Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Insurance Undertaking Counts and Active License Classification

Abstract. This monograph examines the statistical and regulatory meaning of insurance undertaking counts across nationally differentiated insurance markets. It distinguishes authorized legal entities, branches, ownership groups, class-specific permissions, observed writers and institutions administering outstanding obligations. The research combines international supervisory and statistical instruments with selected primary regulatory sources from Australia, the United Kingdom, Hong Kong, the European Economic Area and the United States, alongside Singaporean category descriptions and limited academic literature. A dated Australian general insurer register yields 79 new-or-renewal permission entries and 10 run-off entries, including a repeated displayed name; the exclusion of nine distinct run-off labels changes the combined name-label count by approximately 10.23 percent. Hong Kong's reported 157 authorized insurers and its authorization-event appendix demonstrate the need to separate stock and flow. Standard stock-flow identities, classification bounds and concentration calculations clarify the consequences of alternative counting bases. The study establishes a comparative interpretation and reproducible count logic within the available evidence, without asserting a global census, causal estimate or novel theoretical system.

Keywords: insurance undertakings; insurance licensing; active insurers; foreign branches; insurance run-off; regulatory registers; insurance market statistics; legal entities; market concentration; international comparability

Introduction

Insurance undertaking counts give institutional scale a numerical expression. A jurisdiction with a hundred authorized carriers appears to have a larger population of insurance institutions than a jurisdiction with fifty. Yet the apparent simplicity of the statistic conceals the legal instrument that confers authority, the organizational unit to which authority attaches, the geographical extent of permission, and the economic activity taking place under that permission. Every meaningful count depends upon all four.

This monograph investigates the relationship between regulatory authorization and the statistical population of insurers. Its central question is how insurance markets can be compared when legal entity registration, permission to write or renew contracts, cross-border operation, financial reporting, and continuing responsibility for established contracts define partially overlapping populations. The study treats an insurer as a bearer of contractual obligations within an identified regulatory perimeter and treats authorization as a dated and scoped legal attribute.

The international framework is established by the IAIS (2024) Insurance Core Principles, while the OECD statistical instructions supply a directly relevant reporting convention. The research uses publicly accessible regulatory records from Australia, the United Kingdom, Hong Kong, and the United States, alongside EEA statistical reporting and illustrative Singaporean licensing distinctions. These are purposively chosen institutional cases. They exhibit differences in branch treatment, run-off, financial reporting and category overlap; they do not constitute a random sample of countries. The analysis is comparative, documentary and quantitative where the published record supports recomputation. It does not estimate the total number of insurers in the world.

The inquiry is organized around active insurers and branches, scope and status of licenses, and exclusion of inactive institutions. The empirical Australian example demonstrates the effect of register structure and duplicate labels. The Hong Kong example illustrates the distinction between authorization changes and changes in names. Standard statistical identities and deliberately synthetic counterexamples show how classification choices alter competition and market structure measures. The instruments presented in different jurisdictions are studied at their reported dates, with no assertion that the same national legal effects apply everywhere.

The scholarly contribution is an integrated interpretation of the measurement problem, including explicit denominators, event accounting, and classification sensitivity. Established statistical techniques are used as tools. No priority claim or new named theory is advanced.

Concepts, Institutions and Historical Development

The insured obligation and the institutional unit

An insurance contract creates a conditional financial obligation whose performance depends on events, coverage terms and contractual limitations. Statistical systems study the institution that undertakes that obligation. A licensed insurer can issue many contracts, use multiple distributors, operate through numerous brands and maintain branch offices in several territories. Each of those dimensions has an independent counting unit.

A legal entity is a unit capable of bearing legal rights and obligations under its governing law. An undertaking may encompass a corporate or other permitted legal form. A branch is ordinarily an establishment of a legal entity in a territory. A subsidiary is a separate entity controlled by a parent. A group is an economic or supervisory perimeter embracing several linked entities. An agency may designate an intermediary or an establishment depending on the source's own terminology. Consequently, an analyst must retain both the local classification and a mapped cross-jurisdiction category rather than treating a name containing “agency” as decisive.

A license is a state of permission associated with a legally identifiable subject and defined activities. The same subject may hold several permissions, territorial registrations or class-specific approvals. Counting permission records answers a question about the scope of regulatory authorization; counting unique underlying entities answers a question about the number of institutions. A multi-line insurer can occupy multiple product markets without multiplying the total number of legal carriers.

The OECD reporting questionnaire is especially useful because it separates a country's domestic undertakings from local establishments of foreign undertakings and records foreign-controlled domestic undertakings as a subset. A locally incorporated subsidiary of a foreign group is therefore counted among domestic undertakings under that scheme. Foreign control and foreign incorporation are analytically distinct facts.

Legal authorization and economic operation

The IAIS (2024) ICP 4 on licensing requires supervisory attention to the legal entity, authorized scope and public disclosure. The standard gives a point of reference for comparison across different bodies of national law. A license shows that an institution is legally eligible for specified activity subject to its conditions; actual operation depends on additional commercial, financial and organizational choices.

For a specific reference date, define an authorized stock as the set of eligible undertakings within the stated jurisdiction and coverage. Define the new-business stock using the right to enter into new contracts of the specified class. A renewal-only category, where applicable, is retained independently. Define an observed production population by evidence of new policies or other suitable underwriting activity during a stated period. A liabilities population comprises entities that continue to bear the relevant insurance obligations, including entities that no longer originate new business. These sets are purpose-specific and may have different time dimensions.

