|
Activity Number:
|
291
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business & Economics Statistics Section
|
| Abstract - #300753 |
|
Title:
|
Predicting Corporate Defaults: A Censored Data Analysis Approach for Testing Model Performance
|
|
Author(s):
|
Radu Neagu*+ and Necip Doganaksoy
|
|
Companies:
|
General Electric Research and GE
|
|
Address:
|
1 Research Circle, Niskayuna, NY, 12309-1027,
|
|
Keywords:
|
default ; probability of default ; exponential distribution ; censored data ; Type I error ; Type II error
|
|
Abstract:
|
The progression of a corporation from financial stability to financial distress usually happens over relatively long periods of time, providing the opportunity for identifying these "failing" corporations ahead of time. Consequently, the goal is to provide portfolio managers with early notice of deteriorating financial status for a given corporation so that profitable business decisions can be taken. There are a multitude of techniques available for estimating that a corporation will go into financial default within the near future. We construct an alerting system using equity inferred probability of default (PD) metrics that quantify the likelihood that a corporation will go into financial default within a year. The system was constructed to be optimal from a Type 1/Type 2 error balance point of view. When applying the system on an ongoing basis, as new data get recorded, there is a natural lag in computation of the Type 2 error that may result in an incorrect assessment of the system's performance. Using censored data analysis techniques, we managed to significantly reduce the lag time in computation of the Type 2 error by a factor of four.
|
