By Shirley Birman | January 12, 2022
Are environmental, social, and governance (ESG) factors additive to performance? How well do they really work as stock selection factors? Although there is disagreement on this topic, researchers generally agree that ESG risks ultimately transition into financial risks. Much as a valuation or quality factor measures current financial performance, a material ESG factor measures firm characteristics that will ultimately be reflected in future financial performance. Since ESG is increasingly considered one of the best proxies for the quality of management and other intangible risk factors, it is likely to have an impact on financial factors of a firm.
How Can We Define an ESG Factor?
Having concluded that ESG is an important factor, the question that looms next is how best to define such a factor. Transparency, timeliness, and subjectivity are key challenges inherent in traditional ESG data, and as a result, these models fail to capture material and timely company ESG performance. In this digital information age, investors seek reliable, high-quality data that goes beyond company disclosures and captures information that is embedded in massive volumes of unstructured data. Today’s superabundance of information calls for new ESG analytical models built around the application of artificial intelligence to gain deeper insights into sustainable investing.
Some data providers leverage natural language processing (NLP) and machine learning (ML) to source environmental (E), social (S), and governance (G) signals from unstructured data. This technology can enable documentation and analysis of real-time events, both positive and negative, across multiple ESG dimensions.
One provider that uses artificial intelligence to deliver unbiased ESG insights is Truvalue Labs, a FactSet company. The solution delivers objective algorithmic scoring on ESG factors as identified by the Sustainability Accounting Standards Board (SASB) as having a material impact on company value by industry and sector. Truvalue Labs covers approximately 30,000 companies globally and captures over 100,000 external sources such as local and international news, industry publications, and non-governmental organizations (NGOs) to report evidence of ESG performance across a range of financially material factors.
Creating an ESG Activity Signal
We can create an ESG Activity Signal by incorporating ESG data volume (article volume in trailing 12 months) from Truvalue Labs, along with a measure of the company’s longer-term ESG track record (positive or negative polarity). More specifically, ESG data volume is a strong signal measuring investor and trader attention to a company. Overlaying this investor attention signal with the long-term Truvalue Insight score adds ESG polarity, driven by the degree to which sentiment on ESG issues at a company is positive or negative. Research has confirmed that a lot of positive news leads to outperformance. On the short side, it is the companies with the slow-leaking bad news (i.e., in rumor state) that tend to underperform.
The ESG Activity Signal takes a straightforward approach to combine ranks of ESG volume and sentiment scores. Specifically, the sentiment score uses the Adjusted Insight Score, which is a measure of a company’s long-term ESG track record. Volume is determined by the number of ESG-relevant articles in the past year, relative to expectations. This is measured by normalizing the trailing 12-month article volume by a measure of expectations or size. We chose to use dollar trading volume aggregated over the trailing 12 months as a proxy for attention, as studies have found a positive relationship between news coverage, trading volume, and market liquidity. Alternatively, ESG data volume can be normalized by proxies of company size such as market cap, revenue, and book value.
The measure of information flow (article volume scaled by dollar trading volume) is then ranked across the universe. The ESG Activity Signal is created by summing the ranks for Adjusted Insight and Volume and re-ranking the resulting sum.
Testing the Signal’s Performance
We used FactSet’s Alpha Testing platform to evaluate the performance of the ESG Activity Signal between 2008 and October 2021. Portfolios were constructed on a monthly basis, with no transaction costs. The chart below shows the annualized active returns for Insight, Activity (Volume), and the ESG Activity Signal by quintile for a universe of Developed Europe companies (IEUR ETF). The additive approach of combining Insight and Activity (volume) produces an excellent combination ranking factor (ESG Activity Signal). Specifically, the top quintile of the ESG Activity Signal outperforms the bottom quintile by 6.1% annualized on this universe.
Insight and Activity (volume) complement each other on the positive side, with companies having both strong data volume relative to expectations and high Insight scores demonstrating outperformance. On the short side, however, the combined signal captures those companies with slowly unfolding bad news as these are the companies that are likely to continue to drift downward.
The figures below show the performance of the top and bottom global quintiles vs. an equal-weighted blend of the ETF universe. The results indicate that for both a universe of U.S. (IWV ETF) and Developed Europe (IEUR ETF) equities, the top quintile outperforms relative to the bottom quintile and benchmark.
As the results demonstrate, the ESG Activity Signal has strong potential for alpha generation by bringing together information on ESG sentiment and investor attention. Incorporating the ESG Activity Signal into an investment strategy identifies material risks and opportunities that are likely to impact investment performance. This signal can be used for a multitude of purposes, including long-short quant strategies, in combination with other quantitative factors in multifactor models, and screening for high and low ESG performance companies.