Beyond Greed and Fear

by Hersh Shefrin

  In a pair of papers, Eugene Fama and Kenneth French (1992, 1996) consolidate a large literature and suggest that smaller stocks with low price-to-book ratios are riskier than larger stocks that have high price-to-book ratios. However, Kent Daniel and Sheridan Titman (1997) find that although stock returns are indeed related to characteristics such as size and price-to-book, the characteristics do not seem related to traditional measures of risk. Nevertheless, it is interesting to ask whether we can identify the styles brokerage firms use to recommend stocks.

  Specifically, do the recommended stocks beat the market because the recommendations are for small firms with low price-to-book ratios?

  In a series of articles, John Dorfman provides some insight into many of these questions. When it comes to style, we see a tendency to focus on momentum, the recommendation of “glamour” stocks, with a wide dispersion when it comes to size. Dorfman states:

  [T]he brokerage-house crowd is trying to pick stocks with fast earnings growth, a good “story,” and already-high popularity among investors. The buy ’em-while-they’re-hot method is one legitimate approach to investing, and can work well for nimble traders. …

  [T]he brokerage-house favorites are priced like caviar.

  Start by comparing stock prices with the dividends the stocks pay.

  Over the decades, stocks have usually sold for about 24 times dividends. When the overall market exceeds 33 times dividends, it’s traditionally viewed as a danger zone.

  Currently, the overall market is at 44 times dividends. The Equity Opportunity List at Everen Securities Inc., Chicago, sells for 100 times dividends. The U.S. Priority List at Goldman, Sachs & Co. goes for 93 times dividends. The average for brokerage-house recommended lists is about 64 times dividends, says Rick Chrabaszewski at Zacks.

  Another familiar gauge is the price/earnings ratio, which is a stock’s price divided by the company’s per-share earnings. Over the years, a P/E of about 14 has been average. A P/E below 10 is generally considered quite low, and a P/E above 20 is often considered high.

  These days, the average stock’s P/E is about 18. The average for stocks on brokerage houses’ recommended lists is about 19, and five houses studied have average P/E ratios above 20.13

  Turning to size, we find considerable dispersion. The following excerpt from another column describes the situation.

  In the first quarter, only two firms, both from St. Louis, were able to beat Standard & Poor’s 500-stock index. The “Focus List” at A.G. Edwards Inc.’s A.G. Edwards & Sons chalked up a return of 3.2%, while the “Best Buys” list at Edward D. Jones & Co. was up 3.1%. The S&P 500 was up 2.7%, including dividends.

  The two St. Louis firms picked the largest stocks among the 17 houses, with typical market values of more than $18 billion and $27 billion, respectively. That turned out to be smart: Big-capitalization stocks were in favor, as investors sought their relative safety and stability.

  By contrast, the median market capitalization (share price times shares outstanding) of the stocks recommended by the other 15 firms in the study ranged from $13.8 billion at Salomon Brothers Inc. (a unit of Salomon Inc.) down to $387 million at Wheat First Butcher Singer Inc., according to Rick Chrabaszewski of Zacks. Most brokerage houses fared far worse than the St. Louis pair. Small-capitalization stocks (those with market capitalization under about $750 million) were treated rudely, and technology stocks were bruised. Many brokerage houses have a healthy sprinkling of both on their recommended lists, and they paid for it.14

  So, did the recommended stocks beat the market because they were smaller and had lower price-to-book ratios than the market? Apparently not.

  Momentum

  A common phenomenon in the three types of analyst recommendation studies (Wall Street Journal/Zacks, Wall $treet Week with Louis Rukeyser, and Pros vs. Darts) is momentum. The stocks that get recommended are those that have recently done well. In an important study Narasimhan Jegadeesh and Sheridan Titman (1993) document the existence of a momentum effect.15 Jegadeesh and Titman attribute this effect to the fact that investors underreact to the release of firm-specific information, a cognitive bias.16

  We have already encountered underreaction. In chapters 2 and 4, I discussed the phenomenon of post-earnings-announcement drift—that security analysts underreact to the earnings announcements of firms. As figure 4-2 demonstrates, earnings announcements carry an associated momentum effect. In fact, I devote the next chapter entirely to this issue.

  One thing about momentum is that it requires turnover in order to implement. Do brokerage firms leave their recommendations in place for a long time, or do they change them frequently? With respect to the Wall Street Journal / Zacks study, Dorfman states:

  [B]rokerage houses often act as if most customers were short-term traders. Rick Chrabaszewski of Zacks recently analyzed the turnover rates on brokerage-house recommended lists in the past three years and in the first half of 1997. He found that most brokerage houses are fickle, changing their minds frequently about which stocks are their favorites. Among 16 houses in the study, nine had turnover rates of 100% or more in 1996. Eleven houses are on track to exceed 100% turnover in 1997.

  Such frequent additions and deletions from the recommended list could lead to adverse tax consequences for people who follow the firms’ advice. And it may be emblematic of a general tendency to overtrade—always a temptation for brokers and brokerage houses, which earn a good chunk of their revenue from trading commissions.17

  The study seeks to control for brokerage commissions but not for taxes. Of course, taxes are not an issue for tax-deferred accounts.

  Does momentum represent mispricing? Not all agree that it does. Market efficiency proponents argue that with all those technical analysts looking at relative strength, if it were, mispricing would produce the equivalent of many $20 bills lying on the sidewalk waiting to get picked up. Hence, supporters do not rule out the fact that momentum represents an unobserved risk factor.

  Style and Performance

  The variation in performance across brokerage firms is wide. Figure 7-3 displays the best, worst, and median cumulative performance in the Wall Street Journal/Zacks study, as well as that of the S&P 500. Moreover, the brokerage firms in the study do not display strong herding behavior in the stocks they recommend. In fact quite the opposite. Notably, of the roughly 300 stocks recommended by the firms in the study, only a handful tends to be uniformly recommended by all.18

  Does the difference in performance stem from risk? Did brokerage firms that combined momentum trading with small capitalization and low price-to-book perform best?

  Consider Dorfman’s descriptions of the principles used to pick stocks at the top three firms: PaineWebber, Raymond James, and A.G. Edwards.

  Top performer PaineWebber tended to choose stocks that feature high price-to-book ratios and low dividend yields. John Dorfman wrote: “Its average stock pick sells for 3.3 times the company’s book value, or assets minus liabilities, per share—the highest ‘price-to-book’ ratio among the 10 firms. And the average dividend yield (dividends as a percentage of the per-share stock price) was only 1.6%, the lowest in the field. Edward Kerschner, the firm’s chief strategist, says that’s because the PaineWebber Group unit is emphasizing ‘ruler stocks,’ or stocks whose earnings go up as straight as a ruler. The firm thinks such issues should shine as other companies’ earnings falter this year or next.”19

  Figure 7-3 Performance Range Recommendations by Different Brokerage Firms, January 1993–December 1997

  The cumulative track records for the best, worst, and median performers in the Zacks/Wall Street Journal study, during the five-year period January 1993–December 1997, relative to the S&P 500.

  In August 1994 Dorfman quoted PaineWebber’s research director Ann Knight as saying that her firm is “sticking with two themes”: corporate restructuring and heavy European exposure. He also wrote that PaineWebber tended to pick the most-volatile stocks, stating, “Its selections tended to sh
ow price swings 39 percent wider than those of the overall market.”20

  Second-place finisher Raymond James and third-place finisher A.G. Edwards followed very different strategies than PaineWebber. Both focused on small cap stocks. Those selected by Raymond James were the second most volatile. Dorfman described A.G. Edwards’s approach as follows: “A.G. Edwards tends to select the stocks of relatively small companies, often not widely followed on Wall Street, that meet traditional ‘value’ criteria. For example, it likes stocks that sell at a low multiple of the company’s ‘book value,’ or assets minus liabilities per share. It also favors stocks that pay good dividends. The average dividend yield of its recommended stocks is 3.7%, the second highest in the group after Smith Barney’s 4.1%.”21 In a later article, Dorfman noted: “The St. Louis firm’s “value” orientation has worked well during and after the crash. But it hasn’t always worked. Before the crash, in the final stages of the great bull market, Edwards was an also-ran.22 In terms of price to book and dividend yield, the contrast with PaineWebber could not be more striking.

