Financial Econometrics: Volatility, Dependence & Risk Analysis

This project applies financial econometric models to analyze volatility dynamics, cross-asset dependence, and market risk. It implements GARCH-family models, DCC-GARCH, copulas, and Value-at-Risk backtesting on BMW, Intesa SanPaolo, and ENI stock returns (2005–2024).

---

🔍 Project Summary

1️⃣ Univariate Volatility Modelling

• Explored stylized facts of returns: non-normality, volatility clustering & leverage.
• Estimated and compared: ARCH(20), GARCH(1,1), GJR-GARCH, EGARCH (Gaussian & Student-t).
• Best model: Student-t EGARCH(1,1) — captures leverage and fat tails with smoother volatility.

2️⃣ Multivariate Dependence

• Models: RiskMetrics (EWMA), DCC-GARCH, O-GARCH.
• DCC-GARCH provides the most realistic time-varying correlations.
• Strongest dependence found between ENI ↔ Intesa SanPaolo.
• Student-t copula used to estimate non-linear and tail dependence.

3️⃣ Value-at-Risk Forecasting

• Daily VaR forecasting with GARCH, EGARCH, Student-t GARCH, and GJR-GARCH (rolling estimation).
• Benchmarked against RiskMetrics using:
  • Violations count
  • Total Loss function
  • Diebold-Mariano test
  • MSFE for volatility forecast
• Gaussian EGARCH showed strongest overall performance, especially during COVID-19 and the Russia-Ukraine crisis.

---

🧠 Key Skills

• Time-series modelling (GARCH family)
• DCC-GARCH & copula dependence modelling
• VaR estimation, backtesting & forecast evaluation
• Model comparison (AIC, DM test, MSFE, diagnostics)