In this guest blog, MarketPsych CEO Richard Peterson explores how natural language processing (NLP) can be used to generate ESG data.
- NLP can help quantify references to environmental and sustainable metrics.
- MarketPsych research found that the workplace sentiment of employees influences the company’s stock price.
- The research found that such media-derived data can also be used for risk management.
For more data-driven insights in your Inbox, subscribe to the Refinitiv Perspectives weekly newsletter.
My wife and I moved our family to Bali in 2018. Our children attended the Green School, and my ‘office’ was an open-air coworking loft.
In Balinese ravines, often unseen by tourists, there are mounds of garbage dumped by people unable to afford, or uninterested in, garbage pick-up.
During the rainy season the garbage washes into the ocean and onto the beaches. Single-use plastic water cups comprise a sizable portion of the waste.
I’ve often wondered, is Bali’s waste mismanagement due to ESG factors:
- A lack of environmental (E) respect?
- Inadequate social (S) education or responsibility?
- A failure of government (G) institutions or regulations?
- Something else?
Using NLP to plan for impact
This past week, we and Refinitiv harvested the fruits of this labour by launching the Refinitiv MarketPsych ESG Analytics.
Our company, MarketPsych, hosted our international team to Bali for meetings. Struck by the environmental catastrophe in our midst, we decided to repurpose our natural language processing (NLP) engines to quantify references to pollution and waste mismanagement.
Our NLP engines scan tens of millions of authors in thousands of news and social media sources in real-time. Below is a depiction of the three-stage process of extracting key themes from media text and aggregating into scores.
As a team, we developed NLP tools to start tracking 90 scores for companies (such as Innovation and ProductSentiment) and 150 scores for countries (such as GovernmentTrust and Deforestation). We now track these for 30,000 companies, and 252 countries and territories.
For example, our ‘Sustainable Innovation’ scores are plotted over time in the graphic below. Note that Costa Rica — in third place — may seem surprisingly high, but in fact the country obtains 98 percent of its energy from renewable sources and has set aside 30 percent of its land area in nature reserves.
Since our background is in financial markets, we performed quantitative research to find predictive power in this new form of data.
In our research, we found that the workplace sentiment of employees influences the company’s stock price.
In the below monthly rotational model, S&P 500 stocks were ranked by their average workplace sentiment scores — derived from the media — over the past month. Stock price performance of the top and bottom 5 percent of companies was tracked over the next month.
Investing based on the extremes of workplace sentiment
Month-after-month, companies with happier employees outperform their peers and companies with unhappy employees underperform.
Such data can also be used for risk management — to avoid companies performing poorly on ESG metrics.
In the research below, we examined the stock performance of the top 10 percent of S&P 500 companies with high activist investor activity. It is widely believed that activist investors improve management and unlock company value. The below result suggests the opposite is true.
My family moved on from Bali after a year, but the essential nature of sustainability remains deeply ingrained within our family and our company.
Refinitiv MarketPsych ESG Analytics analyse corporate sustainability-related news and social media in near real-time.
Powered by MarketPsych Data’s AI-based natural language processing engine, the analytics are derived from millions of daily articles in thousands of global news and social media outlets. They provide numerical ESG insights on companies and countries to drive better investment decisions.