by David Orrell
Most information, after all, has a somewhat ambiguous effect on the market. If the U.S. dollar falls in value, this has one impact on oil producers, another on the tourism industry, another on bakers, and so on, so the net effect in an interconnected world is hard to know to complete accuracy. The price of an asset corresponds to a balance struck in a battle between two opposing, almost animalistic, forces: buyers and sellers, bulls and bears. The reaction of different investors to news will depend on their own subjective interpretation of events, and “contradictory opinions about these variations are so evenly divided that at the same instant buyers expect a rise and sellers a fall.” Therefore, not only is the market subject to random external effects, but its own reaction to that news will also to some degree be random.
Furthermore, Bachelier pointed out that the exchange was involved in a kind of narcissistic dance with itself. Because every financial transaction involves a prediction of the future, this means that a speculator on the Bourse cares less about a sober appraisal of an asset’s worth than he does about the opinions of his colleagues. His aim is to evaluate how much the market is willing to pay at some time in the future. He doesn’t mind overpaying, if he thinks that a greater fool will over-overpay the next day.
Bachelier therefore concluded that movements in the exchange were essentially random. Any connection between causes and effects was too obscure for a human being to comprehend. As he wrote at the beginning of his thesis, “The factors that determine activity on the Exchange are innumerable, with events, current or expected, often bearing no relation to price variation.” Mathematical forecasting was therefore impossible. However, he then made a point that underpins much of modern economic theory, which is that one could “establish the laws of probability for price variation that the market at that instant dictates.” To accomplish this, he assumed in his calculations that market prices followed the normal distribution, which seemed reasonable given its popularity in the physical sciences.
The theory implied that there could be no GCM, no grand model of the economy that could predict future security prices. The current prices represented a balance between buyers and sellers, and that balance would not shift without some external cause. All changes are therefore due to random external effects—complications— which by definition cannot be predicted. Any foreseeable future event, such as the impact of the seasons on agricultural produce, would be factored into the price. The net expectation for profit of an investor would at any time be zero, because the price of a security was always in balance with its true value, right on the money. The market was a larger version of a Monte Carlo casino. But like a gambler at the casino, an investor could make intelligent bets by figuring the odds and controlling his risk.
All of this went down like a stock-market crash with Poincaré and Bachelier’s other supervisors. Poincaré might have discovered chaos, but this did not weaken his faith in the scientist’s ability to discern cause and effect. He believed that “what is chance for the ignorant is not chance for the scientists. Chance is only the measure of our ignorance.”15 Bachelier’s thesis was awarded an undistinguished grade, which meant that he couldn’t find a permanent position for twenty-seven years. And his theory remained out of sight until half a century later, when it stumbled back into town as the random walk theory.
RANDOM, BUT EFFICIENT
Interest in Bachelier’s theory revived after a number of studies showed that asset prices did move in an apparently random and unpredictable way—just as he had predicted. In 1953, the statistician Maurice Kendall analyzed movements in stock prices over short time periods and found that the random changes were more significant than any systematic effect, so the data behaved like a “wandering series.” In a 1958 paper, the physicist M. F. M. Osborne showed that the proportional changes in a stock’s price could be simulated quite well by a random walk—like the drunk searching for his car keys.16
This seemed to explain why investors had such difficulty predicting stock movements. In 1933, a wealthy investor called Alfred Cowles III had published a paper showing that the top twenty insurance companies in the United States had demonstrated “no evidence of skill” at picking their investments.17 If market movements were essentially random, it would be impossible to guess where they were headed.
Bachelier’s idea was made manifest in the 1960s by economist Eugene Fama of the University of Chicago in the efficient market hypothesis, or EMH. This proposed that the market consists of “large numbers of rational, profit-maximizers actively competing, with each trying to predict future market values of individual securities.” 18 Because any randomness in the market was the result of external events, rather than the activity of investors, the value of a security was always reflected in its current price. There could be no inefficiencies or price anomalies, since these would immediately be detected by investors.
Of course, no market could be totally efficient—especially smaller and less fluid markets such as real estate, where there may be only a small number of buyers interested in a particular property. But to a good approximation, the large bond, stock, and currency markets could be considered efficient. These involve tens of millions of well-informed investors, and they operate relatively free of regulation or restriction. Different versions of the EMH assume varying amounts of efficiency and take into account factors such as insider trading (where traders profit from information that is not widely available). While the EMH is increasingly being debated, as seen below, it still forms the main plank of orthodox economic theory.19
Where the EMH view of the market differed from Bachelier’s was in the assumption that it was made up of “rational” profit-maximizers. Bachelier had concluded that “events, current or expected, often [bear] no apparent relation to price variation.” According to the EMH, however, an efficient market always reacts in the appropriate way to external shocks. If this wasn’t the case, then a rational investor would be able to see that the market was over- or under-reacting and profit from the situation. The fact that investors could not reliably predict the market seemed to imply that it behaved like a kind of super-rational being, its collective wisdom emerging automatically from the actions of rational investors.
