Statistical analysis
Understanding portfolio tail risk
Using a Vasicek one-factor model and Monte Carlo simulation to study how correlated defaults shape credit portfolio losses.
The question
How do default probabilities and dependence between borrowers change the distribution of credit portfolio losses?
The approach
- Used the public Default of Credit Card Clients dataset as the portfolio basis.
- Modeled defaults through a shared systematic factor and borrower-specific factors in the Vasicek framework.
- Simulated loss distributions under alternative default-probability and correlation assumptions, comparing Value-at-Risk and Conditional Value-at-Risk.
What this explores
The simulations show why expected loss alone does not describe portfolio stress: changing default dependence can alter the tail of the loss distribution substantially.
Scope & limitations
Results depend on the one-factor model, selected scenario assumptions, and a historical dataset. They are a classroom simulation, not a forecast of current credit losses.
Team: Jordan Andrew, Ammy Lin, Farnoosh Memari, and Tonantzin Real Rojas. Completed for MATH 789 at Duke University.
Project summary and attribution are based on the repository documentation.