CEU Electronic Theses and Dissertations, 2025
| Author | Ussenov, Azizbek |
|---|---|
| Title | Predicting Corporate Bankruptcy with Machine Learning: Integrating Financial and Macroeconomic Indicators - Evidence from U.S. Public Companies (1999-2018) |
| Summary | In this thesis, I predict the likelihood of a company's bankruptcy using macroeconomic indicators for American companies listed on the New York Stock Exchange and NASDAQ for the years between 1999 and 2018. Advanced machine learning models, such as XGBoost, are employed to compare them with traditional ones, like Logistic Regression. I implement a rolling window approach to consider time-dependent changes in both firm and economy-level conditions when predicting default. The findings demonstrate that the XGBoost model achieves higher accuracy compared to the logit model, with an pooled AUC score of 0.9443. Even though macroeconomic indicators - such as interest rates, inflation, and real GDP growth - add predictive power to the models, the contribution is not at a significant level. The potential reason can be attributed to the use of fixed and the same macro variables for companies, which fail to provide cross-sectional discriminative power in predicting bankruptcy. Overall, the results demonstrate the advantages of XGBoost in developing data-driven solutions for monitoring the financial health of companies. |
| Supervisor | Robert Lieli and Zsofia Barany |
| Department | Economics MA |
| Full text | https://www.etd.ceu.edu/2025/ussenov_azizbek.pdf |
Visit the CEU Library.
© 2007-2025, Central European University