RESEARCH DOSSIER
Systemic Risk and Contagion
Abstract. Systemic risk arises when shocks, vulnerabilities, contractual obligations, market behavior, and institutional responses jointly threaten the continued provision of financial and economic functions. Its scientific analysis requires an explicit distinction among aggregate exposure to common shocks, statistical dependence in the tails of distributions, and transmission caused by interactions. A financial institution can be highly vulnerable during a systemic event without materially amplifying that event; conversely, a seemingly modest institution or activity can transmit disruption through liquidity obligations, network positions, asset sales, or critical services. This monograph develops an integrated analytical account of systemic risk and contagion, with particular attention to the interpretation of conditional value at risk (CoVaR), marginal and systemic expected shortfall, conditional capital shortfall (SRISK), clearing networks, market liquidity, and funding feedback. It explains why an observed conditional association is not by itself a causal contribution, why connectivity can diversify small shocks yet spread sufficiently large shocks, and why a solvency ratio alone cannot establish short-term liquidity resilience. The insurance analysis combines sector-wide common exposures, reinsurance concentrations, asset–liability mismatch, derivatives and collateral calls, catastrophe claims, run and surrender behavior, and interactions with banking and nonbank financial intermediation. Historical documentary evidence from American International Group during 2008 is assessed against supervisory developments in the International Association of Insurance Supervisors' Holistic Framework, the Insurance Capital Standard, and the Financial Stability Board's macrofinancial work. Transparent stylized numerical cases demonstrate conditional-tail measurement and liquidity amplification, while specifying their assumptions and the limits of inference. The conclusion establishes a source-based comparative interpretation: risk assessment must trace the route from a shock through exposures, behavior, balance sheets, liquidity demand, and economic functions, and must evaluate mitigation under multiple plausible regimes. The work contributes an original synthesis and explicit methodological interpretation without asserting a novel formal theory, model, or empirical discovery.
Keywords: Systemic risk; financial contagion; insurance; tail dependence; CoVaR; marginal expected shortfall; systemic expected shortfall; SRISK; conditional capital shortfall; financial networks; common exposures; macroprudential supervision; reinsurance; liquidity spirals; fire sales; collateral calls; stress testing; intersectoral spillovers.
Introduction
Systemic risk is a problem of interacting exposures and consequences. The unit of interest is a system of financial claims, risk transfers, market infrastructure, operational services, and institutions whose functioning supports insurance protection, payment, credit, investment, and real economic activity. An isolated institutional loss can be substantial yet remain absorbable. A shock can instead induce several institutions to sell similar assets, post collateral simultaneously, withhold funding, curtail capacity, or fail to honor contingent promises. These reactions may enlarge the initial disturbance and disrupt services beyond the party where the shock originated. The distinction requires a description of the transmission mechanism, timing, institutional perimeter, and response of agents under stress.
An essential research problem is the ambiguity of the expression systemic importance. Large assets, large insured liabilities, and extensive connectedness provide useful signals, but a ranking based on any one variable does not identify an institution's marginal causal effect. The Basel Committee's G-SIB assessment methodology deliberately adopts multiple indicators and supervisory judgment for banks; its classification should not be transferred mechanically to insurers. For insurance, the IAIS Holistic Framework treats vulnerabilities and activities at individual and sector-wide levels. These approaches represent different regulatory instruments and populations rather than equivalent scores.
The research method is documentary, conceptual, and quantitative. It compares established formal constructs, checks what the source actually measures, reconstructs selected balance-sheet and liquidity channels, and tests the internal arithmetic of hypothetical examples. The monograph reads foundational research in its institutional context and checks current standards against issuer documentation. Its principal external evidence is drawn from publisher records, supervisory texts, official historical accounts, and standard-setting organizations. Many journal landing pages provide abstracts and bibliographic metadata but not unrestricted full-text access; arguments that require additional proofs or empirical replication are accordingly attributed to the authors rather than presented as independently reproduced findings.
The scope extends beyond banks to insurance and the wider network of intermediaries. Losses arising from catastrophes, climate, cyber incidents, credit deterioration, funding interruptions, and correlated portfolio investments may overlap. A common hazard may generate simultaneous insurance claims while also depressing financial assets held to pay those claims. A second-round problem appears if insurers or counterparties alter behavior in ways that amplify market prices, funding costs, or insurance capacity. Throughout, statistical risk indicators are kept separate from mechanistic simulations and causal identification. This separation determines which claims the evidence can support.
Source and interpretation discipline
The following analysis uses risk-measure definitions under expressly stated loss sign conventions. It distinguishes an observed historical event from a stylized counterfactual and differentiates international supervisory principles from national legal obligations. Whenever an example selects hypothetical parameters, those values are labeled as assumptions. Existing financial-network results, CoVaR, systemic expected shortfall, SRISK, and liquidity-spiral theories retain their original scholarly attribution. The original authorial contribution is the comparative application of these works to insurance, an explicit analysis of their interpretive boundaries, and a set of reproducible cross-sector illustrations.
Systemic Risk Concepts
The object of systemic risk
A useful working description of systemic risk is the possibility that impairment in parts of the financial system disrupts financial services and causes serious economic harm. The IAIS macroprudential standard addresses both exposures that make insurance undertakings vulnerable to system-wide stress and exposures or behaviors through which an insurer or an insurance sector may create disruption outside itself. The concept is inherently relational: it contains a financial-function criterion and a real-economy consequence, not merely a threshold of individual insolvency probability.
Consider the relation among a natural catastrophe, insurer loss, and systemic consequence. A severe storm may generate large concurrent claims for property, business interruption, and reinsurance. Claim aggregation can be costly without automatically causing a system-wide financial breakdown. Whether the event becomes systemic depends on solvency and liquidity buffers, settlement schedules, correlation across firms, reinsurer capacity, funding markets, policyholder behavior, and the economic services that insurance continues to provide. The initial event, joint damage, and subsequent transmission must be represented as distinct analytic stages.
The system perimeter matters as much as the shock. A geographically concentrated catastrophe may be systemic for a regional insurance market and local credit network while leaving global financial intermediation functional. A failure of an internationally important counterparty may be systemically relevant across jurisdictions. The empirical unit may be a legal entity, insurance group, entire sector, or combination of banking, insurance, investment funds, and market infrastructure. A conclusion must specify whose welfare or continuity is at stake and the horizon over which the conclusion is evaluated.
Vulnerability, contribution, transmission, and consequence
Four dimensions clarify empirical results. Vulnerability means the expected harm to entity i from a specified system state. Contribution means the change in system outcomes under a credible counterfactual change in i's exposures, behavior, or failure. Transmission identifies the contract, transaction, pricing interaction, or operational dependency along which a shock travels. Consequence specifies the loss of economic functions, output, insurance protection, or other socially relevant service. A conditional estimate of losses to an insurer in a financial crash measures primarily vulnerability. A model of creditor losses following a failed clearing counterparty can help estimate transmission and contribution if its network and behavioral assumptions are valid. Both are relevant, but they answer different questions.
Let S denote system distress and L_i the loss of entity i. The conditional statistic E[L_i | S] differs fundamentally from E[S | L_i], even before discussing causal intervention. The first asks how much i tends to lose when the system is distressed. The second asks about systemic outcomes in observations where i suffers a loss. Neither equals E[S | do(failure of i)] under a general causal model. Common exposures, reverse causation, anticipatory policy intervention, and sample selection can affect both statistical expressions. Causal evaluation requires intervention design or justified structural assumptions.
