The Importance of Estimation Uncertainty in a Multi-Rating Class Loan Portfolio
Henry Dannenberg
IWH Discussion Papers,
No. 11,
2011
Abstract
This article seeks to make an assessment of estimation uncertainty in a multi-rating class loan portfolio. Relationships are established between estimation uncertainty and parameters such as probability of default, intra- and inter-rating class correlation, degree of inhomogeneity, number of rating classes used, number of debtors and number of historical periods used for parameter estimations. In addition, by using an exemplary portfolio based on Moody’s ratings, it becomes clear that estimation uncertainty does indeed have an effect on interest rates.
Read article
Evaluating the German (New Keynesian) Phillips Curve
Rolf Scheufele
North American Journal of Economics and Finance,
2010
Abstract
This paper evaluates the New Keynesian Phillips curve (NKPC) and its hybrid variant within a limited information framework for Germany. The main interest resides in the average frequency of price re-optimization by firms. We use the labor income share as the driving variable and consider a source of real rigidity by allowing for a fixed firm-specific capital stock. A GMM estimation strategy is employed as well as an identification robust method based on the Anderson–Rubin statistic. We find that the German Phillips curve is purely forward-looking. Moreover, our point estimates are consistent with the view that firms re-optimize prices every 2–3 quarters. These estimates seem plausible from an economic point of view. But the uncertainties around these estimates are very large and also consistent with perfect nominal price rigidity, where firms never re-optimize prices. This analysis also offers some explanation as to why previous results for the German NKPC based on GMM differ considerably. First, standard GMM results are very sensitive to the way in which orthogonality conditions are formulated. Further, model mis-specifications may be left undetected by conventional J tests. This analysis points out the need for identification robust methods to get reliable estimates for the NKPC.
Read article
Estimation Uncertainty in Credit Risk Assessment: Comparison of Credit Risk Using Bootstrapping and an Asymptotic Approach
Henry Dannenberg
IWH Discussion Papers,
No. 3,
2009
Abstract
Bei der Kreditrisikobewertung müssen die Parameter Ausfallwahrscheinlichkeit und
-korrelation geschätzt werden. Diese Schätzung erfolgt unter Unsicherheit. In der Literatur werden asymptotische Konfidenzregionen diskutiert, um diese Unsicherheit bei der simultanen Schätzung beider Parameter zu bewerten. Diese Regionen setzen allerdings eine sehr lange Datenhistorie für eine genaue Bewertung voraus. Als Alternative bietet sich bei kurzen Datenhistorien Bootstrapping an. Diese Methode ist allerdings deutlich rechenintensiver. Im vorliegenden Beitrag wird untersucht, ab welcher Anzahl historisch verfügbarer Perioden Bootstrapping und eine Wald-Konfidenzregion zu einer vergleichbaren Bewertung des Kreditrisikos gelangen. Die hier genutzten Methoden führen zu ähnlichen Ergebnissen, wenn über 100 historische Perioden zur Verfügung stehen.
Read article
Evaluating the German (New Keynesian) Phillips Curve
Rolf Scheufele
IWH Discussion Papers,
No. 10,
2008
Abstract
This paper evaluates the New Keynesian Phillips Curve (NKPC) and its hybrid
variant within a limited information framework for Germany. The main interest rests on the average frequency of price re-optimization of firms. We use the labor income share as the driving variable and consider a source of real rigidity by allowing for a fixed firm-specific capital stock. A GMM estimation strategy is employed as well as an identification robust method that is based upon the Anderson-Rubin statistic. We find out that the German Phillips Curve is purely forward looking. Moreover, our point estimates are consistent with the view that firms re-optimize prices every two to three quarters. While these estimates seem plausible from an economic point of view, the uncertainties around these estimates are very large and also consistent with perfect nominal price rigidity where firms never re-optimize prices. This analysis also offers some explanations why previous results for the German NKPC based on GMM differ considerably. First, standard GMM results are very sensitive to the way how orthogonality conditions are formulated. Additionally, model misspecifications may be left undetected by conventional J tests. Taken together, this analysis points out
the need for identification robust methods to get reliable estimates for the NKPC.
Read article
Schätzunsicherheit oder Korrelation, Welche Risikokomponente sollten Unternehmen bei der Bewertung von Kreditportfoliorisiken wann berücksichtigen?
Henry Dannenberg
IWH Discussion Papers,
No. 5,
2007
Abstract
Die Bewertung der Ausfallwahrscheinlichkeiten von Ratingklassen, basierend auf historischen Daten, ist mit Schätzunsicherheit verbunden. Zur Bewertung dieser Unsicherheit werden in der Literatur Konfidenzintervalle diskutiert. Diesen liegen allerdings Annahmen bezüglich der Abhängigkeiten zwischen einzelnen Forderungen zugrunde, die im Widerspruch zu den Annahmen der gängigen Kreditportfoliomodelle stehen. Im vorliegenden Beitrag wird anhand von Simulationsstudien gezeigt, dass eine Berücksichtigung von Schätzunsicherheit in kleinen Portfolios gerechtfertigt sein kann, auch wenn dafür Abhängigkeiten vernachlässigt werden müssen. Die Modellierung der Schätzunsicherheit beruht hier auf der Idee der Konfidenzintervalle und der ihnen zugrundeliegenden Verteilungen. Die Ergebnisse der Arbeit sind vor allem für die Modellierung von Forderungsportfolios in Unternehmen von Interesse.
Read article
