Course Project Latent risk index
Latent Risk Estimation in the U.S. Market Using Representative Variables
Summary
Poster project proposing a latent financial-risk index from representative global signals such as implied volatility, oil, gold, and the dollar index. The workflow combines dynamic factor modeling with penalized least-squares smoothing and Guerrero-style smoothness selection through validation.
$$ X_t=\Lambda f_t+e_t,\qquad \min_{\tau}\sum_t(R_t-\tau_t)^2+\lambda\sum_t(\nabla^d\tau_t-m)^2 $$
Context
Noisy market variables are compressed into a common latent factor and then smoothed into an interpretable risk-monitoring index.
Main contributions
- Built a latent market-risk index from VIX, WTI, gold, and DXY-style financial variables.
- Used a dynamic factor model to extract a common risk component from representative market signals.
- Estimated a structural trend with penalized least squares and Guerrero smoothness selection.
- Analyzed validation behavior, inflation relationships, and stress deviations around crisis periods.