Contrarian Investment Strategies
Page 20
A special technique was designed, using a statistical test called analysis of variance, or ANOVA, to evaluate experts’ configural capabilities. In one such study, nine radiologists were given a highly configural problem: deciding whether a gastric ulcer was malignant.2 To make a proper diagnosis, a radiologist must work from seven major cues either present or absent in an X-ray. These can combine to form fifty-seven possible combinations. Experienced gastroenterologists indicated that a completely accurate diagnosis can be made only by configurally examining the combinations formed by the seven original cues.3
Although the diagnosis requires a high level of configural processing, the researchers found that in actual practice, it accounted for a small part of all decisions, some 3 percent. More than 90 percent came from serially adding up the individual symptoms.
A similar problem is deciding whether a psychiatric inpatient is to be allowed to leave the hospital for short periods. The hospital staff has to consider six primary cues that can be present or absent (for example, does the patient have a drinking problem?) and sixty-four possible interactions. In another study, nurses, social workers, and psychologists showed little evidence of configural thinking, although it was essential to reaching the optimum solution.4 In a third study, thirteen clinical psychologists and sixteen advanced graduate students attempted to determine whether the symptoms of 861 patients were neurotic or psychotic, a highly configural task. The results were in line with the first two examples.5
Curious as to what he would find in the stock market, Paul Slovic tested the importance of configural (or interactive) reasoning in the decisions of market professionals. In one study he provided thirteen stockbrokers and five graduate students in finance with eight important financial inputs—the trend of earnings per share, profit margins, outlook for near-term profits, and so on—that they considered significant in analyzing companies. They had to think configurally to find the optimum solution. As it turned out, however, configural reasoning, on average, accounted for only 4 percent of the decisions made—results roughly equivalent to those of the radiologists and psychologists.
Moreover, what the brokers said about how they analyzed the various inputs differed significantly from what they did.6 For example, a broker who said that the most important trend was earnings per share might actually have placed greater emphasis on near-term prospects. Finally, the more experienced the brokers, the less accurate their assessment of their own scales of weighting appeared to be. All in all, the evidence indicates that most people are weak configural processors, in or out of the marketplace. So it turns out that human minds—as I’m sure you’ve suspected—are not hardwired anything like the way computers are.
Information-Processing Vulnerabilities
As we saw in our first look at information processing in chapter 3, Herbert Simon more than fifty years ago wrote extensively on information overload. Under certain conditions, experts err predictably and often; in fields as far apart as psychology, engineering, publishing, and even soil sampling, all of them make the same kinds of mistakes. The conditions for such errors are as fertile in the stock market as anywhere.
I briefly mentioned that the vast storehouse of data about companies, industries, and the economy mandated by current investment methods may not give the professional investor an extra “edge.” When information-processing requirements are large and complex analysis is necessary to integrate it, the rational system, which is deliberative and analytical, is often subtly overridden without the professional’s knowledge by the experiential system. Inferential processing delicately bypasses our rational data banks. As we’ve seen, ingesting large amounts of investment information can lead to worse rather than better decisions because the Affect working with other cognitive heuristics, such as representativeness and availability, takes over.
In the next chapter, you’ll see that forecasting, the heart of security analysis, which selects stocks precisely by the methods we are questioning, misses the mark time and again. We’ll also see that the favorite stocks and industries of large groups of professional investors have fared worse than the averages for many decades. This is the primary reason for the subpar performance of professionals over time that was witnessed by the efficient-market researchers in Part II, rather than the EMH myth of rational automatons who make flawless decisions that keep markets efficient.
To outdo the market, then, we must first have a good idea of the forces that victimize even the pros. Once those forces are understood, investors can build defenses and find routes to skirt the pitfalls.
How Much Is Too Much?
Under conditions of complexity and uncertainty, experts demand as much information as possible to assist them in their decision making. Seems logical. Naturally, there is tremendous demand for such incremental information on the Street because investors believe that the increased dosage gives them a shot at the big money. But, as I have indicated earlier, that information “edge” may not help you. A large number of studies show rather conclusively that giving an expert more information doesn’t do much to improve his judgment.7
In a study of what appears to be a favored class of human guinea pigs, clinical psychologists were given background information on a large number of cases and asked what they thought their chances were of being right on each one. As the amount of information increased, the diagnosticians’ confidence rose dramatically, but their accuracy continued to be low. At a low level of information, they estimated that they would be correct 33 percent of the time; their accuracy was actually 26 percent. When the information was increased fourfold, they expected to be correct in 53 percent of the cases; in fact, they were right 28 percent of the time, an increase of only 2 percentage points.
