Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

IAIS Global Monitoring Exercise Population Methods and Data Coverage

Abstract. The Global Monitoring Exercise (GME) of the International Association of Insurance Supervisors (IAIS) combines targeted observation of large insurance groups with a geographically broader collection of supervisory information. This monograph examines the statistical populations, inclusion rules, aggregation conventions, and inferential boundaries of the exercise as components of international macroprudential oversight. The research reconstructs the methodological transition from the June 2023 to the November 2025 framework and compares the completed 2025 Global Insurance Market Report with its preliminary 2026 successor. It distinguishes the consolidation perimeter of Individual Insurer Monitoring (IIM) from the legal-entity and jurisdiction reporting perimeter of Sector-Wide Monitoring (SWM). It develops a comparative reading of reporting institutions in Europe and the United States and examines the consequences of size selection, supervisory discretion, nonresponse, reporting classifications, international reinsurance, currency conversion, and indicator-specific denominators. The original analytical contribution consists of a population and measurement reconciliation, descriptive comparisons of reported participation, algebraic sensitivity analysis, and a public-data interpretation protocol. The results establish that premium-based market coverage supports broad market surveillance while leaving exposure-specific representativeness and network completeness to be evaluated independently. Fixed-reference scores, relative shares, and changes in the reporting population answer different statistical questions. The public evidence supports institutional and methodological conclusions; the confidential returns required for independent group-level recalculation and population estimation are outside the observed evidence.

Keywords: IAIS; Global Monitoring Exercise; Global Insurance Market Report; individual insurer monitoring; sector-wide monitoring; insurer pool; sample selection; supervisory reporting; coverage error; systemic risk; macroprudential surveillance; insurance groups; insurance market statistics; measurement comparability.

Introduction

The international monitoring of insurance is a problem of observation before it becomes a problem of inference. The object of surveillance consists of financial institutions whose liabilities, assets, reinsurance arrangements, legal ownership and insured exposures occupy different geographic and organizational spaces. A risk accepted in one jurisdiction may be invested against in another, ceded to a reinsurer in a third, and recorded in the consolidated return of a parent domiciled in a fourth. The statistical unit chosen for analysis determines which parts of this relationship become visible.

The GME is a supervisory intelligence process built to detect the accumulation and transmission of potential systemic risk. Its annual report combines observations collected directly about selected insurance groups with observations and assessments supplied by insurance supervisors. The population problem arises because these collections answer different questions. The first illuminates concentrations within internationally significant organizations. The second supplies context across national markets, including firms and activities outside the large-group pool.

Three research questions organize the inquiry. How do eligibility rules and discretionary decisions create the observed group population? How do legal and statistical institutions determine the coverage and comparability of jurisdiction-level statistics? How can the combined exercise support judgments about systemic risk when key populations and reporting variables remain only partially observable to external researchers?

The study uses official public documents available by November 19, 2026. It reconstructs the measurement process rather than attributing supervisory outcomes to firms absent from published results. Quantitative examples identified as hypothetical are arithmetic demonstrations of the consequences of alternative assumptions. They are separate from official observations. The inquiry concerns macroprudential surveillance and the interpretation of statistics; it does not estimate the probability that an identified insurer will fail, advise on capital adequacy, or create a new regulatory theory.

Conceptual and Historical Foundations

Insurance obligations and systemic transmission

An insurance undertaking accepts contractual obligations contingent on events that are uncertain in timing, occurrence or monetary severity. The economic function of underwriting is to organize loss sharing and finance future performance. A group can support this activity with its investment portfolio, capital, reinsurance, operational services and funding arrangements. The balance sheet is therefore linked both to insured risks and to risks originating outside insurance.

Macroprudential analysis concerns the consequences of correlated exposure and transmission across institutions and markets. An insurer may face a large loss without causing a system-wide disturbance. Conversely, individually manageable exposures can become consequential when many firms respond to the same shock by selling the same assets, seeking the same liquidity, withdrawing capacity or calling on the same counterparty.

Risk exposure, risk contribution and systemic importance refer to related but distinct objects. Exposure is susceptibility to an external disturbance; contribution concerns the incremental consequences of an institution's activities or distress for the wider system. Systemic importance is an assessment of the consequences of its impairment in a particular institutional and market environment. The expected loss of an individual insurer is not the same as the expected external loss imposed on other parties.

Bierth, Irresberger and Weiß (2015) distinguish these empirical questions in their international insurer study. Their abstract reports results for 253 insurers over 2000–2012, including associations of interconnectedness with exposure and leverage with contribution. Those findings concern the variables and period of that study. They do not establish the causal effect of a GME score in 2026, a distinction particularly important when the surveillance target includes mutual insurers and unlisted entities.

Billio et al. (2010) propose market-return measures of financial-sector connectedness in their systemic risk working paper. Their observed return relationships motivate supplementary monitoring of co-movement across insurance and other financial sectors. Market correlations alone cannot identify contracts, collateral obligations or policy substitution, and predictive relationships can reflect shared macroeconomic drivers as well as direct transmission.

