Earnings calls bear complex characteristics:
The sessions include a formal, generally scripted part followed by interactive Q&As – a mix of genres
They have a dual purpose from a Management perspective: to provide an update on financial performance and to promote a company and its management team – easy to reconcile when the results are good, less so when they are below expectations
They involve parties with diverging objectives dealing with information asymmetry: selective disclosure of information for Management vs. maximum extraction of information for sell-side analysts
It makes earnings calls a challenging exercise for its participants who need to approach these events in a tactical if not strategic manner as they are not one-off events. They require that a relationship between corporate officers and the analyst community be built and nurtured. In fact, in ‘Persuasion in earnings call‘, the author notes the existence of a repertoire of expressions that are distinctive to the earnings call genre itself (a sort of ‘earnings callese’) that reflects the existence of a close-knit professional community.
Notwithstanding this linguistic standard, earnings call dynamics are so complex that they contribute to the so-called ‘post earnings announcement drift’, i.e. the evolution of the share price for days and even months on the back of an earnings surprise following a conference call. Theoretically, the share price of any company should reflect the information disclosed during the call right at the end of it or soon thereafter. And yet the price adjustment process drags which indicates that the information contained in the conference calls takes time to be digested by the market.
Indeed, what matters beyond the actual results is not what Management say, but what they do not say. The Diversified Industrials team at UBS has often used a simple word count analysis based on earnings call transcripts to identify communication biases (e.g. emphasis on growth as opposed to profitability, undeserved focus on smaller divisions, lack of references to innovation, etc.). Going further, analyzing the way corporate officers express themselves on calls can provide important cues. Examples of findings on linguistic patterns can be summarized as follows:
Based on ‘Earnings conference calls and stock returns: The incremental informativeness of textual tone’, ‘Earnings conference calls and stock returns: The incremental informativeness of textual tone’, the researchers prove that the linguistic tone (i.e. sentiment) used in earnings call in the Q&A portion is a greater predictor of abnormal returns and trading volume (including during the post earnings announcement drift) than the earnings surprise per se
In ‘Can investors detect Managers’ lack of spontaneity?’, the authors demonstrate that Q&A scripting (which can be detected through word patterns) signals negative information about the firm. And in fact, further analysis indicates that investors’ negative reaction to scripted Q&A is consistent with future firm fundamentals
According to ‘Detecting deceptive discussions in conference calls’, reference to general knowledge (‘you know’, ‘everybody knows’) is a very strong predictor of deception as represents a filler-type phrase which may imply a lack of substance; A low number of non-extreme positive words is also positively correlated with a deceptive behavior. The authors elaborates on differences in linguistic cues between CEOs and CFOs, evidencing the wealth of information which can be contained in transcripts
‘Does silence speak? An empirical analysis of disclosure choices during conference calls’ demonstrates, as expected, that the refusal to answer questions from analyst reflects negatively on a company
It may not take long before a systematic, broad-based linguistic analysis of earnings call transcripts find its way into trading algorithms. Watch your words!