Generating Word Lists for Analyzing and Monitoring Social Good: The Case of Sustainability

Abstract

This paper examines issues associated with the development of bags of words that can be used to analyze the extent to which descriptors of a concept are related to some dependent variable and as an approach to continuously monitor the occurrence of those concepts in text. We focus on generating bags of words using Word2Vec, using two key sources of business text, Form 10-Ks and “earnings calls,” to support issues of concern in social good. As an experiment, we drill down on building bags of words, that describe independent variables, descriptive of the concept of “sustainability,” which could be related is-sues such as firm value measures (profitability), events (release of new products or mergers) or other dependent variables.