RESEARCH DOSSIER
Insurance Sector Systemic Risk Indicators and Macroprudential Monitoring
Abstract. Insurance sector systemic risk arises through the interaction of contractual obligations, investment portfolios, risk-transfer relationships and the capacity of financial markets and institutions to absorb correlated disruptions. This monograph examines the indicators that make these interactions observable and the supervisory institutions that transform monitoring into preventive action. It compares the International Association of Insurance Supervisors' Holistic Framework and Global Monitoring Exercise with European risk dashboards and legal reforms, United States macroprudential initiatives, and United Kingdom insurance stress testing. The analysis uses documentary comparison, explicit measurement definitions, source criticism and reproducible numerical illustrations. It distinguishes the vulnerability of insurers from their contribution to financial instability, stock measures from cash-flow measures, market concentration from common exposure, and monitoring signals from legally operative intervention criteria. The central conclusion is that aggregate solvency and individual capital adequacy are necessary inputs to surveillance, while identification of systemic transmission also requires distributional, counterparty, liquidity, substitutability and behavioral evidence. A credible monitoring process preserves the population, observation date, denominator and methodological version of each indicator and connects escalation to a specific transmission mechanism. Results are comparative and analytical; illustrative calculations are not estimates of actual market distress or calibrated regulatory thresholds.
Keywords: Insurance systemic risk; macroprudential supervision; financial stability; Global Monitoring Exercise; concentration; interconnectedness; liquidity stress testing; supervisory triggers; reinsurance; common exposures; insurance solvency.
Introduction
Insurance transforms uncertain future expenditure into a contractual distribution of financial losses. Its institutions assemble claims liabilities, assets, capital, operational capacity and arrangements for transferring risk. Through these functions an insurer can absorb a shock that would otherwise interrupt household consumption, business operations or investment. Through the same financial architecture it can also transmit disturbances when its capacity to settle obligations depends on markets or counterparties subject to correlated stress.
Macroprudential monitoring concerns these collective consequences. Its object is the vulnerability and resilience of the financial system and the continuity of economically consequential services. An insurer's financial position matters to this inquiry insofar as deterioration affects third parties, changes the behavior of other participants or undermines the availability of risk protection. Supervisors therefore investigate both the probability that insurers experience distress and the consequences conditional on distress. The two questions require different evidence.
This study asks how existing supervisory instruments establish observable systemic risk, how concentration and interconnectedness affect the interpretation of indicators, and how monitoring signals should be translated into escalation. It treats the international coordinating framework, national prudential law, published aggregate data and scholarly findings as distinct evidence classes. The comparison covers international practice, the European Union, the United States and the United Kingdom; these cases illuminate institutional differences without constituting a census of global law. The time boundary is November 19, 2026. Where the newest data are preliminary, their status is preserved. The calculations below use synthetic figures, except for explicitly attributed published observations.
Conceptual and Historical Foundations
Insurance as a shock absorber and transmission mechanism
Risk pooling distributes independent or imperfectly correlated losses across a portfolio. Capital and reserves support claims payments beyond current premium receipts, while reinsurance transfers selected layers to other risk bearers. An unexpected shock becomes systemic when these arrangements interact with the wider economy in a way that interrupts financial services or amplifies aggregate losses. A catastrophe can devastate a region while remaining manageable for the financial sector; a comparatively small initial valuation loss can become consequential when many leveraged participants need immediate cash from the same market.
Three distinctions organize causal analysis. Inward transmission concerns financial or economic distress entering insurance portfolios and liabilities. Outward transmission concerns distress or protective responses originating in insurance and worsening conditions elsewhere. Common-cause dependence concerns simultaneous impairment produced by a third factor. Evidence of high correlation alone cannot identify which mechanism operates. The same distinction applies to an insurer's own fragility and the external consequences of its failure: a small specialist may be replaceable in investment markets yet essential to a line of coverage.
Traditional underwriting produces exposures with diverse settlement schedules. A liability payable upon an accident may differ radically from an investment-linked contract with a surrender option or a derivative requiring daily collateral. Contract maturity is thus insufficient as a proxy for cash-flow maturity. The speed with which a financial commitment becomes payable depends on options, collateral terms, legal conditions, counterparty actions and policyholder behavior. The balance sheet must be read jointly with the contract structure.
In the United States, Cummins and Weiss (2014) examined the sector's potential systemic transmission, distinguishing core insurance activity from channels of vulnerability involving interconnectedness and reinsurance. Eling and Pankoke's (2016) review of 48 studies similarly differentiated conventional underwriting from potentially destabilizing financial activities and identified uncertainty about outward effects. These are abstract-level findings, historically bounded by the publications' research periods. They motivate differentiated monitoring rather than a timeless classification of safe and unsafe business models.
Bierth, Irresberger, and Weiss (2015), in their international insurer study, examined 253 firms over 2000-2012. Their abstract identifies a crisis-period peak and different associations for exposure and contribution, with interconnectedness and leverage respectively important. An association in that historical sample is not a causal estimate for today's unlisted insurers or alternative-asset portfolios. The study nevertheless illustrates why inward sensitivity and outward importance require separate dependent variables.
From designated institutions to activities and exposures
The financial crisis exposed the importance of distinguishing an insurance brand from the diversified group operating beneath it. A designation-based approach concentrates enhanced supervisory attention on selected institutions. It can facilitate accountability and coordination when a particular group is critical, but may leave common exposures among firms outside the list insufficiently visible.
In 2019 the FSB described the Holistic Framework as an integrated architecture of enhanced supervisory measures, monitoring and assessment of implementation. In its November 2025 decision, the FSB reaffirmed its 2022 decision to discontinue annual identification of globally systemically important insurers and to use IAIS assessments. This does not remove individual-group analysis: the statement preserves the possibility of public views on individual systemic importance and distinguishes that issue from resolution planning lists.
The shift has a precise measurement implication. A global assessment can investigate risk concentrated in a single group, a common activity across many participants, or a sequence of actions that generates a market-wide disturbance. The universe of monitored activities is wider than the set of largest institutions. An individual institution's rank is therefore one part of the evidentiary record rather than a complete regulatory diagnosis.
