HONG KONG,CHINA - Media OutReach - 27 August 2020 - It's been called the holy grail of finance. Is it possibleto harness the promise of artificial intelligence to make money trading stocks?Many have tried with varying degrees of success. For example, BlackRock, theworld's largest money manager, has said its Artificial Intelligence (AI)algorithms have consistently beatenportfolios managed by human stock pickers. However, a recent research study byThe Chinese University of Hong Kong (CUHK) reveals that the effectiveness ofmachine learning methods may require a second look.
The study, titled "Machine Learning versus Economic Restrictions: Evidence fromStock Return Predictability",analysed a large sample of U.S. stocks between 1987 and 2017. Using threewell-established deep-learning methods, researchers were able to generate amonthly value-weighted risk-adjusted return of as much as 0.75 percent to 1.87percent, reflecting the success of machine learning in generating a superiorpayoff. However, the researchers found that this performance would attenuate ifthe machine learning algorithms were limited to working with stocks that wererelatively easy and cheap to trade.
"We find that the returnpredictability of deep learning methods weakens considerably in the presence ofstandard economic restrictions in empirical finance, such as excludingmicrocaps or distressed firms," says SiCheng, Assistant Professor at CUHKBusiness School's Department of Finance and one of the study's authors.
Prof. Cheng, along with hercollaborators Prof. Doron Avramov at IDC Herzliya and Lior Metzker, a researchstudent at Hebrew University of Jerusalem, found the portfolio payoff declinedby 62 percent when excluding microcaps -- stocks which can be difficult to tradebecause of their small market capitalisations, 68 percent lower when excludingnon-rated firms -- stocks which do not receive Standard & Poor's long-termissuer credit rating, and 80 percent lower excluding distressed firms aroundcredit rating downgrades.
According to the study, machinelearning-based trading strategies are more profitable during periods whenarbitrage becomes more difficult, such as when there is high investorsentiment, high market volatility, and low market liquidity.
One caveat of the machine-learningbased strategies highlighted by the study is high transaction costs. "Machinelearning methods require high turnover and taking extreme stock positions. Anaverage investor would struggle to achieve meaningful alpha after takingtransaction costs into account," she says, adding, however, that thisfinding did not imply that machine learning-based strategies are unprofitablefor all traders.
"Instead, we show that machinelearning methods studied here would struggle to achieve statistically andeconomically meaningful risk-adjusted performance in the presence of reasonabletransaction costs. Investors thus should adjust their expectations of thepotential net-of-fee performance," says Prof. Cheng.
The Future of Machine Learning
"However,our findings should not be taken as evidence against applying machine learningtechniques in quantitative investing," Prof. Cheng explains. "On the contrary, machine learning-based tradingstrategies hold considerable promise for asset management." For instance,they have the capability to process and combine multiple weak stock tradingsignals into meaningful information that could form the basis for a coherenttrading strategy.
Machine learning-based strategiesdisplay less downside risk and continue to generate positive payoff duringcrisis periods. The study found that during several major market downturns,such as the 1987 market crash, the Russian default, the burst of the techbubble, and the recent financial crisis, the best machine-learning investmentmethod generated a monthly value-weighted return of 3.56 percent, excludingmicrocaps, while the market return came in at a negative 6.91 percent duringthe same period.
Prof. Cheng says that theprofitability of trading strategies based on identifying individual stockmarket anomalies -- stocks whose behaviour run counter to conventional capitalmarket pricing theory predictions -- is primarily driven by short positions andis disappearing in recent years. However, machine-learning based strategies aremore profitable in long positions and remain viable in the post-2001 period.
"This could be particularlyvaluable for real-time trading, risk management, and long-only institutions. Inaddition, machine learning methods are more likely to specialise in stockpicking than industry rotation," Prof. Cheng adds, referring to strategywhich seeks to capitalise on the next stage of economic cycles by moving fundsfrom one industry to the next.
The study is the first to providelarge-scale evidence on the economic importance of machine learning methods,she adds.
"The collective evidence showsthat most machine learning techniques face the usual challenge ofcross-sectional return predictability, and the anomalous return patterns areconcentrated in difficult-to-arbitrage stocks and during episodes of highlimits to arbitrage," Prof. Cheng says. "Therefore, even thoughmachine learning offers unprecedented opportunities to shape our understandingof asset pricing formulations, it is important to consider the common economicrestrictions in assessing the success of newly developed methods, and confirmthe external validity of machine learning models before applying them todifferent settings."
Avramov, Doron and Cheng, Si andMetzker, Lior, Machine Learning versus Economic Restrictions: Evidence fromStock Return Predictability (April 5, 2020). Available at SSRN: https://ssrn.com/abstract=3450322 or https://dx.doi.org/10.2139/ssrn.3450322
This article was first published in the ChinaBusiness Knowledge (CBK) website by CUHK Business School: https://bit.ly/3fX2ydr.
About CUHK Business School
CUHKBusiness School comprises two schools -- Accountancy and Hotel andTourism Management -- and four departments -- Decision Sciences andManagerial Economics, Finance, Management andMarketing. Established in Hong Kong in 1963, it is the first business school tooffer BBA, MBA and Executive MBA programmes in the region. Today, the Schooloffers 11 undergraduate programmes and 20graduate programmes including MBA, EMBA,Master, MSc, MPhil and Ph.D.
Inthe Financial Times Global MBA Ranking 2020,CUHK MBA is ranked 50th. In FT's 2019EMBA ranking, CUHK EMBA is ranked 24th in the world. CUHK BusinessSchool has the largest number of business alumni (37,000+)among universities/business schools in Hong Kong -- many of whom arekey business leaders. The School currently has about 4,800undergraduate and postgraduate students and Professor LinZhou is the Dean of CUHK Business School.
More informationis available at https://www.bschool.cuhk.edu.hk or by connecting with CUHK Business Schoolon: