Behavioural Finance, Black Swans, and Risk Warning Systems
Nick Bullman – Managing Partner CheckRisk
Richard Fairchild, School of Management, University of Bath
In standard textbook finance, academics make the assumption that financial market actors (managers, investors, institutions) are fully rational, unbiased, non-psychological, unemotional, all-calculating perfect maximisers of expected utility or expected profits (homo economicus). Financial Market Risk is seen as a knowable variable, following a well-defined distribution (usually assumed as a well-behaved, bell-shaped, normal distribution). For example, these assumptions are the bedrock of Modern Portfolio theory (MPT: Markowitz 1952) in which fully rational risk-averse investors hold well-diversified portfolios of shares (all investors holding the same market portfolio) in which every company’s specific risks have been eliminated, and only macro-economic (market) risk remains.
MPT decision-making requires all investors to coldly and unemotionally calculate the amounts of each share to hold in the market portfolio, according to the Markowitz mathematical calculus of each share’s expected returns, variances, and, furthermore, co-variances and correlations between each share in the investing universe. In addition to requiring rational investors to possess supreme mathematical ability and deep market knowledge, MPT assumes that stock returns follow a well-defined and fully-knowable normal distribution.
Another major model arising from standard finance is that of the Efficient Market Hypothesis (EMH: Fama 1970), in which the actions of fully rational market participants result in stock market prices instantaneously and fully reflecting all available information. That is, according to EMH, market prices are always very close to companies’ true fundamental values. In the EMH, prices follow a random walk (and are hence unpredictable), and, due to prices following a random walk, EMH does not allow for market trends, such as momentum and reversals, or bubbles and crashes.
The Real World: Behavioural and Emotional Finance
In the real world, financial market actors (as imperfect human beings) are not fully rational. Furthermore, they are emotional, and subject to psychological biases. This has led to an emerging area of research, Behavioural Finance (BF), that incorporates investor (and manager) psychology and emotions into the standard finance models to understand the effects. More recently, an exciting emerging area of study, emotional finance (EF), incorporates investors’ unconscious biases, child-like attachment, and feelings of love and hate for their investments, in a Freudian psycho-analytical framework.
In the context of (the predictability) of financial market risk, what behavioural and emotional finance bring is an understanding that investor psychology, herding behaviour, emotions (both conscious and unconscious) can result in patterns in the markets, and can explain bubbles and crashes, as investor sentiment and herding behaviour, psychology and emotions can turn at crucial tipping-point moments. For example, in a BF framework, Barberis et al (1998) demonstrate theoretically how an investor’s bounded rationality and psychological biases can result in momentum and reversal in financial markets. Similarly, Daniel et al (1998) provide a theoretical model in which investors’ overconfidence and self-attribution bias can result in market under-reaction to good news and over-reaction to bad news.
In the emotional finance framework, Tuckett and Taffler (2008) conceptualise, in a Freudian psychoanalytical framework, how investors’ unconscious emotions (of love and hate) towards their investments can shift over time, creating herding behaviour, and bubbles and crashes. Fairchild (2009; 2014) provides the first rigorous theoretical modelling attempt.
Although behavioural finance, and particularly emotional finance, suggests that there may be patterns of behaviour in the stock market, does this mean that these patterns are predictable? Shelia Dow (2009) argues that the emotional finance framework provides an ex-post justification of historic bubbles and crashes (e.g. tulip mania, South-Sea Bubble, 1990’s Internet Bubble, the 2008 Financial Crisis), but cannot provide ex-ante predictability of the timing and time-frame of future bubbles.
Recall that, in standard finance, risk is knowable, and it is generally assumed that stock returns follow a well-behaved, bell-shaped, normal distribution. Note that, in this view of risk, investors know, ex-ante, all of the possible future states, or events, of the world (e.g. good economy, medium economy, bad economy etc…), all of the outcomes associated with those states, and the well-defined probabilities associated with those events. Hence, risk becomes predictable, using the normal distribution: e.g. we can analyse the probability of share prices falling by x%, using the normal distribution: this provides a basis for such risk-tools as Value-at-Risk (VaR): more on the shortcomings with VaR later.
In contrast, BF and EF recognise that in the real-world, stock returns do not follow a normal distribution. There is considerable evidence of skewness, fat-tails, kurtosis etc. Furthermore, BF and EF academics argue that investors may in fact be subject to (Knightian) uncertainty, rather than risk, in the financial markets. In Knightian Uncertainty, investors do not even know the distribution of returns, or the related outcomes. In this view, investors are making their decisions rather blindly, with little predictable information to work with.
Risk, Uncertainty and Black Swans
The BF/EF real-world view that financial market returns follow non-normal, skewed distributions, with fat-tails, and that risk may even be unpredictable, with investors instead facing Knightian Uncertainty, naturally leads us to consider the very nature of risk and uncertainty. A useful framework for us to consider is Black Swan Theory (Taleb, 2001; 2007; 2010), and Black Swan Events.