Transmission and time
A useful causal narrative identifies an initial shock, the exposed positions, the contractual or behavioral response, the market or counterparty affected, and the final economic consequence. For example, a rate shock may cause derivatives collateral calls; the resulting need for cash may lead to sales of bonds; if other institutions are simultaneously selling, prices may fall; this can generate further valuation losses and collateral requirements. Each link is conditional. Netting and liquidity resources may interrupt the sequence, as may buyers able to absorb offered assets.
A second channel operates through provision of coverage. If multiple insurers reduce catastrophe capacity, enterprises may find essential risk protection expensive or unavailable, hindering financing and reconstruction. The original insurance loss and the subsequent loss of service are distinct. The implications depend on replacement capacity, affordability and the time needed to transfer risk.
A third channel arises through reinsurance. A cedent remains liable under its direct insurance policies according to their terms even where it expects recovery from a reinsurer. A delay in recoveries introduces a cash-flow gap; a counterparty failure can also cause a capital loss. Conversely, well-collateralized, diversified reinsurance distributes shocks. The monitoring problem is to distinguish the balance-sheet relief from the distribution of ultimate risk and timing of settlement.
These mechanisms operate at different frequencies. A severe claim development may be measured over quarters while collateral demands are payable within days. An annual dataset can reveal structural exposure but cannot measure a daily squeeze. The timing of data acquisition must match the speed of the mechanism under investigation.
The academic literature also supplies a methodological distinction that anticipates activity-based supervision. The November 2014 working-paper version of Eling and Pankoke's insurance systemic risk review examines institutional nodes and the economic activities connecting them. Its 43-paper inventory is distinct from the 48-study count in the 2016 journal abstract. A supervisory portfolio of activities can therefore complement entity-level oversight: changing the legal location of an exposure may change the reporting perimeter without removing the underlying transmission channel. Cummins and Weiss's (2013) U.S. insurance working paper further distinguishes conventional underwriting from particular life-insurance activities and noncore financial exposures, and treats reinsurance as a potential internal transmission channel. Its reported statistical associations are evidence about the sample and method employed; they do not identify the causal effect of an intervention. These versions are cited independently of their subsequent journal publications.
Primary Institutional Documents and Comparative Research Design
Evidence hierarchy and unit of observation
The documentary study prioritizes adopted rules, regulator methodologies, official reports and published stress outcomes. Scholarly articles are used within the scope actually available. The source record separates issued dates from reference dates, proposed provisions from applicable law and aggregate observations from microdata. A short link identifies the exact checked document; complete bibliographic identity appears in References.
The reference population varies by question. An insurance legal entity is the appropriate unit for claims-paying obligations and restrictions on local assets. A consolidated group is the appropriate unit for analyzing shared strategy, financing and group-level exposures. The market is the unit for coverage availability and common holdings. A financial network is the unit for interconnectedness. Mixing these units can both understate local restrictions and overcount intra-group assets. In a consolidated metric, internal transactions are eliminated; in a liquidity-transfer map, the same internal transactions can be essential because a subsidiary may depend on cash from its parent.
Comparable observations require an explicit population, currency, valuation basis, reporting frequency and consolidation status. Direct insurance premiums are not equivalent to reinsurance premiums, and the two cannot be added indiscriminately. Nominal premium growth can represent inflation rather than additional real coverage. Paid claims quantify settlement cash flows; incurred claims include changes in expected obligations. Market shares based on written premiums, liabilities, claims or assets answer different questions.
International monitoring
The IAIS Global Monitoring Exercise methodology issued in November 2025 covers assessments in 2026-2028. It combines individual insurer and sector-wide information, analysis, collective discussion and reporting. Its assessment of individual firms concerns insurance groups in the relevant pool, not necessarily single legal entities. Individual analysis is complemented by aggregate market information and supervisory assessments.
The IAIS (2026) mid-year GIMAR used preliminary information collected through June 2026, with 56 international groups across 18 jurisdictions and qualitative sector-wide input from authorities representing over 90% of global gross written premiums. The main quantitative reference date is year-end 2025, with a fiscal-year exception noted in the report. The group count and premium-coverage figure describe different populations. It would be incorrect to interpret them as evidence that all 56 groups themselves write more than 90% of world premiums. The quantitative sector-wide component was reserved for the year-end report.
The distinction is essential for interpretation. A selection of large groups is suitable for investigating global footprint. Domestic firms can nevertheless matter to individual markets. A sample that covers much of the sector through reporting authorities does not establish full reporting of every institution, every exposure or every territory. The analyst must document omitted exposures before estimating aggregate losses.
The 2025 GIMAR provides a useful example of evidence reconciliation. IAIS (2025b), Figure 29, records a 60-basis-point reduction in the total score between year-end 2023 and 2024. The score's Level 3 assets component represents 22.6%, derivatives 13.0% and liability liquidity 11.5%. The prose and pie-chart labels differ on the ordering of the two intra-financial asset and liability shares. Accordingly, those two shares are omitted here; the 79-basis-point rise in the intra-financial assets component and 101-basis-point decline in the minimum-guarantee component are consistently identifiable in the decomposition. A score share is a share of an index, not a fraction of realized financial losses.
The IAIS (2026) interim results report broadly stable aggregate financial conditions and a slight rise in the aggregate footprint score for year-end 2025. The 2025 methodological changes require caution in linking this movement to earlier vintages. Stable financial conditions and an increasing systemic footprint can coexist because an indicator of exposure and scale need not move with current solvency.
European monitoring and legal reform
The European Union's institutional structure connects national supervision, European sectoral analysis and financial-system monitoring. EIOPA's October 2026 Insurance Risk Dashboard uses Solvency II information from insurance groups and individual undertakings. Quarterly company observations refer to Q2 2026, annual observations to year-end 2025 and most market inputs to end-September 2026. Its forward outlook uses responses from 23 national competent authorities. Historical changes, current risk assessments and expected changes are different information products.
EIOPA reports distributions such as medians and interquartile ranges alongside markets-based indicators. These permit the analyst to separate a typical institution from heterogeneous exposures. A median investment grade does not establish the size of the most exposed portfolio. A population may have a stable median even if a small number of firms accumulates large positions in correlated securities.
