Dynamic Initial Margin Estimation based on Quantiles of Johnson Distributions

A new accurate approach to estimate dynamic initial margin (DIM) for general portfolios, based on regression using Johnson distributions fitted to conditional moments approximated with least squares MC (JLSMC), is described.... [ view full abstract ]

A new accurate approach to estimate dynamic initial margin (DIM) for general portfolios, based on regression using Johnson distributions fitted to conditional moments approximated with least squares MC (JLSMC), is described. It is consistent with DIM estimation using nested MC. DIM values produced under the standard approach (Gaussian portfolio changes) diverge significantly from the better JLSMC estimates. This is due to its superiority when portfolio changes are far from Gaussian. We illustrate JLSMC  on the Heston and Hull-White models for Calls, Puts and Bermudan Swaptions. JLSMC only requires quantities that are readily available when computing XVA.