From designation to ongoing surveillance

The institutional shift toward activity-based and sector-wide supervision reflects the need to study both large institutions and collective exposures. The Financial Stability Board's 2019 Holistic Framework description treats supervisory standards, the monitoring exercise and assessment of implementation as complementary elements.

In 2025 the FSB reaffirmed its 2022 decision to discontinue routine annual identification of global systemically important insurers in favor of Holistic Framework assessments. Its communication retains the possibility of expressing views concerning individual systemic importance and addresses the separate reporting of insurers subject to resolution planning.

The resulting institutional arrangement changes the interpretation of inclusion in the monitoring pool. A reporting group supplies an observation to a supervisory process; a mechanical eligibility test is not a determination of systemic significance. Monitoring can lead to further inquiry or a supervisory response, but public research must distinguish those stages from a legally binding decision. This matters for reputational inference as well as scientific validity. Ranking firms by economic characteristics is conceptually different from evaluating whether their disorderly failure would have global consequences.

The multiple populations of global insurance

For analytical clarity the relevant population may be defined in terms of licensed undertakings, economically controlled groups, domestic insurance activity, insurance contracts or exposures. A licensed legal entity may conduct business in more than one territory. A group includes legal entities under its applicable consolidation perimeter, sometimes including businesses outside insurance. A jurisdictional market is a statistical domain determined by the underlying reporting convention. A policy population consists of contracts or insured units and cannot ordinarily be inferred from financial statement totals.

An empirical question about asset liquidation is best connected to the institutions controlling investment portfolios and funding. A question about the capacity of local motor insurance markets to absorb claims may require a legal-entity or business-line frame. A question about disaster protection requires exposed economic assets and households, including those without insurance. Each defines a different universe.

The GME is appropriately interpreted as an assembly of measurements on these different institutional domains. The public analyst must specify the domain for each conclusion, including the period and matching denominator. Without this discipline, the phrase global insurance sector can silently alternate between the set of observed firms and the economic system that those firms are intended to illuminate.

Primary Institutional Documents and Comparative Research Design

Source hierarchy and observation periods

The core evidence comprises two editions of the GME methodology, the completed 2025 annual report and the July 2026 interim report. Regulatory standards and domestic reporting rules establish the institutional environment; academic literature supplies independently developed approaches to measurement. A methodology is evidence about the design of an exercise. An annual report is evidence about published results and stated data limitations. A national statute establishes obligations within its legal jurisdiction. An academic abstract supports its explicitly reported subject and findings, not unidentified full-text details.

The June 2023 GME methodology governed the 2023–2025 cycle. The November 2025 GME methodology defines the 2026–2028 cycle. The IAIS November 2025 revision announcement identifies amendments to indicator definitions, fixed denominators and calculation arrangements, as well as selection criteria.

An exercise year names the data collection and surveillance cycle; a reference date names the point in time described by a measurement. The 2025 annual GIMAR predominantly concerns positions at the end of 2024. The July 2026 interim GIMAR uses preliminary information collected by June 2026, with quantitative reference positions generally at the end of 2025 and one reporting-year exception. Its quantitative sector-wide results were deferred to the annual edition.

Observation Exercise and data period Public interpretation
GIMAR 2025 2025; principally year-end 2024 Completed annual report under the 2023 methodology
GIMAR mid-year 2026 2026; principally year-end 2025 Preliminary analysis under the revised methodology
Framework issued November 2025 2026–2028 exercise years Rules, rather than a survey outcome

A documentary comparison of these releases has a well-defined empirical limit. The published series are evidence of the IAIS's calculations and statements. The underlying observations, the precise set of contributing firms per indicator, and the respondent-level crosswalk needed to reconstruct a harmonized sample are not available in the examined public evidence. The study consequently performs descriptive checks and conditional sensitivity demonstrations in place of an unsupported independent replication.

An institutional crosswalk

For each source observation, a researcher should record the identity of the reporting agent, the legal unit, consolidation basis, geographic rule, valuation basis, original currency, period, reporting component, inclusion rule and coverage measure. This creates a traceable connection between the statistical statement and its economic subject.

A group consolidated return eliminates many transactions internal to the consolidation perimeter. A domestic market total may count activity of domestic subsidiaries of several groups and branches. A direct-premium series records contracts written with original policyholders; an accepted-reinsurance series describes contracts entered into with ceding insurers. Each can be a legitimate observation in its domain, but combining them as unique end-customer premiums duplicates risk-transfer stages.

A useful reconciliation record therefore has separate columns for group domicile, operating jurisdiction, location of exposure and transaction counterparty. These columns should be retained even where disclosure restrictions permit only aggregate publication. A single country identifier cannot express the cross-border structure of insurance.

An analyst should distinguish data unavailability because a legal entity is outside the frame, because a participating entity did not supply a module, because an item was suppressed for confidentiality, and because an exposure is genuinely absent. The same blank cell may correspond to four different economic circumstances. Substituting zero for all of them gives a falsely precise estimate of both exposures and concentration.

Institutional arrangements for legal reporting

The IAIS ICP and ComFrame tool provides the international supervisory context, including group scope, coordination and the evaluation of capital availability. Implementation requires domestic supervisory law, information-sharing arrangements and organizational capacity. International monitoring participation consequently cannot be read as evidence that identical legal requirements apply in every country.

