Description

“Default Intensities in a Network Perspective” (Härdle–Chen–Xu, HU Berlin) 

1️⃣ Default risk is modeled via default intensities ( \lambda_t ), extracted from CDS spreads.
2️⃣ CDS term structures are fitted by a Dynamic Nelson–Siegel (DNS) model with Level–Slope–Curvature factors.
3️⃣ These latent factors ( (l_t, s_t, c_t) ) capture long-, short-, and mid-run default dynamics.
4️⃣ Each factor evolves across 10 G-SIBs via VAR(p) processes, generating cross-bank dependencies.
5️⃣ Generalized Variance Decomposition (GVD) quantifies directional connectedness ( d_{ij} ).
6️⃣ Network measures (To, From, Net, Total) reveal systemic spill-overs between banks.
7️⃣ U.S. banks transmit default risk; European banks absorb — reversing during the EU debt crisis.
8️⃣ Total connectedness: Level > Slope > Curvature — long-run intensity is the main systemic driver.
9️⃣ Macro regressions show TED spread, credit spread, VIX, and CPC variance as dominant risk drivers.
🔟 Network-based DNS forecasts outperform standalone DNS, confirming predictive value of inter-bank links.