The word active is therefore meaningful only with a qualifier. Active authorization describes permission at a date. Active production describes measured contract origination during a period. Active servicing refers to continuing performance of contracts. Broad legal or supervisory presence describes the regulator's applicable institutional perimeter. A commercial market analysis generally concentrates on providers eligible for and participating in the relevant transactions. A prudential analysis must retain entities responsible for outstanding claims even when their new-business status has ceased.

The development of international measurement

National supervision grew around legal authorization and territorial obligations. Cross-border insurance developed through locally incorporated carriers, branches and services supplied without local establishment. Group supervision addressed relationships across those institutions, while statistical systems increasingly linked entity-level returns to national economy and sector accounts.

The OECD insurance company dataset organizes insurer counts by business and ownership-related dimensions. Its descriptive categories offer a harmonization vocabulary, but category labels do not substitute for national reporting notes. The EEA insurance statistics portal separately presents solo and group-based aggregates. These sources illustrate distinct institutional purposes: one assembles international market statistics, the other provides harmonized supervisory-reporting aggregates within a defined geographical and regulatory domain.

Historical comparisons must account for changes in law, thresholds, regulatory scope, reporting forms and institutional structure. A new government classification may create a discontinuity without a new insurer. A reorganization may dissolve one entity while continuously serving policyholders through another. A merger that reduces entity counts can leave market shares largely unchanged if the entities were already under common control. Conversely, an ownership change can alter competitive independence without changing the number of authorized legal persons.

The measurement task involves two forms of continuity. Identity continuity follows the legal subject across renaming, changes of address and group affiliation. Risk continuity follows the outstanding obligations across transfers, restructuring and insolvency. These continuities can separate: the same entity can dispose of its book, and a transferred book can retain economic continuity after the originator exits.

Insurance boundaries and other licensed participants

An intermediary commonly arranges insurance on behalf of customers or carriers; the risk-bearing institution ultimately bears the contractual obligations. Insurer statistics and intermediary statistics should therefore be reported separately even when a register includes both. Captives and reinsurers belong in an appropriately defined risk-bearing universe, while their relevance to retail market access depends on the type and scope of business.

Singapore's MAS insurance entity classification distinguishes direct insurers, reinsurers and captive insurers in the indexed regulator description. The same description distinguishes representative offices from insurance-writing institutions. This evidence is sufficient to motivate separate categories but does not establish the count or complete national conditions for each.

Special-purpose insurers, mutuals, public schemes and statutory social insurance require explicit perimeter decisions. An insurer dedicated to a capital-markets risk transfer may be legally authorized yet have no retail customer-facing activity. A mutual insurer may be economically significant despite a different organizational form from a stock company. The OECD questionnaire excludes statutory social security administered by the state while covering authorized insurance undertakings, including professional reinsurers. This is a reporting convention for that exercise, not a universal definition of every risk-financing institution.

Time and the unit of observation

A register snapshot is a stock at a reference date. A new authorization during a fiscal year is a flow. A financial return describes a measurement period and may remain relevant after a license change. A historical database may retain inactive records. The observed date, the real-world effective date and the source publication date are different fields.

For legal permission, a record must specify the interval during which permission applies. For financial values, a record must specify the period, unit and reporting boundary. Calendar year and fiscal year reconciliation is necessary only where periods actually match or a defensible conversion can be constructed. Entity counts require no currency conversion, whereas premiums used in a cross-country size comparison require transparent currency, inflation and valuation choices.

The fundamental relation is simple: each observation needs a defined unit, a population rule and a reference time. Without them, numerical precision is merely typographical precision.

Primary Market Sources and Comparative Research Design

Authority and source hierarchy

This study prioritizes primary supervisory registers, source methodologies, statutory provisions and official reports. Each answers a distinct question. The licensing instrument establishes legal authority; the register supplies evidence of recognized institutions and status; the financial return records measured business; the statistical publication aggregates observations under its own methodological rules. Conflicts are resolved by identifying the question the source was constructed to answer.

The source population was selected to expose institutional differences rather than to produce an international ranking. The OECD general instructions were examined through extracted text and an Annex I page image. The accessible version contains a tension between the document code in extracted text and the visually retrieved page, so the claims rely only on definitions common to the retrieved passages. The OECD Data Explorer was consulted at the dataset-description level, without downloading the underlying SDMX observations.

The IAIS December 2024 standard was examined at relevant licensing and exit provisions, including a rendered page on cross-border supervision. The APRA register and the Bank of England regulatory lists were retrieved as official HTML. The Hong Kong annual report was consulted through selected extracted pages, including the authorization-change table; rendering of the relevant PDF page failed. The EIOPA portal and NAIC repository description were read as official reporting documentation. Where direct retrieval returned a challenge or maintenance page, indexed original-source text is used only for expressly supported narrow propositions and its access status is recorded in the evidence ledger.

The academic literature is used at the abstract and bibliographic level where full text is inaccessible. It informs alternative interpretations of market structure rather than supplying unverified numerical estimates. A complete census of national records and longitudinal microdata are outside this edition's achieved evidence coverage.

Comparative case matrix

Source system Unit reported or listed Appropriate use Principal caveat
OECD questionnaire Domestic undertakings and foreign branches/agencies National reporting perimeter National deviations; foreign-controlled subset
EIOPA statistical portal Solo undertakings and groups Supervisory aggregates Reporting scope, vintage and provisional figures
APRA general insurer register Named institutions separated by permission status Authorization and run-off distinction Multiple record labels, conditions and separate Lloyd's treatment
PRA regulated firm list Firm records with reference numbers UK supervisory population Includes designated cross-border run-off firms
Hong Kong annual report Authorized insurer stock and authorization events Dated supervisory population and transitions Authorized stock differs from new-business activity
NAIC financial repository Financial reporting insurers Financial surveillance and reporting coverage Filing population differs from state license incidence

Each row refers to a different statistical object. Comparability arises only after a common target definition is chosen and the original coverage is mapped to it. The table deliberately does not rank jurisdictions by their reported totals.