In the European setting, Solvency II Article 35 provides for supervisory information, its quality and proportionate reporting arrangements. In the version displayed by EIOPA, reporting limitations and item-level exemptions are subject to stated conditions. This demonstrates how a highly structured regulatory regime may still contain differences in granularity or frequency across undertakings.

New York's Insurance Law section 307 prescribes annual statements and contains a distinct US-business reporting perimeter for alien insurers. It also provides for reconciliation of certain pooled or combined financial statements. The document illustrates why a legal filing need not represent the same organization or geographic perimeter as the group's consolidated account.

These examples establish different pathways through which information becomes observable. They do not imply a complete survey of every national implementation or the authority of the IAIS to compel individual reporting directly. The comparative legal method identifies the relevant observable mechanisms without extrapolating a domestic rule beyond its territorial scope.

Source resolution and reproducibility

The primary documents were examined through their public text, including rendered pages where retrievable. The supporting academic records are used at the abstract level for their expressly reported research questions and methods. No assertion of full-text academic review is made where only an abstract was accessible.

Each linked source is preserved in the reference list. Source versions and selected claim locations are recorded in the evidence ledger. A reproducible extension of this project would archive each accessible release with its retrieval date and digest, retain the original denominators and indicator versions, and register all revisions rather than allowing a mutable statistics portal to overwrite the earlier research state.

Insurer Sample Selection

Eligibility and organizational scale

The IIM observes selected large insurance groups. Let A denote total group assets measured in billions of US dollars and q the ratio of premium from outside the home jurisdiction to total premium, expressed as a fraction. The automatic eligibility screen, before exceptions, can be written as:

2023–2025: (A > 65 and q ≥ 0.05) or (A > 215 and q > 0).

2026–2028: (A > 70 and q ≥ 0.05) or (A > 235 and q > 0).

The first expression follows IAIS (2023), paragraph 38; the second follows IAIS (2025a), paragraphs 38–40. The criteria concern groups including insurance and non-insurance subsidiaries. The newer edition also identifies a discretionary regional-balance consideration, particularly for groups with A ≥ 55 and q ≥ 0.05.

These are strict inequalities for the automatic asset thresholds. A value of exactly 70 with q = 0.05 does not fulfill the revised first test. A group with 236 billion in assets and any positive foreign-premium share fulfills the second, even if its share is below five percent. A fully domestic group fails both mechanical international-activity tests, regardless of its size. Supported discretion can alter the actual collection boundary.

Consider a hypothetical set of organizations. The table classifies eligibility before supervisory exceptions and assumes exact input values, consistent currency and group perimeter.

Group assets (USD bn) Foreign-premium share Earlier screen Revised screen
68 6% Eligible Outside screen
70 5% Eligible Outside screen
71 5% Eligible Eligible
220 1% Eligible Outside screen
236 0.1% Eligible Eligible
500 0% Outside screen Outside screen
55 5% Outside screen Outside screen; regional discretion may be considered

The composition of an international surveillance sample can change even if underlying economic exposures remain unchanged. The upward thresholds create a narrow set of formerly eligible groups for which entry becomes conditional on discretion or subsequent growth. The comparison is a direct consequence of the two Boolean rules; it is not an estimate of how many real insurers leave the pool.

Why selection depends on the research purpose

Selection by assets favors institutions controlling large financial portfolios. Selection by foreign premium emphasizes organizations whose activity spans jurisdictions. Both criteria are relevant to the monitoring objective because the potential impact of cross-border failure may be associated with scale and organizational reach. Neither variable measures all routes through which disruption can propagate.

A specialist insurer can supply cover that is difficult to replace while holding comparatively small assets. A large domestic insurer can be important to a country's pension-like savings or health insurance arrangements while failing an international premium-share screen. A service organization can affect dozens of firms without writing a single policy. The appropriate response is to consider complementary data on concentrations, market structure, common service providers and alternative risk channels, not mechanically expand a group sample to cover every institution.

The pool is a purposive supervisory collection. It is not a probability sample from a list of all insurance companies. Elliott and Valliant (2017), in their nonprobability sampling article, explain why design-based inference depends on a known random selection mechanism and outline modeling alternatives. Applied here, their result means that conventional survey margins of error cannot be derived from the number of IIM groups alone.

The objective of purposeful selection is surveillance relevance rather than unbiased estimation of the mean of every insurer. Assessing a set of groups with large portfolios is useful when studying concentrations of assets or funding, even when the sample is inappropriate for estimating the average underwriting margin of small local companies. Sample quality must be judged conditional on the estimand.

Threshold sensitivity and movement

The group asset metric depends on accounting valuation and currency conversion. Write A = a × e for domestic-currency assets a and an exchange rate e expressed as US dollars per unit of domestic currency. A group whose balance sheet is unchanged in domestic currency can cross a dollar threshold when e changes. The same can happen after a change in consolidation scope, an acquisition, a disposal, or a reporting correction.

