EconStor >
Technische Universität Dortmund >
Sonderforschungsbereich 475: Komplexitätsreduktion in multivariaten Datenstrukturen, Technische Universität Dortmund >
Technical Reports, SFB 475: Komplexitätsreduktion in multivariaten Datenstrukturen, TU Dortmund >

Please use this identifier to cite or link to this item:
Title:Testing Homogeneity of Time-Continuous Rating Transitions PDF Logo
Authors:Lawrenz, Claudia
Tschiersch, Patrick
Weißbach, Rafael
Issue Date:2005
Series/Report no.:Technical Report / Universität Dortmund, SFB 475 Komplexitätsreduktion in Multivariaten Datenstrukturen 2005,34
Abstract:Banks could achieve substantial improvements of their portfolio credit risk assessment by estimating rating transition matrices within a time-continuous Markov model, thereby using continuous-time rating transitions provided by internal rating systems instead of discrete-time rating information. A non-parametric test for the hypothesis of time-homogeneity is developed. The alternative hypothesis is multiple structural change of transition intensities, i.e. time-varying transition probabilities. The partial-likelihood ratio for the multivariate counting process of rating transitions is shown to be asymptotically c2 -distributed. A Monte Carlo simulation finds both size and power to be adequate for our example. We analyze transitions in credit-ratings in a rating system with 8 rating states and 2743 transitions for 3699 obligors observed over seven years. The test rejects the homogeneity hypothesis at all conventional levels of significance.
Subjects:Portfolio credit risk
Rating transitions
Markov model
partial likelihood
Document Type:Working Paper
Appears in Collections:Technical Reports, SFB 475: Komplexitätsreduktion in multivariaten Datenstrukturen, TU Dortmund

Files in This Item:
File Description SizeFormat
tr34-05.pdf237.62 kBAdobe PDF
No. of Downloads: Counter Stats
Download bibliographical data as: BibTeX
Share on:

Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.