A source-based interpretation therefore avoids converting a high risk statistic into an assertion that the firm caused a crisis. The original CoVaR article studies system risk conditional on institution states and reports predictive associations. Acharya and colleagues' systemic expected shortfall analysis embeds capital deficiency in a model of systemic externalities. These are related but distinct objects. Neither replaces a complete study of insurance exposures or a causal crisis reconstruction.
Events, time horizons, and perimeter
Systemic events involve dynamic processes. For a bank, withdrawal pressures or wholesale funding conditions can change within hours. For an insurer, gross catastrophe claims may be estimable over weeks while cash settlement, reinsurance recoveries, reserve development, and capital restoration unfold over months or years. Market losses can be instantaneous even when insurance liabilities are long-duration. A one-day trading quantile, a six-month capital shortfall scenario, and a multi-year insurance run-off analysis therefore cannot be compared as though they describe the same random event.
The scale of aggregation also shapes the finding. Consolidating a group may cancel intra-group assets and liabilities; a legal-entity analysis must retain them when local policyholder protection or resolution law applies. Gross claims are not net retained claims, and net claims are not immediately available cash. Accounting values, realizable market proceeds, and collateral eligibility are separate dimensions. A convincing systemic assessment must identify measurement currency, netting rules, intra-group constraints, regulatory valuation, eligible own funds, and the speed at which assets can be turned into unencumbered cash.
Time-varying correlations add another distinction. The cross-sectional diversification benefit inferred during an ordinary regime may overstate resilience during a regime of common investor deleveraging. Conversely, correlation measured during a crisis may partly reflect concerted official policy, portfolio rebalancing, or shared discount-rate shocks rather than a permanent causal network. Robust analysis reports which channels are observed, assumed, or counterfactually simulated and which time windows support the inference.
Insurance and the system boundary
Insurance allocates risk through premiums, reserves, capital, contracts, and reinsurance. It also holds significant financial assets, participates in derivatives and securities-lending markets, and offers long-term guarantees. The assets may be claims on the same institutions whose failure raises insured losses or impairs liquidity. Thus an insurer can face inward transmission from financial markets and produce outward transmission through asset sales, contractual claims on counterparties, withdrawal of insurance capacity, or collapse of a critical risk-transfer function. The same institution may be a net stabilizer for one shock and a propagator for another.
Insurance specialization further undermines a single sector-wide narrative. Many traditional policies are prefunded through premiums and include payment patterns that allow some management of liquidity. Other products involve surrender rights, collateral requirements, guarantees, concentrated catastrophe exposures, or short-horizon funding commitments. The distribution of exposure, contractual architecture, group structure, and actions under stress determine actual vulnerability. The insurer's legal label supplies the regulatory context but does not settle its systemic channel.
The IAIS insurance market surveillance examines sector-wide assets, liabilities, liquidity and solvency, and evolving themes such as private credit, geoeconomic fragmentation, and operational risks. Its reporting confirms that modern insurer systemic analysis includes the asset side as well as underwriting. This premise directs the remaining chapters: metrics are informative only once their underlying financial functions and loss-generation mechanisms have been described.
Common Shocks and Contagion
Common exposure and endogenous transmission
A common shock causes multiple losses through shared exposure to a single external factor. Suppose two insurers hold correlated sovereign bonds and both face a change in yields. Losses occur jointly even if neither insurer trades with the other. Financial contagion instead denotes a propagation process in which a disturbance in one node, market, or balance sheet contributes to additional distress elsewhere through obligations, price effects, expectations, or behavior. Both may operate in the same event. Observational data can struggle to distinguish them, because a common shock makes prices and failures move together even without direct transmission.
In a simple factor representation, a loss is L_i = beta_i F + epsilon_i, where F denotes a common shock and epsilon_i captures remaining specific variation. With independent residuals, Cov(L_i,L_j) = beta_i beta_j Var(F). Joint tail losses can therefore arise entirely from the same factor. Adding direct exposures or adaptive behavior changes this mechanism; the residual independence assumption then ceases to hold. A model that attributes every covariance to contagion will overcount causal transmission. A model that assumes independent residuals when cross-holdings are present will undercount it.
Separating channels matters for policy. Diversification across business lines with the same macroeconomic sensitivity can reduce idiosyncratic fluctuations while leaving concentration in the factor. Capital requirements may absorb first-round losses; liquidity reserves may meet immediate collateral demands; network exposure limits may reduce second-round losses; operational substitutes may preserve essential services. Each intervention addresses a different segment of the transmission chain, and their effectiveness should be evaluated with the chain explicitly stated.
Obligations and clearing networks
An influential formal treatment is Eisenberg and Noe's systemic clearing model (2001). It represents interlocking nominal obligations and determines a vector of payments consistent with limited liability and creditor priority. Each institution can use its external assets and incoming payments to settle its own nominal liabilities, subject to available resources. The resulting clearing payments are jointly determined; a default can reduce receipts to creditors, which can in turn reduce their capacity to pay. Under stated regularity conditions a clearing vector exists and is unique.
A compact representation uses external assets x_i, nominal total debt bar p_i, and the fraction Pi_ji of institution j's total payment owed to institution i. With proportional repayment and no bankruptcy costs, p_i = min{bar p_i, x_i + sum_j Pi_ji p_j}. The system of equations is simultaneous. This formula is a simplified description, not a claim that every actual insurance or derivatives resolution follows proportional repayment. Collateral priority, close-out netting, policyholder preference, separate-account restrictions, and local resolution laws change the payoff structure.
A two-node example makes the distinction visible. Institution A owes 40 to B and has external resources 30; B owes no money to A. Without other liabilities, A pays 30 and B loses 10 on its receivable. If B required that receivable to meet its own creditors, the initial shortfall may propagate. If B had surplus liquidity and equity, the same contract shortfall could be absorbed with no further default. The contract creates a potential path; buffers and legal priority determine the realized cascade. In practice one must distinguish bilateral gross commitments from enforceable net amounts, and contingent insurance recoveries from immediately enforceable financial claims.
Detailed network topology merits its own specialized research treatment. For this dossier the clearing mechanism establishes why joint payoffs matter without supplying a universal mapping of all insurer–reinsurer–bank links. Missing bilateral data cannot be replaced by arbitrary symmetry assumptions without disclosing model risk. Exposure totals alone cannot recover the direction, seniority, maturities, collateral, and recovery rates needed for an empirically credible clearing estimate.
Connectivity and nonlinear fragility
Allen and Gale (2000) investigate financial contagion across regions, where liquidity sharing through interbank claims can support efficient risk allocation yet expose the system to contagion under incomplete linkages. Acemoglu, Ozdaglar, and Tahbaz-Salehi (2015) extend the study of network structures and demonstrate in Systemic Risk and Stability in Financial Networks that greater network density can diversify sufficiently small shocks but spread sufficiently large shocks. Connectivity has a state-dependent influence on outcomes, so a network statistic without shock size and balance-sheet capacity is inadequate as a policy conclusion.
This result has an insurance analogy, with important institutional differences. Reinsurance diversifies the direct consequences of insured events across capital providers. It also creates recoverable assets, counterparty obligations, and concentration in global reinsurance markets. If a reinsurer's protection is collectible when it is needed, risk transfer provides resilience; if multiple cedents depend on the same impaired reinsurer, recoverability becomes a shared vulnerability. A contract's nominal transfer amount describes intended risk allocation, while effective transfer during an extreme event depends on counterparty solvency, timing, collateral, exclusions, aggregation provisions, and dispute resolution.
