We have argued throughout Dance with Chance that it is easy to explain the past but much more difficult and challenging to predict the future. We would now like to consider this statement in light of how well our book predicted events after it was first published.1 In addition, we focus on a question frequently posed by readers and people attending presentations about our book, namely: what can be done to face financial crises and similar high impact events if we are unable to predict their occurrence? But first, we summarize our claim that we face serious limits to predictability in practically all areas of social sciences.
In the late 1960s, and most of the 1970s, social scientists – including business professors – hoped that computing technology and sophisticated mathematical models would lead to the same level of success in forecasting in the social sciences as that achieved in the physical sciences and engineering. For a variety of reasons, it is now clear that these hopes were unfounded. Instead, empirical evidence has shown:
• The future is never exactly the same as the past. This means that extrapolating patterns and relationships from the past to the future cannot provide accurate predictions.
• There are plenty of statistically sophisticated models that can fit (explain) past data almost perfectly. However, these complex models do not necessarily predict the future so well.
• Conversely, whereas simple statistical models do not fit the past very well, they generally do better at predicting the future than their complex counterparts.
• Empirical evidence has also shown that human judgment is even worse at predicting the future than statistical models.
• Both statistical models and human judgment have been unable to capture the full extent of future uncertainty. People have been often surprised by large forecasting errors and events (both negative and positive) they did not even consider.
• Experts do not predict more accurately than moderately well-informed, intelligent people in the street.
• On a brighter note, averaging the independent predictions of several individuals (whether experts or not) generally improves forecasting accuracy.
• What’s more, averaging forecasts based on more than one model also improves accuracy and reduces the size of errors.
Let’s look at some new developments in the areas of medicine, economics, business, and happiness that occurred after our book was first published that further underline the serious limits to predictability that we face in these domains.
In Chapters 2 and 3 we stressed that medicine is an inexact and evolving science, and that claims to the contrary exist not just because of our susceptibility to the illusion of control but also because of vested interests. Below we discuss cases involving “swine flu” and preventive medical tests to verify in an ex-post manner our claims of inexact science and vested interests.
Swine flu: The origin of swine flu has been traced to Veracruz, Mexico, at the end of March 2009. Not long after that time, there were concerns that a new flu virus had appeared that could lead to a pandemic with grave consequences involving many deaths (some estimates were in the millions). A heated discussion took place as to whether widespread vaccination of the general population was necessary. This continued for several months with many people advocating the pros and cons of large scale vaccination but with no consensus among the experts. At the same time, newspapers carried reports of the number of swine flu deaths and the dangers of not being vaccinated. On June 11, 2009 the World Health Organization (WHO) raised the influenza pandemic alert level from phase 5 to phase 6 (its most severe phase) and declared that “the virus was contagious, spreading easily from one person to another, and from one country to another.” WHO’s concerns were echoed by many medical experts who recommended vaccination to minimize the risk of the flu spreading at a fast pace throughout the world. Pharmaceutical firms rushed to produce large quantities of vaccines and a large number of countries ordered them by the millions to prepare for the pandemic’s arrival.
In early March 2010, the National Pandemic Flu Service was closed in the UK and countries that had bought millions of vaccines were trying to get rid of them (but they could not find any buyers). The number of confirmed world-wide deaths from the pandemic has been estimated at around 15,000. To put this figure in perspective, about 36,000 people die in a “normal” flu season in the USA alone according to the Centers for Disease Control and Prevention (CDC).
As expected the swine flu became a cash cow of the pharmaceutical industry with a handful of companies generating millions of dollars in revenue and making huge profits. But this was not the first time. In 1976 there were predictions of a flu pandemic and 46 million Americans were vaccinated. This pandemic also never happened. There was only one confirmed death but up to 4,000 people who had been inoculated fell seriously ill by contracting Guillain-Barre paralysis and some even died.2 In 2010, the Guillain-Barre paralysis of Alyson Dygnas was reported in the UK after she received a swine flu vaccine.3 Curiously, there was little mention of the possibility of side effects or of the 4,000 Americans who fell seriously ill or died during the long debate about the advantages and disadvantages of vaccination.