The European Parliament and Council's Directive (EU) 2025/2, adopted November 27, 2024 and published in January 2025, adds liquidity-related macroprudential provisions to Solvency II. Its general application date under Article 4 is January 30, 2027, following the transposition deadline of January 29, 2027. At this study's cutoff the new directive therefore provides evidence of adopted regulatory architecture rather than a claim that its future provisions are already uniformly applied by all Member States. Existing national tools may differ.
United States supervisory coordination
Insurance supervision in the United States operates through state authorities coordinated by the NAIC and other institutions. The NAIC's macroprudential supervision overview identifies liquidity, capital stress testing, recovery and resolution, and counterparty exposure and concentration as priority areas. Its discussion describes a liquidity stress testing framework for large life insurers finalized in 2021 and incorporated through revisions to the model holding company framework.
A model law is a template for state-level legal implementation. An international comparison must distinguish the model, state enactment, supervisory guidance and the particular group filing required. This study examines the documented macroprudential program rather than claiming identical obligations across all states. In the US case, the practical focus on cash-flow stress supplies a useful contrast to surveillance based mainly on solvency ratios.
United Kingdom stress testing
The PRA (2025) Life Insurance Stress Test results cover eleven major UK life insurers active in bulk purchase annuities, representing over 90% of annuity liabilities in that defined context. The core scenario applied adverse market conditions to year-end 2024 positions under Solvency UK. Reported aggregate SCR coverage declined from 185% to 154%. Two exploratory scenarios investigated asset concentration and funded reinsurance. The exercise was designed for supervisory insight and disclosure, not as a pass/fail test or a process for setting capital requirements or buffers.
This population is narrower than the entire UK insurance market. Its findings cannot establish the resilience of smaller life insurers, general insurers or every relevant counterparty. It is valuable precisely because the scenario, observation date and activity perimeter are identifiable. Its application to a concentrated segment makes the analysis of shared investment and risk-transfer relationships especially relevant.
Comparative interpretation
The cases have related purposes and different units of analysis. The GME coordinates cross-border risk assessment, the European dashboard organizes sector and market indicators, the NAIC program addresses group-level and state supervisory needs, and the PRA exercise tests selected portfolios under a specified scenario. These functions are complementary: surveillance identifies vulnerabilities, scenario analysis estimates conditional consequences, and supervision decides what action is warranted.
Comparability rests on common questions rather than mechanically harmonized numbers. Which firms are in scope? Which contracts can accelerate payment? What assets are available for sale or collateral? Which institutions ultimately carry the exposure? How are critical services replaced? Which authority has power to intervene? This common inquiry makes differences between regimes intelligible without treating all regimes as equivalent.
Macroprudential Measures
Capital measures and loss-absorbing resources
A standard solvency coverage ratio is eligible own funds divided by the applicable capital requirement. It summarizes the relationship between loss-absorbing resources and a regulatory measure of risk. Its informational value depends on the valuation of assets and liabilities, the eligibility of capital, diversification assumptions, risk model calibration and institutional perimeter. A ratio can decline because assets fall, liabilities rise, eligible resources become restricted or required capital increases.
System-level analysts should examine three summaries together: total eligible funds divided by total capital requirements across a comparable population, the distribution of firm-level ratios, and the share of liabilities held by firms under specified levels of capital pressure. The first is weighted by requirements, the second describes the firm population, and the third reveals potential policyholder exposure. No single summary substitutes for the others.
Consider three hypothetical insurers with requirements of 100, 50 and 10 and own funds of 180, 60 and 9. The aggregate ratio is 249/160 = 155.625%. The firm ratios are 180%, 120% and 90%, giving a median of 120%. These statistics convey materially different information. An aggregate above a specified requirement can coexist with one insurer below that requirement; the number of policies and the insurer's role in a critical market would determine the wider significance.
For the PRA's observed stress test population, the reduction of aggregate coverage from 185% to 154% is 31 percentage points. It is a relative reduction of approximately 16.8% of the initial ratio, not a 31% loss of eligible capital. The published stress results also report a GBP 8.6 billion decrease in aggregate surplus and GBP 12.9 billion of assets downgraded below investment grade. These are distinct magnitudes. To infer the absolute change in own funds one also needs the change in the denominator. Scenario resilience is conditional on assumptions and does not exhaust plausible cash-flow stress.
Asset liquidation and cash-flow capacity
A liquidity assessment asks whether obligations can be honored at the time they fall due without unacceptable losses or disruption. A simple horizon-specific economic coverage ratio compares available, unencumbered resources and credible inflows with stressed outflows. Assets must be discounted for haircut, settlement time and market depth; committed facilities should be assessed for availability under the scenario. The outflows should cover claims, surrenders, collateral and financing obligations with timing explicitly represented.
For a hypothetical 30-day scenario, suppose cash is 20, eligible securities have a carrying value of 60, the sale haircut is 10%, and dependable inflows total 15. Stressed outflows of 100 leave available resources of 89 and a shortfall of 11. At a 25% haircut, resources fall to 80 and the deficit grows to 20. These are simple liquidity arithmetic, not a reconstruction of the IAIS Insurance Liquidity Ratio or any regulatory threshold. A facility of 20 might close the numerical gap only if it can be drawn in the correct currency and entity within the horizon.
Liquidity mismatches should be computed by currency and by legal entity as well as group. Imagine the group reports adequate assets while its subsidiary faces collateral demands and local capital restrictions prevent immediate transfer from another entity. Consolidation may hide the binding constraint. A cash ladder reconciles day-by-day payments, settlement lags and collateral changes. Aggregate current assets alone cannot perform this function.
An insurer might choose to liquidate assets with low transaction costs even when those assets are not the source of the original loss. Many firms making the same rational choice may create an adverse price effect. Monitoring must therefore record which assets are likely to be sold, not merely which assets appear to be impaired. Stress tests that assume an unchanged sale price as aggregate sales expand may understate the consequences of collective action.