  Small/value outperformed the pack over the 11-year period 1988–1998 inclusive. But, small/value stocks did not lead the pack over the 6-year period 1993–1998. In fact, for the longer period, 1988–1998, the three brokerage houses switched position, with Raymond James emerging on top, followed by PaineWebber and A.G. Edwards.

  Be it five years after the 1987 crash, ten years after the crash, before the crash: What do the close performance of A.G. Edwards and PaineWebber tell us? One thing it tells us is just how noisy realized returns are. Indeed during the five-year period 1993–1998, overall winner Raymond James placed seventh in a field of 13. This is important, as investors are prone to ask, “what have you done for me lately?”

  Here is another indication of that noise. Consider whether winners repeat on a monthly basis. At the end of each month, divide the brokerage firms into two groups, the top 50 percent and the bottom 50 percent for that month. If performance were completely random, then the chance of making it into the top half during the next month is a fifty-fifty proposition. But if winners tend to repeat, then a firm that placed in the top half last month should stand a better-than-even chance of making it into the top half the next month as well. Over the sixty months between January 1993 and December 1997, there was a 49.4 percent chance that a winner in one month would repeat in the following month: not only close to even, but on the wrong side.

  Behaviorally-Based Theories

  The proponents of market efficiency hold that there are enough well-informed investors to seize all unexploited profit opportunities. The evidence from behavioral decision-making studies is that people learn slowly. Are there enough quick learners to eliminate mispricing in financial markets? That is an empirical issue.

  In the 1980s, academic studies of security pricing specifically based on the findings described in the behavioral decision-making literature began to appear. These studies provided support for “value investing,” an idea described in considerable depth by Benjamin Graham and David Dodd (1934) in their classic book Security Analysis. The logic behind value investing, as explained by Graham (1959), is as follows: “The market is always making mountains out of molehills and exaggerating ordinary vicissitudes into major setbacks.” (p. 110). In other words, investors overreact to negative news.

  Money manager David Dreman, who began to argue in 1978 that stocks with low P/E ratios were undervalued, provided further support for value investing. Dreman used the term investor overreaction hypothesis to describe the tendency of investors to become unduly pessimistic about the prospects for low P/E stocks. Since the crowd avoided them, investing in low P/E stocks became a contrarian strategy.23

  Academics Werner De Bondt and Richard Thaler (1995) describe how in 1985 (p. 394) they extended Dreman’s reasoning to predict a new anomaly.” De Bondt and Thaler hypothesize that because of representativeness, investors become overly optimistic about recent winners and overly pessimistic about recent losers. Hence, De Bondt and Thaler propose buying past losers and selling past winners.

  De Bondt and Thaler define winners and losers according to past performance over the three previous years. Their study focuses on extreme winners and losers—the top and bottom 10 percent. As I discussed in chapter 4, extreme past losers tend to outperform the market over the subsequent five years by about 30 percent, and extreme past winners tend to underperform by about 10 percent. (See figure 4-1.) According to De Bondt and Thaler, too little smart money takes advantage of the profit opportunities created by investors who are misled by representativeness.

  The evidence supporting the P/E and winner-loser strategies described above are based on realized returns, not on the way that investors formulate return expectations and perceptions of risk. It does appear that low P/E stocks and past losers tend to outperform the market. But do investors actually have low expectations for the future returns on these stocks? Or are low P/E stocks simply riskier?

  A Tale of Two Stocks

  Michael Solt and Meir Statman (1989) wrote that the stocks of good companies are bad stocks. In a 1995 article, Statman and I amplified the argument (Shefrin and Statman 1995). Why do investors appear to cling to the idea that good stocks are the stocks of good companies, and vice versa?

