The History of Financial Crises: 2008
The Mortgage That Broke the World
The 2008 financial crisis wasn't caused by greed alone — it was caused by a mathematical model that everyone trusted and almost nobody understood.
The Idea
Most retellings of 2008 focus on the villains: reckless bankers, captured regulators, predatory lenders. Those actors matter. But the deeper story is about how a single intellectual tool gave the entire financial system false confidence at the same moment. The tool was the Gaussian copula function, a formula developed by statistician David Li that allowed banks to calculate the risk of mortgage-backed securities — bundles of thousands of home loans packaged and sold as bonds. The formula's elegant trick was estimating how likely those loans were to fail *together*, not just individually. Default correlation, it's called. The problem: Li's model was calibrated on historical data from a period when house prices had never fallen nationally, all at once. It couldn't imagine a world where they did. So it consistently underestimated the chance that millions of mortgages would sour simultaneously — which is precisely what happened. Banks loaded up on these securities because the model said they were safe. Rating agencies stamped them triple-A. Pension funds bought them. The whole edifice assumed the model was right. When house prices fell across the United States simultaneously — something the model had never been trained to expect — correlations spiked to near-perfect. Every mortgage started failing at once. The triple-A securities turned to dust. What makes this genuinely unsettling is that the crisis wasn't born from ignorance of risk. It was born from a misplaced certainty that risk had been accurately measured.
In the World
In 2005, Wired magazine would eventually publish a piece calling David Li's formula 'the recipe for disaster.' But in the years just before the crash, Wall Street treated it as something close to magic. Li, a quietly spoken quantitative analyst at JPMorgan, had published his copula paper in 2000. Within a few years, it had become the industry standard for pricing collateralised debt obligations — the instruments at the heart of the crisis. Investment banks hired entire floors of mathematicians to build on his framework. The CDO market ballooned from a few hundred billion to trillions of units of notional value within a decade. Li himself grew uncomfortable with how his formula was being used. He had designed it as a rough tool, not a definitive answer. In interviews after the crash he noted that correlation — the key input — was not something markets had ever truly stabilised around. It shifted with sentiment. In calm times, mortgages in Texas and Florida seemed uncorrelated. In a panic, they moved as one. Bear Stearns collapsed in March 2008. Lehman Brothers followed in September. In between, governments across the world scrambled to prevent a full seizure of the global banking system. Iceland's economy effectively ceased to function. In the United Kingdom and United States, central banks pumped in emergency liquidity on a scale not seen since the 1930s. All of it traceable, in part, to a formula that modelled the past so precisely it had no room left for the future.
Why It Matters
The lesson most people took from 2008 was moral: don't be greedy. That's not wrong, but it's incomplete — and the incomplete lesson is the dangerous one. The deeper lesson is epistemic. Complex systems — financial markets, ecosystems, supply chains — have a habit of producing correlations during crises that simply don't exist during calm periods. The tools we use to measure risk are usually built on data from calm periods. This means our confidence tends to be highest precisely when it should be lowest. This matters well beyond banking. Any time you see a forecast, a risk model, or a confidence interval, the right question isn't 'how accurate is this?' It's 'what does this model assume cannot happen?' The tail risks — the scenarios that fall outside the model's imagination — are often where reality lives. For your own financial thinking, this argues for a particular kind of humility: don't confuse a plan that works under normal conditions with a plan that is genuinely robust. Scenarios that seem uncorrelated — your job, your investments, the housing market — can become very correlated very quickly when things go wrong.
A Question to Ponder
In your own life, where are you relying on a model — a plan, a belief, an assumption about how things work — that has only ever been tested during the calm?
Get a new one of these every morning.
Start learning with Thinkable