Tail Credit Losses in Debt Portfolios: Evidence and Measurement in the Vasicek Framework : Evidence from the U.S. Corporate Bond Market

Abstract

Tail credit losses in corporate debt portfolios are highly relevant for risk management, capital planning, and stress testing, because losses in systemic credit downturns can rise far above average loss levels. Despite the practical importance of this issue, there is limited transparent empirical evidence on how issuer-level credit risk inputs translate into portfolio tail losses in large corporate debt universes. This thesis examines tail credit losses in a large debt-only corporate issuer universe in the United States corporate bond market using the one-factor Vasicek and asymptotic single risk factor (ASRF) framework. It focuses on how issuer-level default probabilities, exposure measures, and loss-given-default assumptions map into portfolio tail risk within a transparent and replicable modelling framework. The empirical analysis combines issuer-level one-year physical default probability data with instrument-level bond and note data to construct a debt-only portfolio and a seniority-based loss framework. The model is then used to estimate expected and tail loss measures under alternative dependence and severity assumptions, while the asymptotic approximation is also compared with Monte Carlo simulation to assess finite-portfolio effects. The findings show a clear gap between average and tail outcomes. Tail losses are driven primarily by dependence assumptions, while LGD mainly affects loss severity. The results also suggest that finite-portfolio effects can be economically meaningful in concentrated portfolios, implying that asymptotic approximations may understate risk when concentration is material. Overall, the thesis shows that the framework remains a useful and interpretable tool for portfolio tail credit risk analysis.