  Is the answer any more complicated than representativeness? In investors’ minds, good companies are representative of successful companies, and successful companies generate strong earnings, earnings that in turn lead to high returns. On the other hand, poor companies are representative of low earnings and disappointing returns. Investors shun the stocks of poor companies as a group, and therefore they come to be underpriced.

  To understand this point, let us examine two companies, Dell Computer and Unisys. Dell has a strategy that is simple to describe. They sell custom-made computers directly to their customers. In June 1997, investors were looking at a company whose sales had risen by 47 percent in the previous fiscal year, and whose earnings per share had doubled. The stock had been soaring, providing a rate of return of 161 percent over the preceding three years. It goes without saying that Dell Computer is representative of a successful company.

  Unisys is the product of a 1986 takeover by Burroughs of Sperry Univac. Both were struggling computer companies at the time, and after the merger they became one large, struggling computer company. In June 1997, the picture did not look as rosy for Unisys as it did for Dell. Unisys had lost billions of dollars since 1990, its market share had fallen, its stock price was depressed, and it had been unable to achieve stable revenue growth. Its CEO at the time, James Unruh, had announced his resignation. In order to enable the company to survive, Unruh had lead four major downsizings, amounting to a 70 percent reduction in the workforce. I would venture to say that Unisys is representative of an unsuccessful company.

  Now the question is, How do investors form their return expectations for these two stocks? What does the capital asset pricing model (CAPM) have to say? In June 1997, the beta for Dell was 1.6, and for Unisys it was 1.9. The three-month Treasury bill was yielding 5.16 percent at that time. If, for the sake of argument, we take the equity premium to be its historical 8.7 percent, then investors would have expected Dell stock to return 19.1 percent, and Unisys to return 21.7 percent.

  From an efficient market perspective, beta may not reflect all the risk to which these stocks are exposed. We would have to check the exposure of the two stocks to the various risk factors. For example, being a value stock, Unisys might be exposed to additional risk because it had an especially high value for book-to-market equity. As for Dell, the price of its stock did not just soar on a three-year basis: the rate of return in the first six months of 1997 was 48.6 percent. Given this momentum, Dell might be exposed to risk captured by a momentum factor.

  In June 1997, I conducted a survey to elicit investors’ expectations about returns for the stocks of eight technology companies. There were twenty-nine respondents, all living and working in Silicon Valley
, and all familiar with technology; most were working in that industry. Their ages ranged from 25 to 40, and the median income of the group was $80,000 per year. All were students in the MBA program at Santa Clara University, were at a stage in the program where they were familiar with standard investment concepts, held individual stocks, and were familiar with the two companies.

  The respondents expected that Dell would return 20.9 percent in the period July 1997 through June 1998, a little larger than the 17.6 percent predicted by the CAPM. But what about Unisys? Here the story was quite different. The respondents indicated that they expected its stock to return 6.3 percent over the period July 1997 through June 1998.

  According to market efficiency, this would suggest that the respondents regard Unisys to be a very safe stock, so safe as to only require a risk premium of just over 1 percent! From the behavioral perspective, investors expect a low return on Unisys stock because Unisys is representative of a bad company, and investors believe that stocks of bad companies are bad stocks.

  Were these return expectations generated by considerations of risk or representativeness? Which explanation sounds more plausible? To me, it seems clear that representativeness is more plausible than risk.

  Evidence from Executives and Analysts

  It may be that the subjects in my survey had it all wrong. But if so, they had plenty of company. Every year since 1982, Fortune magazine has conducted an annual corporate reputation survey. In the Fortune survey sell-side analysts, buy side analysts, corporate executives, and members of boards are asked to rate companies in their industry on a variety of attributes. One attribute concerns the company’s stock in terms of value as a long-term investment (VLTI). Respondents rate each attribute, including VLTI, on a scale of 0 to 10.