Show 74 – Behavioural Finance
Recorded in partnership with CFA UK from their conference in London on the topic of ‘Behavioural Finance in the age of algorithms’, we interviewed the five speakers from the event. Our guests were:
1. Markus Schuller, Founder and Managing Partner of Panthera Solutions
2. Philippa Clough, Portfolio Manager at JP Morgan Asset Management International Equity Group
3. Kristina Vasileva, Senior Lecturer in Finance at Westminster Business School
4. Shweta Agarwal, a member of BlackRock’s Risk & Quantitative Analysis Group
5. Magda Osman, a Reader in Experimental Cognitive Psychology at Queen Mary University of London
Markus is Founder and Managing Partner of Panthera Solutions and his talk was titled ‘Human ambiguity tolerance beats artificial intelligence’. He explained that ambiguity tolerance is about maximising the contribution of intellect and reason to our decisions in a complex environment.
In his talk, Markus explained about how the limbic system in the brain manages and modulates our emotions and that it’s not just what’s happening between our ears but also how much adrenalin is in our veins and how emotionally aroused we are as a consequence, which can be positive or negative. He also discussed a number of other topics including resistance to change and knowledge management. He also had an interesting take on Diversity, because as well as gender, ethnicity and age, he said it’s important to also maximize cognitive diversity, because if we are socialized in a very similar way, then our perspective on the world is similar.
Markus’ view on AI is that we’re far away from the point of it replacing humans in the asset management industry, but that it is technology that can support us to make more evidence based decisions. It’s much faster and more concise when it comes to fundamentally analysing thousands of stocks at once, but it’s not yet ready to replace us when it comes synthesizing qualitative and quantitative factors when interpreting data.
Markus has written a number of blog posts for the CFA UK, which are highly recommended for additional reading and explanation of all the topic areas he spoke about. They are on the topics of:
- The Knowing-Doing Gap in Behavioral Finance
- Ambiguity Tolerance Beats Artificial Intelligence
- Survival Strategy: A Learning Investment Team
- Benchmarking Multi-Asset Portfolios: The Global Capital Stock
All available at the CFA UK Blog
Interview starts at 8min22s
Philippa Clough is a Portfolio Manager at JP Morgan Asset Management International Equity Group. She has a Masters in Mathematics from Oxford University and is a CFA Charter Holder and was presenting at the event on ‘Reflections from 20+ years of behavioural finance investing’.
Philippa said that she works in a team of 40 investors who use behavioural finance insights in order to determine how they direct their investment insights and that they apply behavioural finance insights into two areas:
- to stocks and what impacts how stocks are priced
- to themselves – she explained that if we think other people are biased, it’s probably quite likely that we’re biased ourselves and we try and think about ways to enable us to make more rational decisions.
Philippa talked about earnings calls to explain it a little more. She said that as an investor listening in to the call where the company is talking to their investors and updating them on the company is doing, you will have expectations of how it’s going to go in the future. As an investor, you may say, “OK that’s interesting. Well, company management are biased. They tend to be overly optimistic on their future prospects.” So, as an investor, how do you think about that information that’s just come out about the company and the strategy? How do you balance with the rest of your opportunity set? One way to approach that is to listen to lots of other earnings calls so that you get a base of comparison. However, there are a lot of companies in the world, and Philippa said there are 2000 in Europe alone that they consider investable, and as an analyst, it could take you two years to listen to all 2000 calls, which happen on a quarterly basis, and get through that level of information, which is obviously not feasible, especially if you have certain weeks where they may have 100 calls to listen to.
Instead, Philippa said that they have applied natural language processing techniques to automatically calculate the sentiment for an earnings call. This enables them to tell whether it’s positive or negative, which means they can then direct investors to focus most of their time on the ones where the sentiment is particularly extreme relative to comparable companies in comparable industries. It therefore enables them to focus their time on where they think they are going to be able to gain the most insights that are going to be informative for future stock performance.
Philippa said that technology therefore plays a key part in the job, but not just about looking at stocks, as in the example above, but in themselves too. She said that on average, people are overconfident, and gave a common example to explain it, which is the question whether you think you’re better than average at driving? When asked this question, 90% of people will tell you that they’re better than average at driving, yet it’s a 50:50 statistic, meaning 40% of those people are going to consider themselves overconfident. Philippa said that if you are overconfident and you have more conviction in your insights, then you sometimes discount the likelihood that maybe you’ve got it wrong or what the probability of things are not turning out how you expect. From a portfolio perspective, that means that you can end up owning large positions in companies, which have all very concentrated risk and also turning over your portfolio more than you should.