Preventive medical tests: In Chapters 2 and 3 we discussed the uncertain benefits of periodic tests for prostate, breast and other forms of cancer and the possible negative consequences of such tests. The debate about their value is still continuing among physicians and medical experts. In a recent blog in NBC,4 Shannon Brownlee explains screening as follows:
Screening allows us to look under the water, at the tumors that haven’t yet become symptomatic. We assume they will eventually cause symptoms, but increasing evidence suggests that’s not always the case. Evidence from autopsies, for instance: In one study, postmortem exams showed that nearly 9 percent of women of all ages who died of any cause other than breast cancer had undiagnosed DCIS (Ductal Carcinoma In Situ, the most common type of noninvasive breast cancer). Among women from Denmark, where mammography is not as common as it is here, a whopping 39 percent of middle-aged women who died of other causes had undetected breast cancers. Similarly, says outcomes researcher Dr. Welch, a 1989 study found that 60 percent of men over age 60 have undetected prostate cancer – yet only about 3 percent of deaths in men are due to prostate cancer.
Brownlee then refers to the case of Dennis Fryback, Ph.D., a former member of the US Preventive Services Task Force, a group of experts convened by the federal government to make recommendations about screening. The task force recommends colonoscopy every ten years for people between the ages of 50 and 75, yet the 61-year-old Fryback concluded it did not make sense for him to get screened. He explains that he came to that decision in part because he has no family history of colon cancer. If he did, his chances of getting it would increase, and so would the odds he’d benefit from the test. He also knows that getting the exam requires at least a day of taking laxatives to clean out the colon and then facing the possibility of a perforation from the procedure, a risk that goes up with age. He balanced the possible reduction in his chances of dying of colon cancer against his other health problems. He had a heart attack in the previous year and suspects he will die of heart disease before a colon polyp has a chance to kill him. Clearly, even for experts the decision whether or not to screen is not straightforward, though for self-serving interests many doctors recommend screening for all.
As the arguments for and against testing continue, they leave many confused.5 As a compromise, the new recommendation is to test later (e.g., for breast cancer after 50 instead of 40 years of age) and less often (i.e., every two years instead of annually). Clearly the vested interests are huge. Sharon Begley in a Newsweek article entitled “This won’t hurt a bit: How we can save billions by cutting out unnecessary procedures that kill tens of thousands a year”6 references several studies showing that the huge cost of unnecessary treatment totals hundreds of billions and provides no benefits for patients but brings riches to doctors, hospitals, and pharmaceutical firms.
Given the huge and rising cost of medical treatment and the unsustainable financial burden on individuals and nations, the time has come to re-examine the costs and benefits of different medical practices in a realistic and objective way.
Global wealth fell considerably between October 2007 and March 2009 when the world was hit by the most severe recession since the 1930s. In addition to huge stock market losses, the number of bankruptcies increased exponentially, the real estate market disintegrated and the unemployment rate reached double digits in many countries. Interestingly, practically no one predicted the financial tsunami that hit almost all nations of the world with great force. Even today, when the worst of the financial crisis is over, there are no convincing explanations of the underlying causes, who should be blamed, and the effectiveness of the measures taken by governments and central banks to deal with the catastrophic consequences.
Then, in the midst of the gloom when many economists, including the head of the IMF, were predicting a forthcoming depression, things suddenly started to improve. Within fourteen months from March 2009 when the market was falling rapidly, instead of the Dow Jones Industrial Average (DJIA) dropping to 5,000 as had been predicted by some, it increased to above 11,000. As with the strong fall in the market, few predicted the even stronger recovery that increased stock market valuations more than 100% in many emerging markets (e.g., the Russian stock market increased more than 150%). This means that those who stayed out of the stock market missed the greatest boom in history as, even in advanced economies, the values of the stock markets grew between 60% and 80% from March 9, 2009 to a little more than that a year later. The prophets of doom who had been hailed for predicting the recession were unable to explain how they lost their prophetic powers and completely missed the recovery. Lately, Bloomberg and the other financial networks have stopped asking the opinions of those predicting catastrophe for the economic system while the bulls have taken a more prominent role.
