Managing the risks that your portfolio is exposed to can be complicated. It’s easy to lose focus on what you’re trying to do. The constant noise from the media can be deafening, and it can take away your ability to think clearly.
So, it’s a good idea to understand your primary objective. When you’re managing portfolio risk, what’s your goal? Warren Buffet’s Rule #1 and #2 are as good a starting point as any:
Rule #1: Never lose money.
Rule #2: Never forget rule #1.
How can you protect the value of your portfolio? What are the types of risks you have to watch out for? What are companies around the world doing to manage risks using the latest artificial intelligence techniques?
Basic overview of the types of portfolio risks
Here’s a quick overview of the types of risk that your portfolio can be exposed to:
Operational risk: This category includes things like losses from inadequate procedures and policies. Could a fraud or a mistake by a single employee threaten the entire firm? Rogue trader Nick Leeson left a £862 million hole in Barings Bank’s balance sheet. A Samsung Electronics employee made a $105 billion “fat-finger” error.
Market risk: Covers the possibility of losses due to factors that affect the overall performance of financial markets. Sources of market risk include recessions, political turmoil, changes in interest rates, natural disasters, terror attacks, pandemics like coronavirus, etc.
Technology risk: In the increasingly interconnected world we live in, a data breach or a cyberattack can cripple a company. The 2017 Equifax data breach resulted in a $700 million penalty for the firm.
Liquidity risk: Can you sell your investment when you need funds? Star stock picker Neil Woodford’s investment empire crumbled when it was discovered that he had deployed funds in illiquid assets.
Credit risk: Chasing high returns can put your capital at risk. But doing the opposite by playing safe with AAA-rated investments can lower returns. So, do you take a high-risk, high-return approach or settle for mediocre returns?
Currency risk: Familiar with the term “peso problem?” Back in the 1970s, the Mexican peso was pegged to the US dollar, and Mexican banks offered much higher rates than US banks. It was logical to put your funds in Mexican banks and reap the benefits. But not many investors did that. Why? Because they anticipated a “peso problem” – they expected the peso to crash against the dollar. That’s precisely what happened in August 1976.
Types of data used for risk evaluation
Artificial intelligence and machine learning have been successful in helping to manage portfolio risk. But much of their success depends on the availability of the right type of data. If the correct data is at hand, AI and ML can use it to spot trends that can then be extrapolated to arrive at actionable risk mitigation strategies.
Which are the types of data that AI and ML find most useful?
Historical prices: Studying historical stock or bond price patterns can provide insights into what could happen next.
Market-related news: The challenge here is to separate the wheat from the chaff.
Real-time prices: This category of data for a stock or bond is distinct from the historical price. AI/ML can use real-time data to provide investment managers with an invaluable risk management tool.
Social media data: There are 350,000 tweets per minute – 200 billion per year. And that’s only one social media site. Is there any way to use this ocean of data to manage portfolio risk? AI/ML can provide an answer.
Company-specific data: Company’s financials, executives & decision maker’s profiles, and their track record, etc.
Laws and regulations: AI/ML can sift through massive amounts of government-issued rules to identify the relevant bits.
Role of Big Data and Machine learning for risk evaluation – specific examples/use cases
Here are five use cases for AI/ML in portfolio risk management:
1. AI in financial risk
Research by Chartis and IBM highlights new emerging themes in how Artificial Intelligence (AI) is used in regulatory technology to help financial institutions address risk management and regulatory compliance. Some specific use cases include:
- Use of Natural Language Processing(NLP) to convert unstructured data into structured data. For example – bond terms and conditions into transactable and analyzable time series.
- Use of ML/Deep Learning models for scenario generation and stress testing, as well as curve construction and validation. For example – yield curves, smile curves, etc.
- Several approaches have married Monte Carlo simulation techniques with ML and data tagging to more efficiently generate automated maps, validation routines, and portfolio strategies.
- Clustering techniques such as topological data analysis and unsupervised neural networks to help set up factory analysis.
2. AI in analyzing historical trade patterns
An article in the Harvard Business Review titled, What Machine Learning Will Mean for Asset Managers, points out that AI and ML can help to identify potentially outperforming equities. It can do this by finding new patterns in data.
Here’s an example of how this could work. The authors of the article, Robert C. Pozen and Jonathan Ruane say that ML techniques could be used to examine the “substance and style” of the statements that CEOs make in quarterly earnings calls. What are the specific areas that ML could identify? These could include:
- The trustworthiness of the forecasts that the CEOs make.
- Correlation of companies operating in the same sector.
Do these techniques work? The authors state that ML has been 10% more accurate than prior models in predicting bond defaults.
3. 33% of hedge funds use ML for risk management activities
Hedgeweek, an online publication that covers hedge fund industry news, found that ML tools are extensively used by hedge fund managers in the investment process.
Which are the tools in use? BlackRock’s Aladdin Risk Platform uses ML algorithms to monitor the risks that investment portfolios are exposed to. BlackRock says that its portfolio management software can analyze over 2,000 risk factors per day.
- TCA & Trade Optimization: reducing costs plays a part in both increasing returns and reducing losses; both of which are exercises in risk management. AI techniques help optimize the timing of entering and exiting trades.
- Churning through real-time and historical data, AI techniques help identify liquidity patterns in different circumstances to ensure trades are carried out as planned.
4. Using alternative data to devise hedging strategies
The use of non-traditional techniques to devise hedging strategies is gaining popularity. What are the non-traditional or alternative datasets which can be used?
- Fund managers use AI to scour weather forecasts and container ship movements to identify trends.
- Another AI technique involves keeping track of search engines for specific words and topics, search trends, etc.
5. Investment idea generation
Can AI be used to generate investment ideas? The CFA Institute’s AI Pioneers In Investment Management report describes an interesting use case.
When the UN issued its report on smart cities, a fund manager used this information to see which companies would benefit from the implementation of the UN’s proposals. The search for the companies involved a three-step process:
- A search for articles that mentioned “smart cities” and “future.”
- Extracting the names of the companies mentioned in the articles.
- Using an algorithm to create a “force-directed” graph. In this graph, similar articles would be grouped together.
The result of the exercise was a “news map” that provided the names of companies that were most likely to gain from the UN’s initiative.
The bottom line
Improved efficiency, accuracy, improved speed of investments, cloud computing and availability of data are driving forces behind the wide adoption of AI/ML in portfolio risk management. With the benefit of AI, risk teams are operating in a more integrated fashion with portfolio managers and pushing risk management further into the front-office.
Increasing data speeds, the availability of “oceans of data,” cheaper data storage, and more powerful computers will only lead to an acceleration in this trend. The years ahead are likely to see AI will transform how fund managers manage risks.