Philippa said that “we use technology to make us make more rational decisions is by having quite a sophisticated risk analysis framework. So, we want to cut the risk of your portfolio or your exposures through many different lenses to say stop, think and make sure that the decisions in the weightings that you’re placing or the amount that you own of a company, can measure it with the risk and your return expectations.”
Philippa said that knowledge in behavioural finance is growing but it’s definitely not what the majority of people are doing. However, she thought that 2017 was an amazing year for the topic area, as Professor Richard Thaler won the Nobel Prize in Economics and in the same year, also starred in ‘The Big Short’ alongside Selena Gomez, which Philippa said is unusual for someone who spent their time thinking about how psychology fits into investing!
Philippa believes that events like the CFA Behavioural Finance conference are brilliant, whether you’re an investor or someone who focuses on other areas of finance, because it really makes you stop and rethink your behaviours and your decisions, even if it’s not in your day to day job.
To finish off, she gave one example of a behavioural bias that might be easier for anyone to understand, which is called ‘Home Bias‘, where people prefer things that they’re more familiar. She used the example where someone who works in her team is a big Arsenal FC supporter and in his Fantasy Football team, has always had a bias towards including players from Arsenal. She said this could be similar to how it affects people who invest in only one country, perhaps just UK companies because you’re more familiar with them. Phlippa said that fine if all your expectations of future payments are going to be in the UK, but what if you have something like the UK referendum and your currency suddenly devalues by a good 10-12%, potentially where you’d allocated your assets doesn’t fit your future returns and isn’t optimal.
Kristina Vasileva – Can we use big data to predict human behaviour?
Interview starts at 18min7s
Our third guest was Kristina Vasileva, Senior Lecturer in Finance at Westminster Business School, who presented at the the conference asking the question “Can we use big data to predict human behaviour?”
Kristina said that in terms of Behavioural Finance, at least on the academic side, there is lack of good data, meaning they’ve had to prove and analyse behavioural finance issues using data which is not necessarily suitable to capture human behaviour. However, they now have all kinds of other resources through which they can capture data, such as Google searches and other actions that people do or how they interact with their financial products, to capture aspects of human behaviour, which previously was just not possible. To explain this, she gave a few examples from different well-known and well-established behavioural biases:
Overconfidence – which Kristina said works on every level. She explained that for every human being is prone to overestimating our skills or our abilities and you capture this trait through several indicators, which can be used that could show each investor that they are standing out from the general average. For example, volume of trading or intensity of trading would be a very good indicator that might show that you might be more prone to overconfidence than your peers and companies would be ideally placed, because they have access to this data, which is very easy for them to process. This would enable them, quite quickly, to see if you are standing out from your peers.
The Disposition Effect – which Kristina said is the tendency for investors to sell winning stocks quicker than they sell losing stocks. She said that there’s a lot of finance theory on why this is happening, and that it’s a combination of Prospect Theory and Mental Accounting, but ultimately, it’s very straightforward to measure – have you sold more stocks out of old stocks that have made a gain vs the same ratio for losing stocks. She said that there is no theoretical expectation why these two would be different, why it shouldn’t be 50/50 in a broad data set, but we observe that there is a difference. So, if this could be some small indicator that is running in the background, it could blend in with the rest of the financial indicators and could serve to perhaps prompt investors or companies, to investigate further if there is deviation from the expectation, why is it happening is it justified and is it something they can address or improve? Kristina said that she is not talking about this being used in a punitive way to evaluate performance for investors, because she think it will have counteractive effects, but it can be used to improve operational efficiency for companies and investors.
Kristina hopes that more of this will be done in the future with active consent from either clients or employees in the name of improved performance and efficiency, because academics struggle with finding unique data sets that are able to demonstrate these issues more clearly are more robustly. However, if companies are able to provide more anonymised data to academics, then they could achieve greater gains. She said that academics would improve methodology and be able to prove or disprove what actually they think is happening vs. what actually is happening and would be able to improve the quality of information they have to evaluate investments or monitor portfolios. In fact, she said it doesn’t have to work just at an investor portfolio level, as it can work just as well with any kind of employee/employer situation.
