Data-Driven Credit Risk and SME Portfolio Optimization: A Practical Framework for Regional and Community Financial Institutions: Evidence from U.S. SBA 7(a) Lending Data and Interpretable Portfolio Analytics

This study develops and tests a practical credit-risk and portfolio-optimization framework for regional and community financial institutions using official U.S. Small Business Administration (SBA) 7(a) loan data. The empirical sample contains 463,775 loans approved between fiscal years 2010 and 2021 and subsequently observed as either paid in full or charged off. Descriptive analytics, portfolio heat maps and an interpretable logistic-regression model are combined to identify risk patterns that are directly relevant for origination, pricing, monitoring and portfolio steering. The observed charge-off rate in the resolved-loan sample is 7.17%. New or startup businesses (8.20%) and unsecured exposures (9.62%) underperform established firms (6.18%) and secured loans (6.21%). Loans with maturities of 60 months or less exhibit the highest observed charge-off rate (17.59%), while 241-360 month credits perform materially better (1.26%). Sectoral differences are economically meaningful: transportation and warehousing, accommodation and food services, retail trade and construction show above-average risk, whereas health care and social assistance is materially lower. The out-of-time model achieves an area under the ROC curve of 0.846 on a 2019-2021 test set, and the top two predicted risk deciles capture 68.23% of observed defaults. The paper translates these findings into a governance-oriented operating model that combines interpretable scorecards, relationship-lending judgement, collateral discipline, sector and vintage monitoring, and early-warning dashboards. The contribution lies in linking SME credit analytics to portfolio governance in a form that is rigorous enough for risk management yet practical enough for regional and community institutions.