Circuitousness in Disclosure Narratives

Abstract

This paper examines circuitousness, which reflects the extent to which related information is spread throughout a narrative as opposed to being grouped together. Circuitousness in the MD&A is negatively associated with the persistence of earnings, especially for firms with negative earnings. Higher circuitousness is also associated with greater ERCs, analyst forecast revisions, abnormal trading volume, and intraperiod price timeliness. Taken together, these results suggest revisiting topics helps inform investors. However, circuitousness also predicts returns over the subsequent year, consistent with it being costly to process. We find that circuitousness is incrementally and more consistently predictive than other textual characteristics that are typically associated with obfuscation, including Fog index and length, which are extremely popular, as well as repetition, boilerplate, and stickiness, which are closely related. Overall, circuitousness appears to help investors integrate disparate information, on average, while generating additional processing costs.

Keywords: disclosure complexity, natural language processing, machine learning, information quality, textual analysis, circuitousness

JEL Classification: D83, D84, G14, G30, M40, M41