Capital is pouring into insurance on the promise of high returns and low volatility. That is the hook. The harder truth is that this calm often arrives before the levee breaks. When risk looks cheapest, it is usually because the system is absorbing it in ways that are hard to see. The Financial Times points to the boom. The question is what fragilities the boom is building.
Cheap risk in a riskier world: The inversion is stark. After two hard years of catastrophe losses, yields in insurance-linked securities and reinsurance surged, drew fresh capital, and then tightened as money chased the trade. Low realized loss activity in recent periods flattered models and produced standout returns. That track record is now attracting generalist capital conditioned by a decade of yield scarcity. Meanwhile, the world has not grown safer. Secondary perils, litigation inflation, supply-chain fragility, cyber exposure, and political risk pull in the same direction: higher loss complexity and fatter tails. When spreads compress as headline risk rises, either underwriting has become much better, or accounting for risk has become more creative. History has fewer examples of the former.
Rating agency arbitrage returns: The warning lights are not subtle. UBS’s chair drew a straight line from today’s insurance structures to 2007’s subprime playbook, calling out rating agency arbitrage in insurance. The BIS adds that private credit held by insurers clusters around grades from smaller ratings firms, a setup prone to inflated assessments. Why does this matter? Capital rules reward paper safety. A higher rating lowers required capital, juicing returns on equity even if the underlying risk does not change. In life and annuity platforms backed by private equity, this can mean loading general accounts with illiquid assets blessed by friendlier ratings, then laying off risk to offshore affiliates to harvest regulatory spread. In property-casualty, it can mean stacking reinsurance, retrocession, and sidecars designed to earn capital relief. The economic risk migrates, but the capital charge shrinks. The resemblance to pre-crisis CDO alchemy is not theoretical. It is an incentive map.
Volatility is not risk: Investors are anchoring to recent realized volatility and extrapolating it. That is the gambler’s fallacy in institutional form. Catastrophe loss series are not stationary; they cluster. A quiet period does not mean the dice have become kinder. It resets the collective memory and reprices risk lower, which expands exposure and multiplies the eventual loss. Models lean on historical catalogs and parameter assumptions that struggle with regime shifts. Warmer oceans, persistent La Niña or El Niño states, urban sprawl into the wildland-urban interface, and aging infrastructure shift probability mass into the tails. U.S. convective storms have produced heavy insured losses in seasons once considered routine. Wildfire seasons swing from zero to catastrophic based on wind and drought that are not independent. A one-in-100 year output from a model may be one-in-30 under new boundary conditions, and the error bars are wide. Pricing built on yesterday’s weather and last year’s case law is not a margin of safety.
Correlation hides in the footnotes: The insurance market is a network, not a set of silos. Cyber is not just a “new line”; it correlates with grid failure, supply-chain stoppages, and geopolitical events. Flood and wind interact through infrastructure outages. Liability severity tracks inflation and social sentiment. Reinsurers hedge with ILS funds. ILS funds park cash in short-term credit. Insurers boost yield with private credit. Private credit relies on optimistic marks and gentle funding conditions. When a large event arrives, these links matter. The same investor may be exposed through a cat bond, a quota share, and an insurer’s general account loan book. Hedge funds are now stepping into risk that insurers find too expensive, a sign of risk transfer to players with faster redemption clocks. In equilibrium, each actor assumes they can sell to the next one or gate redemptions if needed. That is a coordination game with a bad Nash equilibrium when losses hit. Correlations go to one not because nature conspired, but because balance sheets did.
Liquidity is the underestimated peril: Yield seekers talk about attachment points and perils. Few talk about cash. Many structures do not default in the classic sense; they trap collateral, side pocket funds, or gate redemptions after an event. That is solvency by illiquidity. In calm markets, it barely matters. In stress, the difference between reported NAV and spendable dollars becomes a career risk. The AIG episode was not about ultimate losses exceeding assets; it was about collateral calls and the speed of money. The UK LDI shock was not about final pension solvency; it was about margin drains and forced sellers. Insurance faces an analogue after clustered climate events or a major cyber outage: multiple SPVs hold back cash pending loss development, while end investors face unrelated calls elsewhere. Selling what you can, not what you should, is how mispricing flips to fire sale. The mark-to-model era ends at first contact with a binder full of pending claims.
Government backstops are a variable, not a constant: Politics is part of the underwriting environment. State-backed insurers of last resort are growing in catastrophe-prone regions. Rate approvals lag inflation and hazard updates. Court rulings can retroactively expand coverage, as “silent cyber” once did. After a major event, legislatures can rewrite contract terms or pressure carriers on claims handling. That is not speculation; it is on the record in multiple states and past storms. Reinsurers price this as legal and regulatory risk, but in bull cycles the surcharge gets rounded down. If backstops face their own funding strain, they may tap reinsurers more aggressively or lean on assessments that flow back through the system. The protection gap gives the illusion of optionality for private capital; in practice, it is a latent claim on future premiums and taxes. Investors underwriting tail risk on the assumption of stable public policy are making a strong bet on state capacity.
The model-of-models problem: Few appreciate how much the sector relies on common vendor models and shared assumptions. That is efficient, until it is not. If a key model revision re-rates a peril or region, portfolios that looked diversified can re-correlate overnight. Basis risk in parametric and index products can cut both ways, souring investor sentiment after an event if outcomes deviate from lived reality. Private credit embedded in life insurers rides its own models for default and recovery that have not been tested through a deep, inflationary downturn with tight refinancing windows. Small ratings firms may mean well, but they live in the same world of incomplete data and business incentives. When everyone calibrates to the same maps, the error, when it comes, is systemic.
What an antifragile posture looks like now: The defensive response is not to flee insurance. It is to invert the premise. Start with the tail and work backward. If you cannot name who absorbs the 1-in-50 drawdown in cash terms and on what timetable, assume it is you. Demand wider margins of safety precisely because realized losses have been low. Prefer structures with transparent cash waterfalls, conservative triggers, and minimal reliance on mark smoothing. Price liquidity optionality explicitly; it is not free. Discount ratings from smaller agencies when they drive capital relief, not because they are bad, but because the incentive is bad. Stress portfolios for multiple, concurrent events across lines and for legal change risk. Underwrite sponsor behavior under stress, including gating and collateral terms. If your investment committee is benchmarking to last year’s double-digit returns, ask whether the benchmark is the calm before a clustering regime. The world is more dangerous. If risk is cheaper, the discount is coming from somewhere. In financial systems, the cheapest seller is often the future.
