Performance comparison of empirical and theoretical approaches to market-based default prediction models
(2009)Department of Business Administration
- Abstract
- Results are mixed as to whether the contingent-claim approach to credit risk evaluation is superior to other methods. We question the validity of prior research that has attempted to answer this question by applying normal distribution to calculate the implied probabilities of default in their assessment of Black-
Scholes-Merton models. This is because, unlike the research community, the actual Moody's KMV model commonly used in the financial industry uses empirical default frequency to map distance to default (DD) to probabilities of default (PD). Therefore, it seems questionable whether we can infer the
properties of true MKMV model with the theoretical models presented in the literature. Our study aims at confronting the two... (More) - Results are mixed as to whether the contingent-claim approach to credit risk evaluation is superior to other methods. We question the validity of prior research that has attempted to answer this question by applying normal distribution to calculate the implied probabilities of default in their assessment of Black-
Scholes-Merton models. This is because, unlike the research community, the actual Moody's KMV model commonly used in the financial industry uses empirical default frequency to map distance to default (DD) to probabilities of default (PD). Therefore, it seems questionable whether we can infer the
properties of true MKMV model with the theoretical models presented in the literature. Our study aims at confronting the two approaches and evaluating whether we can bridge the gap between the models, or if we should re-evaluate the quality of MKMV default predictions. We contribute by estimating the empirical Expected Default Frequency (EDF) distribution based on a emulated subset of Moody's proprietary KMV database; then performing a detailed assessment of this KMV emulation vs. the common normal distribution approach. We find that the information content does in fact differ between the two models, and, given a sufficiently large sample, the empirical EDF estimation method attempted can provide a closer approximation of true default probability distribution than the theoretical normal distribution approach. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/2171716
- author
- Holley, Matthew and Mucha, Tomasz
- supervisor
- organization
- year
- 2009
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Management of enterprises, Credit Risk, Probability of Default, Moody's, KMV, Bankruptcy Forecasting, Expected Default Frequency, Företagsledning, management
- language
- Swedish
- id
- 2171716
- date added to LUP
- 2009-06-04 00:00:00
- date last changed
- 2012-11-12 11:58:30
@misc{2171716,
abstract = {{Results are mixed as to whether the contingent-claim approach to credit risk evaluation is superior to other methods. We question the validity of prior research that has attempted to answer this question by applying normal distribution to calculate the implied probabilities of default in their assessment of Black-
Scholes-Merton models. This is because, unlike the research community, the actual Moody's KMV model commonly used in the financial industry uses empirical default frequency to map distance to default (DD) to probabilities of default (PD). Therefore, it seems questionable whether we can infer the
properties of true MKMV model with the theoretical models presented in the literature. Our study aims at confronting the two approaches and evaluating whether we can bridge the gap between the models, or if we should re-evaluate the quality of MKMV default predictions. We contribute by estimating the empirical Expected Default Frequency (EDF) distribution based on a emulated subset of Moody's proprietary KMV database; then performing a detailed assessment of this KMV emulation vs. the common normal distribution approach. We find that the information content does in fact differ between the two models, and, given a sufficiently large sample, the empirical EDF estimation method attempted can provide a closer approximation of true default probability distribution than the theoretical normal distribution approach.}},
author = {{Holley, Matthew and Mucha, Tomasz}},
language = {{swe}},
note = {{Student Paper}},
title = {{Performance comparison of empirical and theoretical approaches to market-based default prediction models}},
year = {{2009}},
}