This is a brilliant work by the famous Lebanese mathematical statistician that deals with highly improbable but deeply affecting events. These are the Black Swan events, sweeping changes in an individual’s life or society, that are impossible to predict. The 9/11 episode was one such example. The author gives a clear example of a turkey that has had a good life until the day of butchering. Till the last day, the turkey might simply extrapolate that life is good, full of almonds and leisure. The final day is a complete contrast to all predictions based on an average life story! This is a Black Swan event.
Of course, there is a certain subjective element to the idea of Black Swan events, says the author. The Black Swan event for turkeys is certainly not one for the butcher, just as 9/11 was not one for the deadly terrorists. Human beings experience Black Swan events in many areas, especially finance and stock markets, but we continuously try to give explanations when none is possible. However, humans have an amazing ability to indulge in retrospective explanations for the Black Swan events. We provide an explanation after the event has occurred, a veiled attempt to conceal our incapacity to acknowledge the significant role of ‘unknown unknowns’ in our lives. The author warns that our obsession with causes and effects, trying to fit in some simplistic and reductionist explanations, is foolish and futile.
The author’s main idea is that we cannot do anything to stop the Black Swan events, but we can plan to cope with them when they happen. Investing in securities is one example. We cannot predict the performance of stocks, and we cannot predict how the markets may collapse suddenly. In a world where the financial markets and banks form a deeply interconnected web, a single event can lead to a huge-scale crash. It is the nature of today’s markets. It would be a good strategy to invest 85–90% of the money in low-risk instruments and leave 10% only for high-risk ventures, and even the latter should spread out. This would blunt the impact of a Black Swan event.
The Illusion of History and Modelling
He deals with the human obsession with history. The latter is opaque in terms of cause and effect. We can simply see events come and go in the past, but we can never assign reasons for them. The biggest blunder a historian might commit is to assign retrospective reasons for historical events. Each event is a complex interplay of thousands of factors coming together, and in such complexity, making simplistic explanations is a foolish task. The author rues that the human mind suffers from three ailments as it meets history, which he calls the ‘triplet of opacity.’
- The illusion of understanding, or the way everyone perceives what is happening in a world that is more complex than realized.
- The retrospective distortion, or how we assess it, matters only after the fact, as if there is a rearview mirror.
- The overvaluation of information and the handicap of authoritative and learned people, particularly when they create categories.
The author then deconstructs modelling’s role in making predictions. He claims emphatically that models do not work in real world, especially when related to finance and economics. A top-down approach with force fitting of data sometimes is the hallmark of such financial and economic models. Data needs a model, and it is not the other way down. The major thrust of statisticians is to rely on models to explain complex social behavior. In many areas, models do not work, but again it is the human need to have simplistic explanations for complex phenomena. Gaussian statistics with the bell curve of normality is a major problem in this regard. This is perhaps the cornerstone of all statistical learning, particularly the medical statistics for doctors.
The Normal Curve and Gaussian Statistics
The normal curve in statistics is a probabilistic distribution of a variable. It has a mean around which most values in the sample population would hover. It has also a measure called the standard deviation, and the major rule of Gaussian statistics is that 68%, 95%, and 99% of observations in the sample lie within one, two, and three standard deviations of the mean. Taleb writes that this is mostly all about it and has dominated the world of statistics and design of various models.
The problem is that when it comes to Black Swan events, they fail. The scaling model, which uses power laws to explain events, would be a more effective model. In simple English, scaling implies that things improve or degenerate at a fixed ratio as the size increases. As the size of the city increases to double; crime, unemployment, patents, infrastructure does not double but increases at a fixed ratio. Hence, the increase or decrease in odds is unrelated to the absolute size, but the relative size of the new population as compared to the earlier population. These are where the Black Swan events lie. However, the major problem is to understand what those ratios and power laws are. They exist, but we do not always know what the values are.
As an example, if in a population, the number of people earning a million is 5%, the number earning two million would become 2.5%. According to Gaussian statistics, the proportion of people would drop from 5% to an insignificant amount, or a range beyond three standard deviations from the mean. This is clearly not true in the real world, where millionaires and incomes follow power laws and scaling, and not Gaussian models. Hence, Gaussian normal curves (the author characteristically terms this as GIF-great intellectual fraud!) work in areas called ‘Mediocristan’ or the ordinary world where extremes do not affect. Height of an individual, weight, shoe size are such examples. However, in a world of ‘Extremistan’, where Black Swan events occur and make sweeping changes, power laws and scaling apply more. But as said before, we do not know the exact values of many of these power scales.
Mediocristan And Extremistan
‘Mediocristan’ and ‘Extremistan’- terms to describe events- form the core of the book and he takes pains to explain them in detail. Mediocristan events are non-scalable, and the randomness or deviations from the average is mild. The most typical member is a mediocre, and extreme deviations are very rare (events beyond 3 standard deviations). Our ancestral environment likely harbored them more frequently. They are impervious to Black Swan events and the events are plottable on a Gaussian bell curve. These events allow us to come to terms with their predictability and impact. These events are associated with physical quantities like height, weight etc.; and an important caveat of these events is that a single instance or observation does not impact the total.
