Unveiling the impact of irrelevant answers on analyst forecast errors: A topic modeling approach

Abstract

This study explores the influence of irrelevant answers during earnings communication conferences on analyst forecast errors. Utilizing the LDA method to quantify text-based answer irrelevance pertaining to various topics, we uncover that the degree of irrelevant responses concerning product-related issues positively correlates with analyst forecast errors, while those related to the firm's financial performance and corporate governance do not significantly correlate with them. This causal relationship is robustly confirmed by a comprehensive series of endogeneity tests and robustness checks. Additionally, our cross-sectional analysis reveals that our main findings are more pronounced in firms with higher operational complexity and weaker information environments, supporting our hypothesis that analysts encounter greater challenges in identifying and interpreting irrelevant answers regarding product information, thereby leading to reduced forecast accuracy.