Evaluation empirique de la relation probabilité de défaut et seuil de défaut: une approche structurelle

Abstract

AbstractThe credit risk models derived from the structural approach, introduced by Merton in 1974, attempt to estimate the risk of default from the specific financial characteristics of the firm. The aim of the article is to describe four structural models and to exanimate their capacity to foresee the average default rates and to predict their chronological evolutions in the time, while bringing out, the effect of the default barrier in assessing probability of default.Threshold, value or event, the element that causes default in structural model, can be exogenous indifferent to the motivations of the borrower; or can be endogenous fixed by the strategy of the borrower. So we suggest analyzing two endogenous models Leland &Toft (1996) and Andersean Sundersan and Tychon (1996) henceforth (LT & AST) and two exogenous ones: Longstaff and Schwartz (1995) and Collin Dufresne and Goldestein (2001) hereafter (LS and CDG). We compare the Probability of default produced by models to the actual default rate, on a sample of Canadian borrower’s companies, over a period of six years going from 2000 till 2005. The endogenous models AST (1996)& LT(1996)provide better estimations compared to those found by the exogenous ones, for the Investment grade firms. While in exogenous models, good estimations were revealed only in LS model for the class A. The adoption of a long horizon of time (5 years), do not improve the adjustment of these models. What still justifies the weakness of performance of exogenous structural models estimates.