The ability to explain a concept or subject to someone else is a challenge. It requires you to both understand the topic you have been asked to discuss, but also communicate it in a way that is understandable to the listener or reader. Listening and reading are two of the most important activities in our industry. After all, we have two eyes and two ears, but only one mouth.
We also have one brain, and whether we care to admit it, by definition, our intelligence is likely to sit somewhere slightly above or below the average level for the human population. When reading a text or listening to someone speak, we often require that person to deliver the information in a way that is understandable to someone of average intelligence.
In the 1950s, American businessman Robert Gunning developed the Gunning fog index, which sought to quantify the readability of a body of text. The index measures the length of sentences and the complexity of words used, coming up with a score that can be compared to school grade reading levels. A fog index score of 12 is comparable to the reading level of an 18-year-old student1 and is seen as a standard for texts intended for a wide audience.
The world of investor communications contains many forms, including annual reports, earnings call transcripts and periodic fund manager letters. If an index like Gunning fog classifies the readability of such materials as very high, this could be a clue that requires further investigation. Complexity of language may be used because the writer or speaker doesn’t understand the topic as well as they think they do. Perhaps more disconcerting however, is the language is being used to conceal something from the reader or listener.
A recent research report from the quantitative team at Nomura found that the complexity of language used in earnings calls corresponded with investor returns. Those management teams that used complex language, as measured by the Gunning fog index, averaged a return of 9.5% per year, while companies that used simpler language averaged 15.4%2.
While detecting deception in the investing world has mostly been concentrated in an assessment of the quantitative data provided by companies, a growing body of research is attempting to build on the insights gleaned from basic readability algorithms such as Gunning fog. Lina Zhou of the China University of Geosciences formalised some of this work into nine categories of deceptive linguistic cues3, which have been successfully used to identify fraudulent financial statements4. Examples of these cues include distancing strategies such as the use of third over first-person pronouns (e.g., “they” instead of “we”), and obfuscation methods such as incohesive sentence formation. Together, the cues form a linguistic fingerprint that can help to identify the true intention of the words used in a communication.
An example in practice is the comparison between the annual CEO letters from two, albeit cherry-picked, companies. The thousand-word 2017 CEO letter from fraudulent German fintech business, Wirecard, scores over 195 on the Gunning fog index, a readability level classed as very difficult or confusing for the audience. Similar sized samples from Warren Buffett’s much longer Berkshire Hathaway letter of 2020, scores around 136, slightly above the guidance for mainstream consumption.
Inevitably there will be exceptions to the above, where a complex communication is not concealing anything, and vice versa. Perhaps though, Nomura’s research also highlights that speaking and writing to stakeholders with both simplicity and clarity, builds a foundation of trust between the parties that ultimately delivers better results. While our communication to clients doesn’t directly influence the underlying assets we hold in portfolios for investors, clear and understandable interactions will build a level of trust that gives investors the confidence to remain invested during the more challenging moments that are faced over time.
3Zhou L, Burgoon J, Nunamaker J, Twitchell D. Automating linguistics-based cues for detecting deception in text-based asynchronous computer-mediated communication.
4Humpherys S, Moffit K, Burns M, Burgoon J, Felix W. Identification of fraudulent financial statements using linguistic credibility analysis.