My research makes extensive use of Natural Language Processing (NLP) techniques in both supervised and unsupervised machine learning settings in addition to industry standard statistical methodologies.
My primary supervisor is Professor Richard Taffler. I’m also actively advised by Professor Joanne Horton.
Trust Portrayal & Market Pricing
This paper studies the impact of trust portrayal on market pricing.
Core Hypothesis
Fundamentally, if trust matters to investors, then firms will make active efforts to portray themselves as trustworthy.
In the absence of private information however, there is no way for investors to know ex ante if firms that portray themselves as trustworthy are genuinely so.
Thus, investors likely disregard the portrayal of trustworthiness in its entirety precisely because anyone can claim to be trustworthy, meaning such a portrayal can be seen as mere ‘cheap talk’.
The confirmation and validation of the claim is a counterfactual until and unless there has been an explicitly breach of trust from a firm which claimed to be trustworthy.
And if investors do in fact disregard the portrayal of trustworthiness, then the returns of firms portraying themselves as more trustworthy should not be statistically different from those which do not portray themselves as more trustworthy.
Measuring Trust Portrayal
I measure the trust portrayal by evaluating the choice of words used by Executives of firms in the Management Discussion & Analysis (MD&A) section of 10-K filings.
Specifically, I measure the proportion of “Trust Related Words” and the total number of (cleaned) words in each MD&A to obtain a Trust Portrayal Score (TPS) for each firm j at each time t.
Consistent with expectations, the TPS does not vary considerably over time, and is negatively correlated (albeit a very small negative correlation) with Uncertainty.