The interpretation also depends on direction of causality. A reinsurer can fail because many cedents share a catastrophe shock, or its failure can worsen cedents' positions. An observed clustering of financial distress may contain both mechanisms. Claims that diversification necessarily stabilizes or necessarily destabilizes a network exceed the results of the cited literature. The useful general finding is conditional: compare joint losses, transmission, and mitigation across several regimes and stress magnitudes.
Market-based and informational channels
Market-based contagion does not require direct interinstitutional debt. Institutions holding similar portfolios can sell assets into the same market, and the resulting price impact reduces the mark-to-market capital of other holders. Shleifer and Vishny (2011) analyze fire sales in finance: forced sales depress prices when potential high-valuation buyers are constrained. The mechanisms can occur alongside an otherwise functioning payment network. They depend on buyer capital, price elasticity of demand, balance-sheet accounting, and the cost of holding positions through temporary dislocations.
Funding creates a feedback channel. Brunnermeier and Pedersen (2009) show that market and funding liquidity can reinforce one another; declining market liquidity worsens margin or financing conditions, while tighter financing reduces the ability of traders to absorb sales. Kiyotaki and Moore (1997) demonstrate in Credit Cycles how collateral values and credit constraints can interact to magnify and prolong economic shocks. These models have different structures and calibrations, but jointly establish that the selling price, funding capacity, and state of the system cannot always be treated as exogenous constants.
Informational contagion complicates measurement further. The failure of one institution may prompt investors, creditors, or policyholders to reassess apparently similar institutions. When information about underlying exposures is scarce, a signal about one firm can lead to withdrawals or higher funding costs elsewhere. The response can be rational given correlated uncertainty, excessive given mistaken inference, or a combination. Observed simultaneous withdrawals reveal behavior, but one still needs evidence separating information, contracts, and price effects to determine the relevant causal channel.
Historical interpretation and identification
Financial crises typically include common shocks, direct counterparty exposure, liquidity contraction, and intervention. A historical description of rapid price declines can establish temporal co-movement but not the proportions attributable to each cause. For causal inference, a researcher needs credible timing, contractual positions, event variation, policy records, and models of what would have happened absent specific transactions or official action. Interventions may prevent observed failures and therefore obscure the unobserved severity of counterfactual contagion; ex post stability can reflect effective intervention rather than absence of initial systemic vulnerability.
The international context now includes financial institutions outside traditional banking. The Financial Stability Board's 2025 NBFI monitoring report documents links involving bank deposits, bank lending and repo exposures to nonbanks, and nonbanks' holdings of bank securities. Its 2024-end aggregate figures cover a broad statistical sector, not a set of entities uniformly engaged in risky credit intermediation. Insurance corporations are included within the broad nonbank perimeter; conflating that perimeter with the FSB's narrower risk-based NBFI measure would produce a misleading estimate of insurance-specific systemic exposure.
The most defensible empirical language preserves these boundaries: evidence for common co-movement, evidence for a documented contractual path, and evidence for an economically consequential amplification mechanism are different steps. A strong study places each in a chronology and tests alternative explanations. This chapter therefore establishes the channels needed for the conditional risk measures that follow, while leaving detailed empirical network reconstruction to specialist dossiers and original data-intensive research.
Conditional Value at Risk
Definitions and statistical meaning
CoVaR provides a distributional approach to the relation between distress at one institution and the tail of outcomes in the financial system. Adrian and Brunnermeier's 2016 CoVaR study defines a systemic conditional quantile and considers a difference between that quantile when an institution is distressed and when the institution is in its median state. The approach measures the association between a given institution's condition and the lower tail of system returns, under a fixed empirical specification. Economic interpretation is governed by the definition of the conditioning event, state variables, horizon, and distributional stability.
For clarity use losses, with positive values representing adverse outcomes. Let L_S be the system loss and L_i an institution's loss. Define the ordinary loss quantile VaR_alpha(L) = inf{ell : P(L <= ell) >= alpha}, with alpha near one, such as 0.95. Define a loss-oriented CoVaR for a conditioning state c by CoVaR_alpha^(S|i=c) = inf{ell : P(L_S <= ell | L_i=c) >= alpha}. At continuous conditioning points the equality is understood through a regular conditional distribution or a suitably estimated neighborhood, not the probability of a single exact observation. A loss-oriented Delta CoVaR compares the system loss quantile conditional on i's high-loss state to that conditional on a reference, often i's median.
The original paper commonly formulates the measure with negative return quantiles, in which high distress lies in the lower tail. Translating into positive losses reverses which numerical tail is called adverse and may reverse signs in displayed changes. A manuscript must state that convention before comparing numbers or importing an empirical coefficient. In particular, a positive deterioration in the loss-oriented measure need not share the sign of a negative-return measure in an original figure. The common substantive object is conditional systemic tail exposure, not the arbitrary sign attached to a return or loss variable.
The reference state is also meaningful. Conditioning on a median institution outcome asks how the system's upper-loss quantile differs from a baseline state, given the selected statistical relationship. It does not compare a world where institution i is removed from the system to one where it remains, and it does not show that making i suffer a loss would generate an equal change in the system's loss quantile. That latter statement requires an intervention model or exogenous variation capable of identifying causal transmission.
Quantile conditioning and estimation
Conditional quantiles may be estimated using quantile regression or other methods, but each requires assumptions. A linear conditional quantile Q_alpha(L_S | L_i, Z) = a_alpha + b_alpha L_i + gamma_alpha' Z provides a local summary, where Z contains observed macrofinancial state variables. A positive b_alpha means higher observed institution loss is associated with higher conditional system-loss quantiles in the estimation domain. Whether it captures common shocks, direct propagation, or both depends on which variables are omitted and how the model was identified.
Quantile regression avoids assuming that conditional relationships are fully summarized by a mean and variance. It nevertheless requires enough observations near the target tail, a reasonably specified conditioning set, and stability across forecast and evaluation windows. Two especially important uncertainties concern endogeneity and time variation. The institution's own price may react to the expected condition of the system; conversely, the system index may mechanically include the institution. Macro controls may capture some common shocks while leaving others unmeasured. Changes in funding structure, central-bank policy, and accounting or regulation may make an estimated coefficient from one period unreliable in another.
Statistical inference is additionally complicated by serial dependence, heteroskedasticity, overlapping forecast horizons, and survivorship. Stressed firms can delist or be recapitalized, and system indices can change composition. Measurement in dollars, percentage returns, or leverage-adjusted losses changes rankings. A carefully documented empirical CoVaR report should describe the system benchmark, the entity universe, sample interval, return frequency, tail probability, method of estimation, macro controls, uncertainty intervals, and test of sensitivity to crisis subsamples.
Insurance application introduces a further mismatch. The traded equity of a listed insurer reflects expectations about investment income, claims, shareholder dilution, regulatory restrictions, and future franchise value. It is not an actuarial measure of policyholder liabilities or statutory capital. Many material insurers are unlisted, mutually owned, or embedded in groups. A market-return CoVaR analysis can therefore complement but cannot substitute for supervisory balance-sheet information, reinsurance collectibility, or cash-flow testing.
A transparent conditional-tail illustration
A simple Gaussian example makes the conditional interpretation reproducible. Assume two jointly normally distributed loss variables with zero means, standard deviations sigma_S = 10 and sigma_i = 8 monetary units, and correlation rho = 0.60. Assume no other conditioning variables and use alpha = 0.95. Under joint normality, the conditional distribution of L_S given L_i=x has mean rho (sigma_S/sigma_i) x and standard deviation sigma_S sqrt(1-rho^2). The 95th percentile of the standard normal is approximately 1.644854. The conditional 95th loss quantile is thus 0.75x + 1.644854 * 8, or 0.75x + 13.1588.
