We discuss a model for analyzing and measuring workers' health and psychosocial work-environment on firm productivity. Productivity is measured through the Malmquist productivity index approach using Data Envelopment Analysis. A novel component of the model is that in addition to standard quantity (or tradable) vairables, we incorporate quality (or non-tradable) variables. Specifically we focus on two quality input variables: workers' health status and psychosocial work-environment. The two variables are modeled as latent or unobserved variables using Item Response Theory. Changes over time in productivity are decomposed to asses the contribution from the changes of the input quality variables. The model is illustrated using data from a worksite health promotion program conducted at three large Swedish manufacturing plants (2 paper mills, 1 steel factory) from 2000 to 2003. Over the four years we observe a general improvement in efficiency of 2-5 %, out of which half can be attributed to the improvement in the quality input variables.
Productivity Decomposition of Efficiency Human Resource Management Health Status Psychosocial Work-Environment Ordinal Data Latent Variables