Analysts' daily lives are busy from dawn to dusk, as they scrutinize company financial statements, listen to management calls, visit facilities, and analyze competitors. And the rivers of information flowing in, much of it unstructured, are flowing faster than ever. This means that more reads are required, potentially reducing the time it takes to synthesize data.
Conversations with management teams are a key source of information for fundamental analysts as they delve into companies and their operations. Financial reporting is particularly important. It is a direct communication channel to company leaders that provides insight into a company's financial health, strategic direction, and market trends.
Financial reporting records can be a rich source of insight, from forward-looking statements to key performance indicators, operational updates to market sentiment. Subtle cues are also important. Tone, language, and level of detail help analysts assess management's confidence and potential future performance. And our research suggests that incorporating these factors can lead to more informed investment decisions.
Sentiment: The “atmosphere” of the message is important
Sentiment indicators are intuitive and powerful indicators of the tone of a document: whether it is relatively positive, negative, or neutral.
Sentiment assessed from records of earnings conferences may influence future stock prices. Our analysis suggests that stocks of companies with high sentiment communication scores tend to outperform after a call. However, given the large volume of earnings reports, manually assessing sentiment can be difficult.
In these cases, we believe that natural language processing (NLP) sentiment analysis can be a useful tool for analysts.
“Bag of Words” and “Context-Aware”: NLP comes to the rescue
NLP, a branch of artificial intelligence (AI), can be a powerful resource to help analysts extract insights from mountains of documents. Instead of hiring a team of interns to hunt down clues, you can use NLP to get the job done faster and more efficiently.
We believe it makes sense to capture emotions using two NLP approaches: “bag of words” and “context awareness.” Bag of Words provides a baseline for evaluating text by counting the number of positive and negative words in a document. It's very intuitive, but simplified and easy to work with.
Context-aware approaches, leveraging models such as BERT, GPT, and LLaMA, analyze the composition of sentences and their contexts within an entire document. These models measure aspects such as sentiment by analyzing overall language patterns, making the analysis more accurate and contextually relevant than simply counting positive or negative words. I'll make it.
Applying both bag-of-words and context-aware approaches to the earnings records of large U.S. companies (screen) Analyzing data from 2010 to 2023, we find that sentiment dropped significantly in mid-2020. The pessimism was related to the impact of the coronavirus pandemic and its global economic aftershocks on company finances.
These economic downturns aside, both indicators show that sentiment has improved since 2010, even more so when using the bag-of-words approach. Although we have not yet analyzed this trend in detail, other research suggests that company executives are adapting to the age of AI by communicating with more positive words and phrases. . And the sharp increase in average bag-of-words sentiment suggests that companies are incorporating positive words more frequently.
