AI’s and algorithms still can’t beat humans in a fundamental analysis: Alex Frino
With waves of cognitive computing developments set to disrupt multiple industries, participants in capital markets are already expressing mixed responses to advancements in artificial intelligence (A) and related technologies. This is a development that capital markets research Dr Alex Frino believes that companies in Asia and beyond will inevitably have to grapple with.
In an email exchange between Venture Views and capital markets researcher Dr Frino, we discuss the role of cognitive technologies in the capital markets; the impact of automated trading systems in shaping market liquidity and the potential of the SGX-Nasdaq co-listing partnership, which creates an Asia-North America bridge for businesses.
Interacting with Reuters last year, Chew Sutat, head of equities and fixed income at the SGX, had commented: “This would be a very good East-West bridge for companies at different stages of growth to accelerate going public by choosing Singapore first or if they want to have a dual-class regime, go to the U.S. but still have a secondary listing in Singapore concurrently.”
In May 2018 the Singapore Exchange (SGX) forged an identical agreement with the Tel Aviv Stock Exchange (TASE). This specific partnership aims to promote both technology and healthcare listings.
While New York continues to be an appealing destination for Asian technology enterprises — its size, depth and community of fund managers familiar with investments in pre-profit tech stocks underwrite this — its highly automated stock market poses challenges.
Weighing in on developments in US equities markets and the dominance of automated trading systems, in their stock markets — JP Morgans’ estimated 90% of trades in US equities in 2017 were via automated systems —Frino does not believe this will be the case for the Indo-Asia Pacific (Indo-APAC) equities space.
He explains: “Robo-trading does indeed dominate US equities markets. The fundamental reason why this happens in the US and not across the Asia-Pacific markets is that the US market is much more fragmented. For example, in the US a stock like Microsoft can be traded across 70 different venues, including stock exchanges and the so-called “dark pools” run by stockbrokers, with robo-trading trying to exploit arbitrage opportunities between exchanges and the other venues.”
These developments are mirrored in a 2015 research note, ”Will high-frequency trading practices transform the financial markets in the Asia Pacific Region?” by Robert J. Kauffman, a professor of information systems at the Singapore Management University (SMU).
Frino observes: “In the Asia-Pacific region we simply don’t have as many access points to the markets. Singapore, Hong Kong and Australian stock exchanges don’t have any meaningful exchanges competing against them, so robo-trading opportunities are much smaller and therefore less of a feature of the market than in the US.”
Cognitive technologies & capital markets
With regards to how cognitive technologies are reshaping capital markets, as well as its impact on high-frequency trading — which has faced a slowdown in 2018 — Frino argues that artificial intelligence (AI) still lacks the capacity to outperform human assessments of a company’ s fundamentals.
Asked if he saw greater automation and use of cognitive agents driving more long-term value creation or an increase in short-termism — particularly with the risk of a near-term recession or economic slowdown in the US — Frino elaborates: “The artificial intelligence (AI) of cognitive technologies as it applied to capital markets is focussed almost exclusively on predicting (and reacting to) very short-term price movements, and unrelated to the fundamentals behind the stocks.”
“Most fundamental “clues” about a stock come from unstructured data, such as statements to the market from the CEO. I have not seen any evidence that cognitive technologies are able to out-perform humans in interpreting this kind of information.”
“For this reason, I think that in the short to medium term, let’s say over the next five to 10 years, artificial intelligence will chiefly be deployed in predicting and exploiting short-term price movements.”
And when factoring in the capacity for investor news consumption behaviour to influence stock prices amid greater automation of the finance sector, Frino predicts more information asymmetries emerging, with the knock-on effect of less efficient pricing.
He explains: “The more you fragment markets through introducing alternative ETFs and alternative ways for releasing information about equities, the more likely you are to create information asymmetries — with some parties having to access to more information than others. This, in turn, leads to more volatility in price movements, and less not more efficient pricing.”
