Enhancing SME financial distress prediction by integrating financial and deposit dynamics indicators

Accurate credit risk assessment for small and medium-sized enterprises (SMEs) is essential for financial institutions. However, traditional models relying on financial statements and credit lines face structural limitations, including a lack of timeliness and the exclusion of microenterprises. We argue that dynamics of liquid deposit balances, held by all firms regardless of credit facilities, serve as an early warning signal of cash flow conditions. Using Japanese SME data and machine learning techniques, we empirically investigate the effectiveness of integrating deposit dynamics with financial ratios. The results show that incorporating deposit dynamics yields statistically significant improvements in financial distress predictive ability over standalone financial models. Furthermore, during uncertain periods like the COVID-19 pandemic, these deposit dynamics indicators deteriorate asynchronously with financial ratios, maintaining predictive ability by compensating for distorted financial figures. High-frequency deposit balances provide additional information beyond traditional financial information, offering a practical approach for SME credit evaluation.