The firm-level effects of banking sector disruption

This thesis investigates the firm-level consequences of banking sector disruptions in the United States (US), focusing particularly on the effects of bank failures on small and medium-sized enterprises (SMEs). Despite the crucial role SMEs play in the economy and their dependence on bank credit, understanding how bank failures impact them has been hampered by a lack of granular, large-scale data linking smaller US firms to their lenders. To address this gap, this thesis first constructs a novel firm-bank relationship dataset by collecting and processing nearly 40 million loan documents from six major US states spanning two decades. Natural language processing and machine learning techniques, including text embeddings and efficient similarity search algorithms, are employed to link these unstructured filings to establishment-level data, creating a comprehensive and detailed dataset of US SMEs and their banking partners. Using this unique dataset and studying 231 different bank failures from 2000-2022, the thesis employs a difference-in-differences methodology to estimate the effects of bank failures on firm survival, employment growth, and revenue growth. The findings reveal that bank failures inflict substantial and remarkably persistent negative consequences on affected SMEs. Firms whose banks fail are significantly more likely to exit, with an 8 percentage point higher probability of failure even a decade after the event. Surviving firms experience lasting reductions in growth, with employment growth 5.3 percentage points lower and revenue growth 9.6 percentage points lower ten years post-failure, showing no signs of recovery within the observation window. Quasi-experimental evidence from bank failures triggered by regulatory cross-guarantee powers and large-scale fraud corroborates these causal findings. Furthermore, the analysis uncovers significant heterogeneity: the smallest and youngest firms suffer disproportionately larger negative effects, highlighting steep vulnerability gradients. Collectively, this thesis provides evidence on the severe and long-lasting real economic costs imposed by bank failures on SMEs. It contributes a new longitudinal dataset for studying US firm-bank relationships, offers insights into the mechanisms transmitting financial distress to the real economy via SMEs, and highlights large heterogeneity in the effects of bank failure on firm-level outcomes, informing policies related to bank regulation and resolution.