Shweta Agarwal – Behavioural finance in the man-machine equation
Interview starts at 26min 35s
Shweta Agarwal is a member of BlackRock’s Risk and Quantitative Analysis Group, where she supports the behavioural finance and advanced analytics initiatives of the group. Before joining Black Rock, Shweta consulted on data driven behavioural science projects for FinTech companies and worked as a behavioural quants consultant at Barclays Wealth. She also holds a PhD in Decision Sciences from the London School of Economics at an MSc in Mathematics from the University of Cambridge. Her presentation at the event was on the topic ‘Behavioural finance in the man-machine equation’.
Shweta said that in her presentation, she talked about human decision making through a three part system:
- subconscious brain, which is 200 million years old
- conscious brain, which only 20 million years old
- reptilian brain
Shweta explained that the subconscious brain has evolved over many many years, so it’s in some sense a very sophisticated, almost beyond our comprehension, and has survived the test of time. Although we might think of the subconscious brain as being something that gets in the way of our decisions and our emotions, that biases our decisions, it is in some sense quite essential for decision making. There are studies that show that humans who have the emotional brain paralysed, struggle to make very basic decisions because they can’t get a feeling for whether their decision is good or bad. She added that in financial markets, it’s been acknowledged that emotions actually drive asset prices, which can be explained if we look at the human tendency to herd, which is again something that is an evolutionary construct. As humans are social primates, they tend to make decisions collectively, they mimic each other and Shweta said that this is reflected in modern day society in fashion as well as in financial decision making.
Shweta said then talked about the reptilian brain, which the flight-or-fight system is essentially related to, which is again a part of our subconscious. She explained that way the reptilian brain works is that there are two nervous systems that cross the heart and the activation in these two nervous systems determines the kind of physiological response to the environment:
- sympathetic state – an immediate response or a fight or flight response to the environment we are in. It’s automatic. It is not deliberate. It does not optimise any information. It’s very instinctual.
- parasympathetic state – which moderates this instinctual response to the environment.
Shweta said that it’s the combination of these two, which determines the physiological state of a person.
Shweta also talked about a pilot project that BlackRock is running, where they are monitoring how the stress levels and the heart rates of members of her team impacts on performance. She said that stress comes in many different ‘flavours’:
- Positive stress, when both sympathetic and parasympathetic nervous system are activated
- Stress that is actually detrimental to performance.
With technology and algorithms, Shweta said that they can actually monitor the heartbeat variability, which signals whether the Portfolio manager is in a positive or negative stress level. It is a completely private project and completely voluntary and the Portfolio manager owns their data. It is never used to evaluate if the performance a portfolio manager is good or bad, but it is intended for the purpose of enhancing their performance. The volunteers are therefore wearing devices to help BlackRock track their heart rate variability to help them see whether their physiological state is helping or hurting their decisions.
We were intrigued by the the reaction of those people being monitored in this project as we recently covered Mental Health & Wellbeing on the podcast in Show 72. In that episode, we talked about the fact that many people, in particularly men, don’t like to talk about their mental wellbeing and how they’re feeling at the time, particularly if they’re stressed or anxious, because of a fear of being seen as weak or it being detrimental to their career progression.
Shweta responded to this by saying that it’s not exactly uncommon to use heart rate monitors as a part of training or improving performance. It’s something we see that in sports, where the sports training includes monitoring the heart rate variability, we see it in the military and and in the medical sciences as well. However, it’s only in investing that this may feel like something new. She said that they don’t want it to be intrusive to the Portfolio managers and that it is completely voluntary. She also thinks that there is an inner drive from the portfolio managers to get better at what they do and if they can bring valuable insights to the table, then they feel there is value in participating.
We finished our discussion looking at how much of our decision making will be machine driven in the future.
Shweta said that it is true that machines are doing more and more of what humans were traditionally doing and are probably increasingly taking on more information processing tasks, which she added is the right thing to do because machines are categorically better at processing information than humans. However, she believes they have a long way to go before they acquire the sophistication and intelligence of the human brain and so to think that humans will be out of business soon is probably a myth.
Shweta said that the field of AI is evolving and developing computational models that mirror the way the human brain is making decisions. In fact, there are actually experiments that show that emotions help decisions and we can even program these emotions in to robots and see an enhancement in their performance. However, a lot of the testing and programming is in very simple environments and it remains to be seen if we can actually extend these to the complex environment and life as we know it. Therefore, Shweta thinks that in the short term, the value is in getting the humans to work alongside machines or together with machines. The real trick is to figure out how this collaboration works and how humans biases don’t come in the way of this collaboration, so that we can actually get the best of the two species.
Magda Osman – ‘Our sticky two-minded syndrome: what is the cure?’
