The Informativeness of Accounting Ratios for
Bankruptcy Prediction Through the Economic Cycle
September 18th, 2025
2:30pm – 4pm in Collines Building (L712) & on Zoom
————–
Speaker: Anne D’ARCY – Wu Vienna University, Austria
abstract
We apply multiple interpretable machine-learning models to identify the most important bankruptcy predictors based on accounting information during different phases of the economic cycle. Most bankruptcy prediction models in prior research assume that the predictive power of their input variables— in particular accounting ratios — stays constant over time.
However, literature suggests that the informativeness of accounting data and therefore their predictive power changes during different phases of the economic cycle. Our results show that the three ratios with the highest importance are constant across the different phases of the economic cycle whereas other ratios become more important during a recession or a non-recession period. We are therefore able to generate more accurate predictions and to identify relevant ratios applying interpretable machine-learning techniques considering the economic cycle.
We conclude that the informativeness of accounting information is a more nuanced concept than prior research suggested.