How PanAgora Is Quantifying ESG Investing
Using machine learning techniques like natural language processing, the quant manager says it is possible to measure ESG impacts even when data is hard to come by.
A common headache in environmental, social, and governance investing is that the data for certain metrics are hard to come by.
For example, while it’s relatively easy to come up with predictions of future returns based on information contained in financial statements, it’s more challenging to measure the value of a brand or the level of employees’ happiness. When it comes to environmental considerations, few companies release their carbon emission or water usage results.
But with advanced statistical and machine learning techniques, it is possible to quantify a company’s ESG impact, according to the Boston-based asset manager PanAgora.
Mike Chen, the firm’s head of sustainable investments, said he has been trying to link social factors to companies’ performance for a decade now. When he left Morgan Stanley to work at Google in 2011, he realized employees could work more efficiently — and thus generate more returns — if they were treated well and spoke positively of the management team.
“I didn’t know it was called ESG,” Chen said. “I just thought it was sensible.”
Working for a bank during the financial crisis was stressful, Chen said. When he saw that tech employees were treated with free food and volleyball courts, he realized that the value creators would only “buy into the company’s mission” if they felt positively about their working environments.
In an investment model he later designed, Chen utilized employee reviews from Glassdoor, a website where current and former employees write anonymous critiques about their companies. The model quantified the positivity in each sentence to create a portfolio of companies with good employee benefits.
“We didn’t know if this had any investment implication at the time,” Chen said. “But it seemed that all else equal, I’d rather invest in a company where employees are happy instead of disgruntled.”
The investment implication has become clearer as ESG investors look to add more quantifiable parameters into their strategies — something that has become difficult as the largest companies increasingly draw value from intangible assets like patents, brands, and innovations. Last year, about 90 percent of S&P 500 companies’ market value was made up of intangible assets, up from 17 percent in 1975, according to data from the intellectual property merchant bank Ocean Tomo.
Many of the intangible assets are only disclosed in the narrative sections of the financial reports, Chen and his colleagues pointed out in an April study. With natural language processing techniques, they were able to analyze the texts and create a model that measures corporate alignment with sustainable development goals.
“There’s a lot of innovation” in ESG investing, Chen said. “It is both interesting and challenging. When more people put in the resources, the complications and nuances just begin to multiply.”