firm boundaries and data linkage
A legal entity may share technology, acquisition and investment functions with sister companies, while its statutory statement records expense allocations. A consolidated group can include investment or noninsurance operations that should be separated when the output is insurance protection. The same claims service can be outsourced, centralized or performed within the legal entity; measured payroll costs differ even if economic resources are similar. The unit of analysis must be preserved when merging datasets and interpreting consolidation.
Cross-country data require alignment of currency, inflation, line-of-business categories, paid-versus-incurred claims, asset valuation, reinsurance accounting, local tax charges and solvency requirements. International economic comparisons should use purchasing-power or appropriate cost-price adjustments with their limits stated. Converting nominal expenses using a single exchange rate cannot account for national differences in wage levels, product coverage or loss severity.
Input prices and environmental variables
Labor price can be constructed from personnel costs per employee, but this ratio is affected by labor composition, outsourcing and temporary staffing. Physical-capital cost may require rental rates or depreciation and user-cost assumptions. Financial-capital prices are highly sensitive to risk, leverage and tax conventions. The researcher must specify whether risk is an output attribute, a quasi-fixed environmental covariate, or a jointly chosen input. Regulatory regimes, legal claim environments, distribution structures, reinsurance and product mix enter the model where they genuinely affect feasible costs.
A cost ranking is most informative when it is stable across reasonable alternative measurements and appropriately normalized subgroups. Reporting multiple frontier specifications, distributions and comparison populations is more scientifically useful than publishing a single unexplained percentile. The empirical framework in Eling and Luhnen's cross-country study explicitly compares methods, countries, lines, firm sizes and legal forms, illustrating the breadth of these specification dependencies.
Uncertainty of estimated efficiency
A numerical efficiency index is an estimate subject to sampling error, assumptions about stochastic terms, data problems, and the existence of a suitable reference frontier. Confidence intervals, bootstrapping where appropriate, outlier sensitivity, comparison of fit and alternative environmental controls help communicate that uncertainty. Rankings of firms with overlapping uncertainty intervals should not be portrayed as sharp managerial performance hierarchies. A substantive review ought to state which inference persists after sensitivity checks and which conclusion depends on the original paper's own estimation convention.
Empirical Findings Across Insurance Markets
United States: property–liability cost frontiers
The early Cummins–Weiss study (1993) uses stochastic cost frontiers in three size-stratified samples of U.S. property–liability insurers covering 1980–1988. It reports average efficiency of approximately 90 percent for large insurers, 80 percent for medium insurers, and 88 percent for small insurers relative to the corresponding separate frontiers. The paper also reports economies of scale for the small and intermediate groups and differences in the production of loss-settlement services. Because the frontiers were estimated separately by firm size, the three numerical percentages are not unconditional scores on a single universal production frontier. They are historic sample-specific findings conditional on the study's technology and estimation.
The result demonstrates two broader points. First, claim-settlement service appears as an explicit output in an important early insurance cost study, confirming that insurer performance cannot be represented by policy sales alone. Second, estimated scale benefits may apply to a defined segment while the highest efficiency does not monotonically increase with firm size. Translating these findings to present-day markets would require current data and explicit tests, not a simple extrapolation from 1988.
Europe: structure, scale and managerial efficiency
The Fenn–Vencappa–Diacon–Klumpes–O'Brien study (2008) analyzes European life, non-life and composite insurers using Flexible Fourier stochastic cost functions. Its dataset spans 1995–2001 in fourteen major European countries. The reported results indicate increasing returns to scale in much of the sample, while larger and high-market-share insurers tend to display greater estimated cost inefficiency. These findings can coexist: the estimated cost technology may favor larger scale even while some large incumbent organizations operate farther above the attainable frontier. The distinction between technological scale advantages and realized managerial efficiency is analytically decisive.