Record-level data specification

A reproducible entity dataset should carry a persistent legal identifier, official name, former names, incorporation jurisdiction, host jurisdiction, supervisory identifier, group affiliation, business classes, permission scope, status start and end dates, source URL, observed date and uncertainty flag. Where a branch has a local identifier and a parent identifier, both are retained. Where the source has only a name, the name is a provisional matching key rather than proof of distinct legal identity.

A relational design separates entities, establishments, permissions, financial returns, group-control intervals and raw source observations. Each table has its appropriate key. This structure supports several different totals without destroying information needed for a later reconciliation. Repeated authorizations appear as rows in the permission table rather than duplicate legal entities.

A permissions table requires an explicit legal jurisdiction even if the institution operates digitally. A company may be authorized to cover a risk located in a country without maintaining an office there. Host-market participation can therefore be broader than physical presence. A source with branches alone cannot reveal all cross-border service providers.

Group identifiers should be dated. A company may change owner during a period, and a group may span several regulatory regimes. To measure market concentration at period end, use ownership at period end and align the business flow accordingly. To examine annual behavior under different owners, use a dated attribution rule and disclose how premium flows are allocated across control changes.

Counting functions and missingness

Let E be the set of distinct identified entities. Let L be the set of scoped licensing records. For jurisdiction j, class c and date t, define a Boolean indicator A(e,j,c,t) of valid authority for the selected activity. The number of authorized entities is the sum of these indicators over E. The number of authorization records can be larger because a single entity can possess multiple records within the selected scope.

For a broader territorial-presence statistic, the unit may instead be an entity-establishment pair, with one or more branches attached to a parent. The resulting figure should be presented separately from a global count of unique risk-bearing entities.

Unknown classification is retained as unknown. Let K denote confirmed eligible units, M confirmed ineligible units, and U unresolved units in an otherwise identified candidate list. For that candidate list alone, the true eligible count lies between K and K+U. These bounds require the candidate list to have complete enumeration for the intended scope. If providers outside the list may be missing, even the upper bound is conditional on that list and cannot be advertised as a market-wide bound.

Missingness has several origins: no public data, withheld values, reporting exemption, unresolved identifiers, late filings and logical nonapplicability. Zero signifies a measured absence. Confidential and unavailable values do not signify zero. A source that publishes no branch count cannot be used to infer there were no foreign branches.

Research design and comparability tests

First, define the target population and the reference date. Second, preserve the original source and its definitions. Third, map records into comparable categories with a traceable justification. Fourth, resolve identity and effective dates. Fifth, evaluate counts under alternative inclusion rules. Sixth, validate summed components against source totals and explain any residual. These are standard research operations collected here to make the comparison auditable.

The EIOPA statistics portal distinguishes separate reporting cycles and provisional releases. A quantitative panel should therefore preserve vintage rather than overwrite earlier values. A revised series may be statistically superior for retrospective analysis but should not be used as if an analyst knew the revision at the original decision date.

Market size is a separate explanatory variable. Counts per million residents can support descriptive comparisons where population is appropriate to a locally oriented market. Counts per unit of premium can indicate the number of carriers relative to measured business volume. A reinsurance hub serving international cedants may have large premiums generated outside its resident population, so a per-capita interpretation requires particular care. Currency conversion is required for cross-country monetary ratios expressed in a common currency, not for the count itself. For real growth, the deflator and base year must be specified.

Matching the numerator and denominator is essential. If the numerator covers direct insurance carriers only, a denominator that includes gross reinsurance accepted can produce a structurally misleading average. If the numerator includes run-off entities while the denominator is new-business premium, the ratio measures average production across both originators and non-originators. Label it accordingly or restrict the numerator.

Validation and reproducibility

The Australian example below uses a directly observed HTML register section and reproduces the row-count operation with visible range boundaries. The Hong Kong example uses regulator-reported counts and event categories. These are limited, transparent observations, not a longitudinal panel. The sensitivity illustrations use synthetic data and are identified as such.

A repeatable study should archive the register bytes or a permitted snapshot, record a checksum, and retain normalized rows and the exact transformation code. The current edition records source locations and the extracted count logic, but the original full register bytes were not independently archived. This limits snapshot reproducibility if a live page changes. The research record therefore distinguishes source observation from a future independently repeatable extraction.

Active Insurers and Branches

Domestic and foreign forms of presence

The OECD undertaking classification distinguishes the local institutional base from foreign establishments. Let D be the number of domestic undertakings, F the foreign-controlled subset of D, and B the count of branches or agencies of foreign undertakings within the source's definition. Then the national total within this convention is D+B. Domestic-controlled domestic undertakings are D−F when the control classification is complete. Adding F to D+B counts foreign-controlled domestic undertakings twice.

Consider an illustrative country with 60 locally incorporated insurers, including 18 foreign-controlled insurers, and 12 foreign branches. The reported total is 72. A mistaken sum of 60+18+12 gives 90, an overcount of 18 or 25 percent relative to 72. The foreign-controlled proportion of local undertakings is 30 percent; the share of foreign branches in the reported total is one-sixth. Neither ratio alone measures the proportion of the market controlled by overseas groups, which also depends on premium weights and the characterization of foreign-controlled local institutions.