For illustration, a group valued at USD 69 billion under an initial exchange rate would be valued at USD 72.45 billion after a 5% appreciation of the reporting currency, with local-currency assets held fixed. If q is 6%, the revised first threshold would be crossed. This is a measurement sensitivity, not evidence of an actual GME reclassification.

The foreign-premium ratio also changes with its denominator. If cross-border premiums remain 5 units while total premiums grow from 100 to 105, q falls from 5% to approximately 4.76%. The group's foreign premium has not diminished in absolute value, but its measured international share has declined. Such a shift illustrates the difference between relative structure and absolute exposure.

Near a threshold, exact valuation, FX timing and period consistency matter more than rounding shown in public summary tables. A classification should use the data collection's own definitions. No real firm is classified by the illustrative calculations.

Discretion, regional representation and group changes

Discretion improves the ability of supervision to cover an economically heterogeneous system. It can bring in an institution whose business model is underrepresented or retain an analytically important observation. It can also complicate replication by a researcher who knows only the mechanical thresholds. A statistical record should distinguish eligibility, recommended inclusion, actual submission and verified inclusion in an indicator's final analysis.

Observed pool membership is dynamic. Entry and exit can arise from acquisitions, exchange rates, changes in business mix, data revision, supervisory judgment or changes to the methodology. To interpret a change in the aggregate, the researcher needs a stable identifier and a bridge between predecessors and successors. A new parent after a merger is not necessarily new insurance exposure; a legal transfer of a business line can reallocate exposure without changing its economy-wide magnitude.

The 2025 annual report and 2026 mid-year report record 57 and 56 participating groups respectively, each drawn from 18 jurisdictions. The difference in counts is observable. The identity of entrants, exits and consolidation changes is not established by these counts. The change of one in the net total cannot identify a change of one in actual membership if several firms enter and several leave.

A panel consisting only of surviving firms can assist in measuring comparable periods but may understate disruptions arising from firms that disappeared. A repeated cross-section can capture the current population but confound change in exposure with change in membership. Both descriptions are legitimate when their interpretations remain explicit.

Sector-Wide Statistics

Supervisory coverage and the market denominator

SWM collects aggregated information through insurance authorities. Under the updated framework, the eligibility conditions include FSB membership or hosting an IAIG or a participating insurer group at home; broader participation is possible. The revised eligibility methodology (IAIS, 2025a) estimates roughly 90% of world GWP for the jurisdictions meeting the criteria, using 2024 data. The same methodology sets a minimum local collection expectation of the larger coverage represented by the top three insurers or 60% of the local market.

There are two distinct geographical denominators. The first is the share of global premiums located within participating jurisdictions. The second is the portion of local business actually represented by the reported observations. High coverage of the first does not establish full coverage of the second.

Let G_j be the comparable premium volume of jurisdiction j and let G be the global total. If J is the set of participating jurisdictions, their geographic market share is C_geo = sum(G_j for j in J) / G. Let c_j be the fraction of each local market represented in the relevant collection. Then, under a common definition and with no inter-jurisdiction duplication, the effective reported premium coverage is C_eff = sum(c_j × G_j for j in J) / G.

The distinction has empirical consequences. If the participating jurisdictions collectively constitute 90% of world premiums and each supplies returns covering exactly 60% of its local market, effective coverage under these hypothetical assumptions is 54%, not 90%. If the local coverage is 95%, the corresponding result is 85.5%. Real coverage can differ by jurisdiction, business line and indicator. These values are illustrative boundary calculations, not revised estimates of IAIS coverage.

The jurisdiction aggregates are also affected by the distinction between direct insurance and reinsurance. The OECD gross premium definition explicitly includes accepted reinsurance and recognizes that the total involves duplication of risk-transfer stages. A premium-based coverage statistic must be interpreted on its own gross or direct basis. Global penetration, unique policy exposure and the volume of accepted reinsurance cannot all be obtained from the same numerator.

Observed participation and reporting modules

The 2025 market report (IAIS, 2025b) records SWM participation from 26 qualifying jurisdictions and 32 additional voluntary participants, yielding 58 jurisdictions across at least one reporting component. The same annual report records 43 climate-data jurisdictions, with 36 providing detailed investment categories. Its cyber underwriting comparison in Figure 26 uses 23 groups with submissions at both reference dates.

These figures illustrate differing sets within a common reporting architecture:

Publicly reported observation Count Unit
IIM 2025 57 Groups
IIM home jurisdictions 18 Jurisdictions
SWM 2025, any component 58 Jurisdictions
SWM climate data 43 Jurisdictions
Detailed climate asset categories 36 Jurisdictions
Cyber matched sample 23 Groups

The climate-jurisdiction participation equals about 74.1% of the number of jurisdictions reporting at least one SWM component. The detailed climate subset equals about 83.7% of the climate-jurisdiction count. These are count ratios, not premium or asset coverage estimates. The cyber matched sample equals approximately 40.4% of the 57-group headline count, a descriptive comparison of published counts only. It cannot be interpreted as the fraction of global cyber premium observed or as the response rate of eligible cyber insurers.

