Estimation of the Default Risk of Publicly Traded Canadian Companies
by Georges Dionne of HEC Montréal,
Abstract: Two models of default risk are prominent in the financial literature: Merton's structural model and Altman's nonuctural model. Merton's structural model has the benefit of being responsive, since the probabilities of default can continually be updated with the evolution of firms' asset values. Its main flaw lies in the fact that it may over- or underestimate the probabilities of default, since asset values are unobservable and must be extrapolated from the share prices. Altman's nonstructural model, on the other hand, is more precise, since it uses firms' accounting data--but it is less flexible. In this paper, the authors investigate the hybrid contingent claims approach with publicly traded Canadian companies listed on the Toronto Stock Exchange. The authors' goal is to assess how their ability to predict companies' probability of default is improved by combining the companies' continuous market valuation (structural model) with the value given in their financial statements (nonuctural model). The authors' results indicate that the predicted structural probabilities of default (PDs from the structural model) contribute significantly to explaining default probabilities when PDs are included alongside the retained accounting variables in the hybrid model. The authors also show that quarterly updates to the PDs add a large amount of dynamic information to explain the probabilities of default over the course of a year. This flexibility would not be possible with a nonuctural model. The authors conduct a preliminary analysis of correlations between structural probabilities of default for the firms in their database. Their results indicate that there are substantial correlations in the studied data.
Keywords: Default risk, public firm, structural model, reduced form model, hybrid model, probit model, Toronto Stock Exchange, correlations between default probabilities.
Published in: Canadian Journal of Administrative Sciences, Vol. 25, No. 2, (June 2008), pp. 134-152.