The EMH naturally posed something of a challenge to economic forecasters: not only was there no way to predict the flow of money using fundamental economic principles, but even the weaker notion of forecasting the movements of individual stocks or bonds seemed out of reach. Nonetheless, individuals, banks, insurers like Prudential, investment firms like Merrill Lynch, large companies, governments, giant financial institutions such as the World Bank and the International Monetary Fund, and perhaps the greatest economic oracle of them all, the U.S. Federal Reserve, which twice a year presents an economic forecast to Congress—all collectively employ thousands of economists who claim to be able to foresee market movements. So what is going on?
MAKING A PROPHET
While mathematical models of physical systems are usually based on the same general principles, economic models vary greatly in both aims and approach. Nowhere is this more true than in academia. The business cycle of expansions and recessions has been modelled based on sunspots, from a Marxist viewpoint, from a Keynesian perspective, as a predator-prey relationship between capitalists and labour, using “real business cycle” theory (which simulates the workforce’s reaction to external shocks), and so on. Part of the problem is that unlike colonies of yeast, colonies of humans do not sit still for controlled scientific experiments, so it is hard to prove that any theory is definitely false.
Most economic forecasters fall into one of two camps: the data-driven chartists or the model-driven analysts. Chartists, or technicians, are people who look for recurring patterns in financial records. Perhaps the simplest predictive method is to assume, as the Greeks did, that everything moves in circles, that there is nothing new under the sun. To forecast the future, it suffices to search past records for a time when conditions were similar to today’s. Such forecasts o
ften appear in the financial sections of newspapers— when, for example, plots are produced to show that recent stock prices mirror those that preceded a historic crash or fit some pattern that signals the start of a bull or bear market.20 More sophisticated versions, discussed below, use advanced techniques to detect signals in multiple streams of financial data.
The only problem with chart-following is that, statistically speaking, it doesn’t seem to work, at least not for most people. Analysis has shown that, after accounting for the expense of constantly buying and selling securities, chartists on average earn no more money for their clients over the long-term than investors earn for themselves with a naïve buy-and-hold strategy.21 There are at least two reasons for this. The first is that for the present to perfectly resemble the past, the inflationary environment would have to be the same, interest rates would have to match—in principle, everyone would have to be doing the same thing as before. Which of course doesn’t happen. Atmospheric conditions never reproduce themselves exactly, and Frank Knight made a similar statement about the economy in his 1921 work, Risk, Uncertainty, and Profit. The economy is not, therefore, constrained to follow past behaviour.
The second reason is that if a genuine pattern emerges, it is only a matter of time before investors notice it, at which point it tends to disappear. Suppose, for example, that investors have a habit of selling stocks at year end in order to write off the losses on their taxes. The price of stocks should then dip, rebounding in January. While the January effect, as it became known, may once have been real, if rather subtle, it became much harder to detect after the publication of a book called The Incredible January Effect.22 Everyone started buying in January to take advantage of it, so naturally prices went up and any anomaly disappeared. Like a biological organism, the economy evolves in such a way that it becomes less predictable.
The fascination with financial charts often says more about the human desire for order than it does about the markets themselves. Any series of numbers, even a random one, will begin to reveal patterns if you look at it long enough. Indeed, in an efficient market, price movements would be random and any pattern no more than an illusion. As Eugene Fama put it, “If the random walk model is a valid description of reality, the work of the chartist, like that of the astrologer, is of no real value in stock market analysis.”23
Unlike chartists, fundamental analysts base their stock choices on their estimate of a stock’s “intrinsic value.”24 This relies on a prediction of the company’s future prospects and dividends, as well as the effects of volatility, inflation, and interest rates. If the forecaster’s estimate is higher than the market price, then he predicts the stock will rise; if his estimate is lower, it should fall. The best-known proponent of this approach is Warren Buffet, known as the “Oracle of Omaha” for his canny stock picks.
The challenge of this approach is that it requires predictions of the future that are better than those the market is making. This in turn often involves some form of chart reading. Suppose a company is set up to market a new product. New companies, products, or innovations that have gone on to be successful often follow an S-shaped curve like that shown in figure 6.1.25 Starting from a low level, sales grow exponentially as word of mouth spreads and the idea catches on. Success begets success in a positive feedback loop. The company or product then enters a period of steady and sustainable growth. But nothing can grow forever, and eventually the growth will saturate.
In principle, you could predict a company’s future prospects if you knew where it was on this curve. Unfortunately, this is impossible: the company could be snuffed out at day one, or it could turn into the next Microsoft. Also, the S-shaped curve is not the only possibility. The company might shoot up, then shoot back down when its product goes out of fashion, then make a startling comeback selling something else (Apple). The market itself might change or even collapse. Most companies that were huge a hundred years ago no longer exist, because the demand for typewriters and horse carriages is not what it was. Estimates of future growth are therefore highly uncertain, especially over the long term. They are really a judgment on how compelling a particular investment story is.