Investment valuation and credit deterioration
Exposure to assets requiring substantial valuation judgment deserves scrutiny because observed market prices may be scarce and liquidation costs uncertain. Accounting valuation classifications nevertheless measure input observability, not mechanically the number of days until a security is convertible into cash. A model-valued asset can have contractual cash flows, and an exchange-traded instrument can become difficult to sell at prevailing prices in stress.
The IAIS (2025d) methodology update revised the Level 3 definition alongside changes to other indicators and reference denominators. Thus a lower Level 3 score can arise from a methodological change rather than improved asset liquidity. A credible analysis maintains vintage-specific definitions and computes a bridge where observations can be restated.
Credit monitoring should distinguish rating distribution, exposure-weighted expected credit loss, realized default, downgrade risk and collateral requirements. A common rating downgrade can alter capital charges even if expected contractual cash flow is initially unchanged. It can also cause market value declines, reduce collateral eligibility or restrict investment mandates. A risk dashboard needs both exposure amounts and mechanisms through which worsening credit translates into cash needs or impaired capital.
Private credit requires particular care. The absence of high-frequency quotations is compatible with delayed recognition of losses rather than economic stability. Conversely, long contractual maturities can reduce forced-sale risk if liabilities are well aligned. Loan covenants, leverage, borrower cash flows, sector composition, valuation date and refinancing exposures are more informative than a single illiquidity label. The analysis is strongest when asset-side risk is matched against the insurer's actual liability options.
Derivatives and collateral
Derivative positions may hedge duration, currency or equity exposure and thereby reduce economic vulnerability. They may simultaneously expose an insurer to variation margin, changes in initial margin and counterparty dependence. The sign of the net economic exposure, the settlement mechanics, collateral currency, eligible securities and close-out terms all matter.
Notional amounts quantify contractual scale; current replacement value quantifies a valuation measure; potential future exposure describes uncertainty over a defined horizon; available collateral and netting determine credit exposure subject to legal enforceability. These quantities have different dimensions and should not be aggregated into a single monetary amount without conversion and a defined purpose.
The IAIS (2025c) ancillary risk indicators add information for credit, derivatives and reinsurance outside the core weighted scoring methodology. Multiple indicators are justified because exposures that look similar in a headline measure may have different consequences when cash and legal relationships are taken into account. A large hedging portfolio with ample liquidity can be less vulnerable to collateral stress than a smaller, unhedged or poorly collateralized position.
Underwriting and continuity of protection
The loss ratio, expense ratio and combined ratio describe aspects of nonlife underwriting performance. Their definition requires the basis of earned premiums, incurred losses and expense allocation. A combined ratio above 100% on a consistent basis indicates an underwriting loss before investment income and other items. The ratio alone does not determine the firm's overall profitability or cash balance.
Under correlated catastrophe shocks, gross expected loss, ceded recovery, retention, limits, reinstatement costs and coverage gaps should be tracked separately. For a simple loss of 120 with an excess-of-loss treaty paying up to 70 above a 30 attachment, recovery is min(70, max(120 - 30, 0)) = 70 and retained loss is 50 before other terms and credit effects. If the reinsurer pays late, the cedent may have to finance the gross claim until recovery. The distinction between retained ultimate loss and peak gross cash demand is fundamental.
In a market experiencing rapid increases in claims costs, coverage withdrawal may be visible before capital becomes impaired. Observed changes in quote availability, renewal refusal, deductibles, exclusions and affordability provide a different perspective from insurance firm ratios. Such information should be collected with clear denominators: eligible risks, applications, policies or insured values. Coverage contraction cannot be inferred from nominal premium growth alone.
Monitoring architecture of the GME
The IAIS 2025 assessment methodology groups the active weighted indicators into size, global activity, interconnectedness, asset liquidation and substitutability. Absolute assessment uses a benchmark fixed to the year-end 2024 data exercise; relative assessment uses the denominator of the relevant exercise year. The official table includes a zero-weight financial-guarantee item alongside 13 weighted indicators. The methodology also employs a derivatives-trading reference scaling factor and provides ancillary analysis. These features should be respected rather than replaced with an improvised reproduction of the official score.
To illustrate why fixed and contemporaneous denominators differ, suppose a firm's exposure grows from 10 to 12 while the pool grows from 100 to 150. Its fixed-base ratio increases from 10% to 12%; its relative footprint falls from 10% to 8%. A rising absolute indicator may therefore accompany a declining relative rank. Neither arithmetic result is erroneous; each answers a different question.
Changes in the pool, indicator definition or exchange rates require bridge calculations before a difference is attributed to risk-taking. Where a firm enters the pool, aggregate exposure may rise without any incumbent increasing its position. A fixed definition and a matched panel of continuing firms help distinguish composition from business change. When restatement is unavailable, the series break should remain visible.
Concentration Indicators
Concentration as a family of measurements
Market concentration measures the distribution of an economic activity across participants. Exposure concentration measures the distribution of a portfolio across obligors, assets or risks. Network concentration measures the dependency of institutions on a limited set of counterparties or providers. Common-exposure concentration measures how strongly multiple participants depend on the same underlying factor. These concepts are related, but a high concentration in one dimension need not entail a high concentration in another.
Commonly used summaries include the top-k share and the Herfindahl-Hirschman Index (HHI). For nonnegative shares s_i summing to one, HHI is the sum of squared shares. The effective number of equally sized participants is its reciprocal. The 0-to-1 convention and the 0-to-10,000 convention differ by a factor of 10,000. The denominator must be positive and the categories exhaustive and nonoverlapping.
Suppose five insurers account for shares 0.40, 0.25, 0.15, 0.10 and 0.10 of an illustrative domestic market. Their HHI is 0.265, or 2,650 on the 10,000 scale; the top three share is 80%; the effective number is about 3.77. If the two smallest firms merge without market-share changes elsewhere, the HHI becomes 0.285, a 200-point increase. This arithmetic measures ownership concentration, not a change in the probability of a systemic event. It assumes a market definition and unchanged shares that may be inappropriate after a merger.