The Eling–Schaper European life-insurance study examines the relationship between business conditions, efficiency and productivity across 970 life insurers in fourteen European countries over 2002–2013. Its central relevance is the explicit incorporation of the business environment into performance research. The number of companies and years expands observational coverage but does not establish direct causal comparability of every national institutional setting.
International efficiency and legal form
The Eling–Luhnen international comparison analyzes 6,462 insurers in thirty-six countries and compares frontier methods, organizational forms, insurer sizes and life/non-life segments. It reports average efficiency improvements for 2002–2006, substantial cross-country dispersion, and results inconsistent with a simple presumption that mutuals are generally less efficient than stock insurers. The ranking depends on the firm's economic and reporting environment and on frontier specification. Its observations describe the early 2000s, and national efficiency levels cannot automatically be interpreted as the causal effect of regulatory quality.
Scope, risk management and merger effects
The Cummins–Weiss–Xie–Zi scope analysis addresses conglomeration and specialization over 1993–2006 in U.S. insurance. The Cummins–Dionne–Gagné–Nouira study shows the analytical importance of endogenous risk management and financial intermediation. Together, these studies broaden the comparison beyond a simple operating-cost-per-policy interpretation: joint activity can change the output set and the use of risk-bearing and financial resources. Their empirical conclusions remain attached to their data, sector and method.
The Cummins–Tennyson–Weiss life consolidation analysis and the Cummins–Xie property–liability merger analysis examine productivity and efficiency around consolidation. Corporate mergers combine firms, assets and portfolios; their private benefits depend on whether consolidation actually improves the production process and on how changes in input prices, business mix and underwriting risk are handled. A merger can also influence competition and customer outcomes, which belong to a separate welfare evaluation beyond technical cost efficiency.
What the evidence establishes
The empirical literature firmly establishes that insurer cost and efficiency can be studied with explicit multiple outputs and frontier methods; that scale, scope, legal form and distribution structure matter in context; and that observed efficiency scores vary with technology specifications and environment. It does not support one universal efficiency ranking or a claim that consolidation, mutual ownership, direct distribution or larger premium volume necessarily produces lower total economic cost in every market. This restrained synthesis follows from comparing the research designs and scopes of the cited studies. It is an analytical reading of their established results, not a new empirical estimate.
Comparative Accounting and Supervisory Measurement
IFRS 17 and economic service measurement
The IFRS 17 Insurance Contracts standard applies to annual reporting periods beginning on or after January 1, 2023, subject to jurisdictional adoption and scope. It distinguishes insurance service results from insurance finance income or expenses within the reporting framework and uses defined insurance contract measurements, including fulfillment cash flows and, where applicable, the contractual service margin and risk adjustment. These categories are designed for financial reporting and must be interpreted through the standard's own requirements. They are not identical to the variables of a production function.
Insurance revenue under IFRS 17 is not a generic substitute for cash premiums received, and insurance service expenses do not necessarily equal the economic cost of all risk-bearing activities. Acquisition cash flows, expected fulfillment costs and contract groups may be recognized on a schedule that differs from cash payment. A frontier study using international financial statements must therefore specify the accounting regime and reconcile reporting changes across 2023. A spurious productivity break can result from comparing pre- and post-adoption revenue concepts without a bridge.
The IFRS Foundation's supporting terminology helps establish reporting definitions. Empirical researchers should consult the primary standard and current local implementation rather than treating its measurement conventions as a theory of economic production costs.
U.S. statutory reporting
The NAIC property/casualty annual statement framework reports earned premiums, incurred losses, loss adjustment expenses, underwriting expenses and investment-related items through specified statutory exhibits. Such statements permit line and entity comparisons within defined U.S. regulatory reporting boundaries. They are also shaped by statutory-accounting objectives, allocation conventions and legal-entity structures, which must be documented before estimating a cross-national cost function. The NAIC explanation of loss adjustment expense is particularly useful for distinguishing claim-servicing resources from paid indemnities.
