Traditionally, manipulation of corporate financial statements has been detected through analysis of financial data and other economic indicators. A 2011 working paper by the Rock Center for Corporate Governance at Stanford Graduate School of Business, “Detecting Deceptive Discussions in Conference Calls,” looks at the potential of linguistic analysis of company’s quarterly earnings discussions.
In the study, transcripts of 29,663 CEO and CFO quarterly earnings conference calls were analyzed using financial information in annual reports as well as the results of prior research on “deceptive language.” According to the researchers, the assumption behind analyzing language as a predictor of deceitful accounting is “that CEOs and CFOs know whether financial statements have been manipulated and that the formal and spontaneous narratives of these executives provide cues that can be used to identify lying or deceitful behavior.”
Results of the study conclude that:
- References to general knowledge act as an indictor of untruthful statements. In particular, phrases such as “you know” or “everybody knows” provide a strong indication that language may be deceptive. For CEOs, the effect of an increase in “you know”-type phrases by 29 increases the odds of deception by a factor of 1.91, for CFOs, the effect of an increase by 19 is at least a factor of 2.05.
- When the increase of words pertaining to anxiety increase by 1% from the median, this is associated with at least an 8% increase in the likelihood of financial misrepresentation.
- Overall, the classification models employed in this research “based on verbal cues of CEOs and CFOs perform significantly better than a random classifier by about 6% to 16%.”
Characterizing the success of earlier methods of identifying potential fraud as “modest” at best, the researchers suggest that linguistic analysis offers a new approach to identifying manipulation of financial statements.