Talking isn’t always cheap. In November 2021, Tesla stock lost an estimated $21 billion on valuation after founder Elon Musk invited his 70 million followers to take part in a poll on whether to sell 10% of their own stock in the company — just one episode in a long-running series of seemingly reckless social media communications from Musk that are believed to have contributed to a $175 billion stock price plunge for Tesla towards the end of 2021.
Although it has long been anecdotally assumed that CEO behavior is directly related to company performance, the improbability of getting billionaire executives to take standardized personality tests has meant that personality-based stock prediction has become a virtual pseudoscience.
Analytical research efforts to create predictive models in this regard have so far focused on the Big 5 personality traits, which are used to assess CEO disposition openness, conscientiousness, extraversion*, compatibility, and neuroticism.
It is not easy to extract all these features empirically based on language and text alone. A 2020 Academy of Management study has since been able to analyze only three of the Big 5 traits (Conscientiousness, Neuroticism, and Extraversion). openness and compatibility are more difficult to identify, especially from text.
A new collaboration between universities in Germany and Italy has taken a novel approach to formulating a predictive framework of this type, using publicly available data to rank 32 senior CEOs in the Myers-Brigg Type Indicator Framework (MBTI).
You probably know MBTI even if you didn’t know his name. Based on the work of Carl Jung, MBTI classifies personalities into 4-character units such as ENTJ (‘extrovert’, ‘feel’, ‘think’, ‘judge’) and has become a popular self-analysis tool on social media over the past decade.
The four axes from which characteristics can be selected are: extraversion vs. introversion (EGG); sense vs. intuition (SN); Think vs. feeling (TF); and judge vs. perceive (JP).
The paper’s authors used crowdsourced data to create MBTI profiles of CEOs, and then created a separate dataset and architecture that was able to successfully predict the impact of CEO personality on stock performance, using Elon Musk as the central subject was used.
The new research aims to create an objective method for predicting stock prices based on the estimated personalities of chief executive officers (and specifically what they say in public) in accordance with the Upper Echelons Theory proposed in 1984, which was first discussed was a correlation between CEO personality and company performance.
The authors explain:
“In a risk regression task, we demonstrate that—in agreement with upper echelons theory—the predicted CEO personality is significantly associated with financial risk in the form of stock return volatility. Qualitatively, extroverted, intuitive and thinking CEOs seem to take fewer financial risks.’
The second phase of the project, which aims to correlate utterances with subsequent stock price movements, produces an algorithm that can display segments of text that are likely to have a positive or negative effect.
The paper is titled Influence from top to bottom? Predicting CEO personality and risk impact from voice transcripts and comes from three researchers from the University of Mannheim and the Bocconi University in Milan.
The authors used three sources to collect data for the personality prediction component of the project. To identify and characterize the CEOs, they used the transcripts of 88,000 earnings talks from Reuters’ Refinitiv EIKON dataset; and Compustat Execucomp from Wharton Research Data Services at the University of Pennsylvania, which was used to programmatically match identified CEOs to age and gender data.
To obtain MBTI personality labels, the researchers used the Personality Database, a crowdsourcing platform for user-contributed personality assessments, where reasonable amounts of data were available for 32 CEOs (including Steve Jobs and Elon Musk).
Rather than defining each CEO as a typical MBTI 4-character abbreviation, the authors presented each personality profile as a vector of four continuous variables on a 0 to 1 scale, consistent with the crowdsourced estimates, providing a more accurate mapping between the big ones enabled 5 and MBTI scales.
To validate the crowdsourced estimates, a correlation matrix was constructed between the crowd-based MBTI and Big 5 votes for all eligible individuals identified in the data. It was found that the two systems reached a consensus on the ratings contributed by the users.
(Because opinions differ as to how subjective the psychological assessment is, this systemic consensus means only that the two opposing scales converge across their different criteria for assessing personality; the objective value of this agreement depends on your level of engagement with both systems .)
The authors tested BERT (base), RoBERTa (base), and Support Vector Machine (SVM) with the trigram TF–IDF as potential models. After the training, it was shown that RoBERTa performed best across all axes. In the image below we can see the different personality assessments of the models for Elon Musk:
The authors note that the best indicators occur for extraversion-introversion and the worst for judgment-perception, perhaps because the latter are difficult to infer from the text. They predict that future work involving indicators of language – such as voice modulation and speech intervals – could add new interpretable dimensions to the data for these indicators.
Linking personality and company performance
To facilitate the risk regression component of the project according to upper echelons theory, the authors then merged the income call data with databases from IBES, CRSP and Compustat Execucomp.
To develop useful indices of stock price fluctuations, according to a CEO statement, they included various risk proxies from previous work and also included age and gender to assess possible confounding effects.
Risk returns were calculated based on volatility indices after the week following an announcement. Since RoBERTa performed best in the earlier module, this was used exclusively for the risk regression stage.
The authors comment on the results obtained by using a personality matrix as a method for predicting stock volatility:
“We find that the first three MBTI dimensions are significantly associated with post-call risk. This meaning is [high] for E–I and T–F. The direction of this association is as expected: a CEO who is introverted and communicative is associated with exalted [risk], while an intuitive communication is associated with diminished [risk].
“Remarkably, these results are robust to age- and gender-specific effects.”
In the future, the researchers plan to further develop the model so that a single regressor can output all four MBTI dimensions and also include non-text data such as speech signals.
*Myers-Briggs’ spelling of this word is specific to her study.
First published on January 20, 2022.