Is financial distress recurring? Predicting multiclass distress and recovery dynamics using machine learning

This study addresses a significant limitation in financial distress research, which has traditionally conceptualized distress as a static, binary outcome, by redefining it as a recurring, state-transitional process. Recurring distress is defined as instances where firms repeatedly transition into and out of distressed financial states, rather than achieving a stable recovery following a single episode. Utilizing longitudinal data from publicly listed Indonesian firms this research employs machine learning techniques to develop a framework to capture firm transitions across varying financial conditions and to examine the unfolding of recovery over time. The results indicate that recurring distress, characterized by fragile and unstable recovery, represents a particularly challenging state to identify due to the ambiguity inherent in transitional financial conditions. The analysis also demonstrates that recovery trajectories are heterogeneous, with rapid recoveries generally more distinct than slower, more complex adjustment paths. Firm characteristics, financial conditions, and strategic responses collectively influence these dynamics; restructuring efforts are frequently associated with improved recovery prospects, whereas high leverage often constrains recovery. In summary, this study contributes a more dynamic perspective on financial distress as an evolving process shaped by the continuous interaction between firm-level conditions and strategic adaptation.