Two modules containing equal counts may contain entirely different participants. Count completeness is therefore different from financial exposure completeness. A data dictionary should identify the applicable respondent frame for each figure, including firms with genuine zero exposure or with business models that make an item inapplicable.

The 2026 interim report presents qualitative SWM evidence; direct comparison with a completed quantitative annual module would combine unlike components. The 2026 release's preliminary nature is material to its interpretation. Observations should retain a provisional flag until a comparable finalized result becomes available.

National reporting as the infrastructure of international data

The statistical capacity supporting SWM is built through domestic supervisory returns. EIOPA's insurance statistics portal differentiates solo and group observations and identifies template-specific series. It also distinguishes provisional and finalized information and records revisions. These categories matter to any attempt at cross-market harmonization.

The NAIC Financial Data Repository describes a centralized collection of US multistate insurer filings, covering, according to its own description, more than 98% of domestic written premiums and over 4,500 individual insurers. In contrast to the IIM, its basic collection arises from legal-entity supervisory reporting. Those headline counts illustrate the difference between a national statutory reporting population and an international group-monitoring pool. They must not be interpreted as a numerical comparison of the relative quality of two samples.

A national return may use a statutory accounting measure that diverges from consolidated financial reporting. Market aggregates based on accounting regulations can differ because of recognition, measurement, currency or the perimeter of entities. The reporting systems may each be internally valid while corresponding variables remain unsuitable for direct pooling.

Group capital resources also differ from legally available capital at a subsidiary. The ability to use capital in another legal entity depends on regulatory and contractual limitations, transferability and circumstances. A worldwide consolidated ratio can coexist with a constraint on the resources available in a particular market. The analyst should record whether the relevant variable is an aggregate accounting ratio, a regulatory capital measure, an available cash position or a jurisdiction-level solvency measure.

Measuring scale, growth and performance

Consider a market containing 200 small insurers with an average ratio of 1.2 and two large insurers with ratios of 2.0 and 3.0. An equally weighted average across all 202 firms is approximately 1.213. If the two large institutions account for 70% of the relevant denominator, hold equal shares of that denominator, and the requirement-weighted mean for the small insurers is also 1.2, the aggregate ratio is 2.11. Both results can be mathematically correct; they describe different properties of the population.

If a solvency ratio is defined as R_i = capital_i / requirement_i, the aggregate ratio for comparable reporting units is sum(capital_i) / sum(requirement_i). It is a requirement-weighted mean of the individual ratios when all denominators are positive. The arithmetic mean, median and lower tail answer different questions, such as the typical entity's position or the distribution of vulnerability. The aggregate does not establish the solvency of all participants.

Premium growth similarly depends on price, exposure quantity and product mix. Revenue measures based on recognized services and written-premium volumes are conceptually distinct. Gross and net reinsurance positions reflect different stages in risk allocation. A claim payment records a cash flow; incurred claims include a liability or expense concept determined by the reporting regime. A statistic suitable for describing liquidity may be inadequate for measuring underwriting profitability.

Cross-currency aggregation requires consistency about whether the item is a stock or a flow. Positions at a balance-sheet date should be distinguished from premiums generated throughout a period. An exchange-rate convention used for end-period financial statements may not be suitable for annual flows without adjustment. The analyst must also distinguish nominal currency growth from real growth. A 10% increase in nominal premium with inflation of 7% corresponds to real growth of roughly 2.8%, assuming the deflator is appropriate. The example is conditional and does not estimate a particular national insurance market.

Systemic Monitoring Coverage

Coverage as a function of the variable

Market size provides one reference for evaluating surveillance reach. It is natural to use written premiums to assess the broad coverage of underwriting activity. Yet the distribution of premiums can differ considerably from that of assets, derivatives, collateral, surrendered policies, reinsurance receivables or specialist underwriting capacity.

Define the coverage of a nonnegative additive variable x as C_x = sum(x_i observed) / sum(x_i in the target population), provided the denominator exists and the units and consolidation bases are comparable. This is a standard descriptive ratio. When the denominator cannot be observed, the coverage is unknown or can be bounded using explicit assumptions. Computing C_GWP does not determine C_x.

A hypothetical market has 100 units of premium, 90 written by monitored firms. Suppose those firms have 30 units of a particular risky exposure and the excluded firms hold 70. Premium coverage is 90%, while exposure coverage is 30%. Alternatively, if the monitored firms hold 95 of 100 exposure units, exposure coverage is 95%. Both situations are consistent with 90% premium coverage.

The distinction becomes especially important for an emerging risk that is concentrated in a relatively small segment. In reinsurance, the jurisdiction of incorporation can have a disproportionate significance to globally dispersed underlying risks. In cyber insurance, specialized providers can have small shares of total market premiums but large shares of the relevant segment. A broad survey can reveal the movement of aggregate conditions while an exposure-oriented collection is necessary to assess that activity.

Fixed and relative scores

The updated methodology defines a weighted quantitative IIM assessment across five categories and thirteen active indicators. The November 2025 document describes the fixed-reference and relative approaches in section 5.1. The base for absolute benchmarks is the 2025 collection referencing year-end 2024; relative assessments use the applicable exercise-year totals.