The 95th percentile of L_i is 1.644854 * 8 = 13.1588, while its median is zero. The system quantile conditional on institution distress is approximately 23.028 and the conditional median-state quantile approximately 13.159. The difference is 9.869 monetary units. This is a conditional association implied by the chosen normal covariance structure. The joint normal distribution has zero asymptotic tail dependence for any absolute correlation below one; a large conditional quantile shift at one selected alpha therefore does not establish the extreme-tail behavior of a heavy-tailed empirical financial system.
Change one assumption at a time to show the dependence of the conclusion. If rho were 0, the conditional mean response and Delta CoVaR would both be zero in this model. If rho increased to 0.80, the conditional mean shift at the institution's 95th quantile would be 0.8*(10/8)*13.1588 = 13.1588, even though the conditional residual standard deviation would shrink to 6. A shift in rho alone changes the measure despite the absence of an explicitly modeled contractual transmission mechanism. If the loss distribution were instead driven by common jumps or nonlinear liquidity thresholds, the Gaussian calculation could misstate the stress region.
This demonstration is intentionally artificial: no observed insurer, bank, historical return series, estimated correlation, confidence interval, or calibrated loss tail is claimed. The calculation is valuable because it isolates exactly what the conditional statistic responds to and why the variable used for conditioning can dominate an apparent ranking. Source-based statistical findings such as the reported predictive associations in Adrian and Brunnermeier (2016) require their own data and validation; they cannot be reproduced from this example.
Comparability, measurement error, and causality
A conditional tail score is comparable across institutions only after the common system index, loss scaling, distress quantile, horizon, conditioning variables, and estimation method have been fixed. Larger institutions may mechanically explain more variance in a value-weighted system return. A smaller but highly leveraged firm can have a large change in its equity conditional quantile. Both rankings can be informative while corresponding to different policy questions. Reporting normalized and absolute measures together can show the distinction between intensity of market distress and the estimated monetary consequences of a system shock.
The choice of system must also avoid overlapping definitions when interpreting estimates. Including institution i's own loss in L_S can cause an arithmetic contribution to covariance. An external-system benchmark or an explicit leave-one-out decomposition can reveal how much of the conditional dependence arises from that overlap. A change in CoVaR after such an adjustment is a diagnostic of index construction; it is still not a causal effect. Similarly, conditioning on broad market stress while studying an individual insurer answers a question about exposure to the system, whereas conditioning on the insurer answers a question about the system's measured tail association with that insurer.
The appropriate use of CoVaR is therefore comparative surveillance, scenario prioritization, and the investigation of possible sources of systemic vulnerability. It should motivate questions about contractual exposure, funding dependencies, holdings overlap, guarantees, business interruption, and market responses. A robust decision compares the statistical signal with balance-sheet and institutional evidence, reports sensitivity to assumptions, and declines causal wording unless independent identification is available. The same discipline applies to the shortfall measures considered next.
Systemic Expected Shortfall
From an institution to aggregate capital deficiency
A central question for financial stability is whether the financial system can maintain essential services after losses have depleted capital. An individual institution may satisfy an ordinary loss threshold yet be insufficiently capitalized in states when many intermediaries lose capital simultaneously. Systemic expected shortfall studies this joint undercapitalization within a model of a financially constrained system. It connects tail-loss measurement to incentives and negative externalities that arise when institutions do not internalize all the costs of their distress.
Acharya, Pedersen, Philippon, and Richardson (2017), in Measuring Systemic Risk, analyze the systemic externality associated with a financial sector becoming undercapitalized and show how an institution's systemic expected shortfall relates to its leverage and marginal expected shortfall. Their model explains why losses during economy-wide distress can matter more for systemic policy than stand-alone volatility. The published result is conditional on the model's notion of efficient capital, systemic crisis, and social costs; it is not a universal capital formula for every banking or insurance jurisdiction.
A system-wide capital deficiency is a state-contingent aggregate shortfall relative to a specified benchmark. The benchmark may be a minimum capital-to-assets ratio, an economic-capital standard, a capacity to maintain lending or underwriting, or a supervisory solvency test. These choices are economically different. A shortfall relative to a hypothetical ratio does not prove legal insolvency. An institution may raise new capital, hedge exposures, reduce dividends, use recovery options, receive support, or shrink its balance sheet. Such responses can change the financial shortfall and introduce new spillovers.
Marginal expected shortfall and systemic expected shortfall
Let R_i denote institution i's investment or equity return over a specified horizon and R_M the market return. If adverse market states are defined by R_M below a fixed negative threshold c, a simple marginal expected shortfall statistic is MES_i = -E[R_i | R_M < c]. The negative sign expresses expected equity losses as a positive number. The selected threshold, horizon, benchmark index, and treatment of the institution's own weighting in the benchmark must all be documented. This MES is a conditional expected loss of i in market distress, not its contribution to market distress by itself.
In Acharya et al. (2017), systemic expected shortfall is related to the capital deficiency of an institution when the financial sector as a whole is short of capital. Their framework links leverage and tail sensitivity to a theoretical systemic externality. The precise analytical identity depends on the capital structure and distress definition used in the article. Translating it directly into a regulatory capital requirement would require matching the legal and accounting measure of capital, the admissible scenario, the stress horizon, and policy objective. The research provides a powerful interpretation of the risk externality, rather than a substitution for those institutional choices.
For insurers, a marginal market shortfall may reflect declines in risky assets at the same time that discount-rate changes reduce or increase the market-consistent value of insurance liabilities. Equity prices may anticipate both. The market MES therefore differs from a statutory solvency stress that separately recalculates eligible own funds and capital requirements. Insurer-specific products can also alter the link between market returns and ultimate insured losses. A full insurance interpretation should put the conditional equity result next to cash-flow, reserving, asset–liability, policyholder behavior, and group-capital results.
SRISK and conditional capital shortfall
Brownlees and Engle (2017) present SRISK: A Conditional Capital Shortfall Measure. Their measure examines a financial firm's prospective capital deficiency conditional on a severe market decline and depends on size, leverage, and market risk. In a common simplified representation, with W the firm's current market value of equity, D its debt, k a target capital ratio, and LRMES the estimated proportion of equity value lost during a defined prolonged market shock, the conditional capital shortfall approximation is
[ CS = kD - (1-k)W(1-LRMES). ]
Under the simplifying assumption of unchanged debt and with stressed equity W' = W(1-LRMES), this is algebraically identical to k(D+W') - W', the difference between target equity capital and projected equity. A common firm-level SRISK convention takes the positive part max(0, CS). The precise horizon, stress threshold, LRMES estimation, coverage of liabilities, and any truncation or aggregation should follow the specification in the original source before scores are reported as empirical SRISK.
This formulation is more intuitively informative than a market-return correlation alone because it introduces balance-sheet scale. Two firms with identical projected percentage equity declines can have different dollar shortfalls due to their financing. It also illustrates why assumptions matter: treating all debt as fixed, assuming an unchanged target ratio, and equating market equity with the relevant capital resource simplify complex real-world restructuring and insurance regulation. In an insurer, reported financial liabilities and the equity valuation of promised benefits may move together under interest-rate changes, so unadjusted application can produce a misleading picture of regulated solvency.
A market-based conditional shortfall is not the same as available liquidity. An institution may have sufficient total capital at a modeled horizon while lacking immediately transferable cash to meet margin, surrender, or claim demands. Conversely, cash-rich entities with negative accounting earnings may temporarily withstand liquidity shocks while remaining exposed to future solvency depletion. Capital and liquidity must be examined on related but separately timed paths.
