Benjamin David investigates whether innovative technology such as ChatGPT can play a role in fashioning market-beating exchange-traded funds.
Artificial intelligence (AI) has become increasingly popular in recent years, with applications ranging from self-driving cars to virtual assistants such as Apple’s Siri and Amazon’s Alexa. However, AI is not just limited to the consumer market. It has also made its way into the financial industry, and one of the most notable examples is the emergence of AI-powered exchange-traded funds (ETFs).
Like stocks, ETFs are investment funds traded on stock exchanges, providing investors with diversified portfolios of assets, usually more cheaply than traditional mutual funds. With the mounting prominence of AI, ETFs have started to leverage machine learning (ML), sentiment and natural language processing to fashion investment portfolios with superior performance and risk-management capabilities.
Leveraging the power of AI
AI-powered ETFs have been designed to use machine learning algorithms to identify patterns and trends in the market in order to make investment decisions. These algorithms are typically trained on vast amounts of historical financial data, allowing for data analysis more quickly and accurately than humans can manage. Consequently, from this vast amount of data, AI-powered ETFs can identify investment opportunities and manage risk more efficiently and effectively than traditional investment strategies. This growing information environment offers a “unique opportunity” to leverage big data to produce better investment outcomes, according to Chris Natividad, CIO and co-founder of EquBot.
However, with traditional investment strategies relying on a set of rules or criteria, such as sector, size or quality, to manage a portfolio, investors are “limited” from using more pieces of the global market picture, Natividad says. Crucially, AI-powered ETFs can adjust their investment strategies based on market data. For example, with an increase in market volatility, an AI-powered ETF might increase its allocation to assets expected to perform well in volatile markets. Equally, with the emergence of a new investment opportunity, an AI-powered ETF can quickly analyse the data and determine whether it is worth investing in. Here, an AI-powered ETF can utilise the power of natural language understanding (NLU), says Natividad, for example, to process “millions of news articles and social media each day and couple the metadata with market data to uncover a portfolio of stocks with the highest probabilities of market appreciation”.
“I selected 15 ETFs across different asset classes, regions and sectors, which could potentially provide diversification benefits and improve risk-adjusted returns over the long term.”
A related point is that a lack of rule flexibility to changing market conditions can increase risk. Here, AI-powered ETFs can leverage sophisticated risk management techniques such as portfolio optimisation and diversification to minimise risk. By creating diversified portfolios across multiple asset classes, they can reduce the impact of any asset on the overall portfolio performance. This can help protect investors from losses in any single asset class and provide a smoother and arguably more stable investment experience. It’s important to remember that markets can and do change, says Natividad, and leveraging the power of AI “makes sense” for investors expecting markets to be influenced by unstructured data (industry reports, social media, tweets, etc.,) in addition to traditional fundamental and technical data.
Despite these advantages, AI-powered ETFs are known to have their downsides. The main one is arguably the complexity of the algorithms used. These algorithms are often proprietary, and many might need help understanding them, making it challenging for investors to evaluate the performance of these funds. Additionally, there is a risk that the algorithms may be biased or flawed, possibly resulting in poor investment performance.
Another challenge often raised with regard to AI is the potential for over-reliance on it. While many might consider AI a powerful tool for investment management, many investors have argued that it is not a substitute for human judgement and expertise. Additionally, experts point out that investors should be aware of AI’s limitations and use it with human analysis and decision-making.
ChatGPT and AIEQ
Given the known advantages surrounding AI-powered ETFs, the possibility of asking ChatGPT to fashion a market-beating portfolio has piqued a lot of investor interest. Unfortunately, when asked, the AI chatbot developed by OpenAI and launched in November 2022 purportedly told users that the stock market lacks predictability. Nonetheless, ChatGPT can provide “general guidelines and considerations”.
For example, told “using between 10 and 15 ETFs, create a multi-asset portfolio that will outperform a European 60/40 (fixed income/equities) portfolio over a long-term horizon” the tool provided a sample portfolio of 15 ETFs that should be considered (see table).
When asked to explain its decision, ChatGPT said, “When selecting ETFs for a multi-asset portfolio, investors and wealth managers typically consider a range of factors, including diversification, risk management, cost efficiency, liquidity, and historical performance.
“I selected 15 ETFs across different asset classes, regions, and sectors, which could potentially provide diversification benefits and improve risk-adjusted returns over the long term.”
Beyond the recommendation-based capability of ChatGPT, the AI-Powered Equity ETF (AIEQ), issued by ETF Managers Group (ETFMG) and partnering with EquBot, claims to have created a market-beating portfolio. AIEQ, launched in 2017, uses the power of IBM’s Watson supercomputer to select its portfolio holdings. As a result, the AIEQ became one of the most popular funds in 2017 and raised more than $70 million within a few weeks. To date, it has just under $117 million of assets under management.
Fund flows in February 2023 were $23 million (€26 million) for the AIEQ, with a month-over-month engagement of 84% and a year-over-year engagement of 173%. The top-three industries within AIEQ’s portfolio are health technology (17%), energy minerals (11%) and finance (10%). The fund leverages machine learning, sentiment analysis and natural language processing to select the stocks in its portfolio. The portfolio, meanwhile, analyses companies via four quadrants: financial analysis, news, management and macro. So far, it’s performing well: more than 14% in the year to January 27, according to Morningstar. That’s about double the S&P 500’s year-to-date return of 7%.
AIEQ was launched as a proof of concept that AI can effectively manage a portfolio of US stocks, says Natividad, who points out that in the year AIEQ was launched, it was one of the “fastest-growing actively managed ETFs in history”. He also points out that he continues to see the system evolve with each trade from machine learning algorithms and believes the fund’s best days are still ahead.
So, do they perform better?
Even with the performance of AIEQ, a broader question is: “Do AI-powered ETFs generally perform better?”
An interesting study (‘Machine Learning for Active Portfolio Management’ by Söhnke M. Bartram, Jürgen Branke, Giuliano De Rossi and Mehrshad Motahari) investigated the performance of a sample of active ETFs that use machine learning in their investments and concluded that performance tends to be “mixed”. However, ML methods have several advantages that can lead to successful applications in active portfolio management, including capturing nonlinear patterns and focusing on prediction through ensemble learning. Overall, they concluded that ML techniques are promising for active portfolio management, but investors should be cautioned against their main potential pitfalls.
Another study by Rui Chen and Jinjuan Ren, titled ‘Do AI-powered mutual funds perform better’ and published in June 2022, sought to investigate the effectiveness of the broader AI capability in the mutual fund sphere. The study looked at the prospectuses of 2,133 newly issued funds from 2017 to 2019 from the EDGAR database of the US Securities and Exchange Commission. It cross-referenced them with AI-powered funds using mutual fund data from the CRSP Survivor-Bias-Free US Mutual Fund Database from January 2009 to December 2019. The study found that these funds do not outperform the market per se. However, a comparison shows that AI-powered funds significantly outperform human-managed peer funds. The study further shows that the outperformance of AI funds is attributable to their lower transaction cost, superior stock-picking capability and reduced behavioural biases.
From these studies, investors might be inclined to think it is reasonable to answer the question, “Do AI-powered ETFs generally perform better?” with a “More or less.” Their ability to adjust their investment strategies based on market data and overall risk management advantages will make AI-powered ETFs increasingly popular moving forward. Yet, even with ChatGPT and AIEQ making interesting inroads into the ETF sphere for investors, many will see AI-powered ETFs as still a relatively new, albeit rapidly evolving field, with many uncertainties around their long-term performance. The next few years will no doubt be decisive performance markers.
© 2023 funds europe
