September 9th, 2004, 3:36 pm
Copulae are useful for estimating the joint probability of two (or more) things when you know the marginal probabilities. For example, you might estimate the distribution of returns for stocks A and B, based on historical price changes, implied volatilities from options or other data. You want to price an option that pays the return on the better of the two stocks. You could estimate this using a correlation coefficient, but that has serious problems. Copula models are commonly used instead.