A concentrated market may support diversified well-capitalized insurers, while a fragmented market may expose all firms to the same catastrophe or borrower. The policy question is how the distribution of underwriting capacity interacts with the cost, speed and feasibility of replacing that capacity in stress. Product heterogeneity, geographic scope, exclusion terms and regulatory access shape substitutability.
Look-through and ownership
A holdings register assigned to immediate counterparties can suggest diversification when several funds invest in the same ultimate issuer. Look-through analysis maps assets to underlying borrowers and relevant risk factors. It requires attention to fund structures, seniority, leverage, guarantees and hedges. Missing look-through exposures should remain separately classified rather than distributed equally or treated as zero.
Ownership is another dimension. Multiple insurers with a common parent may share investment managers, financing arrangements, counterparties and operational infrastructure. Legal separateness matters for capital transfer and policyholder rights, while group commonality matters for behavior and governance. An analyst should calculate concentration at legal-entity and group levels, documenting the ownership date and control definition.
Suppose two portfolios each hold four securities equally. If both portfolios own the same four issuers, each portfolio and the aggregate have HHI 0.25. If their issuers are disjoint and each portfolio has the same total size, the combined HHI becomes 0.125. Individual diversification statistics are identical, while sector vulnerability to the failure of any shared issuer differs. This example uses exact ultimate-issuer holdings, and assumes equal exposures and no dependence across separate issuers. Common macroeconomic drivers may undermine the latter assumption, which is why sector and factor concentration must also be examined.
A common-holdings overlap measure for two portfolios can be defined as the sum across assets of the smaller of their portfolio weights. It ranges from zero for disjoint fully observed portfolios to one for identical allocations. The measure is descriptive and symmetric; it does not quantify the impact of simultaneous liquidation. In particular, two overlapping portfolios dominated by deep sovereign markets may behave differently from two portfolios dominated by small private borrowers.
Reinsurance and funded risk transfer
Reinsurance spreads the burden of underwriting shocks but creates contractual dependence between cedents and risk bearers. The relevant network is directional: a cedent relies on a reinsurer for payment and a reinsurer relies on a cedent for premium or collateral flows. Exposure should be measured using recoverables, stressed payouts, collateral quality, legal netting and maturity. Gross treaty notional, ceded premium and expected recoverable are not interchangeable.
Consider a cedent with hypothetical recoverables of 45, 30, 15 and 10 across four ultimate counterparty groups. The recoverables HHI is 0.325. If the first two counterparties belong to one group, the group-level HHI becomes 0.595. This is a change in the mapping of the same economic positions rather than in the underlying claims. It demonstrates why a counterparty registry needs consistent identifiers and beneficial ownership relationships.
Concentration is often most consequential under contingent obligations. A reinsurer with diverse cedents may nevertheless suffer simultaneous claims owing to a catastrophe, widespread cyber event or common financial guarantee. Collateral can reduce credit losses while creating operational or market risks if assets must be realized under unfavorable conditions. Contract terms determine whether assets and liabilities are transferred, remain on the cedent's balance sheet, or return under a recapture arrangement.
The PRA's 2025 exploratory scenarios explicitly considered asset concentrations and funded reinsurance. Their separate treatment is instructive: a cedent can inherit risks from assets associated with a reinsurance arrangement even if the initial transfer improved its solvency position. Recapture analysis should test the availability, quality, currency and timing of assets returned, changes in liabilities and the ability to manage the resulting portfolio. These considerations extend the documented scenario topic; they are analytical implications, not estimates of the actual UK firms' recapture losses.
Network topology and critical services
A contractual network can be represented by a directed weighted matrix E, where E_ij is a specified exposure of insurer i to counterparty j. Outgoing totals measure a firm's gross dependence under that definition; incoming totals measure the counterparties' combined exposure to the institution if the orientation is reversed. Before computing any centrality, the analyst must specify whether arrows represent claims, liabilities, cash or risk transfer.
Network centrality can help identify highly connected nodes. However, a central node's systemic relevance depends on the loss given disruption, substitutability, contingent resources and direction of transmission. A central service provider without large balance-sheet exposures may interrupt claims processing. A financial counterparty with high gross claims may be less risky when exposures are collateralized and legally nettable.
A minimal stress exercise can calculate the first-round loss under a counterparty default and then trace capital or liquidity breaches. Suppose three insurers have unsecured claims of 40, 25 and 10 against a counterparty and a hypothetical recovery rate of 60%. First-round losses are 16, 10 and 4. To infer second-round losses, one needs the institutions' remaining capital, liquidity, obligations and sale behavior. Summing gross positions with first-round losses double counts exposure; summing sequential write-downs without recording recoveries may also double count.
Operational concentration requires its own data. A shared cloud platform, claims administrator or catastrophe model vendor may be critical without holding underwriting liabilities. Monitoring should ask whether outage scenarios affect multiple firms at once, whether backup providers share infrastructure, and whether contractual and technical exit plans are executable within tolerable interruption windows. Financial solvency ratios cannot establish service resilience.
Distributional risk and tail dependence
Average concentration is informative only when the distribution is stable enough for the mean to represent relevant vulnerabilities. A concentration dashboard should therefore include portfolio-weighted aggregates, quantiles, the largest exposures and the fraction of reporting firms exceeding a predefined screening level. If the data are confidential, aggregated tail bins can sometimes convey vulnerability without disclosing firm identities.
Dependence can become stronger under stress. A normal-period correlation matrix estimated from market returns may underestimate coincident losses in crisis if volatilities, liquidity and behavioral responses shift. Estimation of tail dependence requires extensive data, consistent observation windows and careful treatment of structural changes. A stress scenario may be more useful than a poorly estimated tail coefficient when extreme events are rare.
A risk concentration measure should be subjected to sensitivity analysis for classification choices. For an issuer with uncertain economic ownership, compute both a separate-counterparty and a consolidated-group case. For investments in opaque funds, report known categories and unknown shares. A range can be decision-relevant; false precision can hide the uncertainty that a supervisor actually needs to address.
