Integrating dimensionality reduction and clustering for measuring profit efficiency: A stochastic meta-frontier approach in Korean manufacturing sector

This study evaluates profit efficiency in Korea’s manufacturing sector, with a special focus on strategic industries under national industrial policy. To address heterogeneity in terms of technology level and firm-size, we leverage Stochastic Meta-Frontier (SMF) combined with unsupervised learning techniques (dimensionality reduction and clustering). The result of hybrid approach is validated using clustering quality indices and statistical criteria. Determinants of profit efficiency are further examined using panel (quantile) regression. The results yield three policy implications: (1) targeted policy support for strategic industries is empirically justified; (2) financial soundness alone has limited impact on profit efficiency, whereas tangible asset investment can be more effective in certain sectors; (3) with few exceptions, policies should prioritize skilled labor development over capital deepening. This study contributes a novel and replicable framework by bridging machine learning techniques with SMF. The framework not only enhances methodological rigor but also provides practical guidance for policymakers seeking to foster competitiveness in strategic industries.