The distinction matters in market integration research. A branch registration may signify an additional host-market presence without any increase in global legal-entity count. Incorporating a subsidiary creates a distinct legal entity, but a subsidiary and branch can be controlled by the same group. These two routes to territorial operation may result in different capital allocation and legal arrangements depending on local law.

Home and host perspectives

Home supervisors focus on regulated institutions domiciled or otherwise supervised within their authority. Host supervisors care about business affecting policyholders in their market and the legal arrangements under which it is supplied. The IAIS (2024) cross-border licensing provisions address consultation between home and host supervisors for foreign operations and cross-border services.

A home-country entity total is useful for mapping the legal population of a country's insurance sector. A host-market presence total is useful for examining local access or regulatory exposure. Summing host populations across countries counts an internationally operating insurer in several markets; it therefore estimates participations across jurisdictions and not the global number of unique entities.

Consider two countries, with five and four local insurance entities. Suppose two from the first country maintain a presence in the second, while one from the second maintains a presence in the first. The territorial participant counts are six and six, totaling twelve participations. The underlying legal entities number nine. All figures are correct within their own units. A global aggregator must state whether it is aggregating twelve participations or nine legal persons.

Australia: register partition and activity status

The APRA general insurer register, updated September 23, 2026, separates institutions authorized for new or renewal insurance from institutions with run-off-only authorization and an archive of revocations. The first section has 79 named rows in the retrieved representation, from its first entry to the last immediately preceding the run-off heading. The run-off section has 10 rows. Two rows have the identical displayed entity name, Commonwealth Steamship Insurance Company Pty Ltd, with different commencement years. The remaining eight names are distinct.

This yields the following audit of the rendered list:

Scope Row count Distinct displayed names
New or renewal business section 79 79
Run-off-only section 10 9
Current sections combined 89 88
Historic revocation archive Excluded Excluded

The two currently listed sections have no observed identical names across the sections. Deduplicating the repeated run-off label produces 88 distinct name labels. The legal identity or separate historical authorization conditions behind the repeated name were not independently resolved. A legal-entity count is therefore held distinct from the deduplicated label count.

On the label-count convention, excluding run-off records changes the total from 88 to 79, a reduction of 9/88, approximately 10.23 percent. Conversely, treating all current displayed names as eligible for new or renewal business would inflate the restricted count by 9/79, approximately 11.39 percent. Using rows without deduplication would yield 89, an additional one-row distortion. These are calculations on the retrieved listings, not an official APRA aggregate and not a count of observed writers.

The eligibility interpretation remains scoped: an insurer included in the first section is permitted to conduct new or renewal business, subject to its own restrictions. The section does not prove that every listed institution wrote contracts during a period or is available for every product class. Separating professional reinsurers and specialized risk carriers would require further record-level evidence.

The revocation archive is not part of either current-authorization stock. APRA describes the continuation of some former brands through intermediary roles and lists historical withdrawals separately. This is an institutional reason to preserve identifiers: the disappearance of a licensed carrier may coexist with continued public use of its brand.

Lloyd's and alternative placement channels

The Australian register separately explains Lloyd's underwriting organization. The appearance of a distinct market organization poses a counting choice: society, syndicate, underwriting member or relevant licensed establishment may be appropriate under a particular legal and statistical convention. One should not append a syndicate count to a list of ordinary legal carriers without defining a common unit.

APRA's intermediated insurance statistics include business placed with authorized general insurers, Lloyd's underwriters and unauthorized foreign insurers. The term unauthorized here is the regulator's reporting terminology, and its appearance in this collection does not establish that the particular placement is unlawful. It establishes that a placement dataset and a licensed-institution list have different boundaries.

For market-access research, legally available alternative channels can be economically important even when the domestic authorization census is unchanged. For policyholder protection research, the relevant legal treatment of each channel must be assessed separately. Aggregating the channels' premiums without clarifying origin and recipient can also duplicate relationships in the risk-transfer chain.

The United Kingdom and the reporting distinction

The Bank of England's November 2026 PRA regulated insurers list includes UK entities and certain foreign-branch firms, including EEA firms under supervised run-off. Firm reference numbers provide a superior matching field to names. Yet an identifier proves only records refer to a particular registered firm; the scope and date of its permissions still need validation.

The legal population of firms carrying out existing insurance contracts may include institutions that cannot initiate new underwriting in the UK. The same firm may remain active in new contracts elsewhere. Consequently, a global flag of active or inactive loses essential jurisdiction-specific information. A status must attach to the intersection of entity, territory, activity and date.

EEA solo and group observations

EIOPA's solo and group statistics describe two forms of supervised financial information. A group may contain multiple insurers, and group consolidated measures generally serve questions different from solo reporting. Summing solo and group aggregates or counting each group alongside its subsidiaries creates an incoherent population. Group mapping is appropriate when economic control is the unit, whereas entity mapping remains essential for contractual liability and local supervision.

Cross-country comparisons should specify whether independently governed competitors or distinct risk-bearing legal persons are counted. A group controlling three licensed entities could offer several brands and products; the entities remain distinct legal units, while common control may make them a single competitor for a particular competition analysis. Actual economic independence also depends on market definition, agreements and decision rights, so common ownership is a starting point rather than the complete competitive inquiry.

Licensing Status

Authorization as a scoped state

A useful classification stores permission as a multi-dimensional record rather than a single yes/no indicator. It identifies activity, geography, relevant classes, effective dates, conditions and the institution covered. A license may allow direct insurance, reinsurance or captive business; a class-specific permission may prohibit other business lines; an authorization may permit existing contract performance without fresh underwriting.

