Corporate failure prediction models for the U.S. manufacturing and retailing sectors.
Item
-
Title
-
Corporate failure prediction models for the U.S. manufacturing and retailing sectors.
-
Identifier
-
AAI8820905
-
identifier
-
8820905
-
Creator
-
Theodossiou, Panayiotis Theofanis.
-
Contributor
-
Adviser: Salih Neftci
-
Date
-
1987
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Economics, Finance
-
Abstract
-
The objective of this study has been the development and testing of more efficient corporate failure prediction models. Many researchers have identified their models using step-wise techniques or a priori information implied by finance theory. Unlike their approach we have selected the variables in our models by combining statistical criteria and a priori information.;An innovation in this study is the development of models incorporating industry effects and time trends. For this purpose we have used variables standardized by their industry means and standard deviations. Surprisingly, although such models are expected to out perform models including non-standardized variables, their performance over the latter has been found marginal.;The nature of corporate failure studies necessitates the use of non-random samples. Most studies have treated such samples as random thereby resulting in biased probabilities of failure. This issue is further explored and an estimation technique based on Bayes' formula has been used in order to avoid this problem. It is evident from our results that correcting the bias results in lost of efficiency in the sense of mean square error.;Another innovation is the association of the cutoff probability used in the evaluation of the models with the different types of misclassification costs. Using different cutoff points to assess the performance of different statistical models we have found logit and probit to have the same classification accuracy and perform slightly better than linear probability model and discriminant analysis.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
