Benchmarking Machine Learning Models to Predict Corporate Bankruptcy

Paper: arXiv 2212.12051

Authors: Emmanuel Alanis, Sudheer Chava, Agam Shah

Abstract

Using a comprehensive sample of 2,585 bankruptcies from 1990 to 2019, we benchmark the performance of various machine learning models in predicting financial distress of publicly traded U.S. firms. We find that gradient boosted trees outperform other models in one-year-ahead forecasts. Variable permutation tests show that excess stock returns, idiosyncratic risk, and relative size are the more important variables for predictions. Textual features derived from corporate filings do not improve performance materially. In a credit competition model that accounts for the asymmetric cost of default misclassification, the survival random forest is able to capture large dollar profits.

Complexity vs Empirical Score

• Math Complexity: 3.5/10
• Empirical Rigor: 8.0/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper relies on applied machine learning methodologies rather than advanced mathematical proofs, but compensates with a remarkably robust empirical dataset spanning nearly three decades. Its strength lies in thorough benchmarking, variable importance analysis, and practical application via a realistic credit competition cost model.

Research Flowchart

  flowchart TD
    A[Goal: Predict U.S. Corporate Bankruptcy & Optimize Portfolio Returns] --> B[Data: 2,585 Bankruptcies 1990-2019<br/>Financial Ratios, Stock Metrics & Filings Text]
    B --> C[Model Training & Benchmarking<br/>GBT, Survival Random Forest, etc.]
    C --> D[Model Evaluation & Feature Analysis<br/>Permutation Tests & Credit Competition Model]
    D --> E[Finding 1: Gradient Boosted Trees Best for 1-Year Forecasts]
    D --> F[Finding 2: Market Variables Drive Predictions<br/>Text Features Add Minimal Value]
    D --> G[Finding 3: Survival Random Forest Maximizes Dollar Profits]