In a great book “Classics II – Another Investor’s Anthology”, there is a piece by Peter Carman on the psychology of investment decision- making.

“One of the underlying assumptions of the Bernstein investment philosophy is that distortions in value are created by emotions that domi­nate investor decision-making. . . . These distortions occur regularly and provide opportunities for premium performance for those investors who can capitalize on them. . . . Use of the dividend discount model provides a rational, systematic method that can be used to uncover when these distortions occur and measure how large they are. But why should distortions or inefficiencies exist in the first place? After all, most investors are bright, sophisticated, well educated and well in­ formed. It seems, however, that these irrationalities are a normal part of human behavior.

I came across a survey recently that is particularly appropriate within this context. This survey was taken from a random sample of adult males. The men were asked to rate themselves on a number of param­eters: The first was their ability to get along with others. Every single respondent ranked himself in the top 10% of the population on this score; and a full 25% put themselves in the top 1%. Similarly, 70% rated themselves in the top quintile in leadership ability. And only 2% felt that they were below average leaders. Finally, in an area where self-deception should be difficult for most males, 60% said they were in the top quintile in athletic ability and only 6% said they were be­low average.

Investment Decision-Making

It also turns out that there are a number of studies in the field of be­havioral psychology that strongly suggest that all of us have specific glitches in our thinking patterns that are particularly noticeable in com­plicated, ambiguous situations, or during periods of stress or uncer­tainty. Furthermore, these glitches, while not rational, can be measured; they form predictable patterns which tend to confirm, in a systematic fashion, our intuitive beliefs about investor decision-making and pricing in the securities markets. This research leads us to two conclusions.

Ø       First: We don't usually understand the implications of probabili­ties, and when we do, we don't apply them well in day-to-day situations.

Ø       Second: There are some very strong biases in the way we forecast probabilities in the first place.

The cumulative impact of these two factors results in a number of important and predictable decision-making biases.

v      We overvalue certainty.

v      We overestimate the value of small chances of large gains or losses.

v      We have a disproportionate aversion to losses

At this point, I would like to introduce some research done primarily by two behavioral psychologists, Daniel Kahneman and Amos Tversky, in the field of prospect theory. Their research tries to measure and ex­plain patterns of decision-making under uncertainty. But first, I would like to provide a little background on the way investment theoreticians have traditionally addressed this issue.

Risk Aversion

I am sure that it doesn't come as a great surprise to any of us that the most easily identifiable bias in decision-making is risk aversion. In or­der to address this, I would like to explore the idea of utility. Let me ask you a question: Which would you prefer?

A. An 80% chance of winning $100,000?

or

B. An outright gift of $70,000?

The answer that carries the highest expected monetary value is clearly the 80% chance of winning $100,000, which is worth $80,000. Most people, however, opt for the guaranteed $70,000 payment. What happens is that the extra $10,000 that one should statistically earn on that gamble doesn't contribute anywhere near as much to our satisfac­tion or utility as the earlier increments of $10,000 that make up the guaranteed$70,000. . . .

Utility of Wealth

The utility of wealth is not a linear function of money. Within this frame of reference, a gain of $2,000 contributes less than twice as much to utility or happiness as a gain of $1,000. People tend to feel less and less excited about additional increments of wealth and, as a conse­quence, are willing to take less and less risk to increase that wealth. The set of relationships postulated by utility theory can be described mathematically and, in its general form, this theory has served an im­portant role in analyzing how investors make decisions.

Useful as this framework has been in modeling human behavior, Kahneman and Tversky's investigations suggest that there are a number of areas where people's decisions are consistently out of line with what would be predicted from utility theory. The psychologists' elaboration of this work is called prospect theory.

They developed prospect theory empirically by asking subjects about their preferences for things. Some of the questions involved preferences related to life, death and health; others involved purely quantitative is­sues. Many were based on money, sometimes real money. All were gambles of one form or another, thus testing decision-making in an en­vironment of uncertainty. I will now spend the next few minutes talking about some of these choice problems and relating them to investing in general.

General Jones

As a start in exploring this area we can talk about General Jones. De­spite his reputation as a crack military strategist, he has found himself in a tough spot. He commands 600 troops and they are cornered. He has two choices. . . . He can choose Strategy A and save 200 of his troops, or the General can adopt Strategy B and risk a one-third chance that all of his troops will be saved and a two-thirds chance that all will die. Which should he choose?

In studies of many groups, some acquainted with the mathematics of probability and some not, the preponderance of people pick A. The ex­pectation for either strategy is identical. But people clearly prefer A. Their motivation? We don't know exactly, but probably it is a distaste for gambling with lives.

Doctor Baker

On the other hand, what about Doctor Baker? He has been called in to treat 600 very sick people. He also has two choices. . . . [Therapy A will cause 400 people to die. With Therapy B, there is a one-third chance everyone will be saved and a two-thirds chance no one will be saved.] What should he do? Studies of people faced with these choices show an overwhelming preference for Therapy B. Why? When faced with the possibility of the certain death of 400, why not gamble and try to save them all.

Of course a comparison of the General's and the Doctor's situations indicate that they are in exactly the same position. Strategy A and Ther­apy A will have the same result-400 people will die. Similarly, the objective result of alternative B in both situations is identical. Yet when the problem is stated in terms of a certain gain (that is, lives saved) versus stating the problem in terms of a certain loss, i.e., lives lost, people react very differently.

When these kinds of choices are framed in terms of financial gain or loss, the results are the same. People normally require very high odds of winning to offset a relatively small probability of loss. In fact, on average, people require a 65% chance of winning $100 to offset a 35% chance of losing $100. . . .

Let's turn to another choice problem. You are now given a choice between:

A. 1% chance of winning $6,000.

or

B. 2 chance of winning $3,000.%

Both choices have the same expected value and both are remote chances. In general, people go for A in a ratio of about 7 to 3. Again, the bottom line is that in the case of small chances of very large gains or losses, people seem to turn into risk seekers, despite the fact that this result clearly violates utility theory.

Certainty Effect

The next choice problem is an additional example of this process. Think about the choices offered in the following table:

Choose between

In choice A the chance of not winning is miniscule and the expected value of the outcome is a few dollars higher than in choice B. However, studies show choice B to be the overwhelming favorite, 82% to 18%.

This illustrates what is called the certainty effect-that is, the dispro­portionate interest people have in things that are certain versus things of a moderate probability.

Implications for Money Management

We have gone far enough to draw some important conclusions about prospect theory's implications for money management. The first implication is probably the least provocative: Investors clearly undervalue things they perceive as risky relative to those things they perceive as certain. Characteristics that are frequently associated with overvalua­tion on this basis include stability of earnings during recessions and consistency of earnings growth at all times.­

The second implication is more powerful but doesn't occur with great regularity: Investors can be counted on to overreact to events having small probabilities. This comes in two forms. It causes investors to be­come risk takers when contemplating large gains and extraordinarily risk averse when contemplating possible large losses. Thus, people can be counted on to overpay for the prospect of a promising new drug in early clinical trials, the potential of a big oil find or a hot rapidly growing technology company. On the flip side, people can also be counted on to overreact to potential calamities and undervalue the af­fected company, the latest example of which was the Bhopal disaster at Union Carbide.

The third issue is probably most significant for our investment style. It is the disproportionality of people's tolerance for losses as compared to the pleasure of gains. With great frequency, investors find reasons to fear exposure to various sectors of the economy, a particular industry, or individual companies. The prospects for these stocks are then per­ceived by investors as a pair of choices, one of which implies a loss. The reaction to the potential for loss is to undervalue these stocks rela­tive to others where the probability and magnitude of the potential losses appear small. . . .

Biases in Forecasting

The forecasting process, as you are well aware, is not a simple one. As Mark Twain once said: "The art of prophesy is very difficult, especially with respect to the future." A forecaster is confronted with the problem of collecting massive amounts of information, sorting out the important from the trivial and processing the remaining data in a way that will produce a meaningful approximation of what he is trying to forecast. In addressing this task, we attempt to help our­selves by taking shortcuts that allow us to organize and simplify the process. We do so in a way that seems to be consistent from person to person.

Generally speaking we use the three forecasting tools outlined below:

When using representativeness, we find a model which, in our experi­ence, looks like it has the same characteristics as the thing we are try­ing to forecast. This helps us decide what the key variables should be. Availability has us look at history, both quantitatively and qualitatively, to try to determine the probable behavior of the key variables that make up the model. And when we employ anchoring we examine the level of each variable today, to establish a base point, and then forecast the fu­ture based on steps one and two.

Because of the complexity of the forecasting task, these internalized guidelines are quite useful (one could even say necessary) but they can sometimes lead to severe systematic errors. If we were always totally objective in our thought processes and about ourselves, and if, in addi­tion, we had a good intuitive sense for the laws of statistics, the process outlined above would be fine. Unfortunately, objectivity and apprecia­tion of statistical principles do not often characterize the way we oper­ate. Moreover, the data available to us are usually not complete or are of questionable relevance. Taken together with the way people make judgments, biases are almost assured.

Stereotyping

As the next step in this process, I'll outline some of the more important of these biases as summarized in the following table:

First let me pose a question. Steven is very shy, withdrawn, invari­ably helpful, but with little interest in people or in the world of reality. Is Steve a librarian or salesman?

Of course, everyone assumes that Steve is a librarian. Actually, the odds are that Steve is a salesmen because there are 16 times as many salesmen in the workforce as librarians. People are generally aware of this imbalance but disregard it even when the specific information is available. This is stereotyping. . . . In general, stereotyping is extraor­dinarily compelling and normally overwhelms the discipline required to integrate the laws of statistics into the selection process.