The IAIS (2024) license scope requirements support public visibility into such differences. In the EEA, the indexed text of Solvency II Article 14 establishes prior authorization for the taking-up of the covered insurance or reinsurance business. The January 2025 consolidation of Solvency II Article 144 distinguishes withdrawal or lapse from new operations in other Member States. This edition does not make a detailed assertion about the transposition of every provision into every member country's law.

Operational classification

Analytical status Evidence criterion Origination population Outstanding-obligations population
Valid new-business authority Effective permission covers the activity Eligible Include if obligations exist
Valid renewal-only authority Permission covers renewal but not new origination Separate Include if obligations exist
Restricted or conditional authority Operative conditions reviewed by class Case-specific Include if obligations exist
Run-off or servicing-only Permission limited to existing obligations Exclude Include if obligations exist
Withdrawn or lapsed Effective withdrawal or lapse Exclude Include only where residual obligations remain
Pending application No effective permission yet demonstrated Exclude Case-specific prior status
Unknown or disputed Authority cannot be validated Unresolved Unresolved

This table is an analytical crosswalk, not a statutory taxonomy asserted to exist identically in all jurisdictions. Its rows can be generated from legal instruments and a register's status codes. The last column concerns genuine obligations, which require financial or legal evidence. An entity's appearance in a run-off list supports a service-permission classification but does not quantify residual liability.

A firm permitted to renew may differ from a firm permitted to originate new policies, depending on the regulatory instruments and contract mechanics. A firm with new-business authority can choose not to market the relevant line. An entity in the process of seeking authorization can have staff, capital and a brand without the right to issue business. These states belong in different analyses.

UK contractual continuity

The Bank of England's Financial Services Contracts Regime differentiates supervised run-off and contractual run-off. Both allow defined activities associated with existing contracts, while their authorization and supervisory character differ. The regime arose in the context of UK withdrawal from the EU and former passporting arrangements. The operational purpose was continuity or orderly termination of established contractual relationships rather than an unrestricted right to originate new business.

This makes the license indicator particularly informative. A firm may retain rights essential to administering insurance even after the termination of its previous new-business arrangements. A register that includes deemed authorization can therefore expand or preserve its regulated population while the market-access population shrinks.

For empirical research, such an event is a classification break and a potential behavioral intervention. To identify an effect on competition or coverage, an analyst would need dates of changes, relevant market shares, a comparison group, information about anticipatory actions and a strategy to isolate the change from contemporaneous shocks. The documentary distinction by itself does not estimate those effects.

Incentives, entry, and institutional choice

Authorization involves economic costs: capital, arrangements for governance and compliance, business development and supervisory procedures. A firm may find it efficient to broaden an existing permission rather than establish a new entity. A group may centralize underwriting in one carrier and distribute through several brands. These organizational decisions affect the count without necessarily changing the underlying risk supplied in proportion.

The direction of causal relationships is ambiguous. Strict authorization conditions may discourage marginal entry, but sound conditions can support confidence and durable participation. Larger markets may attract entry while also sustaining large efficient incumbents. Corporate acquisitions can reduce independent ownership while preserving legal carriers. A cross-section with fewer entities and stronger prudential indicators is compatible with several explanations, including scale economies, selection and market maturity.

Choi and Weiss (2005) examine market structure and insurer performance using company and group observations from US property-liability insurers. Their study motivates attention to efficiency as a competing explanation for concentration and performance. Cummins, Rubio-Misas, and Vencappa (2017) study competition and insurer soundness in European life markets using a competition indicator that relates performance to efficiency. These studies motivate the distinction between institutional headcounts and measures of competitive conduct. No full-text coefficient or unverified causal claim is imported.

Status transitions and adjudication

A research dataset should preserve changes as events. A condition can be imposed on a license without any change of entity. A class permission can be added, restricted or relinquished. A company name can change while its license persists. Authorization can terminate while the entity's residual obligations require continuing oversight. A portfolio transfer can relocate financial obligations without creating an originator of new business.

A practical identity adjudication uses the official firm identifier whenever available, confirms dates and names, and retains an audit of the mapping. Where identifiers conflict, no automatic merge is justified. A repeated name may refer to different historical permissions; two different names may refer to a single renamed undertaking. Corporate ownership information is relevant to consolidation but must not by itself cause legal-entity deduplication.

Public supervisory registers may be updated after a legal event becomes effective. A study should retain effective date and observation date. Where the effective date is unavailable, a change detected between snapshots is interval-censored: it occurred sometime within the observed interval, subject to publication delay. Exact day-level event analysis then requires additional evidence.

Financial activity and permission

New written premium, policy inception count and binding authority can answer different questions. Premium can be collected or adjusted under previously written contracts; reinsurance settlements and accounting changes can produce financial values without new retail policies. Conversely, a counterparty may have entered contracts whose accounting recognition or first cash receipt occurs later.

An empirical active-writer criterion should specify the observed activity, period, sign convention and threshold. Positive gross direct written premium may be useful, but the definition of premium must be aligned with the policy of interest. A zero reported premium for one class does not establish exclusion from all classes. A missing filing is not a zero and does not establish inactivity.

Licensing and production data should be joined after their time bases are reconciled. A carrier that wrote policies early in a year and entered run-off later can legitimately appear in both the annual writing population and the closing run-off stock. Treating the annual writer as unauthorized for the entire year would erase actual historical activity. The date rule prevents that error.

Inactive Entity Exclusion

Distinguishing irrelevant records from continuing obligations

An inactive record has no useful universal meaning. For market origination, a carrier that cannot write new contracts is excluded from the origination population. For solvency, claims performance or resolution, the same carrier can be material for years. Exclusion is a population-specific operation, not disposal of source information.

