Since the emergence of “Modern Portfolio Theory” and the “Capital Asset Pricing Model” in the late 1960s, institutional investors have taken a quantitatively driven approach to portfolio construction, looking to create portfolio diversification and obtain better risk-adjusted returns by balancing their asset-class exposures. This journey has seen several important advancements in thinking about how to optimally achieve desired results.
Funds that want to take the total portfolio approach first need to get the total portfolio view. To do that they not only need data – and lots of it – but a rock-solid understanding of exactly how they’re going to use it.
A case study in AI for funds management
The term ‘big data’ has not resonated with investors in the way that the term ‘artificial intelligence’ has. Marrying the two helps. Now applying them to the ESG trend helps a lot more.
Andy Moniz, a quantitative fund manager with a pedigree in data science, says he never much liked the term ‘big data’. The search for more and more data and, more importantly, better ways to analyse and use it, is something which has come naturally to quants for years, tens of years.
Moniz is the director of responsible investing in the London arm of Acadian Asset Management. His roles before joining Acadian in March this year included director of applied data science investments at Putnam Investments and a managing director and chief data scientist at Deutsche Bank.
His PhD from Erasmus University Rotterdam, completed in 2015, is in ‘Information Retrieval and Natural Language Processing’. But before that he was a quant equities manager at Citi and Macquarie Group in London.
Some of his recent work at Acadian involves the application of artificial intelligence (AI) to the process of analysing stocks for their genuine commitments to ESG principles, to enable better engagement with company management and identify any green washing.
Acadian has had a long involvement with Australia, having on the ground representation dating back to the 1990s. With the firm’s use of AI to assist its assessment on ESG matters, for instance, it has done some work on the ASX 300 stocks examining their commitment to abide by modern slavery legislation.
Moniz says this work has resonated in particular with Australian institutional clients, with whom he and others have been speaking. Australia has two pieces of legislation, one for NSW and the other Federal, with respect to disclosures. They are among the most demanding in the world, with NSW setting the pace.
Looking beyond the regulatory statement on modern slavery required of the ASX 300 stocks, Acadian’s AI information-gathering process also includes analyses of corporate attitudes to things such as training on slavery issues, which is arguably a better indicator of commitment.
“This has gone down very well with our Australian clients,” Moniz says. “We are also able to look at company supply chains and possible issues. We can identify and map out company supply chains.”
He says: “We’re not experts in human rights but we can let the data speak for itself. We can do network analyses of connections, including what the media are talking about, for example, to do with human rights in China and connections in renewable energy, such as production of solar panels, or the production of cotton goods. AI is also being used in surveillance.”
A report published by JANA Investment Advisers last month (July 22) followed a study of 20 super funds which had reported on their responses to the Commonwealth Modern Slavery Act. It showed that almost all had not looked past ‘tier one’ levels in their investee companies’ supply chains and no examples of slavery had been discovered. The researchers said most of the bad conditions occur in levels three and four.
Nevertheless, JANA’s report says: “From an asset owner perspective, we expect to see superannuation funds carry out deeper and wider analysis within and across asset classes over time.
“Deeper analysis will involve looking further down the supply chain of investee companies and assets to try and identify any instances of modern slavery. Wider analysis will involve understanding the risks of modern slavery across all asset classes rather than focusing on a select few.
“In addition, as the assessment of modern slavery risk develops over time, we expect to see reporting entities actually identify and report on instances of modern slavery within their investment supply chain. As the depth of analysis increases by looking further down the supply chain, so too should the rate of modern slavery identification across all reporting entities…”
With respect to ESG, Acadian’s analyses are used for engagement purposes to filter information and draw tentative conclusions before the analysts speak with the companies. Moniz admits that “even state-of-the-art AI algorithms have an accuracy rate of about 90 per cent”.
The manager also uses variations on this theme as part of its broader alpha model, looking at what companies say in general and applying this to their accounting. How vague the corporate language is can be a hint about how committed it is to a certain goal.
ESG signals are drawn from text mining of sustainability reports, regulatory filings, shareholder proposals and other company data. Signals range from the identification of physical and transition climate risks, discussions of employee well-being and corporate culture during covid-19 as well as the human rights and supply chain concerns.
Acadian’s system cross-checks what companies say in their annual shareholder meetings, filings and earnings calls, versus their actions, as tracked through corporate policies and sustainability reports.
“We’re merging these datasets together to identify companies that talk a lot about sustainability but perhaps don’t actually do much. The premise being that a company’s actions speak louder than its words,” Moniz says.
“In the case of earnings calls, we identify evasive and potentially deceptive talk,” he said. “Our text mining algorithms assess to what extent managers are directly answering sell-side analysts’ sustainability questions or giving a boilerplate response or an indirect answer.”
The greenwashing model rewards answers which include quantitative metrics and targets, forward-looking views or those focused on opportunities. It penalises vague and backward-looking statements and those that emphasised risks.
Moniz says the firm is aware of the risks associated with attempts to “game” the system. “We don’t search for buzz words; we’re looking for a consistency of language, a more holistic picture of the company,” he says.
Andrew Hair, Acadian’s chief executive for Australia and regional head for APAC, said the firm’s data analysis approach had been of strong interest to clients.
“In some common areas of engagement focus, such as carbon-related risks, company disclosures are relatively mature and understood,” he said. “However, in others, such as human rights issues – for example, modern slavery – there isn’t the same level of reporting available.”
