Peering into Executive Compensation Similarity: A Network Perspective

Abstract

This study examines the economic determinants of similarity in executive compensation using a network-based approach. We develop a novel and holistic measure of executive compensation based on a natural language processing algorithm with machine learning—document embedding—that captures the nuanced semantics of compensation disclosure. Using semantic similarity of compensation disclosures as a proxy for similarity in compensation practices and philosophies, we find that firms facing similar conditions, including employer and employee factors and contracting processes, exhibit greater similarity in their compensation practices and philosophies. The holistic network-based approach enables us to identify novel aspects of compensation contracting, such as more convergence among large firms than among small firms. These findings persist after controlling for similarity in quantitative compensation features, suggesting that compensation disclosures provide unique information about compensation contracting beyond observable realized compensation. We also find that compensation similarity incrementally predicts firms’ selection of compensation peers, beyond factors explaining peer selection documented in the literature. When adjusting peers, firms tend to drop peers with lower compensation similarity, especially those that are smaller, while adding new peers that are still dissimilar and smaller but perform better. Our innovative approach also contributes more broadly to the literature on narrative disclosures, as it addresses some of the shortcomings of traditional approaches and opens up new avenues for research into cross-firm comparisons and firm clustering.

Keywords: Compensation disclosure, executive compensation, document embedding, semantic similarity, textual analysis

JEL Classification: J41, M12, M41, M52