Clustering and Determinant Analysis of Corporate Environmental Management Attitudes: Unsupervised Learning Techniques

Abstract

Purpose - The main objective of this study is to cluster companies based on environmental management attitudes using various indicators, and analyze differences in activities among the clusters to gain insights into environmental management performance. Design/Methodology/Approach - The study encompasses panel data from 12 countries, 22 industry sectors, and 3,402 companies from 2013 to 2022. This study uses both machine learning methodology and conventional econometric methodology. The primary methodology involves using an unsupervised learning technique, K-means clustering, to categorize environmental management attitudes. Subsequently, Kernel Density estimation and Multinomial Logistic Regression were employed to conduct comparative analyses of environmental performance indicators within each cluster and identify the determinants shaping environmental management attitudes. Findings - The research revealed the existence of five distinct clusters of companies with varying environmental management attitudes. Through Multinomial Logistics Regression, it identified that variables such as greenhouse gas emissions, energy consumption, total assets, return on assets, and other significantly influence environmental management attitudes. Research Implications - By uncovering common characteristics and determinants within major countries and industries, this study offers valuable insights in promoting domestic environmental management in the context of rapidly changing international trade and investment environments. The findings provide policy implications and suggestions for fostering proactive environmental management practices among domestic companies.