The unpredictability of both the fall and subsequent rise of the stock market prove, one more time, our inability to forecast and our overreaction to both negative and positive developments. Has such overreaction ended? What will be the effects of sovereign debt, in particular among European countries?Will the market increase or decrease in the rest of 2010? Should an investor be buying or selling stocks? That nobody can answer these questions, and that leading experts can have diametrically opposing views, is illustrated by a recent debate in the pages of the Wall Street Journal (WSJ).7
Shiller vs. Siegel: Professors Shiller and Siegel are two prominent economists we referred to in Chapter 4 and, as noted there, authors of two best selling books about stock market prices. A summary of their views is given in the WSJ as follows:
Mr. Shiller worries that the housing market could be turning down after a brief recovery, which could contribute to a decline in U.S. stocks, which already look expensive to him. “I wonder about a return to another break in the market,” he says, though he notes the market is far less expensive today than when he wrote his book.
Mr. Siegel scoffs at his friend’s concerns – and at his numbers. “This is an extremely cheap market,” he says, and its future is bright . . .
The way Mr. Shiller sees it, the problem today isn’t just that the current P/E is above 20. It is that since 1991 it has spent only seven months, in late 2008 and early 2009, below the average level of 16. At the start of 2000, it was above 40. No one can say how much longer the P/E can keep rising or when the past year’s bull market might end, especially with the government providing heavy stimulus. But past trends, and the law of averages, suggest that at some point the P/E is likely to fall below 16, pulling stocks with it. Mr. Siegel, for his part, strongly disagrees with this kind of analysis. He argues that his friend’s use of 10-year average profits works poorly in the current environment, because big financial companies took such heavy write-offs in 2008 and 2009. Such write-offs, he argues, won’t be repeated, so earnings including them shouldn’t be used for forecasting. . . . “My research shows that the common P/E is 18.5” when the economy is coming out of a recession, he says. The way he looks at it, the market now is trading at about 14.5 times forecast 2010 profits, making it cheap compared with the typical P/E of 18.5. If stocks rise to 18.5 times profits, the S&P 500 could rise to 1400 this year, a 23% gain from today’s level, he notes. “We could easily see 10% to 12% stock returns with low inflation” in future years, he says.
What can a layman say when two of the best world experts, without vested interests, disagree about the direction of the stock market in the USA, the biggest and the most stable in the world? Is there any point in trying to forecast the direction of the market? In addition to whether stocks in the USA are overvalued or not, what about valuations in emerging markets where some have experienced triple digit growth in less than a year? Will advanced countries be able to reduce their huge government debt without halting economic growth? What will be the effects of possible double digit inflation similar to that of the late 1970s and early 1980s? What about stagflation? What about the dangers of a real estate bubble in China where properties values have been increasing at a fast pace? What about a “W” recession? Equally important, what is the value of experts when they provide diametrically opposing advice? Whatever the answer to these questions, one thing is clear: the uncertainty in forecasting is huge and we must not fall prey to the illusion of control that fosters predictability and certainty. Perhaps in the long run the market will increase substantially, but this may take a decade or two. So, as we stated in Chapters 4 and 5, it is up to the individual investor to consider the pros and cons of investing in today’s market and the specific types of securities and risk he or she is willing to assume. But remember, there is no such a thing as a free lunch, and no one can accurately forecast the future.
On July 6, 2009, after many years of financial and other troubles, GM was finally officially placed under bankruptcy protection. It would have been inconceivable 40 years ago that the icon of American business would be overcome by what was at the time a small local firm, Toyota Motors, and that it would eventually end in bankruptcy. Yet, one more time, the inconceivable happened: GM went under chapter 11 protection and Toyota became the undisputable world automobile leader. Its rise to fame was further sealed in 2009 when it was ranked number 3, just behind Apple and Berkshire Finance, in Fortune’s annual survey of the most admired companies in the world.
Toyota has been legendary for its perfect quality, attention to detail and for building a culture among its workforce to assure buyers that its cars will be trouble free. Toyota’s manufacturing has been the benchmark to follow while the “Toyota Way” of management has been the source of many cases written and taught in business schools and innumerable articles and books. For example, a search in Amazon for books about “Toyota” lists 4,000 while a similar search for “Toyota Way” shows 249 books (the corresponding number of entries in a Google search is 153 and 48 million). Toyota was until recently the modern icon of the super successful global firm unsurpassed in quality and service for its customers.
