Case Study 2: AIG, AIG Financial Products, and the Governance That Wasn't

A real, public case. Figures are kept qualitative; the magnitudes are a matter of public record (a federal rescue of historic size), but this study does not attach invented precise statistics to them. The chapter's lesson — what happens when a risk-taking function outruns the governance meant to contain it — is the point. This study addresses only the public, documented record; it does not speculate about individuals' intent beyond what that record establishes.

Background

American International Group (AIG) was, in the mid-2000s, one of the largest and most respected insurance organizations in the world — a global insurer with a high financial-strength rating (Chapter 3) and a vast, diversified book of genuine insurance business across property, casualty, life, and specialty lines. Most of AIG was exactly what it appeared to be: a large, sophisticated insurer.

Inside that organization, however, sat a unit called AIG Financial Products (AIGFP). AIGFP was not writing conventional insurance; it was, among other things, selling credit default swaps (CDS) — in effect, guarantees against the default of financial instruments, including securities tied to U.S. residential mortgages. Economically, writing a CDS is a form of underwriting: you accept a premium in exchange for promising to make a counterparty whole if a specified credit event occurs. AIGFP was, in substance, underwriting credit risk on an enormous scale — but it was doing so outside the insurance function, under different rules, and (as events would show) outside the kind of underwriting governance this chapter describes.

The underwriting issue

Read through this chapter, the AIGFP story is a catalogue of the failures the four levers exist to prevent.

The first failure was concentration and correlation. AIGFP's guarantees were heavily exposed to a single, correlated peril: a broad decline in the U.S. housing and mortgage market. This is the independence assumption of Chapter 1 collapsing on a colossal scale — the book was not a diversified pool of independent risks but, in effect, one enormous bet on a single outcome, the same error the LMX spiral made (Case Study 1) and the same error §38.2 warns a leader to monitor as the shape of the book. The exposures looked diversified (many different instruments, many counterparties) but were driven by the same underlying cause.

The second was a risk-taking function outrunning its governance. The volume and the tail risk being written were not contained by an appetite and an audit framework adequate to their scale. In the language of §38.7, the three lines of defense did not hold: the function writing the business was not effectively challenged and constrained by independent risk management with the authority and the understanding to say stop. A profitable-looking, fast-growing line ran — exactly the structural failure §38.7 names — with no one outside it sufficiently empowered, or sufficiently able to see the whole accumulation, to halt it.

The third was the lag and the collateral trap. Like all underwriting, the true cost was not visible while the premiums were flowing in and the housing market was rising; the business looked enormously profitable. When the housing market turned, the losses and the contractual demands for collateral arrived suddenly and on a scale that threatened the entire group — the §38.5 lesson that the reported result describes a book you may no longer be able to survive, compressed from years into months.

What it shows

The most important thing AIG shows the student of underwriting leadership is that underwriting discipline is not optional anywhere a guarantee is being sold — and the governance must match the scale and the tail of the risk, not the label on the unit.

• The four levers were absent where they were most needed. Appetite did not cap the concentration; authority and audit did not constrain the volume or surface the accumulation; the combined-ratio-style discipline of pricing the tail and the capital adequately was not applied to a book whose tail was existential.
• A guarantee is underwriting, whatever it is called. AIGFP's swaps were economically insurance against default. The lesson is not "derivatives are bad"; it is that any function accepting premium for a contingent promise must be governed as underwriting — within an appetite, under audited authority, against a real assessment of the tail.
• Governance must reach the whole enterprise. A magnificently disciplined insurance operation in most of the group did not protect it from one inadequately-governed risk-taking unit, because the unit's tail risk was large enough to threaten the whole. ERM (§38.7, Chapter 28) exists precisely to aggregate and stress-test risk across the enterprise so that no single corner can sink the rest. Here, that aggregation and challenge did not contain the exposure in time.

Outcome

When the U.S. housing market deteriorated in 2007–2008, AIGFP's exposures generated losses and collateral calls of a magnitude that the group could not meet, and AIG — a pillar of the global insurance industry — was brought to the edge of failure. Because of AIG's size and its interconnection with the global financial system, the U.S. government intervened with an extraordinary rescue in 2008, one of the largest of the financial crisis, to prevent a disorderly collapse. The intervention stabilized the company; AIG was subsequently restructured over the following years, wound down or sold parts of the business, and ultimately repaid the support, but the institution and its reputation were transformed by the episode.

It is worth being precise and fair: the conventional, regulated insurance operations of AIG were, broadly, not the cause of the crisis — they were sound insurance businesses. The failure originated in a unit underwriting credit risk outside the governance that the chapter argues every guarantee-writing function requires. That distinction is itself the lesson.

Lesson

Where Case Study 1 (Lloyd's) shows governance being built and rescuing a market, AIG shows what happens when a risk-taking function outruns its governance: even a great institution can be brought down by one inadequately-controlled book whose tail is large enough to threaten the whole. For the underwriting leader, the lesson is the mirror image of this chapter's argument. The four levers are not bureaucracy; they are the difference between a function that makes money and a function that makes a crater. Appetite that caps concentration, authority that is audited, a discipline that prices the tail and the capital — not just the expected case — and a governance structure with independent challenge that can say stop to the people making the money: these are what stand between a profitable book and a catastrophic one. The CUO's hardest and most valuable act is to build, and submit to, the governance that can overrule them — because the alternative, authority and growth with no one empowered to say no, is the most expensive arrangement in the history of the business.

Discussion questions

1. In what sense was AIGFP "underwriting"? Identify the premium, the contingent promise, and the peril, and explain why the chapter insists that any guarantee-writing function must be governed as underwriting regardless of its label. (§38.7)
2. Map the AIGFP failure onto the four levers of §38.1. Which were missing, and how did each absence contribute? Contrast with the levers Lloyd's built in Case Study 1.
3. The chapter argues the combined ratio "tells the truth — about the past" (§38.5). How does AIGFP's profitable-looking pre-crisis period illustrate this, and what leading indicators (concentration, tail exposure, collateral terms) should a governance function have watched instead of the reported profit?
4. AIG's insurance operations were broadly sound; the failure came from one unit. What does this say about why governance and ERM must reach the whole enterprise (§38.7, Chapter 28), and why "most of the book is fine" is not a sufficient defense for an underwriting leader?
5. The §38.7 Compliance Corner says a CUO's true mark of seniority is "the willingness to build and submit to the governance that can say no to them." Apply that standard to the AIG case. What structural change would have most directly empowered someone to say stop, and why is it culturally hard to put in place when a unit is highly profitable?