The financial services industry is beginning to face disruption from companies leveraging the power of technology. This includes asset management where technology is being used to gain an advantage on several fronts. Not least of these is the emergence of A.I. based investment strategies.
Quantitative investing has already established itself as an important part of the asset management industry. Quant investing is now being taken to a new level by the introduction of artificial intelligence (A.I.) and Big Data. These advances offer a new range of possibilities in the endless search for alpha.
- The evolution of asset management
- What is artificial intelligence (A.I.)?
- What is Big Data?
- How do A.I. and Big Data change the asset management industry?
- Advantages of A.I. based trading strategies
- An example of A.I. and Big Data in asset management
- Outlook: The future of asset management
The evolution of asset management
The asset management industry has evolved in three areas; investment vehicles, approaches to portfolio management, and technology. In many cases advances in one area have allowed the next steps to be made in other areas. Before 1950, the only option available to most people was a stock broking account. Either the clients themselves or their broker would select stocks and manage the portfolio.
Although some research on stock behaviour and portfolio management was done in the 1930s and 1940s, the field of portfolio management began to be taken more seriously starting in the 1950s. Over the next three decades, a series of models and theories now known as modern portfolio theory evolved.
The rise of modern portfolio theory
Modern portfolio theory concerns the relationship between risk and total return. It also emphasizes the benefits of diversification and portfolio optimization. Asset allocation models were expanded to include bonds, property and cash to further improve the risk profile of a portfolio. Much of the research that these theories are based on was only possible when computers became powerful enough in the 1960s and 1970s.
Around the same time the first mutual funds were introduced. Mutual funds expanded the potential client base for asset management companies greatly. The first experiments with hedge fund strategies were also done in the 1960s, though they were only popularised much later.
Quantitative investment management
The next phase of the evolution of the asset management industry only really gathered momentum in the 1990s. This was when quantitative investment management became a serious part of the investment landscape. By then desktop PCs could be used to conduct extensive research, and portfolio analysis and optimization. This style of investing uses empirical evidence to create portfolios, rather than the fundamental analysis favoured by active fund managers.
Passive fund management, which in many ways overlaps with quant funds took off in the 1990s when exchange traded funds (ETFs) were introduced. By then it was becoming apparent that most active managers failed to beat their benchmarks. ETFs allowed investors to easily achieve portfolio diversification and earn the market return for very low fees. The introduction of passive investing with index funds formalised the concepts of alpha and beta.
Focussed approach to generating alpha
Client portfolios began to be divided into a portion that aims to cheaply capture the market return (beta) and a portion that aims to generate excess returns (alpha). This shift, along with advances in technology and lessons learned from the world of quantitative investing, led to a more focussed approach to generating alpha.
For much of the last two decades the search for alpha has centred around stock picking, market neutral and long/short stock trading strategies, and algorithmic trading. However, nearly all of these strategies have made use of the same data – company financials, historic stock price data, and economic data.
Advances in financial technology (fintech) have also led to several new types of asset management firms. These firms are built almost entirely around technology. In some ways they are like the Uber or Airbnb of the asset management industry. For example, social trading platforms allow retail and institutional investors to copy the trades of other traders. And, robo investing platforms automatically create and rebalance portfolios.
The next phase in the evolution of asset management is now underway. It takes algo trading to a new level by introducing artificial intelligence (A.I.) and Big Data to the process.
What is artificial intelligence (A.I.)?
A.I. – artificial intelligence – is often used interchangeably with terms like machine learning, deep learning, neural networks and even fuzzy logic. Strictly speaking, machine learning is a subset of A.I. and deep learning is a subset of machine learning. A.I. has no precise definition, but generally refers to the science and engineering behind machines or programmes that can make intelligent decisions.
Machine learning refers to programs that can alter themselves to improve their decision making. Deep learning and neural networks introduce a more scientific or mathematical approach to machine learning. Regardless of the term, most A.I. programs seek out linear and non linear relationships between data points in order to make predictions or decisions. A.I. programs typically use historical data to learn and find patterns. These patterns are then used with live data to make predictions or decisions.
Most algorithms that make use of A.I. run immense numbers of calculations on enormous amounts of data. This means the evolution of A.I. has been heavily dependent on the available processing power of computers. Computing power in turn has grown exponentially over the last few decades. Advanced A.I. programs can therefore now be run on standard desktop PCs.
Regardless of the type of algorithm, or the available processing power, the algorithms predictive ability depends on two attributes of the data being used. Firstly, there needs to actually be a pattern or relationship within the data set. And, secondly, the data sample used for learning needs to be large enough. The volume of data, types of data, and quality of data are therefore of critical importance to A.I. This is where Big Data comes in.
What is Big Data?
The term Big Data refers to the process of gathering, storing, organizing and analysing very large data sets. Modern technology creates ever increasing amounts of data at an accelerating pace. More than 90 percent of the data that exists today was generated in the last 2 years alone. As of early 2018, 2.5 quintillion bytes of data were generated every day, and that number has since increased.
Potentially useful data is generated by social media, Google searches, mobile devices, and other connected devices including vehicles, drones and cameras. These sources of both structured and unstructured data can be used for Big Data analysis to make predictions and decisions in numerous industries.
