Efficiency of Public Social Security Expenditure: A Cross-Country Study Using Factor Analysis and Advanced Machine Learning

Research objectives. Contemporary global challenges, such as demographic shifts, the climate crisis, and rapid technological transformation, necessitate innovative approaches to managing social security systems. This study addresses the urgent need for tools to enhance the efficiency of Financial-Investment Models of Social Security (FIMSS), particularly under constrained fiscal resources and heightened uncertainty. The aim is to develop and validate a comprehensive approach for assessing FIMSS efficiency, incorporating modern challenges and public finance management specifics. Methods. By integrating ratio analysis, factor analysis, and advanced machine learning techniques, including gradient boosting (XGBoost), this study establishes a robust, multi-level framework for efficiency evaluation. The dataset covers 38 Organisation for Economic Cooperation and Development (OECD) countries, Russia, and China over the period 2005–2022, enabling cross-country comparisons, with regression analysis limited to a subsample of 26 countries due to data availability. The scientific novelty lies in introducing the EffCoverSP indicator, which accounts for social protection coverage and employing partial dependence plots (PDP) to uncover nonlinear relationships among socioeconomic factors, extending macroeconomic theories of social system sustainability and social justice frameworks. Results reveal that FIMSS efficiency is driven by moderate budgetary expenditures, public debt below 50% of gross domestic product, a Gini index of 0.37–0.43, urbanization of 63–74%, and fertility rates of 1.55–1.7. The practical significance lies in the potential application of this approach to reform FIMSS, enhancing their sustainability and adaptability to global challenges, thereby informing evidence-based policy decisions.