An actuarial loss ratio, expense ratio or combined ratio can describe underwriting experience, but none alone measures the minimum feasible operating input cost for a risk-adjusted service vector. An insurer may have an excellent combined ratio because of favorable claims experience or higher pricing. A different firm may have a high combined ratio after a catastrophe while operating an efficient claims organization. The statistical relation between accounting profitability and production efficiency therefore requires explicit modeling.
Solvency constraints and risk capacity
The IAIS Insurance Core Principles and ComFrame describe a global insurance supervision framework whose principles are implemented through national legal structures. Capital requirements influence the cost and feasible quantity of insured risk. Regulatory capital, economic capital and market financing costs are connected but different quantities. An empirical comparison should record the applicable solvency regime, risk weights or internal-model conventions, availability of reinsurance and group support, and the insurer's actual claim-paying obligations.
The IAIS ICP 17 capital-adequacy provisions address the ability of an insurer to meet insurance obligations through a total-balance-sheet supervisory approach. That normative objective clarifies why a production study should not automatically label prudently held capital as avoidable waste. The scientifically relevant question is the minimum cost of producing a defined quantity and quality of insurance protection subject to satisfactory financial security.
OECD 2026 market evidence and contextual limits
The OECD's Global Insurance Market Trends 2026, released November 2, 2026, reports 2025 insurance-market information from more than fifty OECD and non-OECD jurisdictions. The report discusses premium growth, claims payments, investment performance, profitability and changing climate, cyber and geopolitical risks. It provides international market context. The report does not estimate an insurer-level multi-output cost frontier, and its cross-country premiums and payments must not be relabeled as firm production-efficiency scores.
Its discussion of non-life premium growth and underwriting performance indicates that pricing adjustments responded to increased claims costs in many markets. This relationship supports the need to separate nominal revenue growth from improvements in real production. Similarly, claims growth may reflect weather, insured values, claims inflation or portfolio mix. National aggregates can illuminate the operating environment while remaining insufficient to identify an individual insurer's minimum feasible cost function.
Comparison across institutional forms
A regulatory comparison must specify jurisdiction, reporting date, entity perimeter and applicable insurance classes. IFRS financial statements, NAIC statutory returns, national solvency records and OECD aggregate statistics serve related but distinct purposes. A coherent production study builds a measurement dictionary and reconciles business definitions before combining observations. Standardization by simple currency conversion or premium ratios does not resolve differences in contract scope, claim settlement rights, capital regulation and national accounting.
Original Analytical Synthesis
The joint unit of insurance production
The central analytical conclusion is that the insurer's economically meaningful unit of output is a risk- and quality-specified contingent service delivered over time. Its observable manifestations include contracts written, protection in force, claims adjusted, indemnities funded, and intermediation services supplied. Every single-volume indicator captures part of this activity, while a complete empirical production vector should represent the dimensions that determine resource requirements. This conclusion follows from the original frontier literature, claims classification, and insurance accounting, rather than from a newly proposed production theory.
This conclusion has an immediate consequence for empirical research. An apparent improvement in cost per policy may be caused by a change in insured sums, risk mix, legal responsibilities or claim frequency. A decrease in cost per premium dollar may be caused by higher premiums. An increase in claim settlement costs may accompany faster or more accurate contract fulfillment. The researcher needs to identify the service quality and risk obligations behind each denominator before treating the ratio as a measure of technical progress.
The distinction between operating technology and risk-financing cost
A second conclusion concerns the boundary of cost. Operating expenditure finances underwriting and administrative activities; expected benefit payments fulfill the contingent promise; capital supports the credibility of future payment; reinsurance reorganizes the financial distribution of risk. Their combined private economic burden is important for pricing, yet the costs are analytically separable and may have different accounting treatments. A frontier built from wages, other administrative inputs and claim-processing costs estimates one production boundary. It should not be advertised as minimizing all future policyholder benefit payments or the social economic burden of natural catastrophes.