Such large datasets need to be organised (a field in itself), and then processed using Big Data analytics software and large amounts of computing power. Well organised data combined with effective A.I. software can reveal insights that were until recently unavailable. The field of Big Data technology is growing rapidly in conjunction with A.I. due to growing number of data sources and the growing number of applications for autonomous machines and programs.
How do A.I. and Big Data change the asset management industry?
A common misconception regarding artificial intelligence and investing is that computers will replace humans and do the jobs they currently do. While this may be true to an extent, it only touches the surface. A.I., especially when combined with new sources of data, brings completely new methods and possibilities to the investment management industry. In this article we are focusing on security selection and portfolio management. However, A.I. is also being used in operations, client relationship management and in assessing the risk profile and needs of clients.
Robo advisors for instance use A.I. to determine a client’s needs. A.I. is also used for data management as asset managers gather ever growing amounts of customer data. When it comes to using artificial intelligence for investing, the primary objective is to find patterns or relationships between stock prices and other factors. Traditionally that meant fundamental and economic data. With Big Data the possibilities are endless.
For example, satellite photos can be used to monitor the number of cars in the parking lot at a shopping mall. Data from news websites and social media platforms can be analysed to gauge sentiment about a security. Internet search trends and website traffic can also be used to potentially predict consumer behaviour. All of this data can sometimes be used to find very obvious relationships. However, sometimes it’s the patterns that are less obvious that are of the most value. This is where A.I. can be used to find those patterns that wouldn’t have expected or thought to look for.
While traditional active managers based decisions on theories about how companies, and their stock prices, would perform, quantitative investing is evidence based. In other words, decisions are based on models that are based on empirical evidence about what actually happens, rather than what people think will happen. Quantitative investing also emphasises the importance of the understanding of probabilities. Nothing is ever certain, but given a large enough sample size, the probabilities are relatively stable. The use of A.I. and Big Data reinforce the use of both evidence-based models and probabilistic trading.
Asset management firms are now finding themselves competing with fintech companies that are bringing a very different approach to the world of portfolio management and capital management. If a data source is proprietary, rather than open sourced, a successful strategy based on it will be all the more valuable. Large asset managers are therefore trying to build valuable proprietary datasets, which themselves have value.
Advantages of A.I. based trading strategies
Traditional active fund managers base their decisions on theories about the way stock prices move and the future prospects for companies and the economy. This is a very subjective approach which cannot be tested. Both A.I. and the broader quantitative investment field are based on evidence-based investment strategies. Evidence based investing introduces a more scientific approach, providing more certainty and a better understanding of the potential relationship between risk and reward.
The introduction of A.I. and Big Data allows investment decisions to be based on far broader sources of data than the data traditionally used by portfolio managers. In addition, many of these data sets can be accessed in real time, unlike traditional financial information that is released only weekly, monthly or quarterly.
Algorithms can process very large data sets very quickly. This means far more data can be analysed, decisions can be made far faster, and the entire process is cheaper on a “per security basis”. The result is that it is feasible for a fund to have a much larger potential universe of securities and opportunities. Finally, the field of behavioural finance has proven that in the real world, investors are not nearly as rational as the efficient market hypothesis would have us believe. This creates anomalies and opportunities which only A.I. learning algorithms can find and exploit.
An example of A.I. and Big Data in asset management
Catana Capital is an example of a company that uses both A.I. and Big Data driven strategies to manage long only and long / short strategies. The Data Intelligence Fund runs a long short strategy based on data that is gathered through natural language processing from financial news sites, and user generated data from social media sites. A.I. based investment strategies depend on the quality and sources of data used. In this regard Catana Capital endeavours to find as many sources of relevant data as possible.
Data is collected in the form of news, research and analysis, calendars of events, and user generated content from social networks. Social media sites, in particular platforms like Twitter, Reddit and StockTwits, generate ever increasing amounts of data which few other asset managers have managed to use effectively.
Besides sourcing data through natural language processing, the algorithm also monitors over 45,000 securities around the world. A machine learning algorithm looks for relationships between what is being said about securities, important events and the price movements of those securities. Predictions are then made based on aggregated sentiment values and security prices, which are in turn used to generate trading signals.
In addition to using A.I. and Big Data for stock selection and market timing, the stock portfolio risk is also managed using Big Data triggers as well as traditional stop losses. Catana Capital is an example of a new generation of innovative fund managers and fintech companies that combine the power of A.I. with new sources of real time data, to generate alpha producing trading signals.
Outlook: The future of asset management
Asset management will no doubt continue to evolve on a number of fronts, including the types of products that are sold, the way portfolios are managed, and the way client relationships are managed. The one thing we can be certain about is that software will play a growing role as the search for alpha grows increasingly sophisticated. As A.I. investment-based strategies attract more assets under management, firms will commit more resources to exploring this new frontier. As a result, we can expect to see more A.I. based investment products coming to the market.
It is likely that some edges will be eroded as more learning algorithms find relationships between data sets. The result will be that only those firms committed to finding new data and new methods of using A.I. will survive. This will be an industry for specialists, while those that treat A.I. as a novelty or a marketing exercise will be found out quickly.