“By using artificial intelligence (AI) and machine learning, new data providers and analyses are popping up to address some of these concerns,” he adds, although acknowledging that “they are a long way away from becoming mainstream”.
ESG is becoming increasingly important in the portfolio investment process, with asset owners rather than asset managers in the vanguard for its promotion, Gerard says.
Many investment managers now routinely advertise the inclusion of ESG analysis into their investment processes, across a widely-recognised range of five styles, namely: exclusionary screening, positive screening, ESG integration, impact investing and active stewardship.
Some managers are also making efforts to disaggregate the specific contribution of ESG factors to a portfolio’s total return.
AI and machine learning techniques should eventually help resolve the divergence between conventional ESG rating agencies.
However, it remains difficult to evaluate the environmental and social impact of companies’ operations, and the quality of governance metrics. Basically, are the tools – that is, the ESG rating agencies – sufficient?
“Most especially, there are ‘intangible risks’ that can’t always be anticipated and so can’t been measured,” Gerard says.
Currently, agencies such as MSCI and Sustainalytics, assign ESG ratings to firms.
However, academics have shown that these ratings diverge substantially between different rating agencies, according to Gerard.
“There is extraordinary discrepancy in the ratings of firms from one company to the next, making the evaluation of social and environmental impact very difficult,” he says.
The consequences are that ESG performance is unlikely to be properly reflected in stock and bond prices, and that empirical research to improve a fund’s ESG performance becomes tougher because the selection of a particular ESG rating agency may skew the research’s results.
These discrepancies can be assigned to differences in scope (whether they are looking at the same things) and aggregation (their respective weightings to each factor) and, most significantly measurement.
“We need a better way to measure the quality of the data to reach accurate assessments,” Gerard says.
In practice, that means analysing data more recently, more frequently, more systematically, and being able to tap into a larger data base.
Artificial intelligence and machine learning can help improve the processes on all these issues, according to Gerard.
State Street said it has developed an ESG portfolio management tool that incorporates the ESG scores assigned by conventional analytic firms such as MSCI with the scores generated by AI vendors, such as Truvalue Labs.
But Gerard acknowledges that that AI and machine learning applications are at a nascent stage. They rely on “web scraping” and natural language processing to access and interpret vast quantities of data, but there can be difficulties distinguishing between valuable and poor sources and understanding the nuances and emphases of prose and its variety of formats.
“Establishing polarity, that is, determining a positive or negative assessment from a particular piece of content, can be hard,” he says.
Nevertheless, AI and machine learning techniques should eventually help resolve the divergence between conventional ESG rating agencies.
“Investors want a platform that can incorporate many perspectives, both traditional analyst-driven views and the emerging technology-driven views of the future. They want to cross-check biases, decrease information disadvantages, and expand their dialogue with their clients and partners,” he says.
For more insight on asset and wealth management in Asia, please click on www.fundselectorasia.com