In this interpretation, the insurer's managerial efficiency cannot be established by the ratio of its indemnities to premiums without an explicit assessment of policyholder benefits, contract terms and risk selection. Indemnities are a delivered financial transfer in the insured state; their expected value is also an essential private liability. The same cash flow can thus be evaluated from different, clearly specified economic perspectives.
Scale depends on the retained risk constraint
A third conclusion follows from capital duality: economies of scale in policy administration and underwriting systems can coexist with increasing marginal capital needs for concentrated catastrophe exposure or correlated liabilities. A firm that doubles its policy count within diversified risk pools may save operating resources; doubling a concentrated exposure can exhaust reinsurance or capital capacity. Therefore an empirical economy-of-scale result has a local interpretation under the risk and capital conditions included in its production set. The joint observation of lower expense per policy and higher portfolio concentration should be interpreted as a change in the operating–risk mix, rather than as unconditional productivity gain.
Performance combines realized efficiency and organizational design
A fourth conclusion separates feasible joint-production savings from managerial achievement. A large insurer may benefit from shared technology yet exhibit high estimated X-inefficiency, a pattern compatible with the Fenn et al. findings. A diversified insurer may share inputs efficiently while facing governance, distribution or risk-financing costs that exceed the savings. Corporate form and size identify institutional choices; they do not directly measure the quality of daily operational decisions. Empirical estimates should report scale, scope, and estimated efficiency components separately.
An explicitly hypothetical arithmetic illustration
Consider two hypothetical insurers, A and B, each reporting annual premiums of $100 million. A reports expected claims of $65 million, administrative and claim-handling operating expense of $16 million, and annualized net capital and reinsurance cost of $8 million. B reports expected claims of $73 million, operating expense of $12 million, and annualized net capital and reinsurance cost of $10 million. On operating expense divided by premiums, B appears less expensive (12 percent versus 16 percent). On the illustrative total private cost sum, A has $89 million ($65+$16+$8) and B $95 million ($73+$12+$10). This numerical exercise is a transparent construction, not observed data and not a general proof that A is economically more efficient: policy quality, insured risk, claims obligations, financing costs and pricing could all differ. It demonstrates why minimizing one expense component can fail to minimize the full cost of equivalent protection.
The example also highlights a required accounting discipline. Capital and reinsurance costs must be defined without double-counting claim benefits, investment income or expected recoveries. The illustration assumes each category is additive and consistently measured; actual accounting and economic models must establish those conditions rather than presume them.
A research-ready interpretation of international comparisons
The combined literature supports a comparison program with four jointly reported objects: the insurance service-output definition; real operating input requirements; risk-bearing and financial-security constraints; and the identification assumptions governing the estimated frontier. These are descriptive requirements for interpreting established methods, not a newly named conceptual model or doctrine. They make the research falsifiable: changing the output vector, capital treatment or environmental controls should reveal whether a scale or efficiency finding survives reasonable alternatives.
Research Implications and Open Empirical Questions
An empirical research design
A contemporary international study should begin with a line-specific, risk-adjusted panel of legal entities and, separately, consolidated groups. For each observation, the dataset should distinguish earned exposures, insured values, contract limits and deductibles, claims handled, paid and incurred claims, settlement complexity, policy duration, capital resources, reinsurance, labor and service expenses, other input prices, and performance-quality indicators. The specifications should document the origin of each variable, especially accounting conversions after IFRS 17 adoption and statutory-reporting differences. Merged and exited entities should be tracked to address selection and survivorship.
Estimation can compare a well-constrained translog stochastic cost frontier with alternative functional forms, environmental controls and a suitably specified DEA robustness exercise. Scale and scope conclusions should be reported at stated output levels, with risk and capital constraints shown. Standard errors, sensitivity to catastrophe years, alternative outputs and input prices, firm-group perimeter, and technical quality indicators should accompany substantive claims. Where credible identifying variation exists, empirical work can estimate the effect of new technology, consolidation or regulation on resource use. Where it does not, the analysis should identify descriptive patterns and refrain from asserting causal improvements.