For interpretive purposes, a stylized normalized share for insurer i and indicator k is x_ik / D_k, where D_k is a stated denominator. A weighted sum of these shares gives an index after applicable scale factors. This simple expression illustrates denominator behavior; actual scores follow the official indicator definitions, currency rules, weightings and reference-value adjustments.

Suppose a group's exposure rises from 10 to 12 units while the corresponding pool aggregate rises from 100 to 150. Its contemporaneous share falls from 10% to 8%; its share against the fixed base of 100 increases from 10% to 12%. Both are accurate. The first measures position relative to peers, and the second measures movement against an earlier benchmark. A decline in relative share cannot establish lower absolute exposure.

If all participants' data are nonnegative and a relative denominator is their total, the relative shares sum to one by algebra. This identity means a relative index can rank contributions within the pool without displaying broad common growth as a larger total. A fixed benchmark can reflect such common growth, although its interpretation depends on the continued relevance of the benchmark. New definitions, changes in weights or a change in the base can produce a discontinuity in an otherwise unchanged exposure series.

The IAIS's 2025 methodological revision notice identifies changes to minimum guarantees, intra-financial indicators, Level 3 assets and other calculations. Its ancillary risk indicator announcement records expanded monitoring of credit, derivatives and reinsurance. These supplementary measures enrich risk assessment without being interchangeable with the headline weighted score.

Report-year comparison and sample composition

The 2025 GIMAR reports a 1.2% fall in the IIM aggregate systemic risk score between year-end 2023 and year-end 2024, using controls for sample stability. The 2026 mid-year update reports a slight rise in 2025, particularly in asset liquidation and interconnectedness, and notes that the revised method yields lower aggregate levels than the former one.

It is methodologically inappropriate to concatenate these two changes and infer an unconditional multiyear growth rate without first reconciling the calculation basis. The 2026 within-edition comparison provides information about change under that reporting framework; a direct comparison of a 2025-edition level with a 2026-edition level conflates exposure and methodological change.

For an additive exposure total, let S_0 and S_1 be the observed sets in two periods. Let M be their intersection, E the new members and X the departed members. The difference in totals decomposes exactly as:

T_1 − T_0 = sum over M of (x_i1 − x_i0) + sum over E of x_i1 − sum over X of x_i0.

The first term measures change among matched observations, the second entry, and the third exit. With a merger, the grouping should first be reconciled to economically equivalent exposure where feasible. FX and valuation contributions require additional decomposition after fixing a perimeter.

For example, suppose an initial set has exposure 100; matched firms grow from 90 to 95; firms exiting the sample had 10; entering firms report 20. The new total is 115. Growth among matched firms is 5 units, while composition contributes another 10 units. An unqualified statement that the preexisting companies grew by 15% would be false. This exercise verifies the arithmetic identity but supplies no actual IAIS firm-level components.

Missingness and sensitivity bounds

Missing observations are an inferential problem for surveillance when the unobserved units differ systematically from observed units. Possible reasons include lack of domestic reporting authority, incompatible templates, limited capacity, lack of applicability, confidentiality protections and timing. An absence of data about private assets should not be interpreted as absence of private assets.

Suppose a nonnegative exposure-to-premium ratio is observed over a comparable 90% premium share. If the observed ratio is r_o and the unknown ratio in the remainder is r_m, the global premium-weighted ratio is 0.9r_o + 0.1r_m. With r_o = 0.20 and an assumed r_m between 0 and 2, the global ratio lies between 0.18 and 0.38. The observed sample ratio is 0.20. The conditional range is deliberately wide because the unobserved exposure intensity is uncertain.

If the missing-case ratio is not bounded by sound economic or accounting evidence, no finite upper bound follows from the premium coverage statistic alone. Choosing an arbitrary upper limit and treating it as a confidence bound would conceal rather than quantify uncertainty. A documented sensitivity range explains the implications of assumptions; it is not an interval estimated from the sample distribution.

A model for nonresponse or sample inclusion may be useful when relevant covariates and a complementary population frame are available. It requires explicit assumptions about overlap and the relationship between selection, covariates and outcomes. In domains without coverage or where the relationship changes in stress periods, the resulting estimate may depend heavily on extrapolation. An official surveillance collection does not become a probability sample by applying weights after selection.

Interconnectedness and the limits of aggregation

Consider a directed network in which a link represents a reinsurance recoverable or another contractual exposure. Aggregate balances can reveal the total exposure in each sector without identifying whether exposures are spread among many counterparties or concentrated on a single one. Suppose a collection of ten insurers has total recoverables of 100 units. Allocating ten to each of ten independent counterparties differs from allocating all 100 to one counterparty, even though the aggregate asset is identical.

Observed network coverage depends on vertex and edge inclusion. Comprehensive observation of insurer assets cannot guarantee that exposures to non-insurance counterparties, retrocession chains or external service providers have been attributed to the correct entity. Counterparty identifiers, contractual seniority, collateralization and legal enforceability influence the consequences of a default.

