Robust Bernoulli mixture models for credit portfolio risk

Jonathan Ansari,Eva Lütkebohmert

2024-11-18

Abstract:This paper presents comparison results and establishes risk bounds for credit portfolios within classes of Bernoulli mixture models, assuming conditionally independent defaults that are stochastically increasing with a common risk factor. We provide simple and interpretable conditions for conditional default probabilities that imply a comparison of credit portfolio losses in convex order. In the case of threshold models, the ranking of portfolio losses is based on a pointwise comparison of the underlying copulas. Our setting includes as special case the well-known Gaussian copula model but allows for general tail dependencies, which are crucial for modeling credit portfolio risks. Moreover, our results extend the classical parameterized models, such as the industry models CreditMetrics and KMV Portfolio Manager, to a robust setting where individual parameters or the copula modeling the dependence structure can be ambiguous. A simulation study and a real data example under model uncertainty offer evidence supporting the effectiveness of our approach.

Risk Management,Mathematical Finance

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are several key challenges in credit portfolio risk modeling:

1. **Sparsity of default events**: Since default is a rare event, the available data is very limited. This makes it difficult to estimate the default dependence relationships among different entities.

2. **Complexity of positive dependence structures**: There are usually positive correlations among default events, which are driven by common exposures to economic cycles, market shocks, and industry - specific risk factors. Due to the curse of dimensionality, accurately capturing these complex dependence relationships is very challenging.

3. **Tail dependence**: Asset returns often exhibit tail dependence, while Gaussian factor models usually underestimate this dependence. Therefore, tail risks may be underestimated in portfolio losses.

4. **Model uncertainty**: Due to data scarcity, some parameter estimates (such as asset correlation coefficients) have a certain degree of ambiguity. In addition, choosing different dependence structure models (such as Copula) also brings uncertainty.

To solve these problems, the author proposes a robust Bernoulli mixture model (BMMs), assuming that the conditional default probability increases with a common risk factor. Specifically, the main contributions of this paper include:

- **Convex order comparison results**: Provide simple interpretable conditions, making it possible to compare credit portfolio losses under the convex order.

- **Copula - based threshold model**: Construct a stochastically increasing Bernoulli mixture model through Copula, and compare credit portfolio losses by comparing the underlying Copulas point - by - point.

- **Risk bounds**: Establish the upper and lower bounds of credit portfolio losses, and rank credit portfolio risks under the convex order, especially under the law - invariant convex risk measure.

These results are helpful for more robustly evaluating and comparing different credit risk models under the premise of considering tail dependence and model uncertainty.