Artificial intelligence: a brave new dawn for investment management
Much of the investment process comes down to the analysis and interpretation of data, so it stands to reason that those who have better data and better methods with which to analyse it can achieve higher returns. For years, the most effective approach available to investment managers was to hire the brightest and best individuals. Now many investment firms are exploring the possibility that robots might be at least as good an option.
Leading fund management groups have made considerable investment in artificial intelligence. BlackRock set up the BlackRock Lab for Artificial Intelligence in Palo Alto, California last year, looking at ways to gather so-called big data and turn it into effective insight to deliver higher returns for their clients.
Schroders and BNY Mellon also have major technology programmes. A report in 2018 by Standard & Poor’s found that 80% of asset managers plan to increase their investments in big data over the next 12 months.
Much of the investment process comes down to the analysis and interpretation of data, so those who have better data and better methods with which to analyse it can achieve higher returns.
Internet of things
Why now? Only in recent years has it been possible to collect sufficient data to draw meaningful insights. Today, a vast and growing internet of things is gathering data on everything from climate patterns to consumer spending and how many steps people walk. These help build repositories of big data, from which information can be sifted and analysed.
At the same time, computing power has increased the extent to which such data can be stored and interrogated. The world’s largest supercomputer – China’s Sunway TaihuLight – can make 93 quadrillion calculations per second. These innovations have created a platform for fund managers to build their artificial intelligence capabilities.
The theory runs that mining the data will offer superior insights. Certainly, those that can build the right infrastructure to trawl through data that is often messy, unstructured and unwieldy should obtain an edge on their peers. Analysing footfall in a major department store might give clues as to whether their sales estimates are overblown; analysing news feeds after a corporate catastrophe can help investors draw conclusions as to whether the brand can survive; the popularity of a new product can be measured through customer reviews, or social media comments.
As the volume of data explodes – Research group IDC predicts a tenfold increase by 2025 – companies that continue to rely solely on human insight may fall behind. A longer-term problem is that these data insights may becoming more widely available and increasingly tend to be priced into markets. The investment industry may be toiling in the foothills of a digital arms race as companies seek to analyse every larger volumes and sources of data to harness differentiated insights.
Nevertheless, doing nothing is not an option. Luke Ellis, the CEO of hedge fund manager Man Group, recently said that a failure to adopt quantitative approaches using big data could prove fatal for fund groups: “If you don’t understand how to treat data with respect, you will get eaten alive.”
The solution lies in ensuring that investment managers and data scientists work together harmoniously. Investment teams need to identify where data scientists should focus their research. Schroders argues that if each group is left to its own devices, data scientists won’t necessarily ask the right questions, while the investment team will fail to understand the technology.
In a recent blog post, the Schroders data insights team wrote: “Effective data science will unearth insight that is unlikely to have been captured by others. The greater the quantity of data that may be relevant to understanding an enterprise, the more combinations and permutations of analysis it becomes possible to conduct.
“By extension, the likelihood of other parties conducting exactly the same analysis diminishes. It seems likely therefore that data science at scale within a large investment organisation will generate insight that is differentiated and hard or unlikely to be precisely replicated.”
The authors add: “Far from creating a level playing field, where more readily available information simply leads to greater market efficiency, the impact of the information revolution is the opposite: it is creating hard-to access ‘realms’ for long-term alpha generation for those players with the scale and resources to take advantage of it.”
Using artificial intelligence in investment management can make humans smarter and more efficient. By reducing manual processes and number-crunching, it frees them up to conduct value-added analysis.
Making humans smarter
Another key part of the appeal of using artificial intelligence in investment management is that it can make humans smarter and more efficient. By reducing manual processes and number-crunching, it frees them up to conduct value-added analysis.
A robot cannot – at least for now – meet a management team and form a judgement on their trustworthiness and capability, but it can do a range of tasks to assist human fund managers in doing it better. Artificial intelligence (AI) should also help fund managers assess and mitigate risk more effectively.
Equally, AI processes should themselves become smarter over time. As data improves, AI can truly ‘learn’ and develop insights based on those assessments. Groups that have put mechanisms for this in place today should find that they become more useful as time goes on.
Potential cost advantage
Given that setting up AI platforms is costly, there is the question as to whether it gives larger fund management groups an advantage. This may be true to some extent, but it also depends on the area of investment. In the financial or consumer sector, big data insights can make a real difference – how much are people spending and on what?
There may be a cost advantage. After the initial set-up cost, technology can help keep costs lower. At a time when fund manager fees are under scrutiny, this is a notable advantage, particularly if investment groups can demonstrate they are adding value above and beyond index benchmarks.
Does artificial intelligence represent a brave new world for investment management? It is certainly an important development. Groups that can strike the right balance between data science and established portfolio management skills could have a significant advantage.
The limits of forecasting
Yet ultimately one should bear in mind that all financial forecasts, whether about the specifics of a business such as sales growth, or predictions about the economy or financial markets as a whole, are informed guesses. Modelling and forecasting are very useful to help businesses plan activities such as production and financing. In portfolio management, they offer a way to allocate money based on realistic scenarios, combined with the wisdom of experience and risk management. But risk and uncertainty remain central to forecasting.
There are three major problems with relying on forecasts. The data is always going to be old because historical data is all there is to go on, and there is no guarantee that conditions in the past will remain the same. It is impossible to factor in unique or unexpected events, or external developments.
Secondly, assumptions are dangerous and black swan events – considered unimaginably rare – have become more common even as dependence on forecasts has grown. Finally, forecasts cannot incorporate their own impact. The very existence of forecasts, accurate or not, itself influences the actions of businesses, a factor that can’t be included as a variable.
This is a conceptual knot. In a worst-case scenario, management becomes a slave to historical data and trends rather than analysing what the business is doing now. Forecasting can be a dangerous art, because forecasts become a focus for investors, mentally limiting their range of actions, by presenting the short- to long-term future as already determined. In addition, forecasts can be simply erroneous from the start or derailed by random elements that can’t be incorporated into a model.