Statistical Techniques for Bankruptcy Prediction

Inhaltsverzeichnis (Table of Contents)

• Introduction
• Bankruptcy Prediction as a Classification Problem
  • Bankruptcy Prediction Models
  • Structural vs. Reduced Models and Explanatory Variables
  • Collinearity Issues
  • Sampling Considerations
  • Misclassification Costs
  • Measures for Model Performance
    • General
    • Performance Measures for Models with Probabilistic Output
    • Rank Correlation Measures
• Discriminant Analysis
  • Discriminant Analysis as a Classification Technique
  • Bayesian Approach
    • Class membership as posterior probability
    • Assumption of multivariate normality
    • Distributions other than multivariate normal
  • Discriminant Functions Approach
    • General
    • Measures for importance of explanatory variables
    • Testing for mean differences
    • Importance and significance of discrimination functions
  • Stepwise Variable Selection
  • Sampling Considerations
  • Misclassifications Costs
  • Strengths vs. Weaknesses
  • Applications for Bankruptcy Prediction
• Conditional Probability Models
  • General
  • Microeconomic Derivation
  • Model Estimation
  • Link to Discriminant Analysis
  • Significance Testing
    • Nested models and hypothesis testing
    • Test for overall significance of logit coefficients (omnibus test)
    • Wald test for linear restrictions and t-tests
    • Lagrange multiplier test
    • Raftery test
    • Confidence intervals
  • Goodness-of-Fit (GOF)
    • Mean loglikelihood
    • Saturated model and deviance
    • Categorical independents: Pearson and Deviance GOF
    • Hosmer-Lemeshow GOF test
    • Box-Tidwell nonlinearity test
    • Quasi-R² goodness-of-fit measures
  • Variable Selection
  • Sampling Considerations
  • Extensions
    • Interaction and quadratic terms
    • Multinomial models
    • Mixed logit model
  • Collinearity Issues
  • Misclassification Costs
  • Strengths vs. Weaknesses
  • Applications for Bankruptcy Prediction
• Survival Analysis
• CUSUM Charts
• Artificial Neural Networks
• Some Other Techniques
• Bankruptcy Prediction Models in Germany
• Bankruptcy Prediction in Ukraine

Zielsetzung und Themenschwerpunkte (Objectives and Key Themes)

This Master's thesis explores various statistical techniques for predicting bankruptcy. It aims to provide a comprehensive overview of different models, their strengths and weaknesses, and their applicability in predicting corporate failure.

• Analysis of different statistical techniques for bankruptcy prediction
• Evaluation of model performance and limitations
• Exploration of applications and practical implications of bankruptcy prediction models
• Comparison of models across different contexts and datasets
• Discussion of the role of data quality and sampling considerations in bankruptcy prediction

Zusammenfassung der Kapitel (Chapter Summaries)

The first chapter introduces the concept of bankruptcy prediction as a classification problem, outlining key considerations such as model types, explanatory variables, and performance measures. Chapter 2 delves into Discriminant Analysis, exploring its theoretical foundation, practical implementation, and limitations. Chapter 3 focuses on Conditional Probability Models, providing an in-depth analysis of their theoretical underpinnings, estimation methods, and significance testing. The thesis concludes with a summary and discussion of the findings, highlighting the implications for future research and practical applications.

Schlüsselwörter (Keywords)

The central focus of the thesis revolves around bankruptcy prediction, statistical techniques, classification, discriminant analysis, conditional probability models, model performance, data quality, and sampling considerations. Key models explored include logistic regression, probit models, and survival analysis. The thesis also examines the relevance of these techniques in the context of German and Ukrainian economies.