Data-Driven Sensitivity Analysis of Structural Equation Modeling: A Case Study of Bayesian Network Analysis

Abstract

In recent years, securing well-being in the workplace is a shared purpose at Mazda to inspire employees working in all departments. With a view to improving the workplace environment, the company conducts an annual Employee Attitude Survey. However, the survey results were not fully utilized in measures to improve the mental health environment of employees. To solve it, this study attempted to find a way to facilitate fundamental improvement in the environment, using the Structural Equation Model (SEM), with employee well-being as the target indicator. The SEM made it clear that job performance (productivity and creativity) is influenced by several factors related to well-being. This outcome was used as an effective tool by department general managers and in labor-management roundtables. In addition, although the initial focus was on understanding the current situation in the workplace behind the individual responses by improving the data analysis methods and visualization of the results, the feedback of the study outcome showed that: (1) Focusing on the workplace environment, rather than individuals, can reduce the psychological burden on those working on the improvement; (2) The company-common format has made it easier to disclose and compare each other's current status and has created secondary value for the field, such as increased communication across departments. This study has resulted in a more positive change in workplace improvement activities throughout the company.