The IAIS (2024) market exit guidance discusses arrangements for insurers leaving the market, including run-off and portfolio transfer. The supervisory purpose is to protect continuing obligations while managing the exit. A count of new-business providers should therefore decrease when permission for that activity terminates, while a liability-based population changes only when the relevant obligation is transferred, discharged or otherwise ceases.

The Australian observations illustrate the difference quantitatively. With 79 new-or-renewal eligible name labels and nine distinct run-off-only labels, the latter make up about one tenth of the combined labels. This does not establish their share of liabilities, premiums or systemic importance. One large long-tail book can be more important to outstanding claims than many small active insurers.

Hong Kong: authorization event accounting

The Hong Kong Insurance Authority's 2024–25 annual report reports 157 authorized insurers in its reporting highlights and lists events for April 1, 2024 through March 31, 2025. Its appendix lists two grants, one for special purpose business and one for general business, and four withdrawals, three for general business and one for long-term business. It separately provides a name-change category.

The observed net contribution of the new-authorization and withdrawal categories is 2−4=−2. This is a balance of those recorded categories; a verified opening count and confirmation of all other reclassification events would be required to assert exact stock reconciliation. If the listed grants and withdrawals are the only population-changing events, the implied opening stock is 159. That is a conditional calculation, rather than an independently observed initial count.

The regulator's indexed March 2025 market description reports 84 general insurers, 51 long-term insurers, 19 composite insurers and three special-purpose insurers at March 31, 2025. The components sum to 157. On that authorization classification, general-only and composite categories together total 103, while long-term-only and composite categories total 70. Adding these two totals and the three special-purpose insurers gives 176 because the 19 composites enter both sides. The categories do not by themselves reveal observed new underwriting or actual customer access.

Observed or calculated item Count Interpretation
Authorized population at reporting date 157 Regulator-reported stock
New authorizations during reporting year 2 Listed events
Withdrawals during reporting year 4 Listed events
Net listed grants less withdrawals −2 Arithmetic contribution
Implied opening stock under closed event balance 159 Conditional, not separately observed

The authorization stock differs from the much larger licensed intermediary population described by the annual report. Adding those figures would conflate carriers with distribution professionals and firms. It would obscure rather than improve a comparison of risk-bearing institutions.

Stock-flow identity and breaks

For a fixed population definition, let N(t) denote the closing count. Let E(t) be entries to that population, X(t) exits, and C(t) net adjustments attributable to scope, classification or identity corrections not already counted as entries or exits. Then N(t)=N(t−1)+E(t)−X(t)+C(t).

The terms must be mutually consistent. For a permission-based population, an entry can be a new license or reactivation, and an exit can be a withdrawal or restriction that removes the relevant right. For a production-based population, an existing authorized firm may become a new observed writer or cease production without a license change. For group-level populations, mergers and disposals may alter control without changing entity identity.

Identity corrections should never be silently folded into entry or exit. A data provider might discover that two records represent one legal entity, reducing the count without any market event. Separate C(t) preserves comparability and enables researchers to reconstruct the original statistical vintage.

Consider a synthetic population of 100 eligible writers. During the year six new entities become eligible, five lose eligibility, and a register cleanup discovers three prior duplicates. Closing count is 98. The true net eligibility movement is +1, whereas the published stock difference is −2. Without the correction term, a researcher would report a contraction in entry dynamics when the recorded transitions imply slight expansion.

Reporting populations and financial coverage

The NAIC (2025) Financial Data Repository describes its reporting universe as representing more than 98 percent of US premiums and over 4,500 individual insurers. This is evidence of extensive coverage for financial surveillance, rather than a census of new-business permissions or a count of state licenses. The Insurance Department Resources Report serves a distinct purpose by providing jurisdictional regulatory information, including insurer counts.

One company may hold authorization in several states and file financial statements under one identity. Counting each state's licenses and summing across the United States measures license incidences unless duplicates are resolved. A national financial reporting count instead measures entities represented in a particular reporting system. Neither can be mechanically substituted for the other.

High premium coverage does not guarantee similarly high entity coverage: a small number of omitted companies can have very small premiums, or many small missing insurers can account for little aggregate volume. It also does not demonstrate coverage of every risk segment. Segment-specific and entity-specific coverage checks are needed before statistical conclusions are generalized.

Concentration and denominator sensitivity

The Herfindahl-Hirschman Index (HHI) is the sum of squared market shares over the defined set of participants. With shares expressed as proportions it ranges up to one. For N equal-share participants it is 1/N. The reciprocal of the HHI gives the corresponding equal-share effective number of firms. It need not equal the legal-entity count.

Take an illustrative market with four entities holding direct-premium shares of 0.40, 0.30, 0.20 and 0.10. The HHI is 0.30 and the effective number of entities is 3.33. Suppose the first and fourth entities belong to the same control group, and the other two are independently controlled. Under a group-level aggregation their shares are 0.50, 0.30 and 0.20. The HHI becomes 0.38, and the effective number becomes approximately 2.63. The distinction results entirely from the counting and aggregation unit.

Now suppose five inactive zero-share records are appended to the first entity-level population. The HHI is unchanged at 0.30 because the records carry zero share, but the nominal count rises from four to nine. Average premium per listed insurer falls from one fourth of total market premium to one ninth, a reduction of 55.56 percent. The artificial decline has no relation to market supply. If the newly included entities hold historical written premium or residual obligations, different denominators and time frames produce different results.

