Structural Credit Modeling and Boundary Sensitivity : Maybe One Half is not Enough!

We analyze the role played by the boundary value for the sensitivity of the creditworthiness predictions in methodologies based on Merton [1974]. We run Monte-Carlo simulations with two various samples of firms - American, European - in order to build confidence intervals for the estimator of the fractional long-term debt component parameter of the default barrier. The robustness of these confidence intervals is tested by a resampling method for an efficient determination of the a-quantile. In addition to this main task, kernel empirical distributions for the built-default probabilities are performed by using the Silvermann's method. This paper can be seen as a model risk study that emphasizes the possible errors induced by the use of a constant fractional long-term debt component part in the actual definition of the barrier offered by the KMV Corporation.