Funded reinsurance illustrates how the concentration measure becomes operational. In its July 2024 policy statement, the PRA's funded reinsurance policy explains the purpose of three distinct exposure limits: loss following recapture from one counterparty, the consequences of recapture from correlated counterparties, and the operational consequences of total exposure. An immediate recapture calculation precedes discretionary management responses. The rationale matters for monitoring because several firms may plan to sell or rebalance the same assets under stress. Their individually plausible actions may become collectively infeasible. A model should therefore identify the gross recapture obligation, the quality and accessibility of collateral, the capital impact before discretionary responses, and the operational capacity required to manage returned obligations. The firm-specific nature of the policy's risk-appetite assessment prevents translation of this historical statement into a universal numerical statutory trigger.
Supervisory Triggers
Signals, legal thresholds and intervention
A trigger is a specified condition prompting a supervisory review or action. Its meaning depends on the authority invoking it. A statistical anomaly identifies a deviation from a benchmark. A prudential breach establishes a condition defined in law or regulation. An intervention criterion permits or requires action under specified facts and powers. The consequences of these three conditions are not equivalent.
Indicators first support triage. A rapid expansion of exposure can prompt validation of data, discussion with a firm and closer monitoring. A stressed liquidity deficit can prompt examination of the business model and liquidity management plan. A legal intervention condition may authorize restrictions, recovery measures or other actions. The choice should follow the identified risk and the actual legal framework, including procedural requirements and proportionality.
The FSB's Holistic Framework description connects monitoring, supervisory measures and the assessment of implementation. This institutional structure favors a feedback process: identify the vulnerability, evaluate mitigation, determine whether more action is warranted and assess the subsequent result. A score published without a path for analysis and response is insufficient for supervisory practice.
Designing interpretable signals
An indicator register should record definition, data owner, observed universe, numerator, denominator, valuation, consolidation, frequency, missing-data rules and version. It should also identify the mechanism through which a high or low value is concerning, expected direction, plausible alternative interpretations and the decision-maker responsible for reviewing an exception.
A two-dimensional approach can distinguish the level of an exposure from its recent change. A small exposure doubling from a modest starting point may remain manageable, while a large stable exposure may be dangerous owing to its concentration. A rate of growth should be paired with level and denominator. For time series with seasonality, comparison with the same season or a suitable adjustment is necessary.
Robust screening methods include medians, percentiles and median absolute deviations. If an indicator distribution is asymmetric or bounded, a normal-theory z-score can be misleading. A median-based scale is undefined or unstable if the median absolute deviation is zero; alternative dispersion measures or clearly labeled exact ranks should then be used. Percentile methods can identify outliers even if the entire sector's risk increases, so they must be paired with fixed-base or economic levels.
In a synthetic quarterly series, a liquidity coverage metric falls from 1.35 to 1.20, then 1.02 and 0.92. A purely illustrative review threshold of 1.10 with a persistence requirement of two consecutive quarters activates on the fourth observation. A separate urgent signal at 1.00 would also activate in the fourth quarter. If the third reading had been 0.95, the urgent rule would have responded earlier, while the persistence rule still waited for confirmation. The values are teaching examples, not supervisory rules.
Persistence reduces noise but can delay response. A high-frequency shock can justify review without waiting for another quarter, while a minor reporting anomaly may merit data verification first. A sensible surveillance process calibrates frequency and escalation to the loss mechanism and uses legally authorized emergency procedures where circumstances warrant them.
Scenario analysis and model uncertainty
A scenario sets conditions and evaluates their consequences; it is not a probability forecast unless a probability model is supplied. The severity and horizon of a shock should be specified together with initial positions, liability cash flows, collateral mechanics, reinsurance recoveries, behavioral assumptions and possible management actions.
A simple interest-rate sensitivity illustrates the importance of assets and liabilities together. With asset market value 1,000 and effective duration 7 years, liability value 900 and duration 10 years, the first-order change in surplus for a parallel yield rise of 100 basis points is approximately -70 - (-90) = +20. A rate fall of equal size produces approximately -20. The computation omits convexity, spread differences, embedded options, derivatives, capital requirements and liquidity. It demonstrates that the sign of economic surplus sensitivity can differ from that of asset valuation. Additional margin calls could still cause a liquidity gap even as modeled surplus rises.
A scenario based on static balance sheets answers what would happen if positions do not change. A dynamic scenario includes actions such as asset sales, new hedges, financing, premium changes and changes in underwriting. Dynamic modeling is informative only if management actions are feasible under the scenario and do not assume every firm can sell the same assets at undisturbed prices. A transaction that improves a single firm's liquidity may impose losses on other institutions.
Reverse stress testing identifies combinations of outcomes that would produce a specified failure condition, such as inability to meet obligations or a sharp reduction in capacity. It helps reveal exposure to assumptions absent from conventional scenarios. It should specify whether the failure condition is a legal breach, an economic liquidity deficit, or service disruption; these distinct objectives may generate different adverse configurations.
The PRA exercise distinctly separates its principal market scenario from exploratory concentration and reinsurance conditions. This is a useful example of varying the stress question rather than reducing all risks to one capital figure. The PRA report makes clear that the test is not a pass/fail process. Its findings may guide further investigation, while the legal source of any actual capital or intervention requirement lies elsewhere.
Exceptional intervention in European law
The amendments introduced by Directive (EU) 2025/2 include articles addressing liquidity risk management and exceptional supervisory intervention. Article 144b provides for temporary restrictions in circumstances of material liquidity risk that may imminently threaten policyholder protection or financial stability. Suspension of redemption rights is framed as a last-resort action in exceptional circumstances, with collective policyholder interests and potential adverse consequences to be considered. Article 4 establishes the future general application date described earlier.
This legal structure illustrates why a numerical liquidity score cannot automatically justify suspension of individual contractual rights. Supervisors must evaluate the legally relevant circumstances and the relationship of the measure to the threat. Restrictions on distributions and restrictions on policyholder redemptions impose different burdens. A measure that preserves assets but damages confidence may affect future policyholder behavior or shift cash demands to other institutions.
The comparison does not assign the amended directive to UK insurers or to US insurers. Any jurisdiction-specific legal conclusion requires the applicable domestic law, implementation date and facts. International standards supply supervisory expectations; enforceable legal obligations arise through applicable legal instruments.
