winners. As Figure 4-2 shows, their confidence rose directly with the amount of information, but the number of winners, alas, did not."

As the above studies demonstrate, people in situations of uncertainty are generally overconfident based on 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 persons behavior. Analysts, for example, frequently gauge company managements through meetings lasting less than an hour. Extensive research indicates that these 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 gamer higher grades. The candidates selected 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, people 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 abihty to reach a solution (accuracy 61%). If the task is impossible, for example, deciphering European from American handwriting, they became "super overconfident" (accuracy 51%)."

A large number of such 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%." Studies of chnical psychologists, physicians," engineers,"* negoriators, and security analysts* showed they were far too confident in the accuracy of their predictions. Clinical psychologists, for example, believed their diagnosis was accurate 90% of the rime, when it was correct in only 50% 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, as this writer can attest, are sometimes not completed at all.

Studies in cognitive psychology also indicate people are overconfident that their forecasts will be correct. The typical result is that re-

spondents are correct in only 80% of the cases when they describe themselves as 99% sure."

The question becomes even more interesting when experts are compared to lay people. A number of studies show that when the predictability of the problem is reasonably high, the experts are generally more accurate than lay people. Expert bridge players, for example, proved much more capable of assessing the odds of winning a particular hand than average players." When predictability is very low, however, experts are more prone to overconfidence than novices. When experts predict highly complex situations, such as the future of the Eastern Bloc economies, 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. Lay people with a very limited understanding of the subject, on the other hand, will normally be more conservative in their judgments.

Overconfident experts are legion on the investment scene. Wall Street places immense faith in detailed analysis by its experts. In-depth research houses tum out thousands upon thousands of reports, mnning to a hundred pages or more, sprinkled with dozens of tables and charts. They set up Washington listening posts to catch the slightest whiff of change in govemment policy, and call scores of conferences to provide money managers with penetrating understanding.

The more detailed the level of knowledge, the more effective the expert is considered to be. In 1992, for example, several of the most respected analysts following Compaq Computer knew, almost to the unit, the number of new orders and shipments that Compaq and its competitors were making both domestically and abroad. Even so, they wrote bearish reports on the stock, entirely missing the dramatic improvement in Compaqs competitive position. Listing the stock as a sell, they watched it go up over 16-fold in the next five years. As with the clinical psychologists and the handicappers, the information available had little to do with accurately predicting the outcome.

This result is, unfortunately, no exception. The inferior investment performance noted in chapter 1, as well as those that we will look at next, were based on just such detailed research. To quote a disillusioned money manager of the early seventies, "You pick the top [research] house on the Street and the second top house on the Street-they all built tremendous reputations, research-in-depth, but they killed their chents." Nothing much has changed.

RULE 2

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 these configural relationships are extremely complex. Stock market investors are not deaUng with 24 or 48 relevant interactions, but with an astronomical number. We have already seen how far fewer inputs overtax the configural or interactive judgment of experts. Because Wall Street experts, as 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 doesnt help their thinking but makes them more confident and therefore more vulnerable to serious errors.

Unfortunately, according to the findings of a number of clinical studies, overconfidence seems to be a cognitive bias. In other words, the mind is probably designed to extract as much information as possible from what is available, but does not realize that the available information is 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 information, the market becomes a giant Rorschach test. The investor sees any pattem he or she wishes. In fact, investors can find pattems that arent there, says recent research in configural processing-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 experimenters. With remarkable consistency, the psychologists saw the cues in the drawings that they expected to see-muscular figures "drawn" by men worried about their masculinity, or big eyes by suspicious people. Not only were these characteristics not stressed in the drawings; in many cases they were less pronounced than usual. Because the psychologists focused on the anticipated aberrations, they missed important correlations actually present.

I offer a rule for investing that is apphcable in almost any other field of endeavor: