文档介绍：Financial and Econometric Models for Credit
Risk Management
Zur Erlangung des akademischen Grades
eines Doktors der Wirtschaftswissenschaften
(Dr. rer. pol.)
von der Fakult¨at f¨ur
Wirtschaftswissenschaften
der Universit¨at Fridericiana zu Karlsruhe
genehmigte
DISSERTATION
von
Dipl. Wi.-Ing. Bernhard Martin
Tag der m¨undlichen Pr¨ufung:
Referent: Prof. Dr. . Rachev
Korreferent: Prof. Dr. M. Uhrig-Homburg
Karlsruhe (2003)
2
Preface
Over the years financial modeling has more and more turned away from the no-
tion that financial returns follow a Gaussian process with independent increments.
Instead properties have been detected in financial returns time series that negate
the classical normal . assumption. This has led to call for new methods.
This thesis work focuses on credit instruments and the behavior of their returns.
It examines four phenomena observed in the time series of credit returns:
• Heavy-tailedness and peakedness.
• Time-varying volatility.
• Long memory (long-range dependence).
• Cointegration.
Additionally, it analyzes the interdependence among these phenomena.
On the contrary to known credit risk models based on the normal assumption,
the model for price returns in this thesis will assume a stable
empirical studies show, the daily returns of a bond and the resulting credit risk
obey a stable law, exhibiting leptokurtic, heavy-tailed, and skewed distributions.
This leads to the application of stable Value at Risk measures as they provide a
better pared to normal ones.
However, recent research has found that credit returns show certain patterns
that give rise to the assumption of autocorrelation, heteroscedasticity,andlong-
range dependence (LRD) which enhances the need for a refinement of stable credit
models applied so far.
ARCH-type processes are known to be capable of describing heteroscedasticity
in financial time series. Although such heteroscedastic pr