How Firms Fail

This thesis examines whether firms that experience severe reporting failures, including financial misstatement or bankruptcy, follow distinct paths of financial deterioration. Using Dutch firm-level financial, bankruptcy, and audit data (2010–2015), I compare the forecasting performance of two modeling approaches for firm-level financial reporting panel data: functional data models and multi-output panel quantile regression. Functional models treat each firm-year observation as a distribution and allow forecasting entire density functions, offering a rich but complex representation of uncertainty. In contrast, quantile regression provides a simpler and more interpretable framework, particularly useful when forecasting tail events, such as misstatements. A simulation study extensively compares both approaches under nonlinear dependence, heterogeneous trajectories, and extreme tail-risk scenarios relevant for bankruptcy prediction and audit-risk assessment.