1. How long, in days, is the gestation period of an Asian elephant?
2. How deep, in feet, is the deepest known point in the ocean?
3. Figure 5-1 provides the share price chart for a particular security over a forty-eight-month period. What is your prediction for the share price value six months beyond this forty-eight-month period?
4. Figure 5-2 provides the share price chart for a particular security over a forty-eight-month period. What is your prediction for the share price value six months beyond this forty-eight-month period?
Figure 5-1 Share Price
Figure 5-2 Share Price
5. Figure 5-3 describes the dollar change in share price for a particular security over a forty-eight-month period. What is your prediction for the average change in the share price, per month, for the six months beyond this forty-eight-month period?
The answers are (1) 645 days, (2) 36,198 feet, (3) $100.30, (4) $30.83, and (5) $0.83. Count an answer as a hit if the right response lies between your low guess and your high guess. Count an answer as a miss if the right response falls outside of the range between your low guess and your high guess. What score did you get?
Most people miss more than one out of the five questions in the preceding quiz. Actually, most miss four or even all five. Someone who is well calibrated should miss no more than one question. But the percentage of people who miss only one question is less than 1 percent. This means that the other 99 percent are overconfident. Overconfidence abounds.
Figure 5-3 Dollar Change in Share Price
Overconfident people get surprised more frequently than they anticipated. Take the strategists’ predictions for 1997.11 On June 20, 1997, the Dow closed at 7796, well above the expectations of all seven analysts. On June 23, Barron’s reinterviewed them, recorded their reactions to how the market had behaved during the first half of the year, and collected their predictions for the remainder of the year.12
So how did the strategists react? In a word, surprise. The article quotes Smith Barney (now Salomon Smith Barney) stock strategist Marshall Acuff, who had the most optimistic prediction for the first half of 1997, as saying: “Certainly I’ve been surprised, everyone’s been surprised. We’ve all been humbled.”13
So, where did the strategists’ June revisions indicate the Dow would close at the end of 1997? At 6995, down 10.3 percent from its June value of 7796. Gambler’s fallacy? The Dow closed 1997 at 7908.
Do strategists learn? Well at the end of 1997, Barron’s elicited the Dow predictions of eight strategists, six of whom were repeats from the previous year.14 The average prediction was that the Dow would close 1998 at 8500, up 7.5 percent for the year. And what about Marshall Acuff, who, despite his optimism, expressed so much surprise in June 1997? He predicted that in 1998 the Dow would close at 8000, both at mid-year and at year-end, a 1.2 percent increase. In actuality the Dow closed on June 30 at 8952, up 13.2 percent, and ended 1998 at 9181, up 16.1 percent for the year. The surprises continued. Clearly, learning is a slow process.
Betting on Trends: Naive Extrapolation, Anchoring, and Under-reaction
Questions 3 and 4 in the quiz on page 48 bear directly on the overconfidence associated with predicting the market. The questions are taken from a study by Werner De Bondt (1993) titled “Betting on Trends.” In both questions, the security depicted is an S&P 500 index fund, although the forty-eight-month time periods differ. Three major findings emerged from De Bondt’s study.
First, people tend to formulate their predictions by naively projecting trends that they perceive in the charts. Second, they tend to be overconfident in their ability to predict accurately. Third, their confidence intervals are skewed, meaning that their best guesses do not lie mid-way between their low and high guesses.
What does skewness mean? It means that when the market has been going up, people think there is only a little room left on the upside in case they guess too low. But if they turn out to be wrong on the downside, then they would not be surprised by a large drop.15
Think back to the comments of Morgan Stanley’s Barton Biggs, that I quoted in chapter 2. In August 1997, Biggs stated, “we’re at the very tag end of a super bull market,” and although it has looked that way for a long time, “it’s never looked as much that way as it does right now.” The situation Biggs faced in 1997 was similar to that depicted in figure 5-1, and his view at the time conforms with the average response to question 3. In the twenty-one months after Biggs made his remarks, the S&P 500 returned more than 41 percent.
In question 3 above, a person’s best guess usually lies closer to his or her high guess than to the low guess. Perhaps the person is saying that since the stock has gone up by so much already, it can’t possibly go up by much more. Peter Lynch (1989) calls this one of the “silliest (and most dangerous) things people say about stock prices.” We could say the same of question 4, where a person’s best guess lies closer to his or her low guess.
Why do people have skewed confidence intervals? De Bondt suggests that their predictions are anchored on the early history presented in the chart, a primacy effect. The effect of an anchor depends on just how salient past history is. Questions 3 and 5 in our overconfidence quiz are actually based upon identical data. Question 3 presents the data in terms of levels, while question 5 presents the data in terms of changes. Most people predict a higher change in share price for question 5 than for question 3. Why? As suggested by psychologist Paul Andreassen (1990), the early low share price values are less salient in figure 5-3, used in question 5, than they are in figure 5-1, used in question 3.