A network constructed solely from public financial statements may also have missing directions and duplicated positions. Assumed and ceded reinsurance can be recorded in liabilities and assets at different values and dates. The exposure of a cedant is not necessarily equivalent to the gross obligation of a reinsurer. In a severe scenario, timing of collections and disputes can matter even if eventual recovery is high.

Monitoring should consequently examine common holdings, common counterparties and contractual dependencies alongside totals. The analyst may use market-return connectedness as an additional signal, provided it is recognized as a different observation process. A statistical association does not, by itself, identify a particular exposure path.

Qualitative evidence and supervisory judgment

Qualitative information gives context on expected developments, business plans, changes in risk management and the relevance of a quantitative outlier. The IAIS's collective assessment combines analytical inputs and judgment; the collective assessment methodology describes feedback with supervisors and stakeholders. A forward-looking observation should be interpreted according to the question asked, respondent unit, horizon and the distinction between expectation and realized outcome.

An outlier can reflect genuine concentration or a data anomaly. Rapid growth from a small initial exposure can be statistically striking while financially immaterial. A low score may still coexist with a material emerging risk omitted from existing indicators. A reasonable analytical process examines level, direction of change, absolute materiality, peer comparison and potential transmission.

External analysts are usually unable to observe the complete supervisory feedback and deliberation. They can evaluate the consistency of public conclusions with public documentation, while recognizing that the publicly accessible data do not exhaust the evidence available to authorities. The reverse also holds: access to confidential data does not automatically remove measurement error, model uncertainty or the need for judgment.

Integrated Scientific Synthesis and Future Research

The division of analytical labor

The results establish complementary functions for the two monitoring collections. IIM concentrates information on groups whose scale and reach make them relevant to international oversight. SWM supplies wider institutional observations and national market context. External population statistics and public financial information assist in assessing the breadth of observation and checking individual definitions. Qualitative supervisory evidence helps interpret changing mechanisms.

The GME's expected market reach is not a combined sampling fraction obtained by adding IIM and SWM shares. These collections overlap economically. A single subsidiary located in a participating jurisdiction can contribute to a sector aggregate while its parent contributes to a group return. In one series the subsidiary is part of a domestic market; in another it is part of a consolidated entity. The overlap is conceptually useful for triangulation, but the observations cannot be summed into a unique exposure without a valid crosswalk.

The interpretation of agreement across the collections rests on their units and definitions. If group exposure and sector aggregates rise under comparable definitions, a common trend may merit attention. If only the group measure rises, concentration among covered institutions is one possible explanation. Alternative explanations include differing geography, a change in the sample, reclassification, FX and nonresponse. If the sector aggregate rises while the large group measure remains stable, growth may be occurring outside the group pool or reflect an unmatched reporting perimeter.

These are discriminating hypotheses, not causal results. Testing them requires comparable period data, sufficient disaggregation and a measure of the residual population. Their practical value is to prevent the analyst from accepting one explanation merely because it is compatible with the observed aggregate.

Documentary reconciliation of the transition

The primary evidence establishes three particularly important distinctions.

First, the threshold revisions restrict automatic eligibility for some groups previously inside the earlier rules. They coexist with discretion aimed at representing economically significant business models and regions. The precise realized membership consequences are not publicly recoverable from the examined aggregate sources.

Second, the 2025 report and the 2026 interim report describe primarily different reference years and use different calculations. A scientifically defensible comparison uses a common definition and benchmark, checks the membership bridge and data revisions, then attributes identified movements to measured contributors. A change of method is documented as a break, rather than attributed to an economic event.

Third, reporting breadth varies by module even during one exercise. The 2025 observations on climate and cyber activity have different populations from the overall group and jurisdiction counts. The evidence supports the distinct counts and their logical implications; it does not establish the magnitude of reporting bias or hidden risk.

These findings are an original synthesis of publicly documented methods and results. They do not assert a novel empirical discovery, propose new regulatory obligations or replace the official GME procedure.

Implications for risk research and source publication

Research users would benefit from a stable population crosswalk reporting, where confidentiality permits, the number of eligible, invited, submitting and validated units for each module; the denominator and reference date; the basis of consolidation; the handling of FX; and coverage or missingness by economically relevant strata. Publication of aggregate bridges across methodological revisions could support historical analysis without exposing confidential submissions.

A scientific study of cross-border risk transmission would require a separate design for counterparties, geography and contractual positions. A study of underwriting capacity would require business-line market shares and data on substitutability. A study of household protection would require the insured and uninsured population and the severity of underlying hazards. Each objective identifies distinct data needs, so the monitoring report is most useful when matched with complementary statistics rather than stretched beyond its own observations.

Future empirical work can evaluate the sensitivity of industry estimates to alternative selection thresholds by using a complete public or accessible regulated group frame. It can test whether common exposure measures vary systematically by firm size, domicile, business model and degree of international activity. It can compare result stability under current and fixed memberships and evaluate whether weighting by premiums, invested assets or technical provisions changes conclusions.

An independent reconstruction of systemic scores would require authorized access to exact templates, definitions, reference values, denominators, actual respondent observations and any sample controls. Validation should use reconciled source data, an independently implemented calculation and transparent comparison of divergences. A public record of methodology can guide such work but cannot substitute for the data.