These examples are exact mathematical illustrations, not country estimates. They demonstrate the need to combine license-status populations with measured segment-specific shares. A concentration calculation derived from consolidated revenues cannot be interpreted as the concentration of direct underwriting unless the revenue measure supports that interpretation.

Unknowns and interval sensitivity

Suppose a synthetic register has 80 confirmed eligible carriers, 12 confirmed run-off-only carriers and eight unresolved records. Conditional on the completeness of the register population, origination-eligible count lies between 80 and 88. The unknown classification rate is eight percent. Treating all unknown records as inactive produces the lower bound. Treating them as eligible produces the upper bound.

If another market has a confirmed comparable count of 84 with no unresolved records, a ranking is indeterminate because both sides of 84 lie within the first market's admissible interval. If a regulatory definition or data source adds entrants outside the original candidate universe, the interval itself must be revised.

Administrative data are intended as population records, not probabilistic samples. Classical sampling confidence intervals do not solve classification uncertainty. A sensitivity interval is more appropriate when the principal problem is unknown membership. Statistical confidence intervals remain valuable for actual samples, such as surveys of market entrants, when assumptions about the sampling design are documented.

Empirical limits and identification

The observed results in Australia and Hong Kong demonstrate different aspects of the same measurement problem. APRA provides an explicit status partition and a repeated displayed name. Hong Kong provides an authorization stock and event categories. Neither source supplies a complete harmonized cross-national panel of legal identities, permissions and actual contract production.

A causal analysis of inactive-entity exclusion over time would require repeated observations linked by stable identifiers, sufficient evidence of the status decision, a defined baseline and the relevant business measures. Removal of an entity is itself often endogenous to financial distress, group restructuring or commercial choice. An event study that treats exit as an exogenous shock would need to defend that assumption or use a research design that addresses selection. No causal coefficient is estimated in the present monograph.

Integrated Scholarly Synthesis and Research Frontiers

What undertaking counts establish

A well-defined legal-entity count estimates the size of the population of regulated risk-bearing organizations. A permission count estimates the incidence of legal access in a particular scope. A production count measures observed providers over a period. A group count approximates the number of distinct control units under an explicit grouping rule. Each provides an answer to a scientifically useful but different question.

An international market dataset therefore needs a family of related measures rather than a single ambiguous headline. At minimum, record domestic legal entities, foreign establishments, independent group count where ascertainable, permission-eligible entities by class, observed writers where suitable production data exist, and run-off or residual-liability entities. These are complementary outputs of standard data reconciliation.

A broad stock of authorized entities can be useful in regulatory workload research, although the complexity of supervision depends on entity size, risk and cross-border arrangements. A count of providers holding permission to write the relevant line can inform entry and accessibility studies, but realized competitive alternatives require further evidence. A liabilities population is relevant to prudential and resolution analysis because exit from new underwriting does not automatically extinguish past obligations.

The relationship between counts and market competition

Market competition concerns alternatives available to customers or cedants, the conduct of suppliers and the economic consequences of entry, substitution and exit. Legal permission is a necessary institutional condition in the relevant regulated segment, yet permission alone does not reveal prices, supply, solvency or terms of coverage. A market with more licenses can be dominated by a small number of groups. A market with fewer carriers can have vigorous rivalry if customers can switch among sufficiently independent suppliers.

Measures of HHI, top-firm shares, entry and exit and the response of profitability to efficiency answer different questions. The contrast between company-level and group-level analysis in the insurance economics literature reinforces the need to align unit and interpretation. Choi and Weiss (2005) and Cummins and colleagues (2017) are used here only for their documented research questions and approaches, not for extrapolating numerical results to the present cases.

An empirical researcher should define the product and geography before calculating market structure. Mixing life, non-life, captive and professional reinsurance may be useful for a whole-sector institutional map, but an HHI over that mixture does not necessarily represent a market whose participants compete for the same contracts.

Interpretation of international differences

Differences in published counts can reflect at least five processes: actual entry and exit; changes of scope and legal status; the organization of cross-border operations; corporate consolidation or deconsolidation; and statistical revisions. An explanation that attributes all changes to economic growth or regulation omits these mechanisms.

A jurisdiction with many specialized reinsurers may have a high count relative to population. A jurisdiction with extensive imported insurance services may have a small local incorporation count despite substantial coverage. A market with legacy liabilities may continue to supervise a considerable run-off population. These features describe different institutional contributions to insurance activity and risk transfer; they should not be reduced to simplistic rankings of market development.

A cross-national time series is strongest where the researcher can reconcile original supervisory figures and identify common definitions. If that reconciliation is impossible, present source-native counts and clearly indicate the unresolved dimensions. A qualified comparison is more informative than a synthetic international total without a stable statistical unit.

Governance of longitudinal evidence

Public registers should be preserved as dated snapshots where permitted. The archive should include retrieval timestamp, source version, content fingerprint, transformation scripts and the normalized record schema. Record linkage decisions require provenance and separate status fields. A correction discovered in a later period should be reflected in both a revised history and a reconstruction of the originally published vintage when relevant.

A cross-border identity system should attach jurisdiction-specific permissions to stable legal-entity identifiers and distinguish branches from incorporated subsidiaries. Global identifiers are useful when available but should be cross-checked with national records. Absence of a global identifier cannot be treated as absence of a company.

A statistical publication should make the stock date visible alongside the population definition, record any data-exclusion conditions, and distinguish official totals from analyst-derived estimates. An entity-level classification should be traceable to a dated permission or document, especially where a status is ambiguous or subject to revision.

