Working Paper: NBER ID: w27950
Authors: Sean Cao; Wei Jiang; Baozhong Yang; Alan L. Zhang
Abstract: Growing AI readership, proxied by expected machine downloads, motivates firms to prepare filings that are friendlier to machine parsing and processing. Firms avoid words that are perceived as negative by computational algorithms, as compared to those deemed negative only by dictionaries meant for human readers. The publication of Loughran and McDonald (2011) serves as an instrumental event attributing the difference-in-differences in the measured sentiment to machine readership. High machine-readership firms also exhibit speech emotion assessed as embodying more positivity and excitement by audio processors. This is the first study exploring the feedback effect on corporate disclosure in response to technology.
Keywords: Corporate Disclosure; AI; Machine Readability; Sentiment Analysis
JEL Codes: G14; G30
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Higher expected machine downloads (C59) | Increased machine readability of corporate filings (G38) |
| Higher expected machine downloads (C59) | Increased vocal positivity during conference calls (G40) |
| Higher expected machine downloads (C59) | Reduction in use of LM negative words in disclosures (G33) |
| Publication of the Loughran and McDonald dictionary (G10) | Adjustments in sentiment of disclosures (G38) |