Evaluating safety performance in manufacturing sector: An enhanced super-efficiency data envelopment analysis approach

• Developing a super-efficiency DEA model for systematic safety performance assessment. • Incorporating historical incident data and proactive safety measures for performance analysis. • Addressing challenges of infeasibility in DEA models caused by zero input values. • Enhancing the discriminatory power of conventional super-efficiency DEA models. • Demonstrating the applicability of the proposed model for safety evaluation in manufacturing. Workplace safety in the manufacturing sector is a critical concern, with high numbers of lost-time injuries and fatalities impacting overall productivity and economic stability. Traditional safety performance evaluation methods rely on lagging or leading indicators, yet these approaches often fail to provide a comprehensive assessment of workplace risks. Data envelopment analysis (DEA) is a valuable tool for systematically evaluating safety performance, considering both lagging and leading indicators. However, the presence of zero values in input data presents a significant challenge for the feasibility of DEA efficiency evaluation models. This study aims to develop a super-efficiency DEA model that effectively addresses these infeasibility issues and manages zero input values. Besides, the model incorporates time-series data to assess safety performance within the manufacturing sector. The proposed model also formulates input savings and output surpluses to ensure a stable and consistent evaluation of workplace safety. This offers a practical tool for policymakers and managers to improve safety performance. The effectiveness of the developed DEA model is demonstrated through its application to safety performance data from the Canadian manufacturing sectors. The findings indicate that Ontario, Alberta, and Quebec achieved the highest levels of workplace safety efficiency in specific years compared to the other provinces. The results also imply that the proposed model provides a more precise evaluation and discriminatory power compared to previous approaches.