Heuristic Diversity
Oh, how straightforward the world would be if investors committed just one type of error in predicting the market. Alas, the world is not that simple. Those who bet on trends extrapolate: they bet that trends continue. Those who commit gambler’s fallacy predict reversal. And both predictions stem from representativeness. For those who bet on trends, continuation is representative of what they perceive, while for those who commit gambler’s fallacy, reversal is representative of what they perceive. The same heuristic, but different perceptions—that is what leads to these different predictions.
Interestingly, there are systematic differences in perceptions across groups, leading to systematic differences in their predictions. De Bondt (1998) points out that Wall Street strategists are prone to committing gambler’s fallacy, whereas individual investors are prone to betting on trends.16 The result is that in rising markets, we get to see headlines such as the following, which appeared in the Washington Post on November 25, 1998: “Joe Investor Beats the Mavens; Roller-Coaster Market Confounds Wall Street’s Experts.”17
Technical Analysis versus Fundamental Analysis: Addressing Sentiment
Ralph Acampora has served as Prudential Securities’ director of technical analysis since 1990. In a November 20, 1998, appearance on the television program Wall $treet Week with Louis Rukeyser, Acampora described technical analysis in the following terms: “Well, it’s what I do for a living is I’m basically a trend follower.” In effect, betting on trends lies at the heart of technical analysis. It therefore should come as no surprise that the predictions of technical analysts and those of fundamental analysts often conflict, sometimes dramatically. Here is a short case that highlights some of the differences.
In early 1997, Acampora accurately predicted that the Dow would close 1997 at 8000 after hitting a high of 8250.18 He continued to be bullish throughout 1997, predicting in August that the Dow would cross 10,000 by June 1998. At the same time, he expressed concern that a bear market was in store for the second half of 1998. His next Wall $treet Week with Louis Rukeyser prediction, made on January 2, 1998, was for a high of 8600 on the Dow, a low of 6000, and a close of 7300.19 In making this forecast, Acampora had stated: “I think the sentiment out there is there’s too much optimism, and I think this is the year for contrarians.”
Now sentiment is a very important concept in behavioral finance. A consistent t
heme in this book is that sentiment is the reflection of heuristic-driven bias. As we shall see, fundamentalists and technicians both address issues of sentiment, but in quite different ways.
The first half of 1998 followed Acampora’s general script. By July 20, the Dow had reached 9367. But then it began to tumble, closing at 8883 for July.
Keep in mind that Ralph Acampora is a technical analyst. By and large, technical analysts predict the continuation of trends until a clear reversal pattern develops. That is, they follow the maxim “the trend is your friend,” qualified by “trees don’t grow to the sky.” In late July 1998, chart patterns were changing. Support lines associated with the Dow’s movement in 1998 were being approached. Was sentiment changing? The number of stocks whose prices were on the decline was growing, leading technical indicators of breadth to signal that a reversal pattern was underway.
During an August 3 appearance on CNBC, Acampora warned of a “stealth bear market” in which small to midcap stocks would be more vulnerable to a downswing than blue-chip stocks. The very next day he eliminated the word stealth, as the Dow declined by nearly 300 points to 8487. In fact, Acampora was in the process of being interviewed by CNBC, predicting that the Dow would fall by 20 percent from its recent high, when it fell by 60 points. The interview ended abruptly when Acampora declared: “This is going down. It’s going as I’m talking. Got to run.”
Interestingly, the CNBC interview with Acampora, which took place at 2:45 P.M. Eastern time, appears to have induced the media to attribute the market decline to his pronouncements. The August 10, 1998, issue of Barron’s described the events as follows: “Prudential Securities’ Ralph Acampora, a once-obscure practitioner of the voodoo art of technical analysis, basked in the national limelight for more than the requisite 15 minutes after he was credited/blamed for helping trigger Tuesday’s thrashing by bearish comments he made on CNBC late in the trading session. (Critics dub him ‘Ralph Make ’em Poorer.’)”20
It seems that a gauntlet had been thrown, and it was time to duel. In an apparent attempt to counterbalance the concerns stemming from the drop in the market and the negative pronouncements emanating from Acampora, Abby Joseph Cohen emphasized to the Goldman Sachs sales force, and later to the press, that the underlying fundamentals had not changed. Specifically, she pointed out that she saw no deterioration in second-quarter earnings for 1998, a justifiable price/earnings ratio given the unusually low inflation and interest rate environment, and the prospect of stabilization in Asia during 1999.21 Strategists Edward Kerschner of PaineWebber and Thomas Galvin of Donaldson, Lufkin & Jenrette reinforced this analysis.
