Formalizing this kind of intuitive linguistic knowledge, two Stanford professors have set out some interesting parameters for detection of CEO and CFO deception in just that venue.
The parameters are based on an analysis of 29,663 transcripts of quarterly earnings calls that U.S. public companies held between 2003 and 2007: an overall 17,150 instances of CEO-speak, and 16,032 CFO statements in management discussions and Q&A segments. The result is a model that the researchers find to be as much as 16% more predictive of deception than would be found in a random approach.
"The language composition of truthful narratives differs from that of false narratives," write David F. Larcker, of Stanford Graduate Business School's Rock Center for Corporate Governance, and Stanford Graduate School's Anastasia A. Zakolyukina, in a .
The paper is a must-read for those of us who like appendices listing all kinds of words. More importantly, it offers some useful tips for investors seeking to detect falsehood, and executives looking to burnish credibility.