ETD

• ccp
• counterparty credit risk
• ewma
• filtered historical simulation
• gas
• initial margin
• procyclicality
• score models
• value at risk
• volatility

This thesis studies how the Exponentially Weighted Moving Average (EWMA), the cornerstone of many initial margin models used by central counterparties (CCPs), can be improved through score-driven dynamics. Building on the Generalized Autoregressive Score (GAS) framework, the Score-Driven EWMA updates volatility estimates using the information content of the predictive distribution, allowing the model to react more precisely to market conditions and to reduce undue sensitivity to outliers.

The analysis spans from single assets to multivariate portfolios and employs Filtered Historical Simulation as a central tool for estimation. Empirical evidence on multiple financial instruments with heterogeneous risk profiles indicates that Score-Driven EWMA models provide relevant advantages over the traditional RiskMetrics benchmark, improving robustness and interpretability while preserving operational simplicity.