Interview starts at 35min 48s
Our final guest of the day was Magda Osman, a reader in experimental cognitive psychology at Queen Mary University of London. Magda’s main research interests concern understanding the underlying mechanisms involved in learning decision making and problem solving in complex dynamic environments. Her closing keynote at the conference aimed to look at an alternative thinking to the popular two system of decision making outlined in Daniel Kahneman’s book ‘Thinking Fast and Slow’, where System 1 is fast instinctive and emotional and System 2 is slower, more deliberate and more logical.
Magda began our chat by saying that there are lots of different flavours of this two system type way of describing the way we think. Some people will attribute more sophisticated mechanisms to System 1. She said that the Kahneman approach is one of many, but overall people using it are making the distinction which is, there is one system which is automatic and fast on some level and another, which is slow, potentially more rational and often these two systems don’t actually speak to each other. So, the way we intuitively make sense of what we do is that we feel like something comes to mind very quickly and then if we actually spent more time reflecting on that, we might end up coming up with a different decision to the first one.
Magda can see the appeal of thinking about the way we make decisions and judgements and reason along those lines, but her talk was about saying, while it has intuitive appeal and we tell ourselves stories about the fact that our mind is divided into something that’s rational and irrational, emotional and logical, and so on, the theoretical landscape, at least in the literature, which she said spans psychology, behavioural economics, neuroscience, as well as the empirical evidence, over a quite a long period of time now, seems to challenge and conflict with these very general kinds of claims. Therefore, if we took these claims to their logical conclusion, what are they implying? She therefore aimed to look at some of the evidence base to see how well supported some of these claims are and ask what other better methods are there or frameworks for understanding the mind.
A good everyday example that Magda gave was by relating it to something that everyone does have some experience of, i.e., driving, where if you are well practised in it and you do it every day, you often find people say ‘I got to the the end of the route and I felt like I did this automatically. I have no idea how I got there.‘ and then ‘I hadn’t even intended to get to that point I wanted to go somewhere else’. Magda said that this is a really good example of automatic processing. She added that if you look into the experimental literature, what that involves is getting people in driving simulators in the lab and training them up to drive a particular route so that it’s well practised and highly familiar. Then, when you get them to actually do something else at the same time, something quite cognitively effortful, which mimics some of the things that we sometimes do while we’re driving as often we’re not just driving but doing other things even though we aren’t necessarily supposed to do that, the empirical findings will tell you that actually when this happens, they can’t do both things well enough that the performance of both is optimal. So something has to give. Often, what you find is that if you if you engage people in a cognitively effortful task and you monitor how they’re driving the route, they increase their chances of crashing. Magda said that what that tells you is that while we feel like we’re not actually paying attention consciously to the things that we’re doing even where we’re doing something that’s credibly familiar to us, we actually need to spend some portion of our conscious cognition monitoring the environment, tracking our speed, checking what’s going on around the traffic and so on, and you can’t do that without attending consciously to the environment.
Relating this to the business, Magda shared her experiences of working on secondment to the government and said that one of the interesting issues she looks at is how we raise compliance of business. Generally businesses are compliant but how do we increase that and what methods can we use? She said that one useful approach is rather than saying businesses reviews have a System 1 approach, the alternative is to break down and characterise the processes that are involved – what are their main incentives for compliance and non-compliance, how are they incentivised in ways personally that conflict with regulation?
Magda said that the motives of the enforcer and what they consider to be of interest to incentivise business don’t perfectly align with the way businesses operate and so it makes sense to try to deconstruct and characterise on a much more robust level how businesses operate, what kind of risk appetite they have, how that interacts with their incentive structures and their value systems in order to build up a better understanding of what their main operations are to then figure out ways of how to target compliance and raise that. So for example, do we need to inform them more or do we need to train them more? Do we need to make them aware of the consequences more? All of that can be understood from frameworks that characterise decision making from a value based approach, which MAgda said have nothing to do with making the very simple division of it’s fast or slow or unconscious or conscious or automatic, etc.
Her point was that this is a kind of framework is a very convenient way of trying to characterise biased decision making, but when you understand the conditions under which decisions seem to look as if they’re biased, there’s a lot more going on than just something that is automatically kicking in. In order to be able to try to debias decision making processes, you need to characterise in more detail and profile the context in which people make decisions and for Magda, that’s not something that these dual process type accounts say much about – there are other alternative models which do a better job of that.
For more information on this topic, visit Magda’s website.
For more information from CFA UK, visit www.cfauk.org
Finally, if you subscribe to the show, please can you give it a positive rating and review on iTunes in particular as this helps it up the charts!