Identifying Key Predictors of Firm Performance: An Analysis Using Machine Learning Models

Abstract

This study explores the importance of worker bodies in combination with 67 other features on firm performance using the data from the European Company Survey (ECS) 2019 dataset. The scope of this study is limited to the Germanic cluster of countries, including Austria, the Netherlands, and Germany. Firm performance was measured based on a subjective variable rated by the management of the establishments based on their profit-making situation. The main research question of the study is “What are the most influential factors on firm performance?”, and the sub-question is “How important is the role of worker bodies in predicting firm performance?”.

We used Random Forest, LightGBM, and XGBoost models using both classification and regression approaches to find the feature importance and SHAP values of the features. The results showed that worker body existence is the least important factor across all other features, while changes in production level, employment status, and motivation of employees are the most important features. At a higher level, firm characteristics, skill, and training factors demonstrated the highest level of importance, whereas collaboration and external factors, such as product market strategy, had the lowest importance values. This study is of value to econometricians and management researchers as it gives them an integrated and holistic overview of multiple features while focusing on a subset of them in their fields of interest.