Comparative numerical cases
Consider two hypothetical groups evaluated with the same capital ratio k = 8%, the same market scenario, and W and D measured in millions. Group A has W = 20, D = 180, and assumed LRMES = 40%; stressed equity is 12. Its conditional capital shortfall is 0.08180 - 0.9212 = 14.4 - 11.04 = 3.36. Group B has W = 30, D = 120, and LRMES = 20%; stressed equity is 24, and its conditional shortfall is 9.6 - 22.08 = -12.48. If SRISK is calculated as the positive part, A contributes 3.36 and B contributes zero under this exact simplification.
Now alter only A's assumed stressed equity-loss fraction to 20%. Stressed equity rises to 16 and the shortfall becomes 14.4 - 14.72 = -0.32, so the positive part is zero. The ranking is sensitive to the conditional loss estimate; an apparently precise dollar result cannot be more reliable than its model of the tail. Another diagnostic is to change k, the hypothetical benchmark ratio. Raising k from 8% to 10%, with the original A stress, gives 18 - 0.9*12 = 7.2. Such changes are sensitivity exercises, not new estimates of actual regulatory capital requirements.
| Hypothetical case | Current market equity W | Debt D | Assumed LRMES | Stressed equity W' | Conditional shortfall at k=8% | Positive part |
|---|---|---|---|---|---|---|
| A | 20 | 180 | 40% | 12 | 3.36 | 3.36 |
| B | 30 | 120 | 20% | 24 | -12.48 | 0 |
| A with milder stress | 20 | 180 | 20% | 16 | -0.32 | 0 |
Every value in this table is assumed or derived arithmetically; it is not an observation of a real financial institution. The calculation illustrates the comparative meaning of a capital-shortfall construct and demonstrates that a claim of systemic contribution still requires a model connecting that shortfall to real-economy disruption and to other institutions. Cross-sector use additionally requires care: insurers' gross policy obligations, technical provisions, and credit debt need not belong to a single homogeneous D category.
Taken together, CoVaR, MES/SES, and SRISK reveal complementary aspects of systemic risk. CoVaR focuses on the system's conditional quantile given an institutional condition. MES conditions the institution's loss on adverse system outcomes. SES ties undercapitalization to the system-wide crisis within an economic externality model. SRISK approximates a conditional equity-capital shortfall using market and balance-sheet variables. Their direction of conditioning, units, and assumptions differ; direct rank comparison without normalization or narrative is conceptually invalid. A sophisticated insurance analysis retains each measure's strengths while anchoring conclusions in the underlying contract and cash-flow mechanisms.
Capital and Funding Feedback
Leverage, collateral, and liquidity spirals
Systemic losses can be amplified when an initial price change modifies the balance-sheet constraints of market participants. Leveraged institutions may be compelled to deleverage after asset prices decline or borrowing conditions tighten. Forced sales depress market values, and the subsequent price decline can generate another round of deleveraging. The Brunnermeier–Pedersen liquidity mechanism formalizes a feedback between funding access and the ability of intermediaries to provide market liquidity. The Shleifer–Vishny fire-sale analysis explains why prices can be distorted when natural buyers are also distressed.
Margin calls add a contractual route. When collateral values move against a derivative holder, an increase in variation margin transfers cash to the opposing side. This transfer can reduce counterparty credit risk while creating an immediate cash need for a participant that must source eligible assets. Initial margin increases or haircuts may further raise liquidity requirements. The FSB's 2024 margin preparedness recommendations explicitly recognize the possibility of market-wide cash demands and address insurers among the nonbank entities within scope. They are policy recommendations rather than globally uniform statutory requirements for every institution.
Liquidity shocks can spread even if each firm remains solvent under its own expected ultimate cash flows. Sale timing, the currency of eligible collateral, contractual settlement calendars, repo rollover access, trading-market depth, and constraints on intra-group transfers can make liquid-asset availability binding. Portfolio holdings matter because institutions that invest in the same securities can suffer price effects from sales by others. A stress test limited to the accounting value of total assets and liabilities will miss this timing mechanism.
A minimal feedback calculation
Consider a hypothetical insurer holding 100 units of an asset with initial market price 1.00, so the position is worth 100. An adverse market event first lowers the price to 0.95; the mark-to-market loss is 5. The insurer then must raise 10 units of immediately available cash. Suppose, only for illustration, that it sells q units and that sales of q units cause an additional linear price impact of 0.001q on the remaining position. If all sold units transact at the post-impact price p=0.95-0.001q, the proceeds satisfy q(0.95-0.001q)=10. The small positive root is approximately q=10.65 and p is approximately 0.93935. The first-round mark-to-market price decline is therefore enlarged, and the total position value at this final price is about 93.94 before accounting for cash proceeds and any use of that cash.
This approximation deliberately does not treat the 10 units of cash raised as an economic loss. Selling an asset converts an investment into cash; the price impact redistributes or destroys value relative to a frictionless benchmark depending on the underlying market structure. If the insurer uses the cash to satisfy a contractual liability, the liability and cash decrease together; if it posts collateral, the relevant cash becomes encumbered and a corresponding derivative receivable or contractual right may arise. The valuation and liquidity consequences must be recorded consistently rather than counted twice.
A second institution holding an identical 100-unit position without selling would observe an additional price-driven valuation decline from 95 to about 93.94, a reduction of roughly 1.06 under the invented impact function. This is a simplified illustration of a market-price externality in the absence of a direct bilateral loan. The size of the effect depends entirely on the assumed price-impact coefficient, order execution convention, market demand, and whether the price shock persists. It provides no estimate of any actual insurer's fire-sale contribution.
The logic remains useful when alternatives are introduced. A committed liquidity facility could change the insurer's need to sell, but facility creditworthiness and drawdown conditions matter. Additional eligible collateral could postpone or avoid asset sales, but pledged assets would cease to be freely available for unrelated obligations. Coordinated sales by several institutions could create more pronounced price effects, unless other buyers enter. These alternatives must be modeled as conditional actions rather than treated as risk-free remedies.
Capital, cash, and the timing of claims
Insurance liabilities generate multiple liquidity profiles. Catastrophe losses may arrive in waves of notifications and settlements, while recoveries from reinsurers may be subject to proof, adjustment, coverage interpretation, and payment schedules. Life-insurance and savings products can face correlated lapse and surrender behavior. Derivative margin can be due more quickly than the associated liability cash flows, and securities-lending collateral may require rapid redemption. The interaction of these obligations can leave a group with positive economic net worth but a temporary deficit in transferable cash.
A disciplined stress test creates separate time buckets: intraday or daily collateral, weekly funding or redemptions, monthly claims, and longer-run reserve and capital effects. For each currency and legal entity, available liquidity at time t should reflect unencumbered starting cash, contractually reliable inflows, realizable asset sales net of estimated haircuts, and cash outflows due by t. It should also record restrictions on intra-group transfers, including ring-fenced policyholder assets and local capital requirements. A consolidated surplus does not automatically settle a legal-entity shortfall.
Long-term guarantees introduce valuation effects distinct from immediate cash. A fall in interest rates can increase the economic value of some promised benefits and raise hedging collateral demands, even as fixed-income assets rise in value. A rise in rates can reduce bond valuations and induce policyholder surrenders in certain products, despite reducing the present value of some liabilities. The direction of net effect depends on duration, option exercise, hedges, and accounting conventions. Hence an insurer's net interest-rate exposure cannot be inferred by looking at asset-price declines alone.
