Contrarian Investment Strategies Page 5
Next we’ll look at four important forms of Affect, all of which have bloodied investors’ portfolios over time: (1) insensitivity to probability, (2) negatively correlated judgments of risk and benefit, (3) the Durability Bias, and (4) Temporal Construal.
1. Insensitivity to Probability
There are a number of key ways in which Affect leads to errors in judgment. The most important is that it causes us to be insensitive to the true probability for investments to increase (or fall) in price while not factoring in the reasons this should happen. When a potential outcome, such as a major gain from a stock purchase, carries sharp and strong affective meaning, the actual probability of that outcome, or changes in the probability due to changing circumstances, will tend to carry very little weight.6
Insensitivity to probability is backed by strong research findings.7 Professors George Loewenstein, Elke Weber, Christopher Hsee, and Ned Welch8 conducted a fascinating study that showed that if people think they are going to win a state lottery, their bets and their expectations about the chances of winning are likely to be similar whether the probability is 1 in 10,000 or 1 in 10 million. If people feel they are going to win, they can be willing to pay up to 1,000 times more for the same lottery odds! Interestingly, these figures are in line with what speculators paid to buy the hot stock of the day in a stock mania. Loewenstein and his coauthors note that what is going on here is that gamblers are more moved by the possibility, rather than the probability, of a strong positive consequence. The result is that very small probabilities carry great weight.
Another interesting study found that investors apparently did not care what price they paid for an exciting initial public offering (IPO) in a bubble market. The distinguished economist Robert Shiller demonstrated that if an investor wanted to buy a hundred shares of a company (say, at $10), it didn’t matter to him if the company had one million shares to sell or, having split just prior to the IPO, five million shares to offer. Shiller found that quintupling the price of the outstanding stock was immaterial to the IPO buyer, who still wanted to buy the same one hundred shares at $10 a share even though they gave him only one-fifth of the previous value, because he was convinced the price would go higher.
Insensitivity to probability is backed by other strong research findings. Yuval Rottenstreich and Christopher Hsee9 demonstrated that if the potential outcome of a gamble is emotionally powerful, its attractiveness (or unattractiveness) is relatively insensitive to changes in probability as great as from .99 to .01, or 100-fold. These findings go to the heart of the overvaluation in a bubble. If we have very strong feelings about the prospects of a stock or another investment, we will sometimes pay 100 times its real value or more. This finding captures the major reason why stock prices are driven to astronomical heights during a bubble.
Figure 2-1 shows just how dead-on these findings are. If anything, as we’ll see, overvaluations can be even greater than 100 times earnings. From the beginning of 1996 through the peak in March 2000, the NASDAQ 100 made up primarily of the largest dot-com and high-tech stocks increased 717 percent. From the high-water mark of March 2000, the index dropped 83 percent to its October 2002 low, the worst decline of a major U.S. market index since the 89 percent drop in the Dow Jones Industrial Average in the 1929–1932 period. Near the height of the bubble, the NASDAQ 100 had a P/E multiple of over 200 times earnings. Virtually all of the companies in the index had products or services that projected images of rapid growth, often far removed from any realistic chance of being attained.
Two fairly typical examples of the extent of investors’ enthusiasm for these stocks follow.
Example A: America Online (AOL) had a P/E of 200 times trailing earnings in March 2000. The company had shown spectacular earnings growth in the previous six years, and analysts believed the growth rates were likely to accelerate as millions of new customers signed up annually to use its popular online services. Using a standard earnings discount model, I calculated that to justify its then-current price, it would need approximately 18 billion subscribers, or roughly three times the population of the earth. My conclusion at the time was that a large extraterrestrial population needed to be discovered quickly to meet these “modest” growth goals.
Soon after, AOL merged with Time Warner. Much heavier than anticipated competition worldwide, a severe slowdown in online growth, and accounting methods that dramatically overstated earnings and had to be subsequently sharply revised downward, resulting in a major drop in price of the shares of the merged company. From a price of $100 in the first quarter of 2000, the value of the AOL portion of the company fell about 90 percent to its low.
Example B: eToys.com was an exciting new dot-com merchandiser that sold a large variety of toys online. eToys.com’s concept was not only to provide shoppers with the largest variety of popular toys available but also to be a substantial time-saver for users of the site. Moreover, users would benefit from discounts that would underprice almost all its other major competitors.
The fact that a number of large competitors already had or were constructing powerful online sites was dismissed by most analysts and investors, as was the fact that eToys.com’s discounting policy resulted in large losses for the firm because it did not have the sales volume to receive the large discounts that its competitors did from the major toy manufacturers. At its peak in October 1999, eToys.com had a market value of $10.7 billion, more than triple that of Toys ‘R’ Us, the largest toy retailer, which had many hundreds of stores nationwide. eToys.com’s sales were less than 1 percent of those of Toys ‘R’ Us. The company, as noted, operated at a loss, whereas Toys ‘R’ Us had a long record of profitability. eToys.com was also conceived and operated by a middle-level retail executive, and its management depth was limited. More accurately but less kindly, its management was so-so at best.
