Case Study 1: AIG (2008) — When the Capital Behind the Promises Ran Out

A real, public event, told with public facts only. No precise loss, capital, or bailout figures are reconstructed here; where this study would otherwise cite a number it keeps the magnitude qualitative, in keeping with the book's accuracy rule. The point of the case is the mechanism, not the arithmetic.

Background

American International Group (AIG) was, in the years before 2008, one of the largest and most respected insurance organizations in the world — a sprawling group of insurance companies operating across property- casualty, life, and specialty lines on nearly every continent, carrying the highest financial-strength and credit ratings the agencies award. Its core insurance subsidiaries were, by the ordinary measures of the trade, sound: they held reserves and surplus against the policies they wrote, they were regulated as insurers by the states and countries in which they operated, and they paid their claims. If the only thing that mattered to an insurer's survival were the disciplined underwriting and capital management this book has spent twenty-eight chapters teaching, AIG's insurance operations were not the problem.

The problem lived in a corner of the holding company that was not, in the traditional sense, doing insurance at all. A unit of the group — AIG Financial Products — had written enormous volumes of credit default swaps, contracts that functioned economically like insurance on the value of complex mortgage-linked securities: in exchange for a stream of payments, AIG promised to make counterparties whole if those securities lost value. The exposures were concentrated, they were correlated to a single underlying bet (the U.S. housing market), and — critically — many of the contracts contained provisions requiring AIG to post collateral if the securities were downgraded or if AIG's own credit rating fell. The parent company had written, in substance, a vast quantity of catastrophe cover on a single peril, with terms that demanded cash precisely at the moment the peril struck.

The insurance / underwriting issue

Read AIG's collapse through this chapter's lenses and it stops being a mysterious financial-crisis story and becomes a textbook violation of everything Chapters 1, 27, and 28 teach about capital and correlation.

The independence assumption failed — at the level of the whole enterprise. Chapter 1 warned that the law of large numbers stabilizes a pool only when the losses are independent. AIG Financial Products had built a book whose losses were not independent of one another at all: they were all bets on the same housing market, so when that market turned, they did not fail one at a time in a statistically smooth trickle — they failed together. This is the §28.5 lesson in its purest form. No single contract looked ruinous; the aggregate was a single concentrated wager that an enterprise risk-management function, had it been governing the whole picture, existed precisely to catch.

The capital was the wrong kind, in the wrong place, against a correlated event. The insurance subsidiaries' surplus was held against insurance risk and was, in large part, ring-fenced by insurance regulators for the protection of policyholders — it was not freely available to meet the parent's swap obligations. Meanwhile the swap book's terms turned a ratings downgrade into an immediate demand for cash collateral. So the very event that signaled trouble — the downgrade of AIG's credit — simultaneously triggered a liquidity demand the holding company could not meet. This is §28.3's covariance warning made flesh: a model that treated asset risk, credit risk, and the firm's own rating as separable would have badly understated the danger, because in the crisis they all moved at once and in the same direction.

Liquidity, not just solvency, was the killer. This case sharpens a distinction the chapter draws between solvency (holding enough capital to meet obligations) and the closely related problem of liquidity (having cash available when obligations come due). An institution can be arguably solvent on a balance sheet measured over time and still fail if it cannot produce cash on the day the collateral call arrives. The collateral provisions on the swaps were, in effect, a percentage-deductible-and-trigger that demanded payment at the worst possible moment — the structural opposite of the patient, reserved, capital-backed promise an insurance policy is supposed to be.

The deepest irony for an underwriter: AIG's insurance companies largely did the things this book teaches. They were brought to the brink by an affiliated unit that had written insurance-like risk without the insurance discipline — without independence in the pool, without capital matched to the correlated event, and without an enterprise view that could see the whole bet. The failure was not of insurance underwriting; it was of the capital and correlation governance this chapter exists to teach.

What it shows

• Capital must be matched to the correlated tail, not the average year. AIG Financial Products collected steady payments in the calm years and looked profitable on anything resembling a combined ratio. The capital it should have held was against the once-in-a-generation correlated event — exactly the 1-in-200-year tail Solvency II sizes to and the severe-event test a rating-agency model and an ORSA stress test demand. (§28.4, §28.5, §28.7)
• Enterprise risk management is not optional, and silos are dangerous. No single desk's view revealed the aggregate bet. ERM and the ORSA exist precisely to govern the risk that lives between functions and in the aggregate. (§28.5)
• The covariance/independence assumption is the soft spot of every capital model. Crises are defined by correlations breaking — by things that "never move together" suddenly doing so. A capital framework is only as good as its treatment of correlation in the tail. (§28.3)
• Ratings and collateral terms can transmit trouble instantly. A downgrade was not merely a verdict; it was a trigger. The interaction of the rating with the contract terms turned a slow problem into an immediate liquidity crisis — a reminder that the rating-agency relationship (§28.7) is woven into a carrier's survival in ways beyond capacity.

Outcome

To prevent a disorderly failure that authorities feared would cascade through the financial system — because AIG's counterparties were themselves major institutions — the U.S. government intervened with extraordinary public support, taking a large ownership stake and extending financing on a scale without precedent for a single company. Over the following years AIG was restructured: it sold off large parts of the group, wound down the problematic financial-products exposures, and ultimately repaid the public support. The insurance subsidiaries, for the most part, continued to pay claims throughout. The episode became a defining example in the subsequent debate over systemic risk — the risk that the failure of one large, interconnected institution can threaten the whole system — and contributed to a wave of regulatory reform aimed at identifying and supervising systemically important institutions and at strengthening the governance of aggregate, correlated risk.

Lesson

For the underwriter, AIG is not a story about exotic derivatives; it is the most expensive possible demonstration of this chapter's core claim. A promise is only as good as the capital behind it, and that capital must be sized to the correlated catastrophe, governed across the whole enterprise, and available as cash when the obligation comes due. Every shortcut AIG Financial Products took has an analog on your desk: writing a pile of risks that look independent but share one peril; pricing for the calm year and ignoring the capital the tail consumes; trusting a model whose comfort depends on a correlation assumption that the catastrophe will break; and forgetting that the rating which lets you do business can also be the trigger that undoes you. The reason your CUO refers a catastrophe-heavy account, the reason the rating agency caps coastal exposure, the reason an ORSA stress-tests the whole book against one storm — all of it is the institution remembering AIG so that you do not have to relive it.

Discussion questions

1. AIG's insurance subsidiaries largely followed sound underwriting and capital practice, yet the group nearly failed. What does this say about the limits of risk-by-risk underwriting and the necessity of enterprise-level capital governance (§28.5)?
2. Explain how the collateral-on-downgrade provisions turned a solvency concern into an immediate liquidity crisis. Why is liquidity a distinct problem from holding adequate capital over time?
3. The RBC covariance adjustment (§28.3) assumes risk charges are largely independent. Using AIG, explain why that assumption is most dangerous in exactly the scenarios capital is supposed to protect against.
4. AIG Financial Products' swap book would have looked profitable on a combined-ratio-style measure in the calm years. Apply §28.6: what was it failing to earn, and against what capital?
5. Insurance regulators ring-fence subsidiary surplus to protect policyholders. In the AIG case this both protected policyholders and limited the capital available to the parent. Discuss the tension between protecting the policyholders of one entity and the solvency of the whole group.
6. After AIG, regulators moved to identify and supervise "systemically important" institutions and to strengthen aggregate-risk governance (the spirit behind ORSA adoption). Do you think a single capital formula like RBC could ever have caught AIG's risk? Why or why not?