Yet during the last two years, Toyota has fallen from grace. It has suffered a large reduction in its market share and a loss of $5.5 billion for the fiscal year ending in March 2010. Uncontrollable acceleration in some of its cars, problems with the brakes of others, some design faults and defects in construction have allegedly caused deaths and serious injuries to drivers and passengers. At present, there are thousands of individual and class action law suits against Toyota that threaten its financial stability. (Moody’s cited litigation risks when it warned in February 2010 that it might downgrade Toyota’s credit ratings.) Worst still, its reputation with consumers has been seriously damaged as up to ten million vehicles have or will be recalled worldwide. No more can its cars be considered of the best quality vis-à-vis those of its competitors.
In Fortune’s 2010 list of the most admired companies, Toyota fell to the seventh position from third in 2009. In all likelihood, it will not be part of the top ten in 2011. Once again, we observe that outstanding firms of the past can fall into serious financial and operational problems. In fact, Toyota reminds us of Dell mentioned in Chapter 6. Dell was the most admired company in Fortune’s ranking in 2005, fell to the eighth position in 2006 and then it slid out of the 2007, 2008, and 2009 surveys altogether. A massive battery recall in its laptop computers tarnished its high quality reputation, making it just another computer company. By the middle of 2010, Dell has still not recovered. It seems that a first rate reputation built over decades can be lost overnight. This fact should be taken seriously by business school professors and business gurus praising certain companies for their unique achievements and recommending that others should imitate them in order to succeed. Past success is no guarantee for the future.
Does success breed its own failure? The answer according to Toyota’s critics is yes. The firm’s managing team started believing their own PR slogans. They were the best, superior to all others, and nothing they did could be wrong. They never accepted, for instance, that the unrelenting cost reduction demanded of their suppliers could affect the quality of their vehicles or that other car manufacturers would catch up with them in terms of quality and manufacturing performance. Their verdict is that Toyota’s problems were inevitable. Others criticize Toyota’s management for its unwillingness to accept that there were problems with its cars although some of these problems had been known to them for more than a decade.
Our purpose here is not to analyze how Toyota fell from grace but to re-emphasize that it is difficult to predict the future. As we showed in Chapters 6 and 7, Toyota is not an exception and the Toyota Way is just another “method” that may have been useful in the past but carries no guarantee of success in the light of changing environmental, business and management conditions. Will Toyota follow the path of GM? Well, whereas we are the first to say that accurate forecasting is not possible, the possibility always exists because success breeds its own failure and creative destruction never ends.
The financial crisis has created a phenomenal degree of hardship. In addition to those who lost huge amounts of money in stock markets, large numbers of people defaulted on their homes, many more were fired from their jobs and have remained unemployed for long periods of time, and still others saw their incomes falling substantially. Under these circumstances the dissatisfaction, if not unhappiness among those affected increased. As we saw in Chapter 13, once people pass a certain level of income, they soon get used to extra money with no corresponding increase in their happiness. However, it probably takes much longer to get used to income decreases and more so when such decreases bring people’s income to levels below the minimum required for life’s necessities. There are interesting implications in the asymmetry between income increases/decreases and life satisfaction or happiness that needs to be further explored.
Recent research has shown that more important than money is the ranked position of people’s income within a comparison group.8 Such position predicts general life satisfaction much better than absolute income and reference income which have no effect. At the same time, individuals weight upward comparisons more heavily than downward comparisons. Thus, increases in individuals’ incomes will increase their utility only if their ranked positions also increase. However, this will reduce the utility of others who will lose rank. This explains (partially) why people are outraged at the huge salaries and bonuses of the golden boys (and women) in banks and financial institutions who make millions at a time when the income of most people is reduced. This is especially the case when these people believe that the banks are responsible for the financial crisis.
All this suggests that it is worth exploring alternate models of “well being” and progress in societies rather than those which are primarily anchored on economic numbers like GDP.
The serious limits to predictability in all the important areas of our lives raise vital issues for anyone making decisions. What is the best way of taking care of your health, wealth, and career, not to mention your happiness in a context of high uncertainty and possible futures that you cannot even imagine?
Physical scientists are generally very good at making predictions. But the scientific community is equally good at knowing its own limits. And earthquakes are a good example. Scientists accept that it’s impossible to predict the timing and location of earthquakes. Indeed, current understanding of the processes that produce them leaves no doubt that no one is able to pinpoint their occurrence or exact location in advance. Yet, the intensity and frequency of earthquakes exhibits a remarkably consistent pattern in the long run.