With no real business plan, the company continued to generate large losses and went into bankruptcy in 2001. All those facts were readily available at the time; however, analysts, money managers, and investors did not process the information because of the very powerful positive affective image the company conveyed, and continued to believe that the stock was an outstanding investment opportunity almost all the way to the funeral. Once again, it seems clear that when positive Affect is in high gear, rational and logical analysis is often overwhelmed by the experiential system.
Although the two stocks had unique individual characteristics, both were part of the NASDAQ Composite Index. Along with many other high-tech and dot-com issues, they were markedly overvalued through this time period. The sensitivity to the possibility of a major gain, rather than the probability of one, appears to have played an important role in the stupendous overvaluation of dot-com and high-tech stocks in the 1996–2000 dot-com bubble.
Table 2-1, which I originally constructed in November 1999, near the height of the Internet bubble, shows the price-earnings ratios (P/Es) of ten of the most popular Internet stocks at that time. The classic two-tier market of 1971–1974, which focused entirely on large, rapidly growing companies, had an average P/E of 51 for the fifty leading growth stocks at its peak, well above the normal P/E of such growth stocks of 25–35х earnings. After the stocks dropped drastically in the 1973–1974 bear market, they were long held up as the leading example of investors’ paying far too much for prospective growth.
Not so in the 1996–2000 dot-com bubble. The price-earnings ratios in Table 2-1 were as high as 1,930 times earnings; the average P/E of the group was an amazing 739. The magnitude of the excesses of this bubble can be gauged from the fact that the average P/E of the ten dot-com companies shown in Table 2-1 was fourteen times as high, on average, as the average P/E of fifty-one of the “Nifty Fifty” stocks of the two-tier market of 1971–1974. The 1996–2000 bubble companies were not small or unknown but had market capitalizations ranging from $1.6 billion to $30.1 billion, larger than the average company in the S&P 500.
We decided to find out what the fundamental value of the stocks i
n the table really was in October 1999, near the top of the bubble. The analysts’ consensus earnings estimates for 1999 were used as the starting point, and then the highest earnings growth rates in U.S. company history were applied for the next twenty-one years for each company in the table, after which the normal earnings growth rate of the S&P 500 was applied. The earnings growth rates attributed to the exciting concept companies were almost farcically high (see the assumptions in Table 2-1), demonstrating that even if earnings had met the almost impossible targets, which almost no company had ever achieved since the founding of the nation, the stocks would still be wildly overpriced.
The stock prices we created in column 3, which showed somewhat lower but still wildly bullish evaluations, were derived by using the discounted earnings of each company. To arrive at these prices, we used a simple earnings discount model, one of the basic methods of valuing a stock’s price.*4
Compare column 3 with column 1, the then current price of each company. The stocks in column 1 were trading from fourteen times the highest valuations the model gave a company in column 3 to slightly over two times as much on the lowest. Although the prices we gave the column 3 group, based on the hyperexpanding earnings, were preposterously high, the actual company prices in column 1 almost doubled again by March 2000, catapulting the prices of dot-com stocks sharply higher again.
Column 4 of Table 2-1 (the one column added since the 1999 presentation) shows the prices of the stocks on August 31, 2002, after the bubble imploded. The average decline of the group was 79.1 percent. Only one stock, eBay, remains well above the present-value model estimates established in column 3. One stock is a little less than 1 percent above the prices in column 3, while the remaining eight are below these estimates, some very substantially. As noted, an estimated $7 trillion was lost in the bubble in high-tech stocks and the market. As a comparison, the loss from the market’s 1987 crash from its peak to its nadir was $1 trillion.
2. Judgments of Risk and Benefit Are Negatively Correlated
Does Affect also blind us to the risk of a security or of our entire portfolio? Our investment teachings clearly state that it doesn’t. After all, risk theory has existed for fifty years, and there are untold numbers of antirisk defenses out there to protect risk from attacking our portfolios. The efficient market hypothesis (EMH) and the great preponderance of modern risk theory, used by investors, their advisors, and their mutual funds, believe that risk is solely volatility. Unfortunately, recent work on Affect provides strong evidence that the defenses most people use will not do their job.
EMH argues that the greater the risk taken, the higher the perceived rewards of an investment will be. Affect theory discovered that it doesn’t quite work this way. Professors Baruch Fischhoff, Paul Slovic, Sarah Lichtenstein, Stephen Read, and Barbara Coombs10 found that the judgments of risk and rewards are negatively correlated. That is, the greater the actual risk, the smaller the perceived gain, while the smaller the perceived risk, the greater the perceived gain. In a word, modern risk theory is turned upside down. And research into the role of Affect appears to explain why.
These findings were supported in numerous follow-up papers. Researchers asked subjects how risky various activities were. Repeatedly, subjects answered that for many potentially dangerous or hazardous situations the greater the perceived benefit or reward, the lower the perceived risk. In virtually every case the risk is there but is perceived differently depending on the magnitude of the benefits. Conversely, the lower the perceived reward or benefit, the greater the perceived risk. Researchers, for example, have found that alcoholic beverages and food additives are believed to be low in benefit and high in risk, whereas vaccines, antibiotics, and X-rays tend to be seen as high in benefit and low in risk. In the investment world higher risk is believed to have a close correlation to higher return. This psychological behavior also simply cannot exist according to current investment teachings.
