FD Fengnan Deng

Research Project

Sharp Large Deviations and Gibbs Conditioning for Threshold Models in Portfolio Credit Risk

Threshold-model analysis of portfolio losses with rare-event asymptotics, sharp risk estimates, and conditional structure for large default events.

arXiv 2025 Credit Risk Large Deviations

Overview

  • Analyzes portfolio-level default losses under threshold models in credit risk.
  • Establishes asymptotic distribution theory and large deviation principles for rare loss events.
  • Derives sharp approximations for Value at Risk and Expected Shortfall.

Methods and contribution

The project focuses on the tail behavior of portfolio credit losses when defaults are driven by threshold-type mechanisms. Rather than stopping at coarse exponential rates, it develops sharper asymptotic descriptions that are more informative for extreme-risk assessment.

A key result is a Gibbs conditioning theorem that characterizes the distribution of each obligor conditional on a large portfolio-loss event. This adds an interpretable structural view to rare-event finance problems.