The point we’d like to make is that statistical regularity does not equal predictability. For example by studying the occurrences of past earthquakes statistically, we can predict that during the next 35 years they will be roughly 44 earthquakes with an intensity of 7.5 to 7.599 on the Richter scale. But seismologists have no clue as to when or where they’ll occur – apart from being in one of the world’s earthquake-prone zones and that they will be accompanied by aftershocks. Will these zones be populated or unpopulated? Will there be a tsunami after the earthquake? Will they cause large-scale death and destruction? No serious scientist pretends that he or she can answer these questions.
How, then, does the world cope with earthquakes? The answer is simple. Instead of relying on prediction, the strategy focuses on preparation. For example, engineers can and do construct buildings capable of withstanding very strong tremors, and quick-response emergency services are considered crucial. The recent earthquakes in Haiti and Chile underline this point. The earthquake in Haiti measured 7.0 on the Richter Scale, while the earthquake in Chile was significantly more powerful and measured 8.8 on the Richter Scale. Yet, the number of deaths in Haiti has been estimated to be as high as 230,000 while those in Chile has been estimated around 700. Both these earthquakes were unpredictable but two factors distinguish the situations. First, the earthquake in Haiti struck an area with a much higher population density than that in Chile. But, second, in Chile there was much greater awareness of the possibility of earthquakes and hence greater preparedness in terms of building codes and emergency services (and this despite accepting the inability to predict the timing and exact locations of earthquakes). Of course, the fact that Haiti is one of the poorest nations on earth made it difficult if not impossible to be well prepared. Nonetheless, the point remains that a strategy focused on preparation rather than prediction is key.
Hurricanes such as Katrina can wreak just as much devastation as big earthquakes. But even ordinary storms can cause significant damage. From a statistical point of view the number and intensity of hurricanes is pretty much constant over the long run. Like earthquakes, the number of storms and hurricanes in a given period decreases exponentially, the higher the wind speed. Yet unlike seismologists, meteorologists can usually predict where hurricanes will strike a few days in advance. Once they start developing, they can be tracked and their path forecast. If you’re out at sea with a safe harbor within easy reach, this is extremely useful. On land, however, it’s more like coping with an earthquake. The key is to be prepared: stay home, cover your windows, and fasten down. In some cases, there may be time and resources for a mass evacuation, as occurred in August 2008, when 1.8 million people were moved from the coast areas of South Louisiana. But Gustav only changed its course, thus once again highlighting the inaccuracy of predictions even when hurricanes can be monitored with computer age technology and high definition satellite images.
By analogy with natural disasters, think of the enormous number of small businesses or new ventures that start or fail worldwide. The precise figures vary from year to year, but there is a continual process of businesses entering and leaving the market – with some regions of the world more prone to both start-up and failure than others. Those companies that don’t depart early in their existence may go on to be hugely successful, while many more simply survive for decades. Sticking with the earthquake analogy, the small-business failures can be seen as minor movements of financial tectonics, while the Lehman Brothers, Enrons, and WorldComs are larger tremors. Clearly, the large 2007–2009 recession is the equivalent of a major earthquake or a catastrophic hurricane, shaking the global economy to its core and sending huge aftershocks rippling across the world as well as towering ruinous economic tsunamis. Yet we must be prepared to face the next financial crisis, or other unpredictable events that will certainly hit us again some time in the future, although when and with what intensity, we don’t know.
In terms of what to do, we remind readers of the “triple-A” approach (accept, assess, augment) presented in Chapter 10 as a way to help us never to forget that we live in an uncertain world and that we must take steps to face its uncertainties as realistically and effectively as possible.
This is the first and crucial step. Psychologically it’s tough, we know, but ignoring uncertainty is not an option. In fact, whether your interest is in tomorrow’s oil price, next quarter’s sales data, next year’s stock price, earthquakes, or simply getting to work on time, you can’t be realistic about assessing the chances of a given event occurring unless you first confront all the other possibilities that might come true instead.
