Originally, behavioural finance was overly focused on the construct of ‘rationality’. This was demonstrated by the contrast often set between the hyper-rational and logical Mr. Spock from the popular sci-fi series Star Trek and pop culture sensation Homer from the animated series “The Simpsons”, who regularly fell prey to his baser instincts.
The result is over 200 listed biases on Wikipedia providing examples of irrational behaviour.
Meier Statman released a paper in 2019 that charts the way forward for behavioural finance. Statman proposes that being labelled as irrational is perhaps not useful. If what the investor wants from his or her portfolio is financial security and a certain event, like COVID-19 for example, causes severe market turbulence, it is perfectly rational to move money to less volatile assets. Behavioural finance should therefore focus on real people, what they want from their investments, and indeed what they get.
This is where artificial intelligence (AI) plays a key role: In understanding investors. Take for example machine learning that is a foundational discipline in AI. There are two components of machine learning that are very useful in understanding the behaviour of investors. What is termed ‘unsupervised’ machine learning algorithms are used to identify patterns or commonalities between variables, such as risk and return in the context of investments. They are termed ‘unsupervised’ because once the algorithm is equipped with the knowledge that risk and return is important to investors, it will search for and reveal the underlying behavioural patterns in switching (for example).
Interestingly, when studying the behaviour of over 35 000 investors over 16 years from 2006 to 2022, without any prior knowledge of ‘behaviour tax’ or value eroded from switching between funds, the algorithm finds four patterns of behaviour that are organically split into four different average behaviour tax levels. The algorithm clearly groups investors into those that are loss averse (don’t necessarily mind risk but at the first sign of volatility, de-risk their portfolios), risk averse (those who avoid risk) and those that chase past performance relentlessly. The fourth grouping is a combination of these three groupings and what we call the ‘Market Timer’.
Unsupervised algorithms are useful in revealing past patterns but provide little indication of the future. To predict behaviour, we need to turn to supervised learning. Here the algorithm is ‘supervised’ or provided with many data features to determine the relationship between inputs (market, fund, and personal features) and outputs (behaviours like switching or staying invested). Here, algorithms such as artificial neural networks (ANNs) or random forests based on a decision tree architecture are used to predict investor behaviour based on this relationship of inputs to outputs. When taking insights from the behavioural patterns discussed here, the Market Timer archetype is shown to have consistently the highest probability to switch when using over 12 million observations. The algorithm here was deliberately deprived in one test case of past switch behaviour from investors and still predicted this group to switch (which they indeed do). This shows the algorithm is effective in correctly identifying what makes these investors ‘tick’.
The essence is that AI is helping us to gain a deeper understanding of investors, so it becomes possible to engage with the right person at the right time with the right message.
The essence is that AI is helping us to gain a deeper understanding of investors, so it becomes possible to engage with the right person at the right time with the right message. From here, behavioural science techniques such as nudges may be targeted and delivered more effectively. The result should be better investment outcomes – such as staying invested to avoid the dreaded behaviour tax.
More information on Momentum Investments’ explorations into the realm of machine learning are available in the annual Sci-Fi report, available here:
Momentum Investments is part of Momentum Metropolitan Life Limited, an authorised financial services (FSP 6406) and registered credit (NCRCP173) provider.
