Gaussian copula

The Gaussian copula is a statistical function used to model the dependence structure between multiple random variables under the assumption that their joint distribution can be described by a multivariate normal distribution. In finance, it became the dominant tool for pricing collateralized debt obligations and mortgage-backed securities during the 2000s because it allowed complex portfolios to be reduced to a single correlation parameter that could be estimated from historical data.

The fatal flaw was not mathematical but epistemological. The Gaussian copula assumes that correlations are stable and that extreme events in different variables are independent. In the 2008 financial crisis, these assumptions collapsed: housing markets across regions were far more correlated in distress than the model predicted, and the tails of the distributions were fatter than the Gaussian assumption allowed. The model did not fail because it was wrong; it failed because it was treated as a description of reality rather than as a provisional fiction. The crisis was not a surprise to statisticians who understood tail dependence. It was a surprise to a network that had locked into a single model and forgotten that models are maps, not territories.