Know these first
- L2.6.05: Diversification and its limits
Math Atoms
By the end you will be able to
- 1Use Sklar's theorem to separate margins from dependence, and turn returns into pseudo-observations.
- 2Write down the Gaussian, t, Clayton and Gumbel copulas, their link to Kendall's τ and their tail dependence.
- 3Fit the four copulas by maximum likelihood, read standard errors and AIC, and check a fit by simulation.
- 4Judge a copula by the joint tails it implies, and explain why the t copula wins for GBP/USD and EUR/USD.
Why it matters
On 19 of 5,118 days, GBP/USD and EUR/USD were both in their worst 1%. If they were independent there would be about half a day like that; a bivariate normal with their correlation says about 12. L6.1.08: Correlation and its limits showed that correlation cannot describe this corner and L2.6.05: Diversification and its limits that it is where diversification fails. A copula can: it is the dependence on its own, and different copulas put different weight in the corners. Fit four, and let the data choose.