FIGURE 6.1. The S-shaped development curve goes through three stages: an initial stage of exponential growth, a period of steady growth, and finally a period of saturation. This plot shows market share of a hypothetical company as a function of time.
Once again, the future need not resemble the past. In fact, the record of most analysts is not much better than that of chartists. Their predictions routinely fail to beat naïve forecasts, and funds that hold all the stocks in a particular index routinely outperform managed funds, at least after expenses. Some do much better, but the occasional success story may be simple luck: every investment strategy has to win some of the time. As the economist Burton G. Malkiel wrote, “Financial forecasting appears to be a science that makes astrology look respectable.”26
The mediocre performance of most stock pickers seems to confirm the EMH: in an efficient market, stock selection should be a waste of time because the true value of any asset is always reflected in the price. It is impossible to make better forecasts than the market itself, at least on a consistent basis. Price fluctuations are a random walk, so they are inherently unpredictable. According to Fama, “If the analyst has neither better insights nor new information, he may as well forget about fundamental analysis and choose securities by some random selection procedure.”27 Throwing darts at the financial section of the Wall Street Journal is said to work quite well as a stock-picking technique.
Prices of basic commodities such as oil are especially difficult to forecast, because both supply and demand are subject to shifting political, economic, and geographic factors.28 The export quotas of OPEC countries are determined by their oil reserves, so there is an incentive to inflate the estimates. Many of the OPEC nations involved are notoriously unstable (the Middle East), unpredictable (Venezuela), or at risk of terrorism. As much as a fifth of the 2006 oil price is made up of the so-called political-risk premium, which spikes every time there is a perceived threat against an oil supplier or the transport network. Hurricanes are also a factor, as seen by the fluctuations in oil price as the storms duck and weave their way towards Gulf coast refineries.
Even markets that are not that efficient are hard to predict—like real estate, where there are few buyers per property, and the process of buying or selling is relatively slow and expensive. Some housing markets are believed to follow cyclical trends.29 When house prices are at a low level relative to the cost of renting, new buyers enter the market because it is affordable. As prices begin to rise, speculators join the party, driving the price up with positive feedback. When house prices grow too high relative to rents, the supply of first-time buyers is cut off and the market reaches a plateau—negative feedback. If prices begin to fall, more sellers will try to off-load their properties, driving prices down: positive feedback in the other direction. The cycle therefore continues. However, there are many other factors to take into account, and even if such a cycle does exist, the pattern constantly varies. It is hard to know when turning points will occur, and even harder to beat a buy-and-hold strategy once transaction fees are taken into account.
If no individual person can predict the future, perhaps several people can. In 1948, the RAND Corporation came up with a method of making decisions based on an ensemble approach. A number of experts were polled on a series of questions. The questions were refined based on their input, and the process repeated until a group consensus was obtained. The method, known as Delphi after the Greek oracle, was first used by the United States Defense Department to investigate what would happen in a nuclear war, but it was soon adopted by businesses for making financial decisions.
Imagine that you are a theoprope who has time-travelled from ancient Greece. You have heard of this place called Delphi. You show up at an office in a high-rise in downtown New York. You hand the bemused-looking secretary a goat. She sh
ows you to a room. Inside are . . . a group of management consultants. “Hello, Mr.Theopropous,” says one. “We’ve come to a consensus. We will have the Mediterranean lunch special, all round.” You run out screaming. Not only did the new-version Delphi lack a certain mystique, but it wasn’t very good at predicting the future. A 1991 study by the experimental psychologist Fred Woudenberg showed that the Delphi was no more accurate than other decision-making methods, because “consensus is achieved mainly by group pressure to conformity.”30
So if financial analysts of all stripes cannot predict the future, why are there so many of them, and why are they so well reimbursed? Would the market work just as well without them? The fact is that the market augurs exist because they do make a lot of money—for themselves and their employers. Buying and holding may be good for the client, but the fastest way to generate commissions is by buying, selling, buying again, and so on. Also, there is a strong market for predictions, and accuracy is of secondary importance. It is always easy to generate data that makes it look as if a method has been highly reliable in the past. Finally, as discussed below, some stock pickers really do manage to beat the market, at least for a while.
Of course, if all the chartists, analysts, and consultants were replaced with chimps armed with darts, the market would disintegrate pretty rapidly. If the market has any semblance of efficiency, it is because of the combined efforts of predictors who make up the investor ecosystem. You could argue that if the system has parasites, it is those who invest in index funds. These funds tend to purchase the stocks that have been picked by active managers, so investors benefit passively from their decisions.
