Beyond the Numbers! The Role of Equity Analyst and Management Perspectives in Shaping Stock Returns in Developed and Emerging Markets: An Exploratory Study Using Semantic Text Analysis

Management communication through investor conferences have been studied in
various contexts like shareholder value, managerial opportunism, measurement of innovation quotient, measurement of marketing mindset, etc. In this project we propose to study the textual articulations of company management (investor call transcripts, referred to as the management perspective in this study) and equity analysts (equity research reports, referred to as the equity analyst perspective) in the context of company performance. Company performance will be examined through two lenses – value and growth. The study will examine value in terms of a valuation measure like market capitalization and examine growth in terms of a profit measure like profit before tax.
The study intends to answer the following questions.
a. What is the extent to which the management perspective explains the variability in company performance in terms of value and growth?
b. What is the extent to which the equity analyst perspective explains the variability in company performance in terms of value and growth?
c. What is the extent of overlap and divergence between management perspective and equity analyst perspectives in explaining the variability in company performance in terms of value and growth?
d. What is the nature (like, financial information vs strategic information, tangible vs intangible value, etc.) of overlap and divergence between management perspective and equity analyst perspectives in explaining the variability in company performance in terms of value and growth?
e. How do US-listed companies (Dow 30) and India-listed companies (Nifty 50)
compare with each other when examining the above questions?
Exhibit 1: Overview of the dimensions of research questions
From a methodology perspective – we propose to use Artificial Neural Network (ANN) based semantic-language models to analyze the text content and decipher perspectives of company management and equity analysts. One of the most popular and effective language models is BERT (Bidirectional Encoder Representations from Transformers) which has been found to be effective for several downstream Natural Language Processing (NLP) tasks. We will use language models based on BERT, like Sentence�BERT, FINBERT, etc. The unit of information will be a sentence, i.e., every sentence in every document (a call transcript or an equity analyst report) will be treated as the unit of information. These sentences will be converted to vector representations using a BERT-based language model, which will then be used in the analysis. We will finalize
the choice of a specific language model based on their suitability for the tasks relevant to this research and their techno-economic feasibility.