This aspect of capital markets is particularly pertinent in Asia, given the growth of ETFs in Asia. While popular for their low cost and transparency, they are also limited by information gaps in developing countries.
Spence Johnson, a research firm that specialises in asset management, reported that ETFs held US$314 billion worth of Asian assets at the end-2016, an 11% from $283.9 billion at end-2015.
At the beginning of 2018, ETF industry consultant ETFGI reported that assets invested in ETFs and ETPs (exchange-traded products) listed in the Asia Pacific ex-Japan increased by US$40.2 billion (31.4%) during 2017, reaching US$170.3 billion. This growth is more than triple the $12.5 billion it reported in 2016 and is forecast to reach US$170.3 billion at the end of December 2018.
Algo-Traders, liquidity & Singaporean developments
As the investment public of jurisdictions like Singapore, Korea, Hong Kong and Japan age, creating various financial pressures, this could also impact the liquidity of capital markets.
A report by the Financial Stability Board released in November 2017, “Artificial intelligence and machine learning in financial services Market developments and financial stability implications”, indicates many institutions are already using AI and machine learning (ML) techniques to optimise allocation of capital, coupled with the use of back-testing models to analyse the market impact of trading large positions.
Hedge funds, broker-dealers, and other institutional investors are also employing AI and are using AI and ML to discern signal for higher (and uncorrelated) returns and optimise trading execution, with organisations in the public and private sectors employing these technologies for regulatory compliance, surveillance, data quality assessment, and fraud detection.
However, these applications can also create new learning — and unexpected — forms of interconnectedness between financial markets and institutions, such as through previously unrelated data sources. Risks also arise from the “lack of interpretability or “auditability” of AI and machine learning methods”, which could become a macro-level risk, as well as the widespread use of opaque models.
Asked if the adoption of automated trading systems could possibly mitigate this develop and assisting in maintaining liquidity and turnover velocity in their stock markets, Frino elaborates: “Whilst robo-traders do increase traded volumes, this should not be confused with increased liquidity. Liquidity means being able to buy and sell large volumes of shares without moving prices, and there is strong evidence that robo-trading actually impairs liquidity.”
“The best evidence demonstrates that when a large trading institution enters the market to buy, robo-traders attempt to predict the impact on price by buying ahead of the institution then selling back to the institution at a high price. Research suggests that this type of parasitic trading has increased as a result of robo-trading.”
Meanwhile, despite the prediction that the city-state is set to overtake London as an offshore financial sector in 2020 amid its struggle to retain its status as premier financial hub post-Brexit, the city-state’s securities market is seemingly unable to leverage these large pools of private capital held under management there.
Speaking on this, Frino notes that while the while the Singapore bourse is perceived as “an international broker of capital with international clients who want international opportunities”, local equities markets possess “risk profiles, industry exposure and geographical exposure” that are less appealing to international investors.
However, in a 2015 report, “The efficiency in pricing of initial public offerings: A comparison of Australian and US markets”, and a subsequent 2016 report, “The efficiency in pricing of initial public offerings: A comparison of SG and US markets”, Frino also noted the greater efficiency of the Australian Securities Exchange (ASX) and SGX — despite their significantly smaller securities markets — in supporting the capital raising needs of small-cap and mid-cap issuers compared to the US equities market.
Reflecting on the Nasdaq-SGX co-listing agreement inked in October 2017 and permits companies to explore serial or simultaneous public floats on both the SGX and NASDAQ, Frino elaborates: “It makes sense for the SGX to develop partnerships with US mid-caps that do not make the SMP500.”
“Our research demonstrates that mega-cap companies are better off listing in US markets where their visibility in the S&P 500 Index guarantees them access to enormous pools of liquidity.”
“Stocks require a threshold of $US3 billion to make the S&P 500, and stocks outside the S&P 500 are virtually ignored in the US marketplace. Yet these stocks can still be relatively very large stocks for Asian markets. They would make local (Asian) markets and therefore would get considerable attention and liquidity from those markets.”