Most risk analysis, and risk appraisal tools such as value-at-risk, is based on the convenient but misguided assumption that event-probabilities follow a normal distribution: thus, for example, it is assumed that stock returns in the financial markets are randomly drawn from such a distribution. This distribution does not allow for prediction of ‘black swan’ events. These are dramatic extreme and sudden events, outside the prediction of the normal bell-curve. According to Wikipedia “The theory was developed by Nassim Nicholas Taleb to explain:
- The disproportionate role of high-profile, hard-to-predict, and rare events that are beyond the realm of normal expectations in history, science, finance, and technology.
- The non-computability of the probability of the consequential rare events using scientific methods (owing to the very nature of small probabilities).
- psychological biases that blind people, both individually and collectively, to uncertainty and to a rare event’s massive role in historical affairs.”
Taleb recommends not attempting to predict future black swan events, but to build robustness against their occurrence and provides “Ten Principles for a Black-Swan-Robust Society”, He also recommends the use of other types of distribution, such as fractal, power law, and scalable distributions in tempering expectations. He also recommends “the use of counterfactual reasoning when considering risk.” (This paragraph draws heavily from Wikipedia’s summary of Taleb’s book).
Black Swans: Predictions in Practice
Black Swan events are by definition, unpredictable. However, there is increasing evidence that indicates that the likelihood of a Black Swan occurring can be forecast. In other words, the nature of the event may remain unpredictable, but the probability of such an event happening is within our reach.
The problem with legacy risk systems such as Value at Risk (VaR) are well known and lies in the fact that they project the past forward to anticipate what will happen next. This results in a beautifully symmetrical bell curve that bears no reality to the real world. The tails of this bell curve severely underestimate risk and therefore so-called “fat tail risks” make the risk system virtually unusable as a forward-looking risk tool. Essentially the tail risks occur more frequently and with a larger scale than the model predicts.
Most importantly VaR fails to address the issue of Risk Clusters. It is counterintuitive, but it turns out that sometimes a risk event generates more risk. Usually when a risk is known it may be dealt with, but sometimes a risk exposes weaknesses in whatever system is being considered, and unzips the system as a whole.
This unzipping effect CheckRisk has called “bridging” and bridging of risk most approximates Black Swan events. The charts below show how risk clusters. The study ran from January 2008 until October 2010. Using Table 2. it can be seen that VaR underestimated daily stock market returns over an 80 year period significantly.
Table 1: Daily S&P Returns January 1928 to October 2010
Sample number: 20,788
Standard Deviation (σ): 1.195%
Max Daily Return: 16.61%
Min Daily Return: -20.47%
Source: CheckRisk LLP
The number of days recorded was 20,788 and the maximum daily return was 16.61% and the worst -20.47%.
Table 2: Standard Deviation of Returns VaR versus Actual
Source: CheckRisk LLP
As can be viewed in the table, traditional Risk systems (VaR) only predicted one day where the market would fall -4.78% to -5.97% in fact that event occurred 43 times. The model predicted zero 5 standard deviation (SD) and 6SD days, whereas reality saw 18 and 22 of those events occurring over the period.
Compare Table 3 with Table 4. Table 3 uses VaR to predict what should happen over the measured period. Table 4 shows what happened. It is clear that risk clusters. There are benign low-risk periods, and there are high-risk cluster periods. This single fact has led CheckRisk on a quest to place risk at the centre of investment as opposed to a postscript. The central issue with investment risk is simply this; are you being paid to take the risk or not? CheckRisk believes it is possible to identify the risk regimes that answer that question.
Table 3: VaR prediction on Risk
Source: CheckRisk LLP
Table 4: What actually happened
Source: CheckRisk LLP
Given that VaR remains the mainstay of Central Bank, Commercial Banking, and Insurance Company modelling perhaps it is not surprising that the world remains so exposed to Black Swans. Using VaR alone we are doomed to be perpetually surprised by events that could otherwise be forecast.
To take a more realistic approach, it is necessary to cast the risk net far wider. Using modelling like Network Risk Analysis, Early Warning Risk Systems, Cascade modelling and Nowcasts it is possible to build a much more complex picture of the way risk spreads and to pick up changes in the risk environment with sufficient warning. Also advances in Behavioural Finance theory, particularly in the field of Emotional Finance, shows us that risk is not equal for everyone. Thus our perception of risk matters, and may well lead to some of the vagaries of how deep a correction or long a bull market run is, for example.
There will always be risk events that cannot be predicted. Real Black Swans, however, are rarer than the financial market currently believes. Most Black Swans are not true Black Swans; they can be predicted and the mechanisms and models to do so are already available.
