Credit Risk Modelling for Small and Medium-Sized Enterprises in Zimbabwe

Purpose: The main aim of the study was to determine a context-specific credit risk assessment framework that integrates both traditional financial metrics and alternative data sources to better evaluate the creditworthiness of Zimbabwean SMEs. It also identified key factors affecting credit risk for SMEs in Zimbabwe by incorporating both financial and non-financial data. Design/methodology/approach: We employed machine learning algorithms which were Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and AdaBoost (AB). The data was obtained from the loan database of an SME banking division of a commercial bank of Zimbabwe consisting of 52,750 loan applicants over the 5-year period from 2018 to 2022. The features of the credit dataset included capital structure, financial history, profitability, liquidity, growth potential, industry characteristics, management quality, social media engagement, macroeconomic environment, customer concentration, credit history, firm age, collateral availability and whether the applicant defaulted on the loan or not. We performed data preprocessing and cleaning, feature development, hyper parameter selection and cross validation. A split ratio of 80% for the training set and 20% for the testing set was used, followed by an evaluation of the model based on the following performance metrics: classification accuracy, precision, recall (sensitivity), F1-score and the Receiver Operating Characteristic - Area Under the Curve (ROC-AUC). Findings: We find that traditional financial indicators such as profitability, liquidity and leverage, non-financial factors—including collateral availability, cash flow stability, management quality, and macroeconomic conditions—play a significant role in shaping credit risk profiles. Non-traditional data sources such as firm characteristics, supplier–buyer relationships, and social media activity can provide deeper insights into SMEs’ operational performance and risk exposure. Incorporating these data sources alongside traditional financial information can significantly enhance the prediction of defaults. Research limitations/implications: This study was confined to one Zimbabwean bank, this represents a narrow focus since the Zimbabwean banking industry has 343 players as at 30 September 2025, (Reserve Bank of Zimbabwe (2025)). Also, since we base our research on Zimbabwe, it implies that the findings of this study cannot be generalised to all developing countries. Originality/value: This study contributes to the theory by providing an enhanced credit risk assessment framework that integrates traditional financial indicators and alternative data sources for Zimbabwean banks when determining the credit risk of SMEs. This will improve access to credit by SMEs in Zimbabwe and in jurisdictions with similar economic environments as those found in Zimbabwe. By providing reliable credit risk assessment methods it increases the financial inclusion of SMEs.