Research frontiers and remaining gaps

The most valuable extension is a multi-jurisdiction panel linking dated license records to company financial returns and group ownership. It would permit independent analysis of entry, exit, branch strategies, mergers and transitions into run-off. Such a panel would need explicit evidence of product permissions, entity identifiers and reporting vintage; a raw consolidation of register names would be insufficient.

A second extension concerns non-established service providers. Physical establishment is visible in many regulatory lists, while nonbranch cross-border provision may require access to separate notification or reporting systems. Comparison of territorial supply requires accounting for that mode in a defensible manner.

A third extension concerns residual claims and run-off. Matching closure status to outstanding technical provisions and claims settlement would quantify the economic relevance of institutions excluded from new-business counts. Classification uncertainty should be propagated into the reported results, rather than being erased through default coding.

The present edition establishes documentary and analytical foundations and limited recomputations. Its empirical scope does not encompass all OECD member observations, EEA register-level entity matching, a complete independently archived APRA dataset or a long national panel. Full-text academic access remains limited. These are factual boundaries that prevent an assertion of complete empirical coverage or comprehensive independent replication. No worldwide active-license estimate is offered.

Conclusion

Insurance undertaking counts are institutional statistics whose meaning is established by their unit, legal scope, time and observation rule. An authorized legal carrier, a host-market branch, an individual permission, an observed writer, a group and a run-off obligation represent different dimensions of the insurance sector.

The OECD reporting distinction between domestic undertakings and foreign establishments, the IAIS treatment of license scope and exit, APRA's partition of general insurers, the PRA's run-off arrangements, and the distinction between Hong Kong authorizations and intermediary licenses establish concrete evidence of those dimensions. The Australian count demonstrates the sensitivity of a published register to scope and repeated labels. The Hong Kong event table illustrates the requirement to separate authorization movements from changes of name and unverified stock adjustments. Financial reporting coverage, as illustrated by the NAIC repository, measures a further population.

Three consequences follow. First, statistical comparisons must retain legal identity and dated permissions independently of actual business flows. Second, exclusion of inactive institutions depends upon the research question: origination and residual liability require different populations. Third, numerical comparisons involving premium, market share or concentration must align the business measure with the underlying entity or group population. Even exact arithmetic cannot rescue mismatched institutional boundaries.

The original analytical synthesis identifies the consequences of these established distinctions and demonstrates them using a primary register observation, a reported annual authorization stock and illustrative sensitivity calculations. It supplies reusable definitions and a research design without claiming a new theory or empirical validation beyond the material examined.

Reliable undertaking statistics arise from transparent reconciliation of legal authority, economic operation and the institutional obligations that survive changes in business activity. This relationship is essential to the scientific interpretation of the scale, structure and development of insurance markets.

References

  • OECD. (2024/2025 document version; accessed 2026-11-19). OECD Global Insurance Statistics: General Instructions. Annex I, Section 1 and questionnaire Section 1. Publicly served PDF; extracted text and Annex I page image. Source.

  • OECD. (n.d.). Insurance companies and employees. OECD Data Explorer, DSD_INS@DF_NB_COMP. Dataset description indexed. Source.

  • IAIS. (2024). Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups. December 2024 edition, ICP 4 (4.6–4.9), ICP 12 (12.1 and 12.14). Selected extracted provisions and page 38 image. Source.

  • EIOPA. (2026). Insurance statistics. Official statistical portal, update of September 1, 2026 as retrieved. Source.

  • APRA. (2026). List of registered general insurers. Register updated September 23, 2026. Sections on new or renewal business, run-off, revocations and Lloyd's. Source.

  • APRA. (2026). Intermediated general insurance statistics. Official collection page, June 2026 release. Source.

  • Bank of England. (2026). Which firms does the PRA regulate? Insurers we regulate, November 2026 list. Source.

  • Bank of England. (n.d.). Financial Services Contracts Regime. Official description of contractual and supervised run-off. Source.

  • NAIC. (2025). Financial Data Repository. Insurance Topics, updated June 2, 2025. Source.

  • NAIC. (n.d.). Publications. Insurance Department Resources Report description. Source.

  • Hong Kong Insurance Authority. (2025). Annual Report 2024–25. The Year Under Review and Changes of Authorized Insurers, appendix p. 109. Selected extracted text and table. Source.

  • Hong Kong Insurance Authority. (2025). Conduct in Focus, issue 11, article 3: Understanding the Risks of Offshore Coverage and the Importance of Broker Disclosure. Indexed original article excerpt; direct open returned HTTP 403. Source.

  • Monetary Authority of Singapore. (n.d.). Types of Insurance Entities. Official regulatory page, indexed original passages; direct open returned maintenance page. Source.

  • European Parliament and Council. (2009). Directive 2009/138/EC (Solvency II), Article 14, as displayed by EIOPA Single Rulebook. Indexed provision. Source.

  • European Parliament and Council. (2009). Directive 2009/138/EC (Solvency II), Article 144, consolidated text dated January 17, 2025. Indexed provision; direct retrieval bot challenged. Source.

  • Choi, Byeongyong Paul, and Mary A. Weiss. (2005). An Empirical Investigation of Market Structure, Efficiency, and Performance in Property-Liability Insurance. Journal of Risk and Insurance 72(4): 635–673. DOI 10.1111/j.1539-6975.2005.00142.x. RePEc abstract and bibliographic record. Source.

  • Cummins, J. David, Maria Rubio-Misas, and Devin Vencappa. (2017). Competition, Efficiency and Soundness in European Life Insurance Markets. Journal of Financial Stability 28: 66–78. Indexed publisher/repository abstracts. Source.