In contrast, Extremistan events are scalable and do not confirm to Gaussian bell curves. The deviations are wild, and the most typical member is an extreme value, either a giant or a dwarf. The extreme events will determine the total, and there are no constraints on what a number can be. These events are also more likely in our modern environments. It takes a long time to know what is truly going on. In contrast to the Mediocristan events where there is tyranny of the collective, here there is tyranny of the accidental. The past information is completely inadequate to predict the Extremistan events. Market crashes, bank collapses, twin-towers falling come to mind as examples of this world. The world of economics (despite elegant models, equations, and formulae of many Nobel laureates) is Extremistan on which the author focusses the most. History makes jumps of unexpected magnitude in Extremistan as compared to Mediocristan where it ‘crawls.’
Complex domains are characterized by interdependence between its elements or variables temporally (past changes), horizontally (variables depending on one another), and diagonally (one variable depending on the past changes of another variable). Here, positive, and reinforcing feedback loops in the linked mechanisms causes ‘fat tails’, characteristic of the Extremistan events. Complex domains are impossible to predict, but economists do not simply stop. The author goes especially for the jugular of many Nobel prize winners in economics to show how they failed miserably when it came to the real world of finance. Their models did nothing to predict or salvage huge market and banking collapses despite their personal income not getting dented. He feels particularly angry when he says that economic policies pursued by governments is the worst of both socialism and capitalism. Strangely, there has been a socialization of losses and privatization of gains when it should be the other way around. He takes his ire against financial planners who have no dents in their own earnings despite bad decisions. He is especially angry at the bailouts given by taxpayers when companies sink due to faulty planning. He rues their inability to deal with a Black Swan world.
Decision Making
While making decisions, there are two types of exposures one need to consider. There are ‘binary’ exposures: true or false; yes or no; medical results for a single individual; life or death outcomes; the outcome of a soccer game bet; and so on. Binary outcomes are not very common in real life existing mostly in research labs. These events do not depend on high-impact events with their limited payoff. The author gives an example (which can perhaps come from him alone). Either one is pregnant or not. One cannot be very pregnant or very non-pregnant. The idea is that ‘very true’ does not have any added advantage over true.
The second type of decision is more complex with open-ended exposures and variable payoffs. Here apart from the probability of event, one cares about the impact as well. A war or an epidemic can be mild or severe. In investments, one does not care about the number of times of wins or losses, but more important is the product of the number of times multiplied by the amount made or lost. This sum is more significant. These are complex domains where the impact factor has also to included.
Taleb states that when one looks at the generator of events, one can tell a priori which events can deliver large deviations (Extremistan) and which environment cannot (Mediocristan). This is the only assumption one needs to make while dealing with predictability. He creates a four-quadrant map based on the type of decision and the event generator.
Mediocristan’s simple binary payoffs are listed in the first quadrant. Forecasting is safe, models work, but these work in laboratories and games. They are rare in real life with examples of medical decisions involving a single patient or casino bets or results of a soccer game. Payoffs in economic decision making are rare in this quadrant. In Mediocristan, complex payoffs form the second quadrant. Statistical methods work here with some risks though.
In Extremistan, simple payoffs form the third quadrant. There is little harm in being wrong here because extreme events do not impact the payoffs. In the final fourth quadrant, the complex payoffs in Extremistan mark the Black Swan domain. This true problem domain encompasses most of the economic world. Unfortunately, it is also the world of the unknown unknowns because many times we do not know about the nature of Black Swans too. Payoffs undoubtedly are extremely complex to predict in this quadrant.
Dealing With Black Swan Events
So how does one deal with the Black Swan domain? He lists some ideas. The first is to have respect for time. Things which have worked for a long time are preferable. Short-term windows do not work when long-term impacts are more important. Second, redundancy is vital in investments rather than optimization. Money in savings and even cash under the mattress can spread the impact of a Black Swan. There should be avoidance of prediction of small probability payoffs. One should be extremely suspicious of statements regarding ‘atypicality’ of remote events. Predictions and conventional risk metrics based on Gaussian curves are well-suited to the Mediocristan world. And finally, the most important caveat, do not confuse absence of volatility with absence of risk. Absence of evidence does not imply evidence of absence. This mostly sums up the book.
A good part of the book is to denounce the Gaussian statisticians and the economists who base their models on them. His language is sometimes blunt and also indulges in a bit of personal mudslinging. Perhaps, one can still make a point without bringing other persons down. But this is his unique style of passionate writing. There is a vicious slaughtering of a few Nobel economists. But he is rich; he has plenty of important things to say; he is right (or rather, not proven wrong); and people love a blunt style- a combination allowing him to sit high up on a tree and throw stones at people. The book does come out a bit rambling, and one needs constant focus to understand the truly relevant points, and to sift the useful from occasional verbiage.
Rarity, extreme impact, and retrospective predictability are the hallmarks of Black Swan events. The author says, “A small number of Black Swan explain almost everything in the world, from the success of ideas and religions to the dynamics of historical events, to elements of our personal lives. Ever since we left Pleistocene, some ten millennia ago, the effect of these Black Swans has been increasing. The effect began to accelerate during the industrial revolution, coinciding with the world’s increasing complexity. Meanwhile, ordinary events, which we study and attempt to predict through newspaper reading, have become increasingly inconsequential. We need to adjust to their unpredictable nature rather than naively trying to predict.”
The fun-style writing is a mask to some serious talk regarding the subject; and the rambling tone of the author should not fool the reader. There is a serious discussion of critical issues, and it is, to say the least, a most enlightening and thought-provoking book. But one does have to sift.
Comments