What we have here is a classic confrontation between those whose market predictions are based on technical analysis and those whose predictions are based on fundamental analysis. In fact, on August 6, 1998, the Wall Street Journal concluded its coverage of Cohen’s remarks by stating: “As for others’ more pessimistic views, she professed to pay ‘very little attention to what others may be saying,’ explaining that she focuses on fundamentals instead of the charts and other statistical material that form the basis for work by technicians like Prudential’s Mr. Acampora.”22
Remember the moral of the pick-a-number game described in chapter 1? Technical analysts seem to attribute a more important role to sentiment in their market predictions than do fundamental analysts.23 Nevertheless, there is one point on which both agree. Sometimes markets reflect changes in sentiment quite apart from any change in fundamentals. Tuesday, August 4, 1998, appears to have been such a time. A day later the Wall Street Journal pointed out that there appeared to have been no apparent news about changing fundamentals, specifically stating “And the sell-off suggested a fundamental shift in mood among investors, not because of its magnitude or breadth, but because there was no precipitating bad news.”24
As I noted in chapter 4, sentiment can influence market prices for prolonged periods. After the Dow declined to 8051.68 on Friday, August 28, Barron’s rounded up the usual suspects to interview. In the August 31 issue, Abby Joseph Cohen is described as having “swiftly pronounced last week’s market slide a sentiment-propelled overreaction.”25 But on the next Monday, the Dow dropped by 512.62 points to 7539.07, 19.5 percent off its July 17 peak and within a whisker of the “bear market” threshold of 20 percent that Acampora had predicted several weeks earlier.
Illusions About Randomness: Lessons to Be Learned from Coin Tosses
Most people have a poor intuitive understanding about the character of random processes and how to predict the future behavior of these processes. Therefore, let me discuss some key lessons about the character of coin tossing and then apply those lessons to predicting the market.
I usually divide students in my MBA classes into two groups. I ask everyone in the first group to take a coin, toss it one hundred times, and record the sequence of heads and tails that result. I ask everyone in the second group to imagine that they are tossing a coin, and to record the outcome of an imaginary sequence of one hundred tosses.
Then I collect the responses of each group, and analyze each student’s response according to the number and length of the runs they have generated. What is a run? Here is a quick example. Imagine a family in which the first three children were boys, and the fourth child was a girl. This family had a run of three boys. For coin tosses, a run is defined as a sequence where consecutive tosses result in the same outcome. So, if someone started out tossing three heads, and had a tail on the fourth toss, then the first three tosses would constitute a run of three heads.
What is the point of asking half the group to toss a real coin and the other half to do so only in their imagination? Simply this: The imaginary tossers do not generate enough long runs. In a hundred real tosses it is unusual to experience as many as thirty runs of length one, and likewise no run longer than four. Yet, the records of most imaginary tossers feature too few runs of length five or more. That is, the records of most imaginary tossers feature too many runs of length one or two.
The real tossers are often surprised to see long runs during their tosses because their experience is at odds with their intuition about random sequences of coin tosses.26 Rather, their intuition conforms to the law of small numbers. People believe that most strings of coin tosses feature about the same number of heads and tails; hence, they expect short runs. That is why most people fall prey to gambler’s fallacy. I think the situation with the market is analogous. Through 1998 the S&P 500 has experienced four consecutive years of gains in excess of 20 percent—quite a run!
For most, an equal proportion of heads and tails corresponds to what they view as the representative toss of a fair coin. Representativeness, in this case synonymous with stereotype, is one of the most common and widespread heuristics. It frequently serves as the basis for how people make predictions. This sometimes works well, and sometimes not.
Consider the following experiment. Suppose that I plan to toss a fair coin one hundred times. Before each toss, you have an opportunity to bet on whether the coin toss will turn up heads or tails. I agree to pay you $1 for each correct prediction and nothing for each incorrect prediction. So, what procedure would you use to arrive at your prediction?
When I play this version of the game in MBA classes, students’ predictions usually feature about fifty percent heads and fifty percent tails, in accordance with what they imagine randomness to have produced. Then we play a second version, where the coin is a little worn on one side, thereby changing the odds slightly: I tell them that the probability of heads has increased from 50 percent to 51 percent. When I ask how people would change their prediction patterns, most say that they might change one or two of their tails to heads, but nothing more. Then I ask how people would change their prediction pattern if the probability of heads moved from 51 percent to 55 percent. Most change a few more tails to heads. But is this the right way to predict?
The optimal prediction pattern for all versions of this game is to predict heads every time. This surp
rises most people, because their instincts are to make their prediction representative of the process they are trying to predict. However an optimal forecast is much less variable than the process being predicted. The key to optimal forecasting is to minimize the likelihood of mismatching; yet a variable forecast does just the opposite.
De Bondt (1991) emphasizes that for all but the shortest-term predictions, statisticians find it difficult to do more than extrapolate the historical rate of growth.27 Remember Ralph Acampora’s eighteen-month prediction for the Dow, made in August 1997? Bullish until June 1998 and then bearish, big time. The Dow has behaved like that in the past. Not this time. The Dow closed 1998 at 9181, up 16.1 percent for the year. But that is not the point. The point is that an efficient forecast does not exhibit as much volatility as Acampora’s forecasts.28
Technical analysts are prone to making excessively volatile predictions because they are like generals who continually fight the last war. Here is an example to illustrate what I mean by that. The August 31, 1998, issue of Barron’s quotes Richard Russell, respected editor and publisher of the Dow Theory Letter, as having said: “Past history suggests that most bear markets wipe out at least half of the preceding bull market.”29 You may ask, what’s wrong with that? The point is that markets behave a lot like coin tosses. Coin tosses produce interesting patterns, but past patterns provide little if no guidance about how to predict the patterns of the future.
Beyond Greed and Fear Page 10