Coupling Occupational Safety and Labour Productivity in Construction: A Multi-City Empirical Analysis and Predictive Optimization Framework

¹ Professor & Head, Department of Civil Engineering, KAAF University, Accra, Ghana
² Programme Manager, Senior Lecturer, Department of Civil and Mechanical Engineering, Middle East College, Muscat, Oman.
³ Senior Lecturer, Department of Civil and Mechanical Engineering, Middle East College, Muscat, Oman.
⁴ Professor, Department of Midwifery, KAAF University, Accra, Ghana.

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

Construction Safety Management and Labour Productivity are frequently viewed as mutually exclusive in construction management. However, there is an increasing body of research indicating that occupational safety contributes to operational stability and better employee performance. This paper assesses the quantitative association between safety and labour productivity by using a multi city dataset which comprises 1275 observation for construction projects across six metropolitan areas. Indicators for both safety and operationally were evaluated to determine how they contributed to the labour productivity through use of key indicators including Personal Protective Equipment (PPE) compliance, hours of safety training, frequency of accidents, number of near misses, number of hours worked at overtime, hours lost due to work stoppages, and number of skilled workers. The predictive modeling approach utilized included Quantile Regression, Recursive Least Squares, Generalized Estimating Equations, and Weighted Least Squares with five fold cross validation to test for robustness. The results show that all safety compliance variables positively correlated with productivity, specifically PPE adherence and training intensity. The models indicated that disruption related variables such as accident frequency, hours worked overtime, and hours lost due to work stoppages negatively correlated with productivity. Overall, the models explained around 72-74% of the variation in productivity in unseen data and therefore have significant predictive capabilities.