Validating warning indicators
An early warning indicator should be evaluated against a prespecified event definition. Candidate outcomes include a cash-flow shortfall, delayed payment, supervisory action, disorderly failure, sustained reduction in essential coverage or measurable adverse spillover. Using only firm failures may miss systemwide instability in which authorities prevent failure or institutions collectively curtail services.
A validation set should use historically available data, preserve the dates on which observations became available, and prevent data leakage. Threshold selection should occur on a training or calibration period; out-of-sample evaluation should test stability across periods and jurisdictions. Cross-validation that randomizes adjacent observations may overstate accuracy when financial series are autocorrelated. Rare crisis events also require precision-recall analysis and disclosure of uncertainty.
Consider a synthetic panel of 1,000 observations, with 20 independently identified adverse outcomes. A candidate rule flags 50 observations and catches 12 outcomes. Its precision is 12/50 = 24%; its recall is 12/20 = 60%; there are 38 false positives and 8 false negatives. Specificity is 942/980, approximately 96.1%. Overall accuracy is 954/1,000 = 95.4%, whereas a rule always predicting no outcome would achieve 98%. The higher naive accuracy does not establish better supervisory usefulness because missing all adverse events can be costly. This illustration shows why an event-rate-weighted objective is required rather than raw accuracy alone.
Supervisory intervention complicates retrospective validation. If a warning causes an effective remedial action and the adverse event never occurs, treating the observation as an ordinary false positive can punish useful signals. Evaluation should preserve action dates and describe how outcomes changed after intervention. Establishing a causal effect requires an identification strategy addressing nonrandom treatment; it cannot be inferred simply from pre/post comparisons.
Cross-border coordination and reporting
A group that receives conflicting instructions in multiple jurisdictions may find timely response difficult. Supervisory colleges can coordinate group-wide evaluation while preserving the role of local legal-entity supervision. The IAIS ICP 25 describes responsibilities for coordination, sharing information, assessing group and local risks and preparing for crisis management.
A useful communication record distinguishes preliminary concerns from established breaches and includes the facts shared, time of transfer, applicable confidentiality basis and intended decision. Cross-border data exchange must respect legal restrictions and confidentiality. The group supervisor's view of consolidated resources should be reconciled with local supervisors' knowledge of ring-fencing and claim obligations. Timeliness matters especially for collateral calls and counterparty distress.
Integrated Scientific Synthesis and Future Research
Connecting observations to a transmission mechanism
The evidence reviewed supports a practical conclusion about interpretation: monitoring should connect the observable indicator to the balance-sheet or service mechanism through which distress could become collective harm. Exposure scale identifies the potential magnitude; concentration identifies dependency; liquidity identifies timing; capital measures absorption; and substitutability identifies consequences of interruption. These are ordinary dimensions of existing financial analysis, not a proposed new theoretical system.
Take the hypothetical example of several insurers holding long-duration bonds and hedging interest rate sensitivity with derivatives. An adverse price movement might affect valuations and margins simultaneously. A supervisor would need the net duration exposure, collateral obligations by currency, liquidity buffers by legal entity, assets available for sale, overlap in sale candidates and capacity of intermediaries to absorb aggregate sales. The same observed derivative notional could describe portfolios that are economically stabilizing, destabilizing or both at different horizons.
The surveillance record must also contain mitigating factors. A diversified counterparty structure, collateral held in an enforceable arrangement, a reliable committed source of liquidity and capacity to transfer policies can interrupt transmission. Indicators should measure these protections without assuming that they remain effective under every scenario. A credible conditional conclusion names the conditions under which the mitigating resource is accessible.
Comparative case findings
International monitoring and domestic supervisory programs exhibit a division of functions. The IAIS methodology assesses both sector activities and concentration in major groups. European dashboard analysis combines prudential and market information at different observation dates. The NAIC's program includes liquidity testing and group-related concerns. The PRA's stress exercise illuminates the conditional balance-sheet behavior of a particular annuity-oriented population.
A general lesson follows from these documented differences: the adequacy of monitoring must be evaluated relative to the method's purpose. A broad public dashboard may lack the bilateral data needed for network simulation. A confidential group-level stress test may have considerable detail but limited information on smaller firms. A global index may describe comparable footprint while omitting local legal restrictions. Each is a necessary component of the larger monitoring process and each supplies different evidence.
The 2025 and preliminary 2026 IAIS publications demonstrate that the risk-score series should be accompanied by methodology and sample vintages. A numerical score combines reporting definitions, weights and denominators. An empirical conclusion about the direction of underlying risk requires information on these components. In particular, the revision of the Level 3 classification, reporting pool and base reference values is analytically significant. The study therefore avoids calculating a continuous growth rate from different report vintages.
The UK scenario supplies a contrasting example: the initial and stressed SCR ratios are comparable within the specified exercise but are not directly equivalent to the IAIS systemic footprint. The former measures capital relative to modeled requirements, the latter combines activities and exposures. Relating them requires a causal account, not a conversion factor.
A reproducible data collection design
For future longitudinal work, the observational unit should be defined before constructing a panel. Each record would identify the insurer legal entity and group, reporting jurisdiction, activity classification, period end, filing availability date, currency, valuation basis and data version. Exposure records would contain both immediate counterparty and ultimate owner where known, instrument type, contractual maturity, collateral, valuation classification and liquidity horizon. Policy records could be aggregated by line and geography, with explicit definitions of coverage and claims.
A separate catalog of legal instruments should record enacting authority, instrument type, adopted date, commencement date, geographical reach, relevant population and operative trigger. A supervisory event register should identify intervention time, trigger, type of action, authority and the observed outcome. Confidential microdata can be protected through authorized access, aggregation and carefully documented disclosure control. Public analysis should remain reproducible at the aggregate level where possible.
Missing observations should be assigned reasons: unavailable, not applicable, not reported, confidential, restated or not comparable. Treating every missing exposure as zero can bias lower the footprint and concentration estimates. For concentration, an unknown share can be used to produce scenario bounds. For liquidity, unknown cash outflows may preclude a credible lower-bound assessment until data have been obtained.