Slovic, MacGregor, Malmfors, and Purchase11 surveyed members of the British Toxicological Society and found that these experts, too, produced the same inverse relation between judgments of risk and benefit.*5 12 The behavioral researchers state that a negative correlation between risk and reward occurs even when the nature of gains or benefits is distinct and qualitatively different from the nature of the risks. What is difficult for investors to understand is that psychologically risk and return appear to often be negatively correlated. In markets we believe that rationally they have to be positively correlated.
A research paper by Ali Alhakami and Paul Slovic13 indicates that the perceived risk and perceived benefit of an activity (e.g., using pesticides) were linked to the positive or negative Affect associated with the activity. The result implies that the more strongly people like an activity, such as the use of an untested treatment for cancer or the purchase of a dot-com stock, the more they judge the risks to be low and the benefits high. Conversely, the more they dislike activities, such as using coal as a source of energy, drinking alcoholic beverages, or buying stocks with so-so returns, the higher the risk levels they attribute to these.
According to Alhakami and Slovic, this result implies that people base their judgments of the risk and benefits of an activity or a technology not only on what they think about it, but also on how they feel about it. If they have an idea or concept they strongly like, they are moved to judge that the risk is low. The more they dislike an idea or concept, the higher they judge the risk. So Affect again enters into the picture, this time allowing our feelings to tamper with and alter our rational decision making and choices on risk. In the realm of finance, Yoav Ganzach14 found support for concluding that if stocks were perceived as good, they were judged to have higher returns and low risk, whereas if they were perceived as bad, they were judged to be lower in return and higher in risk. However, for familiar stocks, perceived risk and return were positively correlated, rather than being driven by this attitude.
This work is important in explaining the strong role of Affect in the perception of risk in Figure 2-1 and Table 2-1. In both cases the Affect about the investments was very positive. As indicated, it appeared that stocks were priced at many times their real values; in the face of widely followed investment principles, the risk of investing in them was probably overriden by the strong positive Affect for them. Analytical or logical comparisons, which showed how overvalued top NASDAQ stocks were in late 1999 (Table 2-1), failed to change investors’ opinions that they were low-risk holdings.
In making their evaluations, swayed by Affect, the analysts probably really did not believe the risk factor was unduly high relative to the expected returns. In my reading of large numbers of analysts’ reports near the height of the dot-com bubble, I saw very little analysis of the major risks involved. This point is important in gauging the effectiveness of modern security analysis, and the rational-analytic processes more generally, when emotional influences such as Affect are involved. The risks were often obvious, as in the case of eToys.com and many hundreds of similar stocks of companies that had limited finances, questionable management, and poor business plans but skyrocketing stock prices. Almost all these factors could be checked quickly by any competent researcher but were not, or the information did not register on the researchers. Instead, the great majority of reports discussed the enormous potential returns ahead and detailed why they should occur.
Conversely, in the 1996–2000 bubble, value stocks, which were far less risky by standard valuation yardsticks, had very negative Affect attached to them. As a result, the influence of Affect on most investors at the time made them believe that those stocks were far more risky than their valuation standards would indicate. They were often thought to be significantly more risky than IPOs or dot-com and high-tech stocks, both of which subsequently collapsed. The risk-reward situation had obviously been turned upside down during the dot-com mania. But as we’ll see, though a bubble brings these relationships out with far greater clarity, they are always there, opening the road to majo
r opportunities.
When the Internet bubble imploded in the spring of 2000, the supposedly significantly more risky, low-reward value stocks showed stunning positive reappraisals as the carnage in dot-com and other favorites was inflicted. Once again in the investment world, we see a clear case of Affect overpowering the rational-analytic approach, and probably to a degree that has not been observed elsewhere as yet in experiential experiments. Shortly thereafter, more normal evaluations began to be applied again to both dot-com and value stocks, resulting in the reversal, to a major extent, of the valuations during the 1996–2000 bubble.
Unfortunately, before that, the strong influence of Affect on many investors, including large numbers nearing retirement, resulted in enormous losses. The outcome, reported repeatedly in the press, is that millions of people have had to continue to work for extended periods. Introducing findings from the research on Affect appears to provide a plausible explanation of this harmful phenomenon.
Though the linking of strong positive Affect with limited risk in investors’ minds, and of negative Affect with high risk, appears far clearer in a bubble environment, it also seems to be present in many other types of market circumstances, as will be discussed. As we saw on page 39, the risk-Affect correlation also has important implications for contemporary risk management and financial theory.
3. The Durability Bias
Another way that Affect tends to skew our assessments in the market arises because we tend to overestimate how long a positive or negative event or earnings surprise will have an impact on a stock and industry or the entire market itself. Professors D. T. Gilbert, E. C. Pinel, T. D. Wilson, S. J. Blumberg, and T. P. Wheatley observed a consistent tendency of market participants to overestimate the length of time a positive or negative Affect would last after experiencing a pleasant or unpleasant event.15 This effect is called the “Durability Bias.”