PREDICTING RETIREMENT MOTIVES USING MACHINE LEARNING TECHNIQUES IN THE ZIMBABWEAN PENSIONS SECTOR

Zimbabwean pension fund stakeholders face challenges in managing retirement schemes and planning policies due to the limitations of traditional statistical methods in predicting retirement motives. These limitations contribute to mismatches in retirement timing, misaligned policies, and inadequate preparedness. Accurate, data-driven prediction models offer a promising opportunity to enhance retirement planning and policy alignment. This study investigates the application of supervised Machine Learning (ML) techniques to predict retirement motives among members of Zimbabwe’s pension sector. Understanding retirement behavior is especially critical in emerging economies to design inclusive and sustainable social security systems. Using pension member data, the study evaluates the performance of various ML models, including Logistic Regression (LogReg), Support Vector Machine (SVM), Random Forest (RF), Multilayer Perceptron (MLP), and Extreme Gradient Boosting (XGBoost). Model performance was assessed using cross-validation with metrics such as accuracy, precision, recall, and ROC-AUC. SHapley Additive exPlanations (SHAP) were employed to interpret model outputs and identify key features influencing retirement decisions—such as age, health status, job embeddedness, service length, financial literacy, and family responsibilities. Results show that ensemble models, particularly XGBoost and Random Forest, outperformed neural networks and linear models in predicting early and delayed retirement motives. The study highlights the potential of socially-informed ML approaches in low-resource contexts, offering valuable insights for policymakers and pension administrators. By integrating AI into pension forecasting, the research supports smarter fund management and policy development, contributing to more resilient and responsive pension systems.