Version control enables an independent analyst to distinguish a corrected report from a new economic observation. Where exchange rates translate local exposures into a reporting currency, retain both the native and translated amounts. Nominal valuation movements may reflect currency fluctuations, repricing or actual acquisitions. A decomposition should separate these when data permit.
Hypotheses for subsequent empirical testing
Several questions are available for research using adequate microdata. Does overlap in holdings predict coincident sales after controlling for common market shocks and firm size? Do contractual features of cash-value policies explain surrender sensitivity after accounting for market rates, duration and penalties? Does funded reinsurance concentration explain variation in modeled recapture losses once collateral and contractual terms are considered? Does independent identification of ultimate counterparties materially alter concentrations relative to immediate-counterparty data?
These are testable research questions, not established findings. The appropriate methods depend on data availability. For market-return analysis, regressions and tail measures may characterize co-movement but require careful identification before being interpreted as transmission. For balance-sheet networks, scenario simulation can examine propagation conditional on assumptions. For policy design, an event-study or quasi-experimental design would require comparable treated and untreated populations and a plausible identification strategy.
Policyholder protection can be evaluated in parallel with financial stability. An intervention restricting cash outflows may reduce forced sales but limit access to policy funds. Tightening underwriting can preserve capital while increasing the protection gap. The research should record who bears these costs, their duration and whether alternative service providers are available. An action's aggregate benefit is not established by a single institution's improved ratio.
Limitations and scholarly boundaries
The dossier draws on documented institutional procedures and public observations. It does not contain confidential individual-insurer GME submissions, bilateral exposure matrices, nationwide US statutory implementation checks, individual policy contract data or a replicated econometric study. The older scholarly materials are used within the scope of accessible abstracts and bibliographic records. The European directive is analyzed as adopted text with future general application, not by surveying national transposition. The mid-year IAIS assessment remains preliminary. These constraints determine the strength of the conclusions.
The numerical examples demonstrate arithmetic and identification problems. They cannot estimate actual risk distributions or validate a supervisory threshold. A real quantitative comparison would require aligned observation dates, matched populations, restated variables and model validation. The research has not established a new systemic-risk measure, novel causal model or regulatory doctrine.
Conclusion
Insurance sector macroprudential monitoring is an exercise in translating heterogeneous contractual and institutional information into evidence about collective financial vulnerability. The reviewed international and domestic programs show why surveillance must include both individual institutions and the activities they share. A credible assessment distinguishes capital available to absorb losses, liquidity available to meet obligations, the distribution of counterparties, common exposures and the replaceability of economic services.
Concentration is a property of a defined population and denominator. Its significance depends on market structure, balance-sheet composition, ultimate ownership, contractual terms and adverse contingencies. Aggregate capital strength can coexist with severe vulnerabilities in a segment, and stable individual ratios can conceal collective exposure to the same asset or service provider.
A supervisory trigger obtains practical meaning from the mechanism it identifies and the authority under which action is taken. Statistical anomalies invite investigation, conditional stress results describe vulnerability and legal thresholds govern enforceable measures. Treating those distinct classes as interchangeable undermines both empirical understanding and accountable supervision.
The primary scientific contribution of this study is the comparative interpretation of these dimensions and their measurement boundaries. The strongest evidence supports integrated, version-controlled monitoring that retains distributional information and makes assumptions about contractual cash flows, counterparty exposure and management actions explicit. Subsequent research using confidential, harmonized and longitudinal data can test the proposed empirical questions. The present conclusions remain bounded by the public documentary record and the stated analytical design.
References
FSB (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
International Association of Insurance Supervisors (IAIS) (2025a). Global Monitoring Exercise. November 2025. Basel: IAIS. Source
IAIS (2025b). Global Insurance Market Report 2025. December 2025. Basel: IAIS. Source
IAIS (2026). Global Insurance Market Report 2026: Mid-year update. July 2026. Basel: IAIS. Source
IAIS (2026). IAIS mid-year Global Insurance Market Report 2026 reflects insurance sector stability amid global uncertainty. July 2026. Source
IAIS (2025c). IAIS introduces additional ancillary risk indicators for its Global Monitoring Exercise. November 25, 2025. Source
European Insurance and Occupational Pensions Authority (EIOPA) (2026). October 2026 Insurance Risk Dashboard. EIOPA-BoS-26-495. Frankfurt: EIOPA. Source
European Parliament and Council (2024). Directive (EU) 2025/2 of 27 November 2024 amending Directive 2009/138/EC as regards proportionality, quality of supervision, reporting, long-term guarantee measures, macro-prudential tools, sustainability risks and group and cross-border supervision, and amending Directives 2002/87/EC and 2013/34/EU. Official Journal of the European Union, L, 8 January 2025. Source
Prudential Regulation Authority (PRA) (2025). Life Insurance Stress Test: 2025 Results. Bank of England, November 24, 2025. Source
National Association of Insurance Commissioners (NAIC) (2025). Macroprudential Supervision. Updated December 19, 2025. Source
Bierth, C., Irresberger, F., and 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. Source
Eling, M., and Pankoke, D. A. (2016). Systemic Risk in the Insurance Sector: A Review and Directions for Future Research. Risk Management and Insurance Review, 19(2), 249-284. DOI: 10.1111/rmir.12062. Source
Cummins, J. D., and Weiss, M. A. (2014). Systemic Risk and the U.S. Insurance Sector. Journal of Risk and Insurance, 81(3), 489-528. DOI: 10.1111/jori.12039. Source
IAIS (2024). Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups: ICP 25, Supervisory Cooperation and Coordination. December 2024 standard, online version. Source
IAIS (2025d). IAIS publishes updated Global Monitoring Exercise (GME) document. November 25, 2025. Source
Eling, M., and Pankoke, D. (2014). Systemic Risk in the Insurance Sector: Review and Directions for Future Research. Working paper, November 2014. Institute of Insurance Economics, University of St. Gallen. Source
Cummins, J. D., and Weiss, M. A. (2013). Systemic Risk and Regulation of the U.S. Insurance Industry. Working paper, Temple University. Source
Prudential Regulation Authority (PRA) (2024). PS13/24 – Funded reinsurance. Bank of England, July 26, 2024. Source