Regulation, competition, and consumer protection
For supervisors and policymakers, the efficiency question is whether insurers can supply reliable, appropriately priced protection using reasonable resources while satisfying claims-paying obligations. Scale can reduce costs but may also concentrate market power or common risk. Outsourcing can improve measured internal efficiency but transfer costs and operational dependencies outside the insurer. Digital automation can lower handling expenses but require continuing model governance, security and fair claims treatment. These effects call for evidence on both the cost function and the security and quality of the insurance service.
The IAIS supervisory principles make financial capacity and governance relevant constraints, while the IFRS 17 reporting framework provides one important set of reporting distinctions for applicable insurers. The best comparative research uses these frameworks for their defined functions and avoids conflating legal compliance with minimum economic cost.
Open research directions
Several substantive questions remain suitable for original empirical investigation. They include measuring the effect of automated claim settlement on both real handling resources and customer outcomes; quantifying the marginal capital requirements of geographically correlated new business; estimating whether modern cloud-based policy systems change the minimum efficient scale; identifying when broker distribution creates service value that offsets acquisition expense; comparing mutual and stock insurers with matched product and risk portfolios; and distinguishing true cost savings from outsourcing shifts and accounting-reclassification effects. Each requires actual data, clear treatment assignment or justified econometric identifying variation, and transparent uncertainty.
Further work could examine the interaction of rising catastrophe concentration, reinsurance availability, risk modeling and insurer cost frontiers as an intertemporal process. The OECD's 2026 overview provides contemporary macro-sector context, while insurer-level research would be required to identify the production response. The scientific value of these questions lies in resolving empirically distinct mechanisms with primary observations.
Conclusion
Insurance firm production consists of interrelated underwriting, contractual administration, loss settlement, risk-bearing and financial intermediation services. The quantity and quality of contingent protection must be specified before a cost function can be meaningfully estimated. Premium volume, realized indemnity and policy counts supply useful observations, but they combine prices, risk composition, claims experience and service units in different ways. The distinction between an insurance output and an input or financial transfer is essential to the interpretation of every estimated efficiency measure.
Production duality supplies a rigorous language for cost minimization, marginal cost, scale and scope; translog and related flexible forms support empirical approximation; stochastic and nonparametric frontiers provide different relative-efficiency measures. The published literature documents heterogeneous economies of scale and scope, the role of distribution and organizational form, and material cross-country and methodological variation. Its findings are strongest when read alongside their actual samples, output definitions, risk environments and estimation choices.
Economic analysis further requires the separation of operating resource costs from promised indemnities and the financial resources needed to honor them. Claims quality, reinsurance, solvency and regulatory requirements belong to the feasible production problem. International comparison requires explicit reconciliation of IFRS, statutory and supervisory quantities. The original synthesis of this monograph therefore identifies coherent conditions for interpreting established production-cost research and an agenda for testing them in contemporary insurer data. The work does not claim a new theory, independent empirical estimation or universal numerical efficiency rule.
References
Aigner, D., Lovell, C. A. K., & Schmidt, P. (1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics, 6(1), 21–37. Publisher DOI.
Berger, A. N., Cummins, J. D., & Weiss, M. A. (1997). The coexistence of multiple distribution systems for financial services: The case of property-liability insurance. The Journal of Business, 70(4), 515–546. Institutional bibliographic record.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2(6), 429–444. Publisher DOI.
Cummins, J. D., & Weiss, M. A. (1993). Measuring cost efficiency in the property-liability insurance industry. Journal of Banking & Finance, 17(2–3), 463–481. Publisher DOI.
Cummins, J. D., & Weiss, M. A. (1998). Analyzing firm performance in the insurance industry using frontier efficiency methods (Working Paper 98-22). Wharton Financial Institutions Center. Working paper record.