Unlike in the physical sciences, where the scientists focus on “nature” minus the human beings, social scientists must necessarily include human beings in their models. And it is not difficult to imagine that reducing the possible behavior of human beings to some few universal laws as in Physics is simply too much of a blind leap of faith. In Chapter 10, we discussed two types of uncertainty: subway and coconut. Subway uncertainty is what we can model and hence assess fairly well. Payoff on a roulette wheel in a casino, demand for electricity in a given period of time in a major city, and number of phone calls in a given period of time in a network, are examples of uncertainty we can model from fairly well to extremely well. On the other hand, coconut uncertainty is something which is difficult if not impossible to model and hence difficult to assess accurately. One source of coconut uncertainty are the “unknown unknowns,” or what might be called Black Swans, which by definition are extremely rare events with disproportionately large consequences and are events that we can’t even imagine beforehand. However, a much more common source of coconut uncertainty is the class of events that are “known unknowns,” events we know can happen and will happen, events that are much more frequent than Black Swans, but which we can’t model or predict with any degree of accuracy. For example, will there be another asset bubble in the financial markets that will subsequently burst at some point? The answer is “almost surely yes” but we don’t know when this will happen and in what form. All we can predict is that it will happen fairly frequently and to different degrees. In fact, Alan Greenspan, who was the Chairman of the Federal Reserve of the United States from 1987 to 2006, remarked in the Financial Times on March 27, 2009 that “we have never been able to model successfully the transition from euphoria to fear,” highlighting the difficulty of predicting when a bubble in the financial markets forms and then bursts.
The overall uncertainty in most real life situations is a mixture of subway and coconut uncertainties. And in the socio-economic domains, compared to most physical sciences, the coconut uncertainty is a much greater proportion of the overall uncertainty, thus making predictions much more difficult. This is what we have to accept.
In the assessment of uncertainty in a given situation, first deal with the subway uncertainty which can be measured. Evidence has shown that in such an effort simple statistical models can be very beneficial and do much better than human judgment. This brings out another point that might be misunderstood in some of our discussions. All models are not incorrect or useless, but are very often incomplete, as these tackle only the subway uncertainty in a given situation. The models then have to be complemented by the possibility of coconuts in order to get a more realistic level of overall uncertainty, by using judgments.
Consider the staffing decision for an emergency ward at a hospital. It is fairly easy to model the uncertainty regarding the number of incoming patients on a “normal day”, without any extreme events. This would be subway uncertainty. This however must be complemented by the possibility of coconuts, such as a big traffic accident, a major fire or an earthquake, a start of an epidemic, and the like, which can’t be predicted accurately but which can’t be ignored either. The best response to such uncertainty is not building huge amounts of overcapacity, but includes taking steps to create contingent staff who can be called to duty in a short time, creating pools of multi-skilled staff who can perform emergency procedures in addition to their normal tasks, increasing staff during the snow seasons when more accidents are likely, and so on.
Take the sales of a first novel by an unknown author. It sounds like a unique case. But our suggestion to publishers is to ignore the uniqueness. Instead, look at the track record of the sales of first-time authors in general. You have no valid reason to believe that the uncertainty surrounding your new author differs from the wider population of new authors to which he or she belongs – especially if you’ve used an industry-standard process for collecting reader feedback (also known as human judgment). By now, you should have a reasonable estimate of just how low or high the sales might go. This range probably covers 95% of all possible outcomes. Done that? Well now take the estimated range . . . and increase it! Hence the next step: augment.
You can be sure that you’ve just underestimated the range of uncertainty you estimated, no matter how realistic you thought you were when you assessed the uncertainty. Extensive empirical evidence shows that people consistently underestimate uncertainty – perhaps because their powers of imagination are usually less than their powers of mathematics. As a rule of thumb, we suggest that you double the range of uncertainty estimated at the second stage above. If this seems excessive to you, we suggest that you collect data about forecasts made in your own organization. We think you will be quite surprised!
To sum up, we must learn to live with uncertainty in most domains of our lives. Falling into the trap of the illusion of control and underestimating uncertainty has very high potential costs. Of course, we must continue efforts in trying to predict the future more accurately, as this will always be an integral part of human nature. However, in doing so, we should not lose sight of the fact that we can never eliminate the uncertainty completely and in fact in most socio-economic domains the level of uncertainty will remain persistently high. The implication is that we should shift resources from being focused on prediction to being prepared for the unexpected, building resilience to live through negative events, and at the same time being nimble enough to leverage unexpected good luck.