Propagation, counterfactuals, and model risk
Feedback models require rules for behavior. A mechanical leverage target may predict one set of asset sales; management discretion, access to private capital, supervisory forbearance, or a market-wide official liquidity facility may produce another. A network model may impose creditor priority and instantaneous clearing; real systems may have litigation, settlement delays, or emergency resolution powers. The model should describe how responses are generated, why they are plausible in the jurisdiction, and how sensitive outcomes are to alternative rules.
The FSB's 2025 work on NBFI leverage identifies leverage through markets and links with systemically important institutions as potential amplifiers, and promotes an integrated monitoring approach. For quantitative analysis, this points to a relevant information set: who supplies leverage, which positions secure it, the haircut and collateral rules, and how liquidity shocks interact with market participants' capacity to absorb trades. Sectoral aggregates alone cannot reveal all these features.
A productive methodological comparison uses at least three states: baseline conditions, an exogenous common shock with fixed behavior, and the same shock with plausible endogenous responses. The difference between the second and third represents modeled amplification under the chosen behavioral specification. Adding policy interventions can show avoided modeled losses, but policy costs and moral hazard should also be considered. None of these differences is an observed historical fact unless calibrated and identified against real data. The practice is an analytical synthesis of established mechanism-based research, not a new proprietary systemic-risk doctrine.
Insurance Sector Implications
Sources of insurer systemic exposure
Insurance may stabilize the economy by transferring losses across time, regions, and risk-bearing institutions. Yet the same insurance system holds financial assets and links counterparties through derivatives, securities lending, reinsurance, banking deposits, credit products, and critical operational services. Systemic relevance consequently depends on the actual exposure profile. Three questions organize the empirical inquiry: what losses could reach insurers from the wider economy, what actions might insurers collectively take under stress, and which failures of insurance protection would impair other sectors' economic activity?
On the liability side, a single hazard can affect multiple otherwise diversified insurance lines. A cyclone can produce property claims, business interruption, infrastructure damage, and agricultural losses. An unusually large correlated cyber event may cross liability, property, business-interruption, and operational coverages, with uncertainty about policy wording and aggregation. A pandemic may affect mortality, health costs, business operations, and financial markets on distinct timelines. The common-shock component of these scenarios is not identical to contagion. Propagation becomes relevant when delayed recoveries, scarce adjustment services, concentrated reinsurance, withdrawal of coverage, or disruptions to payment and funding mechanisms induce additional losses or functional breakdowns.
On the asset side, insurers may hold concentrated bonds, structured credit, private debt, real estate, or other market-linked positions. Their valuations can deteriorate simultaneously with a rise in claims or with a policyholder response that demands cash. The IAIS 2025 Global Insurance Market Report identifies private-credit growth and other emerging structural themes requiring surveillance, while describing an overall resilient global sector within its monitored perimeter. This balanced observation matters: emerging exposures warrant attention without establishing that the insurance sector was in systemic crisis at the reporting date.
Reinsurance adds an especially instructive duality. Risk transfer can expand primary underwriting capacity and distribute extreme losses across diverse balance sheets. The same arrangements establish contingent receivables and concentrations among risk carriers. A cedent's gross loss, retained loss after reinsurance, cash available before recovery, and final recovered amount may diverge. Counterparty default, collateral structures, legal disputes, and time to payment determine whether a nominal recovery supports resilience when needed. The volume of reinsurance alone cannot establish that diversification has succeeded under the relevant extreme joint scenario.
The liquidity channel is likewise heterogeneous. Traditional insurance arrangements can have substantial resources accumulated before claims, but particular business models contain liabilities that can be accelerated by surrenders, policyholder options, securities-lending recalls, or collateral calls. Derivative positions may hedge economic risk while creating settlement-time demands for cash. The FSB's collateral liquidity guidance thus has a direct insurance application, especially where hedging, securities finance, and cash management intersect. The intended hedge benefit and the funding needs it creates must be estimated together.
AIG and the limits of institutional labels
The failure risk at American International Group in 2008 offers an unusually well-documented example of an insurance conglomerate's links to financial-market disruption. The Federal Reserve Bank of New York's AIG intervention account identifies mortgage-related investment losses, collateral calls on credit default swaps, ratings downgrades, and severe liquidity pressure in September 2008. The U.S. Government Accountability Office's review of AIG assistance distinguishes two major sources of financial strain: the securities-lending program operated by insurance subsidiaries and the credit default swaps written through AIG Financial Products. The documented interaction therefore extends beyond ordinary insured catastrophe claims.
According to the GAO, the government's sequence of support measures ultimately authorized $182.3 billion in federal aid. This historical figure describes cumulative aid authorization within the report's scope; it should not be interpreted as a single loss, realized government cost, or direct measure of a counterfactual contagion effect. The official Federal Reserve retrospective describes the feared spillovers through financial counterparties, commercial paper, insurance obligations, and the wider credit supply. Those descriptions support the inference that authorities judged a disorderly collapse to pose serious systemic dangers, while the precise losses that would have occurred absent intervention remain counterfactual.
This episode supplies three analytically transferable lessons. First, an insurance group contains multiple types of exposure, and the relevant stress may originate in a nontraditional activity or in an investment mechanism rather than in insured losses. Second, an institution can face urgent liquidity needs even while questions of ultimate asset recovery and solvency remain contested. Third, an assessment of systemic harm should document counterparties and economic functions rather than relying on the corporate name of a firm. These are interpretive conclusions grounded in the historical record, not a claim that every modern insurer has AIG's 2008 business structure.
The episode also warns against unqualified causal language. Official emergency decisions reveal policymakers' contemporaneous risk judgments under extraordinary uncertainty. They do not provide a controlled experiment isolating every systemic effect of the contemplated failure. Evaluating how changed derivatives collateral rules, capital standards, group supervision, or resolution procedures would have affected AIG requires a separate, historically calibrated analysis. The present dossier uses the case to identify mechanisms, not to produce a novel estimate of the rescue's macroeconomic benefit.
Supervisory architecture and the international perimeter
International insurance supervision has moved toward activity- and sector-sensitive monitoring of systemic risk. The IAIS adopted its Holistic Framework in 2019, implemented from 2020, combining supervisory measures, a Global Monitoring Exercise, and implementation assessment. The IAIS implementation assessment record explains how enhanced material was integrated into the Insurance Core Principles and ComFrame. In 2022 the Financial Stability Board discontinued annual identification of global systemically important insurers in favor of the Holistic Framework assessments; in November 2025 it reaffirmed that decision.
This development changes the interpretation of public systemic-risk lists. The absence of a contemporary annual G-SII designation does not show that all insurers are non-systemic, nor does the presence of an insurer on an unrelated list establish an identical supervisory requirement. The FSB separately maintains information on insurers subject to resolution planning, within the scope of participating authorities' reporting. Resolution-plan coverage and a discontinued global identification procedure are distinct. Each list must be understood in relation to its legal and supervisory purpose.
The International Association of Insurance Supervisors adopted the Insurance Capital Standard in December 2024. The ICS framework provides a comparable risk-based group capital measure for internationally active insurance groups under ComFrame, while jurisdictional implementation and supervisory use require attention to the applicable framework. A globally comparable capital language improves communication and can strengthen surveillance; it does not by itself measure every intraday liquidity risk, contagion path, or functional disruption.
The FSB's Key Attributes for resolution state the core characteristics of effective resolution regimes for financial institutions and include sector-specific insurer guidance. They address orderly management of severe distress where traditional insolvency may interrupt important functions. The document is an international policy standard; actual resolution powers, insurance guarantee arrangements, creditor priority, and policyholder protections depend on the legal framework in each jurisdiction. A monograph with international scope must not equate the FSB's recommendations with immediately enforceable domestic law.
