Narrative Innovation, Real Innovative Activities and Earnings Persistence: Evidence from Word2Vec
Posted: 15 Dec 2018
Date Written: September 30, 2018
Whether the narrative innovation disclosure in financial report captures incremental useful information content beyond traditional real innovative activities (i.e., R&D intensity and patent)? To what extent the narrative innovation disclosure is related to real innovative activities? Is there any association between narrative innovation disclosure and earning persistence in the future? To investigate these research questions, we employ emerging text mining techniques to explore potential answers. Assisted with Chinese Natural Language Processing tool and a neural network algorithm termed as Word2Vec, we extract and analyze the narrative innovation disclosure from annual financial reports of listed companies in Chinese A-share market from 2007 to 2016. We find that (1) narrative innovation disclosure is positively related to the level of current real innovative activities; (2) narrative innovation disclosure is positively related to current earnings; (3) narrative innovation disclosure relates positively to future real innovation activities; (4) narrative innovation disclosure also relates positively to future earnings persistence. From the perspective of large-scale text analysis, we provide novel empirical evidence for the information usefulness of narrative innovation in corporate disclosure.
Keywords: Narrative innovation, Real innovative activities, Earnings persistence, Text analysis, Artificial intelligence
JEL Classification: M41; G14; O31; D82; D21
Suggested Citation: Suggested Citation