Cummins, J. D., & Weiss, M. A. (2013). Analyzing firm performance in the insurance industry using frontier efficiency and productivity methods. In G. Dionne (Ed.), Handbook of Insurance (2nd ed.). Springer. Publisher chapter DOI.
Cummins, J. D., & Xie, X. (2008). Mergers and acquisitions in the US property-liability insurance industry: Productivity and efficiency effects. Journal of Banking & Finance, 32(1), 30–55. Publisher DOI.
Cummins, J. D., & Zi, H. (1998). Comparison of frontier efficiency methods: An application to the U.S. life insurance industry. Journal of Productivity Analysis, 10(2), 131–152. Publisher DOI.
Cummins, J. D., Dionne, G., Gagné, R., & Nouira, A. (2009). Efficiency of insurance firms with endogenous risk management and financial intermediation activities. Journal of Productivity Analysis, 32, 145–159. Publisher DOI.
Cummins, J. D., Tennyson, S., & Weiss, M. A. (1999). Consolidation and efficiency in the US life insurance industry. Journal of Banking & Finance, 23(2–4), 325–357. Publisher DOI.
Cummins, J. D., Weiss, M. A., & Zi, H. (1999). Organizational form and efficiency: The coexistence of stock and mutual property-liability insurers. Management Science, 45(9), 1254–1269. Publisher DOI.
Cummins, J. D., Weiss, M. A., Xie, X., & Zi, H. (2010). Economies of scope in financial services: A DEA efficiency analysis of the US insurance industry. Journal of Banking & Finance, 34(7), 1525–1539. Publisher DOI.
Eling, M., & Luhnen, M. (2010). Efficiency in the international insurance industry: A cross-country comparison. Journal of Banking & Finance, 34(7), 1497–1509. Publisher DOI.
Eling, M., & Schaper, P. (2017). Under pressure: How the business environment affects productivity and efficiency of European life insurance companies. European Journal of Operational Research, 258(3), 1082–1094. Publisher DOI.
Fenn, P., Vencappa, D., Diacon, S., Klumpes, P., & O'Brien, C. (2008). Market structure and the efficiency of European insurance companies: A stochastic frontier analysis. Journal of Banking & Finance, 32(1), 86–100. Publisher DOI.
IFRS Foundation. (n.d.). IFRS 17: Key terms and implementation materials. Official guidance.
International Accounting Standards Board. (2017, as amended). IFRS 17 Insurance Contracts (effective for annual periods beginning on or after January 1, 2023). IFRS Foundation. Official standard.
International Association of Insurance Supervisors. (2024). Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups (ComFrame), December 2024 compilation. Official PDF.
International Association of Insurance Supervisors. (n.d.). Insurance Core Principles and ComFrame. Official standards directory.
Kumbhakar, S. C., & Lovell, C. A. K. (2000). Stochastic Frontier Analysis. Cambridge University Press. Publisher DOI.
National Association of Insurance Commissioners. (2025). Property/Casualty Annual Statement Blank. Official reporting form.
National Association of Insurance Commissioners. (n.d.). Glossary of Insurance Terms: Loss adjustment expense. Official glossary.
Organisation for Economic Co-operation and Development. (2026). Global Insurance Market Trends 2026. OECD Publishing, November 2, 2026. Official report.
Organisation for Economic Co-operation and Development. (2026). Global Insurance Market Trends 2026: Premium growth and underwriting performance in the non-life insurance sector. OECD Publishing. Official report section.
Rubio-Misas, M. (2022). Analysis of insurers’ performance using frontier efficiency and productivity methods: The great contributions by David Cummins and Mary Weiss. Risk Management and Insurance Review, 25(4), 445–489. Publisher DOI.
Toivanen, O. (1997). Economies of scale and scope in the Finnish non-life insurance industry. Journal of Banking & Finance, 21(6), 759–779. Publisher DOI.