Rating momentum in the macroeconomic stress testing and scenario analysis of credit risk

摘要

With a focus on multi-horizon macroeconomic credit loss projection models in stress testing and impairments, it is of interest to understand how, under stressed and best-estimate economic projections, different model assumptions can affect such projections. We focus on the popular factor model of credit risk with an added rating-momentum feature, which violates the Markov property of the model. While in retail credit models it is obvious that past delinquency is an important predictor of state path, commercial models are often implemented as Markovian, conditional on the macroeconomic paths. We find that models that do take into account the stylized fact of rating momentum can accelerate the timing of the loss significantly, compared with the non-rating-momentum case. The exact effect depends on the scenario's time horizon, severity and portfolio quality. In general it takes longer for differences to materialize for good quality portfolios, while the effect on lower rating quality portfolios can be almost immediate, with significant loss underestimation. The factor model sensitivity to the explained part of the macroeconomic factors versus idiosyncratic effects is well known but must be recognized in practice by regulators, as models with a small portion explained by the macroeconomic factors can protect the portfolio loss significantly.