The role of personal credit in small business risk assessment: a machine learning approach

Zilong Liu and Hongyan Liang

Abstract

Accurately predicting default risk among small businesses is critical for lenders and policy makers. However, traditional credit risk models often rely on extensive financial statements that many small enterprises lack. This study explores the value of integrating the personal credit bureau data of business owners, along with business-level and tradeline variables, within a machine learning framework to improve default prediction. Using a large data set from the Gies Consumer and Small Business Credit Panel, our baseline models relying solely on fundamental business attributes achieve an area under the receiver operating characteristic curve (AUROC) of approximately 0.78. Incorporating business tradeline information (such as active accounts and delinquency patterns) raises performance only slightly (AUROC ≈ 0:79). In contrast, adding personal credit features substantially boosts accuracy, pushing the best-performing gradient boosting models (XGBoost, Light- GBM and CatBoost) above 0.83. Feature importance analyses underscore the intertwined nature of business and owner finances: variables capturing personal credit scores, outstanding balances and recent inquiries rank among the strongest predictors, alongside measures of business debt (eg, Uniform Commercial Code (UCC) filings and open tradeline balances). These findings reveal that personal credit factors can fill critical information gaps when formal business records are scant, thereby strengthening credit risk assessments and enhancing lending decisions in the small business sector. In addition, our results highlight the critical importance of validating risk models using alternative data sources, ensuring greater robustness and reliability in predicting small business defaults.