For example, CheckRisk’s proprietary early warning risk system (CREWS) indicates that the period 2017-2020 as being a period of intensified financial market risk, with 2018 and 2019 of particular concern. Of course, Central Bankers can decide once again to extend and pretend, whether it be via helicopter money, more QE or direct investment, however, eventually the debt level consuming the world like a creeper will suffocate the very system it feeds off.
To prepare oneself for the imminent risks it is required to have a change of mind set. Regarding investment risk CheckRisk calls it “Investing for Risk” as opposed to investing for return. Putting the risk horse in front of the return cart has always made more sense to us. Secondly, throwing your risk net wider is an essential step in preparation. Thirdly, be prepared to use outside risk advisors or third parties, their perception of risk may be very different to your own. Fourthly, give freedom to a proper risk culture in your organization. Too often CheckRisk sees lip service but no real ownership of risk, as a consequence group risky shift leads to much greater risk being taken at a corporate level than would be taken as an individual.
Black Swan events are rare, and must have a true element of surprise to fit the standard definition. An undetected meteor strike that changes the world’s weather patterns would be a good example. Events like 9/11 would be another example, however, there is even debate as to whether such an event was predictable given the warning signs that could have been captured by big data.
Events like another global financial crisis, the break-up of the EU, an expansionary Russian policy toward the Ukraine and elsewhere, a Trump Presidency in the USA, China’s economy becoming unstable, a cyber attack or an extreme weather pattern are all examples of risk’s that financial commentators call Black Swans but in fact are not. Indeed, the risks above form part of the risk forecasting CheckRisk conducts on a daily basis.
For example, CheckRisk believes another global financial crisis is increasingly likely in the period 2017-2020. This can only be delayed or muted, but not avoided. Global debt has increased by over $60 trillion since 2008.
Unless the EU takes Brexit as a cause for radical reform of its mandate, the collapse of the EU and the Euro in its current form is also quite probable and predictable. The EU may survive but it will be radically different. At present, the reaction to Brexit has been ever closer union the opposite of what the people of Europe desire.
Russia under Putin will continue to be expansionary on a geopolitical scale. This is a risk but not a Black Swan event as Putin’s behaviour patterns are relatively predictable.
Trump may well be elected President. The US and other equity and bond markets around the world are underestimating this risk. However, CheckRisk believes there is a good possibility, near a 50/50 chance that Trump will win, particularly following Clinton’s recent health scares. Irrespective of what one might think about Donald Trump in the White House, markets have underestimated the probability in the same way that the Remain camp did for Brexit.
China’s economy is so indebted that there are serious risks of a correction in financial markets. China would recover over time but the impact to the world economy would be serious. The end of a massive credit cycle is no Black Swan.
Cyber attack or solar electromagnetic events are less predictable but again entirely possible. A solar electromagnetic event would cause chaos in systems such as supermarket ordering systems. Governments believe that humanity in the Western World is just three meals from anarchy. Current resupply systems for the supermarkets operate on an 18-hour time scale. If those systems fail, it will not take long for basic law and order to fail too. It is, however, very difficult to plan for such events.
Extreme weather events: rather than a storm or hurricane or earthquake CheckRisk is concerned about global water supplies which can lead to geopolitical shifts of significant proportion. China currently has very little clean potable water sources, and may eventually have to divert water currently used by their neighbour’s to satisfy their own people. Such action will unleash civil unrest in countries bordering southern China in particular.
The point of all of these risk events is that they are predictable and so are the tools to model the outcomes. Investors need to place risk in front of return, and begin “Investing for Risk” in doing so they will not only avoid the predictable risks but be better equipped for true Black Swans when they occur.
Barberis, N., A. Shleifer, and R. Vishny (1998). “A Model of Investor Sentiment.” Journal of Financial Economics 49: 307-343.
Daniel, K., D. Hirshleifer, and A. Subrahmanyam. (1998) “Investor Psychology and Security Market Under- and Overreactions.” Journal of Finance, vol 53, no 6. 1839-1884.
Dow, S. (2009). “The Psychology of Financial Markets: Keynes, Minsky, and Emotional Finance.” Book Chapter in “The Elgar Companion to Hyman Minsky.”
Fairchild, R. (2009). “From Behavioural to Emotional Corporate Finance: a New Research Direction.” SSRN Working Papers.
Fairchild, R. (2014). “Emotions in the Financial Market.” Book Chapter in “Investor Behavior: The Psychology of Financial Planning and Investing.” H. Kent Baker and Victor Ricciardi, editors, 347-364, Hoboken, NJ: John Wiley & Sons, Inc., 2014
Taffler, R., and D. Tuckett (2008): “Phantastic objects and the financial market’s sense of reality: A psychoanalytic contribution to the understanding of stock market instability”. International Journal of Psychoanalysis.
Taleb, N.N. (2001). “Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets.” New York: Random House.
Taleb, N.N. (2007; 2010): “The Black Swan: the Impact of the Highly Improbable.” Penguin books.