Research limits and disciplined interpretation

The empirical cutoff is November 19, 2026. The primary documentary evidence includes the completed 2025 annual report and the July 2026 interim update. This research has no access to confidential IIM or SWM submissions. It provides selected domestic institutional comparisons and not a jurisdiction-by-jurisdiction legal census. Public portal descriptions establish reporting architecture, not independent validation of regulator microdata or individual company filings.

The supporting academic findings are explicitly confined to accessible abstracts and their bibliographic identities. The numerical sensitivity examples are deterministic and hypothetical. Neither out-of-sample estimates of systemic risk nor causal claims about the effectiveness of supervision follow from them. This scope is sufficient for a documentary analysis of the design and inferential implications of the exercise; it leaves empirical replication and estimation as distinct future investigations.

Conclusion

The IAIS Global Monitoring Exercise is an international supervisory observation process organized around heterogeneous statistical populations. Its ability to illuminate systemic risk depends on matching each question to the scale, reporting perimeter and exposure represented in the underlying data. The observed insurer pool is deliberately selected for supervisory relevance, and the sector-wide collection extends the perspective to jurisdictional market conditions and common activities.

The revision of the eligibility and assessment methodology establishes a concrete version boundary between the 2023–2025 and 2026–2028 regimes. A change in an observed score or participant count has scientific meaning only after the relevant reference period, membership, definitions, weights, valuation and denominator have been established. An absolute benchmark captures movements differently from a current relative share.

Premium coverage, respondent coverage and coverage of any specific exposure are separate measures. They can diverge even in a collection that encompasses the dominant insurance markets. National regulatory systems, group structures, reinsurance chains and reporting exemptions explain why population boundaries matter throughout the GME. A high observed fraction of one variable supports surveillance of that variable and calls for independent evaluation of coverage of other risks.

The central contribution of the inquiry is to reconcile institutional purpose with the limits of statistical inference. International insurance surveillance gains explanatory strength through complementary sources of information, version-controlled comparisons, explicit treatment of selection and missingness, and careful distinction between evidence, interpretation and unresolved uncertainty.

References

  • IAIS (2023). Holistic Framework for Systemic Risk in the Insurance Sector: Global Monitoring Exercise. June 2023, paras. 37–50. Source

  • IAIS (2025a). Holistic Framework for Systemic Risk in the Insurance Sector: Global Monitoring Exercise. November 2025, paras. 38–90. Source

  • IAIS (2025b). Global Insurance Market Report 2025. December 2025, §§ 1.1, 3.4.1.1, 4.1–4.2, 5.2, 6.1. Source

  • IAIS (2026). Global Insurance Market Report 2026: Mid-year Update. July 2026, §§ 1, 4–5. Source

  • Financial Stability Board (2019). Holistic Framework for the Assessment and Mitigation of Systemic Risk in the Insurance Sector. November 2019. Source

  • Financial Stability Board (2025). The FSB Reaffirms Its Decision to Use IAIS Holistic Framework Assessments Instead of an Annual Identification of Global Systemically Important Insurers. November 25, 2025. Source

  • IAIS (2025c). IAIS Publishes Updated Global Monitoring Exercise (GME) Document. November 25, 2025. Source

  • IAIS (2025d). IAIS Introduces Additional Ancillary Risk Indicators for Its Global Monitoring Exercise. November 25, 2025. Source

  • IAIS (n.d.). ICP and ComFrame Online Tool. Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups, ICPs 9 and 23 and assessment introduction. Accessed November 19, 2026. Source

  • European Insurance and Occupational Pensions Authority (EIOPA) (n.d.). Insurance Statistics. Accessed November 19, 2026. Source

  • European Parliament and Council (2009). Directive 2009/138/EC (Solvency II), Article 35, Information to Be Provided for Supervisory Purposes. Text reproduced in EIOPA Single Rulebook, accessed November 19, 2026. Source

  • New York State Legislature (2014 revision). New York Insurance Law § 307: Annual Statements; Audited Financial Statements. New York State Senate, Open Legislation. Accessed November 19, 2026. Source

  • National Association of Insurance Commissioners (NAIC) (n.d.). Financial Data Repository. Accessed November 19, 2026. Source

  • Organisation for Economic Co-operation and Development (OECD) (n.d.). Gross Insurance Premiums. OECD Data Indicators. Accessed November 19, 2026. Source

  • Bierth, C., Irresberger, F., & Weiß, G. N. F. (2015). Systemic Risk of Insurers Around the Globe. Journal of Banking & Finance, 55, 232–245. DOI: 10.1016/j.jbankfin.2015.02.014. White Rose Research Online record. Source

  • Billio, M., Getmansky Sherman, M., Lo, A. W., & Pelizzon, L. (2010). Econometric Measures of Systemic Risk in the Finance and Insurance Sectors. NBER Working Paper No. 16223, July 2010. SSRN record. Source

  • Elliott, M. R., & Valliant, R. (2017). Inference for Nonprobability Samples. Statistical Science, 32(2), 249–264. DOI: 10.1214/16-STS598. Institute of Mathematical Statistics, issue PDF. Source