Know these first
- L2.5.03: Correlation and diversification
Math Atoms
By the end you will be able to
- 1Extend GARCH so that falls and rises move volatility differently (GJR and EGARCH), fit both to GBP/USD and read the asymmetry honestly.
- 2Let correlations move with EWMA covariance and DCC, and estimate DCC on GBP/USD and EUR/USD by maximum likelihood.
- 3Explain why EWMA and DCC matrices are positive definite by construction, and why matrices stitched together pair by pair need not be (M-E04: Eigenvalues and positive definiteness).
- 4Repair a broken correlation matrix and say how far the repair moved it.
Why it matters
A risk system holds thousands of positions whose risk depends on how their factors move together. Over twenty years GBP/USD and EUR/USD have a correlation of 0.65; in March 2020 their DCC correlation fell to 0.07, and in the gilt crisis of September 2022 it was 0.80. A model that fixes one correlation, or one volatility, will be wrong on exactly the days that matter. This lesson lets both move, and makes sure the matrix that comes out is still a valid covariance matrix.