Chapter 59: The Collapse of Investment Banking
Core idea
The 2008 investment banking collapse was not a single event — it was a cascade of incentive failures that each made the next worse. Rating agencies rated the products their clients paid them to rate. Institutional investors bought those ratings without looking under the hood. Investment banks sold insurance (credit default swaps) against defaults they knew were possible but modeled as unlikely. Each actor was following its own incentive, and none of them held the aggregate risk. When housing defaults arrived en masse, every link in the chain snapped simultaneously, and the world’s largest financial institutions became insolvent overnight.
Authors’ framing: For the investor, CDOs with credit default swap protection seemed like the perfect investment. For the investment banks, they were making money hand over fist selling the CDOs and then again charging for the CDSs. There was only one small problem.
Why it matters
The principal-agent problem at industrial scale
A principal-agent problem occurs whenever one party (the agent) is hired to act on behalf of another (the principal), but the agent’s incentives don’t fully align with the principal’s interests. This problem appeared simultaneously at multiple levels of the CDO machine:
- Rating agencies were hired and paid by the issuers of CDOs — not by the investors who relied on the ratings. An agency that gave low ratings would lose business to competitors. Upward bias in ratings was the predictable outcome.
- Portfolio managers at pension funds and insurers were managing other people’s money. As long as they were buying “AAA-rated” instruments and following written policy, they bore little personal liability when those instruments failed.
- Bank executives were compensated on short-term profits from CDO sales, not on long-term performance of the underlying loans. Maximizing compensation meant maximizing volume, not quality.
When multiple principal-agent problems stack on top of each other in the same system, the errors compound rather than cancel out.
Moral hazard and the “too big to fail” dynamic
A moral hazard arises when insurance, or the expectation of a bailout, encourages more risk-taking than would otherwise occur. The implicit guarantee that large financial institutions would be rescued — because their failure would harm millions of unconnected people — encouraged exactly the excessive risk-taking that made bailouts necessary.
Bear Stearns was rescued via a JPMorgan acquisition engineered by the New York Fed. Policymakers hoped this would calm markets. Instead, it confirmed that large institutions had a backstop, which encouraged remaining firms to stay in their risky positions rather than deleverage. When the government didn’t rescue Lehman Brothers, the shock was amplified precisely because markets had priced in a rescue that didn’t come.
The CDS market was unhedged insurance at systemic scale
A credit default swap (CDS) is essentially an insurance contract: the seller agrees to compensate the buyer if a referenced debt instrument defaults. Unlike regulated insurance, CDS sellers were not required to hold adequate reserves against their potential obligations. AIG sold hundreds of billions of dollars in CDSs without maintaining reserves proportionate to the risk — assuming, as its models did, that mass correlated default was essentially impossible. When correlated default is exactly what happened, AIG was suddenly liable for losses it could not pay.
Key takeaways
Key takeaways
- CDOs were divided into tranches: high-quality loans at the top (lower yield), riskier loans at the bottom (higher yield). The blended structure earned high ratings that misrepresented the underlying risk.
- Rating agencies were paid by CDO issuers — a classic principal-agent conflict that produced systematically inflated ratings across the industry.
- Credit default swaps (CDSs) functioned as insurance against CDO default, but sellers were not required to hold adequate reserves — so when mass defaults arrived, CDS sellers like AIG faced insolvency.
- Bear Stearns was the first major Wall Street casualty; its rescue by JPMorgan (facilitated by the NY Fed) temporarily calmed markets but set up Lehman Brothers' unrescued failure as a greater shock.
- The government bailed out AIG because it was judged 'too big to fail' — its failure would have triggered immediate losses at every institution that held its CDSs.
- Moral hazard is self-reinforcing: each rescue makes future risk-taking more likely, because actors learn that extreme downside risk is partially socialized.
- Credit Suisse's 2023 collapse showed that the structural vulnerabilities — overleveraged balance sheets, loss of client confidence, and contagion — resurface in new institutions even after reforms.
Mental model
Read it as: Four groups — originators, packagers, insurers, and buyers — each acted rationally within their individual incentive structure. No one held the system-level view. When the mass default trigger (yellow) fired, the failures at the bottom (red) were simultaneous and mutually reinforcing. The credit freeze was not an external shock; it was the natural conclusion of the logic already baked in.
Practical application
Recognize principal-agent problems before relying on third-party ratings
- Identify who pays the evaluator. Before trusting a rating, an audit, or a certification, ask who is funding it. If the entity being evaluated is also the one paying the evaluator, the independence of the assessment is structurally compromised.
- Check what the evaluator is not rating. The rating agencies rated the top tranche of each CDO. They were not rating the quality of the underlying mortgages in the lower tranches that determined how quickly the top tranche would absorb losses.
- Ask what happens to the evaluator if they’re wrong. Accountability structures matter. Rating agencies faced no financial penalty when their inflated ratings proved wrong — they had already collected their fees.
Think about moral hazard before advocating for bailouts
Bailouts of systemically important institutions are sometimes the least-bad option in a crisis. But each one sets a precedent. The question worth asking, both in finance and in policy generally, is: what behavior is this rescue rewarding, and what behavior will it encourage next time? Policymakers who ignore this question are, as the authors note, themselves creating a moral hazard.
Example
Imagine a small town has one insurance company — Valley Mutual — that insures all 400 homes against fire. Valley Mutual charges premiums based on estimated fire probability and sets aside reserves accordingly.
Now Valley Mutual’s executives discover they can earn large upfront fees by selling “neighborhood fire insurance” contracts (analogous to CDSs) to remote investors who want income from a low-risk instrument. To earn more fees, they sell far more contracts than their reserves can cover — modeling “simultaneous widespread fire” as negligibly unlikely. For ten years, everything is fine and Valley Mutual’s executives earn large bonuses.
Then a drought and arsonist combine to burn 60% of the town simultaneously. Valley Mutual owes far more in claims than it holds in reserves. The federal government — unwilling to let 400 families lose their homes and watch the regional economy collapse — steps in to cover the gap. Valley Mutual’s executives are not penalized because they acted within written policy. The executives at Ridgeline Insurance, the competing firm that also sold these contracts, draw the obvious lesson: if the government backstops the downside, the strategy was rational all along. Next year, Ridgeline sells twice as many contracts.
That is the moral hazard loop. Each crisis both justifies the bailout and seeds the conditions for the next one.
Related lessons
Jump to…
Type to filter; press Enter to open