The banking comparator helps illuminate the scope distinction. The Basel Committee's global-systemic-bank indicators incorporate measures of size, interconnectedness, substitutability, complexity, and cross-jurisdictional activity in a bank-specific assessment. Insurance market structure, contractual liquidity, and policyholder protection require different data and institutional analysis. Applying a bank market-based score to an insurer without checking the population and liabilities could generate an apparent precision with no appropriate supervisory interpretation.
Stress-testing and reinsurance scenarios
Stress testing can connect the statistical and mechanistic approaches when scenarios are constructed around actual balance-sheet constraints. In its 2024 insurance stress test, the European Insurance and Occupational Pensions Authority considered the effects of geopolitical and macrofinancial stresses on European insurers' capital and liquidity. The exercise used 48 participants from 20 countries, representing roughly three-quarters of the European Economic Area market in its specified sample, and assessed severe but plausible conditions. It was a non-pass/fail exercise whose conclusions belong to that sample, design, and shock calibration. It should not be described as a forecast or as proof that no insurer elsewhere is vulnerable.
An insurance systemic scenario requires simultaneous treatment of asset prices, insured events, liquidity, and counterparties. Consider a fictional regional shock that causes 100 units of gross catastrophe claims to three insurers, a 12% decline in certain investment portfolios, and a 30-day disruption to the settlement of eligible reinsurance recoveries. Even if eventual net catastrophe losses are only 40, claims may require 70 of cash before the scheduled reinsurance reimbursement. An insurer with 45 of liquid unencumbered funds then faces an immediate 25-unit gap unless it can sell assets, obtain dependable credit, defer payments legally, or receive accelerated valid reinsurance recoveries. These numbers are assumptions, and the example describes a liquidity timing mechanism rather than a simulated sector-wide failure.
The second-round question is what happens if all three insurers must sell the same liquid securities during a period of thin demand. Market price impact may reduce their own balance-sheet capital and that of unrelated holders. Counterparty downgrades or collateral disputes may delay recoveries further. If primary insurers restrict new coverage, local firms can also lose access to essential protection and financing. A scenario should measure that economic function explicitly; a negative net income figure alone cannot quantify the effect on business formation, supply chains, or postdisaster recovery.
A more complete simulation would compare four states. In the first, the disaster occurs with stable markets and prompt reinsurance payment. In the second, the disaster coincides with a market shock but normal funding and settlement. In the third, delayed recoveries and common asset sales create the feedback described above. In the fourth, plausible cash buffers, collateral management, contingent liquidity, and recovery planning mitigate the feedback, allowing an evaluation of cost and feasibility. The incremental modeled damage across these cases would reveal the assumptions driving the result. An empirical estimate would require contract-level, asset, market-depth, and cash-flow data absent from this documentary dossier.
Climate, cyber, and geopolitical risks pose further questions of dependence and scenario horizons. Natural catastrophe claims can be spatially correlated and may cause persistent affordability effects. Cyber incidents can cause operational disruption and correlated insured exposures across different policy wordings. Geopolitical fragmentation can affect both foreign assets and access to reinsurance or payment infrastructure. These are reasons to specify exposure pathways and joint distributions, not grounds for asserting a single universal tail law or a common loss multiple for every line of insurance.
Risk governance, data, and interventions
Systemic-risk governance should maintain several related views rather than a single synthetic score. The first is a cross-institution exposure map including concentrations, reinsurance recoverables, derivatives, securities finance, bank funding, and infrastructure dependencies. The second is a consolidated and legal-entity cash-flow analysis that distinguishes capital from immediate liquidity. The third is a conditional-tail assessment using multiple market and actuarial indicators with consistent horizons. The fourth is a set of coherent stress and reverse-stress scenarios, including plausible behavioral responses, counterparties' simultaneous demands, and limitations on asset monetization. The fifth records consequences for the provision of insurance and other financial services.
Data quality determines whether these analyses can be used. The Basel Committee's principles on risk data aggregation were formulated for banks, especially globally important ones, in response to deficiencies exposed by the 2007–2009 crisis. Their emphasis on timely, accurate, comprehensive exposures provides a useful general data-governance lesson, but the document should not be represented as an insurer-specific binding international standard. Insurance data requirements must be derived from the applicable supervisory and accounting regimes and from the economic channels being tested.
For macroprudential supervision, the IAIS Insurance Core Principles and ComFrame supply international standards including macroprudential monitoring and intervention. The 2024 amendments, described in the IAIS standards announcement, advanced recovery, resolution, capital, and supervisory material. A financial-stability analysis can use these instruments to structure a discussion of vulnerabilities, while recognizing that the legally operative rules depend on national adoption and application. Supervisory instruments should be assigned to the jurisdiction and date under review, not pooled into an invented uniform global legal regime.
A practical intervention evaluation tests the mechanism addressed by the instrument. Risk-based capital may strengthen capacity to absorb ultimate losses. Liquidity planning and high-quality liquid assets may manage collateral and early cash demands. Diversification of assets or reinsurance counterparties may mitigate common concentrations. Recovery and resolution planning may protect critical functions when an institution approaches failure. Exposure reporting may allow supervisors to detect common positions and institutional linkages. None is costless; capital and liquidity requirements can affect underwriting supply, asset duration, pricing, and opportunities for risk transfer. The proper comparison examines the chosen policy's benefits, costs, and effects on other parts of the network.
Another important distinction is between policyholder loss protection and wider systemic stability. A guarantee scheme, priority of insurance claims, or special resolution procedure may protect covered claimants while changing the allocation and timing of losses among creditors, reinsurers, and other intermediaries. A measure that is favorable to one class of claimants can create liquidity effects for others; the design must specify its beneficiary and economic function. Systemic risk research cannot substitute a vague reference to the public interest for a traceable accounting of those effects.
Methodological synthesis and research limits
The evidence supports an integrated diagnostic sequence. Begin with the initiating shock, identify exposures and their direct valuation consequences, map contractual claims and time-to-cash, specify behavioral responses under constraints, determine the second-round effects on prices and counterparties, and evaluate interruptions of financial and insurance functions. Apply statistical tail measures at the stages they actually estimate, and apply mechanism-based models where contracts or price impacts are specified. Report uncertainty from missing data and alternative assumptions. This is a structured synthesis of established methods rather than a newly named model.
Five recurring analytical errors deserve explicit attention. The first conflates common shocks with contagion. The second infers causality from CoVaR or conditional correlations. The third treats a projected capital ratio as a direct measure of immediate cash. The fourth treats a legally intended reinsurance recovery as contemporaneous liquidity. The fifth reads international supervisory standard-setting as uniform national law. Each error can change the stated policy response and the apparent ranking of risk. Preventing them is part of the scientific task, not a merely editorial refinement.
This dossier is bounded by its source base. It does not estimate a real-world CoVaR, SES, or SRISK time series; replicate journal results; access undisclosed contract-level positions; conduct new interviews; observe trade-level market impact; estimate causal effects of regulatory reforms; or provide a jurisdiction-by-jurisdiction legal survey. Published article abstracts and official landing pages were verified, while full text remains restricted for some journals. The quantitative cases are independent hand-checkable illustrations, with explicitly assumed parameters. Any future data-intensive extension must specify reproducible datasets, selection criteria, methodology, estimation uncertainty, and competing hypotheses before claiming empirical discovery.
Conclusion
Systemic risk is an interaction among exposure, transmission, institutional response, and impairment of economic functions. Its quantitative analysis must preserve the difference between joint loss from common shocks and additional loss generated by contagion. Conditional financial statistics are valuable tools for identifying vulnerabilities, but their interpretation depends on direction of conditioning, measurement units, tail assumptions, market composition, and time horizon. A causal attribution requires a credible transmission mechanism and justified counterfactual evidence.
