Female Equity Analysts and Corporate Environmental and Social Performance
As a key capital market intermediary, sell-side equity analysts are known for their information discovery and production roles. Equity analysts also play an important monitoring role in scrutinizing management behavior. Yet none of the existing governance research has taken a gender lens to explore the role of female analysts in monitoring corporate environmental and social (E&S) performance.
Motivated by survey evidence indicating that women, compared to men, tend to place greater emphasis on the well-being of others, their communities, and the environment, in a paper titled “Female Equity Analysts and Corporate Environmental and Social Performance,” Management Science forthcoming, we examine whether female equity analysts are more likely to monitor a firm’s environmental and social (E&S) practices than their male counterparts and whether there are gender differences in their equity research approaches, thus shedding light on the origins of gender differences in skills within the equity analyst profession.
To investigate whether and how female analyst coverage influences corporate E&S performance, we first hand collect the gender information of equity analysts in the U.S. based on their online bio. We also make use of textual data that cover analyst research activities in the form of analyst research reports and analysts asking questions during earnings conference calls, and apply machine learning to those textual datasets to capture gender differences in their research approaches.
Our empirical investigation proceeds in three steps. First, we show that there is a positive and significant association between the number of female analysts covering a firm and that firm’s E&S performance. For identification, we exploit broker closures as a quasi-exogenous shock to female (male) analyst coverage and show that following such an event, firms losing female analysts experience significant declines in E&S ratings relative to firms losing male analysts, suggesting a causal impact.
Second, we apply machine learning tools to detect discussions of E&S topics in analyst reports and questions raised by analysts in earnings call transcripts. Because E&S-related discussions encompass a broad range of topics and linguistic expressions, conventional keyword-based textual analysis methods are inadequate. We develop a new active learning approach to efficiently search for and annotate E&S-related discussions in analyst reports (earnings call transcripts). We then fine-tune the FinBERT model, a large language model trained on financial text, to create two tailored E&S text classification models that capture analysts’ writing (in analyst reports) and raising questions (during earnings calls) about E&S issues.
We find substantial differences between female analysts’ and male analysts’ E&S discussions in both the intensity and thematic content. For example, we note that female analysts discuss E&S topics more frequently and emphasize sustainability-relevant themes such as regulatory compliance, stakeholder welfare, and the environment, whereas male analysts focus more narrowly on financial considerations such as operational efficiency and performance. We further show that in terms of cognitive and linguistic approaches to E&S issues, female analysts produce more readable analyses about E&S issues in their reports and employ more sophisticated cognitive processing in their E&S questions during calls. All these traits suggest that compared to male analysts, female analysts monitor broad E&S issues more closely and communicate E&S-related research more persuasively and clearly, which helps enhance the accessibility and impact of their E&S-related equity research.
Finally, we examine whether female analysts “walk the talk” by examining analysts’ actions following their E&S-related research and/or firms’ E&S incidents. We find that compared to male analysts, female analysts are more likely to lower their stock recommendations and target prices following negative E&S discussions in their reports. Similarly, female analysts are more likely to lower stock recommendations following firms’ E&S incidents. We further show that investors take note and they react more strongly to female analysts’ negative tones in discussing a firm’s E&S performance in their reports, which suggests that the market participants recognize the male-female skill differences in detecting E&S issues. Importantly, female equity analysts’ research on firms’ E&S issues is incorporated into stock prices via the price reaction to the release of their reports.
We conclude that female equity analysts play an important monitoring role in enhancing corporate E&S performance through their research activities – writing analyst reports and asking questions during earnings calls, and/or taking actions following firms’ E&S issues (or E&S incidents).
Our paper makes three contributions to the literature. First, our study contributes to the gender and finance literature. We establish that female equity analyst coverage causally improves a firm’s E&S performance. In doing so, we show that gender diversity among equity analysts serves as an impetus for firms to adopt more environmentally and socially responsible policies.
Second, our study contributes to the analyst literature, by taking a gender lens and identifying the specific mechanisms through which female analysts influence corporate E&S performance and by providing an explanation for the observed gender differences in analyst impact – female analysts are more effective at uncovering firms’ E&S issues than their male counterparts.
Finally, our study contributes to the finance and accounting literature that employs computational linguistic methods to analyze large, unstructured data sets, particularly in the context of corporate environmental exposure. By incorporating the principles of data centric-AI, which emphasize that high-quality training data set is just as critical as new modeling techniques, our study introduces a novel active learning approach that identifies domain-specific training examples from substantially larger and more diverse datasets than previously explored. Our approach, when combined with a pre-trained large language model such as FinBERT, proves to be an effective strategy in accurately classifying text, particularly in situations when there is limited training data due to specialized language and terminology in diverse contexts. We find that gender differences in analyst impact stem from female analysts’ greater propensity to monitor corporate E&S performance, their focus on more general E&S themes/topics, as well as their superior skills at persuasively and clearly communicating E&S issues compared to their male counterparts.
