Understanding and Predicting the Resolution of Financial Distress

In this study, we empirically investigate the determinants of the process utilized to resolve financial distress (resolution process) and also the outcome of the financial distress (resolution outcome). Specifically, we separate firms that utilize a private (work out) versus a public (filing for bankruptcy) resolution process and then we further separate the firms by outcome - liquidation or reorganization. Various qualitative dependent variable models are estimated and compared: ordered logistic regression (OLR), local regression models (LRMs) and feed forward neural network (FNN). We select several accounting and economic variables, measured at the time of default, which are expected to influence the resolution process and the resolution outcome. Estimation results reveal the OLR specification achieves the best balance between in-sample fit, consistency with financial theory, and out-of-sample classification accuracy. We find that larger firms with higher liquidity and more secured debt in their capital structure are more likely to follow a public resolution process. Firms with higher Z-scores and more total leverage are less likely to follow a public resolution process and attempt to resolve the financial distress privately. For resolution outcome, we find that firms with greater liquidity, more secured debt and lower cumulative abnormal returns are more likely to be liquidated rather than reorganized. And firms with more leverage, more intangible assets and filing a prepackaged bankruptcy are more likely to be reorganized.Model performance is assessed on the dimensions of discriminatory power, predictive and classification accuracy. The former two are measured by implementing standard tests (power curve analysis and chi-squared tests), while classification accuracy is assessed according to alternative categorization criteria (expected cost of misclassification, minimization of total misclassification and deviation from historical averages) as compared to nayacute;ve random benchmarks. While in- and out-of-sample performance along these dimensions exhibits wide variation across models and criteria, the OLR and LRM models are found to perform comparably, while the FNN model is found to consistently underperform. The statistical significance of these results is rigorously analyzed and confirmed through a resampling procedure, yielding estimated sampling distributions of the performance statistics, confirming these observations