Managerial horizon: an LLM approach

This paper develops a novel, large language model (LLM)–based measure of managerial horizon using managers’ language in earnings calls. Leveraging a hybrid GPT–BERT framework, we classify millions of transcript segments into performance, strategy, and activity content and construct a firm-quarter measure of managerial short-termism based on the relative emphasis on performance. Firms with more short-term–oriented managerial language exhibit lower R&D and capital investment, weaker innovation output, slower earnings growth, and poorer long-run stock performance. The effects are economically meaningful and strongest among opaque firms, demonstrating that narrative disclosure reveals managers’ intertemporal preferences and has real valuation implications.