The mathematical literature makes the distinctions concrete. Eisenberg and Noe show how interlocking obligations produce jointly determined payments. Allen and Gale and Acemoglu, Ozdaglar, and Tahbaz-Salehi explain why connectivity can share risk while generating fragility under particular shocks. Adrian and Brunnermeier's CoVaR analyzes system-tail outcomes conditional on institutional distress. Acharya and coauthors connect expected capital deficiency in systemic crises to economic externalities. Brownlees and Engle's SRISK supplies a market-based conditional capital-shortfall approach. Liquidity and fire-sale research shows how market prices, collateral, and funding can amplify initial disturbances. These results form a complementary methodological literature; each retains the assumptions and interpretive boundaries of its original formulation.
The insurance-sector conclusion is equally specific. Underwriting losses, asset-market losses, derivative collateral calls, correlated surrenders, delayed reinsurance recoveries, and failures of critical counterparties operate on different balance-sheet and cash-flow schedules. They may coexist and reinforce one another. The 2008 AIG experience demonstrates the relevance of group-wide financial activities and liquidity pressures, and subsequent international standards establish supervisory attention to both individual and sector-wide vulnerabilities. Contemporary sector monitoring provides evidence of resilience alongside changing concentrations and emerging risks; one cannot infer either universal stability or universal fragility from an institutional label.
Sound scientific assessment consequently requires a defined perimeter, a coherent shock and timing model, contract-sensitive mapping of direct exposures and propagation, separate evaluation of solvency and liquidity, realistic behavioral responses, and explicit functional consequences. Multiple scenarios reveal where conclusions depend on assumptions. The original contribution of this monograph is the systematic comparison and insurance-oriented synthesis of these established findings, including reproducible numerical interpretations and a source-to-claim structure. No new authorial theory, empirical discovery, or regulatory doctrine is asserted. The work is complete as an internally reviewed documentary and analytical research dossier within the limitations stated above.
References
Acemoglu, D., Ozdaglar, A., & Tahbaz-Salehi, A. (2015). Systemic risk and stability in financial networks. American Economic Review, 105(2), 564–608. https://doi.org/10.1257/aer.20130456
Acharya, V. V., Pedersen, L. H., Philippon, T., & Richardson, M. (2017). Measuring systemic risk. The Review of Financial Studies, 30(1), 2–47. https://doi.org/10.1093/rfs/hhw088
Adrian, T., & Brunnermeier, M. K. (2016). CoVaR. American Economic Review, 106(7), 1705–1741. https://doi.org/10.1257/aer.20120555
Allen, F., & Gale, D. (2000). Financial contagion. Journal of Political Economy, 108(1), 1–33. https://doi.org/10.1086/262109
Bank for International Settlements / Basel Committee on Banking Supervision. (2013). Principles for effective risk data aggregation and risk reporting (BCBS 239). https://www.bis.org/publications/201301-guidelines-principles-effective-risk-data-aggregation-and-risk-reporting
Bank for International Settlements / Basel Committee on Banking Supervision. (2026, accessed November 14). Global systemically important banks: Assessment methodology and additional loss absorbency requirement. https://www.bis.org/committees/bcbs/gsib
Brownlees, C., & Engle, R. F. (2017). SRISK: A conditional capital shortfall measure of systemic risk. The Review of Financial Studies, 30(1), 48–79. https://doi.org/10.1093/rfs/hhw060
Brunnermeier, M. K., & Pedersen, L. H. (2009). Market liquidity and funding liquidity. The Review of Financial Studies, 22(6), 2201–2238. https://doi.org/10.1093/rfs/hhn098
Eisenberg, L., & Noe, T. H. (2001). Systemic risk in financial systems. Management Science, 47(2), 236–249. https://doi.org/10.1287/mnsc.47.2.236.9835
European Insurance and Occupational Pensions Authority. (2024). Insurance stress test 2024: Results, scenario specifications, and accompanying materials. https://www.eiopa.europa.eu/insurance-stress-test-2024_en
Financial Stability Board. (2024a, April 25). Key Attributes of Effective Resolution Regimes for Financial Institutions: Revised version 2024. https://www.fsb.org/2024/04/key-attributes-of-effective-resolution-regimes-for-financial-institutions-revised-version-2024/
Financial Stability Board. (2024b, December 10). Liquidity preparedness for margin and collateral calls: Final report. https://www.fsb.org/2024/12/liquidity-preparedness-for-margin-and-collateral-calls/
Financial Stability Board. (2025a, July 9). Leverage in nonbank financial intermediation: Final report. https://www.fsb.org/2025/07/leverage-in-nonbank-financial-intermediation-final-report/
Financial Stability Board. (2025b, November 25). The FSB reaffirms its decision to use IAIS Holistic Framework assessments instead of annual global systemically important insurer identification. https://www.fsb.org/2025/11/the-fsb-reaffirms-its-decision-to-use-iais-holistic-framework-assessments-instead-of-an-annual-identification-of-global-systemically-important-insurers/
Financial Stability Board. (2025c, December 16). Global Monitoring Report on Nonbank Financial Intermediation 2025. https://www.fsb.org/2025/12/global-monitoring-report-on-nonbank-financial-intermediation-2025/
Financial Stability Board. (2025–2026 access). Insurers subject to resolution standards. https://www.fsb.org/work-of-the-fsb/market-and-institutional-resilience/crisis-management-and-resolution/insurers-subject-to-resolution-standards/
Federal Reserve Bank of New York. (2010–2026 access). Actions related to AIG. https://www.newyorkfed.org/aboutthefed/aig
Board of Governors of the Federal Reserve System. (2010–2026 access). American International Group (AIG), Maiden Lane II and III. https://www.federalreserve.gov/regreform/reform-aig.htm
International Association of Insurance Supervisors. (2019–2026 access). Holistic Framework for the assessment and mitigation of systemic risk in the insurance sector. https://www.iais.org/activities-topics/financial-stability/holistic-framework/
International Association of Insurance Supervisors. (2020–2026 access). Holistic Framework implementation assessment. https://www.iais.org/activities-topics/implementation-assessment/holistic-framework-implementation-assessment/
International Association of Insurance Supervisors. (2024, December 5). IAIS adopts Insurance Capital Standard and other enhancements to its global standards. https://www.iais.org/2024/12/iais-adopts-insurance-capital-standard-and-other-enhancements-to-its-global-standards-to-promote-a-resilient-insurance-sector/
International Association of Insurance Supervisors. (2024–2026 access). Insurance Capital Standard. https://www.iais.org/activities-topics/standard-setting/insurance-capital-standard/
International Association of Insurance Supervisors. (2024–2026 access). Insurance Core Principles and ComFrame online tool. https://www.iais.org/icp-online-tool/
International Association of Insurance Supervisors. (2025). Global Insurance Market Report 2025, official report portal. https://www.iais.org/activities-topics/financial-stability/gimar/
Kiyotaki, N., & Moore, J. (1997). Credit cycles. Journal of Political Economy, 105(2), 211–248. https://doi.org/10.1086/262072
Shleifer, A., & Vishny, R. W. (2011). Fire sales in finance and macroeconomics. Journal of Economic Perspectives, 25(1), 29–48. https://doi.org/10.1257/jep.25.1.29
U.S. Government Accountability Office. (2011, September 30; released October 31). Financial crisis: Review of Federal Reserve System financial assistance to American International Group, Inc. (GAO-11-616). https://www.gao.gov/products/gao-11-616