In the stock market, the impact of stereotyping also makes itself felt. For example, when investors were most concerned with problems facing electric utilities with unfinished nuclear facilities, virtually all of those utilities were thrown into the same category. On the other hand there were significant differences among those companies, including relative financial exposure to a plant writeoff, the character of regulation in the state and specific company experience in building nuclear facilities.

Misconception of Regressions

Of greater significance to our business is what is called the effects of misconceptions about regressions. A regression is normally made up of a large number of equally weighted data points that reflect historical relationships. Frequently, forecasters tend to bias historical results by overweighting the most recent data. The effect of this bias is particu­larly powerful. In our business, for example, investors show a very strong tendency to overreact to either very good or very bad quarterly earnings. They normally act as if such developments are highly unusual, contain new information and frequently decide that something of per­manence has occurred. Estimates are changed and elaborate analytical explanations and rationalizations are devised. Instead, the more likely explanation is that such performance is statistically possible, will occur in the future with some frequency, and the next data point will move back toward the average performance of the company or the industry in question.

Law of Small Numbers

Among people's intuitions about statistics is something called "local representativeness." In the case of flipping a coin, people regard get­ ting three heads in a row as much less likely than it really is. As a consequence, they attach much more significance to that event than they should. Because of this intuitive belief, people have far too much con­fidence in forecasting models inferred from small samples of data that seem to manifest a pattern. The most pernicious of these patterns, from a stock market perspective, is consistency of earnings growth. This kind of bias produces excessive confidence and high stock prices relative to future prospects and is generally associated with stocks that appear overvalued within the framework of a dividend discount model.

Failure of Initiative

Failure of initiative reflects an inclination to believe that existing con­ditions are unlikely to change and that the world is generally static. To frame this idea in the context of investments: People tend to assume that certain difficult economic or industry-specific problems are intractable and that forecasts should be made accordingly. Actually very few prob­lems are really intractable.

One of the most colorful and impressive illustrations of failure of initiative was the Club of Rome report issued by a large distinguished group of international economists and social scientists in the early1970s, which concluded that we were running out of natural resources. It totally missed the ability of the world economy to conserve and to find new sources of raw materials as prices rose. However, the forecasts that now seem silly in light of the current world-wide commodity glut were then taken very seriously, forming the basis of a significant num­ber of investment strategies.

Availability

Now, let me ask you another question. Do more people die of homicide or of diabetes? The answer is diabetes by a ratio of two-to-one, al­though people normally vote for homicides. . . .

We just went through an exercise in estimation based on the idea of availability. I'm sure that very few people here carry around death-rate data in their heads, so your judgments on the subject come from im­pressions that you have gathered from your exposure to the issue. The problem with exposure as a measure of probability is that its frequency may have no relationship to the actual value or mathematical odds of the variable being estimated.

Here's where journalism comes in. News reporting is a business geared towards maximizing market share. As a result, things that attract attention such as death, destruction and financial disaster are reported in great detail and with high frequency.

In addition, all too frequently, the issues that control a forecast have no measurable statistical history. You can look up death rates but you can't look up the probability of an Argentine default or the conse­quences of that default.

In the absence of concrete data, our judgment tends to be colored by how often we hear and read about a default. Equally important, the more provocative and dramatic the stories are, the more they are re­membered and the greater the impact they have on decision-making. It is in this way that availability leads to some of the largest distortions in forecasts and stock prices. It also feels like the hardest bias to fight. After all, we have to read The Wall Street Journal every day.

Anchoring

Finally, let me spend a moment on anchoring and the "as if" phenom­enon. Anchoring is a term used to describe the inherent conservatism that people seem to have in making single-stage estimates, which is to say that estimates of the future value of a variable will be biased to­wards its initial value. . . .' In a number of tests by several different investigators, this tendency was seen to be quite strong. I think that it is another example of people's intuitions about statistics: We imagine most processes to be less volatile than they really are, and, despite one sur­prise after another, we never really learn.

"As If" Phenomenon

Lastly, there is the "as if" phenomenon, a concept proposed by psy­chologists that relates to how people process multistage estimation problems, or in investment terms, scenario forecasts. It appears that once people collect enough information about one stage of a problem to reach a decision with some confidence, they then move on to the next stage" as if" the prior stage is known with certainty. . . . In elaborate scenarios with many stages, the real probability of any given outcome becomes very small. The confidence in the outcome, on the other hand, does not fall nearly as fast as the cumulative probability.

Market Impact

Given the nature of the biases surrounding both the forecasting and decision-making processes, we should expect to find a more or less pre­dictable pattern of distortions in the stock market. Stocks that are cheap, those that have high expected returns, should have the following characteristics:

v      Problems will look hopeless if the stock is really cheap because failure of initiative will apply.

v      Current financial performance of these companies is likely to be well below what it has been in the past, and their operating situation is likely to leave something to be desired.

v      The environment in which these companies operate will appear to have shifted permanently for the worse-the underpriced stocks will feel inherently risky, the possibility of substantial loss will remain high.”