Interestingly, the finding seems universal: only a marginal improvement in accuracy occurs as increasing amounts of new information are heaped on. The same results were obtained using racetrack handicappers. Eight veteran handicappers were progressively given five to forty pieces of the information they considered important in picking winners. One study showed that their confidence rose directly with the amount of information received, but the number of winners, alas, did not.8
As the studies demonstrate, people in situations of uncertainty are generally overconfident on the basis of the information available to them, believing they are right much more often than they are. One of the earliest demonstrations of overconfidence involved the predictive power of interviews. Many people think a short interview is sufficient for making reasonable predictions about a person’s behavior. Analysts, for example, frequently gauge company managers through meetings lasting less than an hour. Extensive research indicates that such judgments are often wrong. One interesting example took place at the Harvard Business School. The school thought that by interviewing candidates beforehand, it could recruit students who would earn higher grades. In fact, the candidates selected that way did worse than students accepted on their academic credentials alone. Nevertheless, superficial impressions are hard to shake and often dominate behavior.
Overconfidence, according to cognitive psychologists, has many implications. A number of studies indicate that when a problem is relatively simple to diagnose, experts are realistic about their ability to solve it. When the problem becomes more complex, however, and the solution depends on a number of hard-to-quantify factors, they became overconfident of their ability to reach a solution (accuracy 61 percent). If the task is impossible—for example, distinguishing European from American handwriting—they became “superoverconfident” (accuracy 51 percent).9
A large number of other studies demonstrate that people are consistently overconfident when forming strong impressions from limited knowledge. Lawyers, for example, tend to overestimate their chances of winning in court. If both sides in a court case are asked who will win, each will say its chances of winning are greater than 50 percent.10 Studies of clinical psychologists,11 physicians,12 engineers,13 negotiators,14 and security analysts15 have all shown they are far too
confident in the accuracy of their predictions. Clinical psychologists, for instance, believed their diagnosis was accurate 90 percent of the time, when in fact it was correct in only 50 percent of cases. As one observer said of expert prediction, “[It] is often wrong but rarely in doubt.”
The same overconfidence occurs when experienced writers or academics working on books or research papers estimate the time of completion. The estimates are invariably overconfident; the books and papers are completed months or years behind schedule, and sometimes they are not completed at all.
Studies in cognitive psychology also indicate that people are overconfident that their forecasts will be correct. The typical result is that respondents are correct in only 80 percent of the cases when they describe themselves as 99 percent sure16—not what you’d exactly want in a stress test or another health-critical test result.
The question becomes even more interesting when experts are compared with laypeople. A number of studies show that when the predictability of a problem is reasonably high, the experts are generally more accurate than laypeople. Expert bridge players, for example, proved much more capable of assessing the odds of winning a particular hand than average players.17 When predictability is very low, however, experts are more prone to overconfidence than novices. When experts predict highly complex situations—for example, the future of troubled European Union countries such as Portugal and Greece, the impact of religious fundamentalists on foreign policy in the Middle East, or the movement of the stock market—they usually demonstrate overconfidence. Because of the richness of information available to them, they believe they have the advantage in their area of expertise. Laypeople with a very limited understanding of the subject, on the other hand, are normally more conservative in their judgments.18
Overconfident experts are legion on the investment scene. Wall Street places immense faith in detailed analysis by its experts. In-depth research houses turn out thousands upon thousands of reports, each running to a hundred pages or more and sprinkled with dozens of tables and charts. They set up Washington and international listening posts to catch the slightest whiff of change in government policies or economic conditions abroad. The latest turn here, now being investigated by the SEC, is to hire former corporate executives from major companies they follow to give them “the real scoop” on “what’s going on” inside. Scores of conferences are naturally called to provide money managers with this penetrating understanding. All too often they have proved to be “wrong in depth,” as one skeptic put it.
The more detailed the level of knowledge, the more effective the expert is considered to be. Despite widespread concerns from 2005 through 2007, major bankers and investment bankers, including Citigroup, Lehman Brothers, and Goldman Sachs, stated emphatically that there was no sign of a bubble in the housing market. They continued to sell subprime toxic waste, including part of their own inventory, by the tens of billions of dollars to their own clients, until the subprime markets completely dried up in mid-2007. To survive, many were forced to ask for a bailout, which was organized by Hank Paulson, the Treasury secretary and ex-CEO of Goldman Sachs. As with the clinical psychologists and the handicappers, the information available had little to do with accurately predicting the outcome.
The inferior investment performance we’ve seen, as well as more that we will look at next, was based on just such detailed research. To quote a disillusioned money manager of the early 1970s, “You pick the top [research] house on the Street and the second top house on the Street—they all built tremendous reputation, research-in-depth, but they killed their clients.”19 Nothing much has changed.
I offer a Psychological Guideline for investing that is applicable in almost any other field of endeavor:
PSYCHOLOGICAL GUIDELINE 7: Respect the difficulty of working with a mass of information. Few of us can use it successfully. In-depth information does not translate into in-depth profits.
I hope it is becoming apparent that configural relationships are extremely complex. Stock market investors are dealing not with twenty-four or forty-eight relevant interactions but with an astronomical number. As we have already seen, far fewer inputs can overtax the configural or interactive judgment of experts. Because most Wall Street experts, like those elsewhere, are unaware of these psychological findings, they remain convinced that their problems can be handled if only those few extra facts are available. They overload with information, which doesn’t help their thinking but makes them more confident and therefore more vulnerable to making serious errors. The difficulty of configural reasoning is unfortunately almost unknown by both investors and EMH theorists.
Overconfidence Takes a Bow
As we saw earlier, the phenomenon of overconfidence seems to be both an Affect and a cognitive bias. In other words, the mind is probably designed to extract as much information as possible from whatever is available. The filtering process, as we saw in chapter 3, is anything but a passive process that provides a good representation of the real world. Rather, we actively exclude information “that is not in the scope of our attention.”20 It thus provides only a small part of the total necessary to build an accurate forecast in uncertain conditions.
Evaluating stocks is no different. Under conditions of anxiety and uncertainty, with a vast interacting web of partial information, it becomes a giant Rorschach test. The investor sees any pattern he or she wishes. In fact, according to recent research in configural processing, investors can find patterns that aren’t there—a phenomenon called illusory correlation.
Trained psychologists, for example, were given background information on psychotics and were also given drawings allegedly made by them but actually prepared by the researchers. With remarkable consistency, the psychologists saw in the drawings the cues that they expected to see: muscular figures drawn by men worried about their masculinity or big eyes by suspicious people. But not only were those characteristics not stressed in the drawings; in many cases they were less pronounced than usual.21 Because the psychologists focused on the anticipated aberrations, they missed important correlations that were actually present.22
Investors attempt to simplify and rationalize complexity that seems at times impenetrable. Often they notice facts that are simply coincidental and think they have found correlations. If they buy the stock in the “correlation” and it goes up, they will continue to invest in it through many a loss. The market thus provides an excellent field for illusory correlation. The head-and-shoulders formation on a chart cuts through thousands of disparate facts that the chartist believes no one can analyze; buying growth stocks simplifies an otherwise bewildering range of investment alternatives. Such methods, which seemed to work in the past, are pervasive on Wall Street. The EMH theorists’ search for the correlation between volatility and returns that the theory demands seems to be another example. The problem is that some of the correlations are illusory and others are chance. Trusting in them begets error. A chartist may have summed it up most appropriately: “If I hadn’t made money some of the time, I would have acquired market wisdom quicker.”
Which brings us to the next Psychological Guideline, one that may at first glance appear simple but is important and will prove harder to follow than you may think!
PSYCHOLOGICAL GUIDELINE 8: Don’t make an investment decision based on correlations. All correlations in the market, whether real or illusory, will shift and soon disappear.
This is important for the investor: if analysts are generally optimistic, there will be a large number of disappointments created not by events but by initially seeing the company or industry through rose-colored glasses.
The late Amos Tversky, a pioneer in cognitive psychology, researched expert overoptimism and overconfidence in the stock market. According to Tversky, “In one study, analysts were asked such questions as what is the probability that the price of a given stock will exceed $X by a given date. On average, analysts were 80% confident but only 60% accurate in their assessments.”23 The study was repeat
ed numerous times.
In other studies, analysts were asked for their high and low estimates of the price of a stock. The high estimate was to be a number they were 95 percent sure the actual price would fall below; the low estimate was the price they were 95 percent sure the stock would remain above. Thus, the high and low estimates should have included 90 percent of the cases, which is to say that if the analysts were realistic and unbiased, the number of price movements above and below this range would be 10 percent. In fact, the estimates missed the range 35 percent of the time, or three and a half times as often as estimated.
Tversky went on to note that “rather than operating on rational expectations”—with total logic, unaffected by behavior, as efficient-market theory assumes investors do—“people are commonly biased in several directions: they are optimistic; they overestimate the chances that they will succeed, and they overestimate their degree of knowledge, in the sense that their confidence far exceeds their ‘hit rate.’”24
Tversky was queried about overconfidence at an investment behavioral conference in 1995 that I attended and also spoke at. The questioner asked what he thought of the fact that analysts were not very good at forecasting future earnings. He responded, in part, “From the standpoint of the behavioral phenomena . . . analysts should be more skeptical of their ability to predict [earnings] than they usually are. Time and time again, we learn that our confidence is misplaced, and our overconfidence leads to bad decisions, so recognizing our limited ability to predict the future is an important lesson to learn” (italics mine).25
He was asked at the same conference if analysts and other professional investors learn from their experiences. He replied that “unfortunately cognitive illusions are not easily unlearned. . . . The fact that in the real world people have not learned to eliminate . . . overconfidence speaks for itself.”26