Predicting Bankruptcy for Government Contractors
Abstract
This thesis attempted to build a bankruptcy prediction model for evaluating potential and current government contractors. The study addressed two general research questions: (1) What financial distress models have been developed and how reliable are these models for predicting bankruptcy for government contractors? (2) Can new models be built that can reliably forecast bankruptcy for government contractors? Review of the literature found a multitude of previous research efforts on predicting financial distress. Two of the models developed by other researchers were based on government contractor data; one was developed using discriminant analysis (Dagel and Pepper) and the other using a combination of discriminant analysis and univariate analyses (Moses and Liao). Inputting the financial information from the samples of bankrupt and nonbankrupt firms into four other models (Altman, Dagel and Pepper, Moses and Liao, and Zavgren) showed that the models were less successful predicting bankruptcy than reported in these studies. None of these four studies appeared to take prior probabilities of differences in misclassification costs into account when reporting their accuracy/error rates. These can have a tremendous impact on a model's reliability. Two techniques--logistic regression and discriminant analysis--were used to build models based on the sample data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1990
- Accession Number
- ADA229417
Entities
People
- Irene C. Godfrey
Organizations
- Air Force Institute of Technology