In the Fortune survey published in March 1997, Unisys stock received a very low score of 4.0 on VLTI. In contrast, Dell received a 6.57.24 Is some general phenomenon at work here? In examining the VLTI responses for the full range of Fortune magazine surveys, Meir Statman and I (Shefrin and Statman 1998) find the following: Respondents expect that past winners will continue to be winners and that past losers will continue to be losers. Respondents also expect that high P/E stocks will outperform low P/E stocks.25
It is noteworthy that analyst recommendations feature the same patterns. In examining the behavior of recommendations tracked by First Call, Meir Statman and I find the same effects. Analysts recommend the stocks of past winners (high P/E stocks) more highly than they do the stocks of past losers (low P/E stocks).
Advocates of market efficiency maintain that losers are riskier than winners, and low P/E stocks are riskier than high P/E stocks. Is this reflected in investors’ risk perceptions? To address the issue, I included some risk perception questions in the investor expectation surveys that I disseminated. My subjects did perceive that losers are riskier than winners. And they perceived that low P/E stocks were riskier than high P/E stocks.
On its own, my survey results offer support for market efficiency—but not when return expectations are taken into account. Taking account of return expectations, I find that investors believe riskier stocks to have lower expected returns.
The Best of Stocks, the Worst of Stocks?
The Dell-Unisys story offers additional insights. Like the Fortune magazine survey respondents, analysts have been more favorably disposed toward Dell than Unisys. In June 1997 the consensus recommendation for Dell was between a buy and a strong buy, but closer to buy. For Unisys, it was between sell and neutral, but closer to neutral. However, since most analysts are known to mean sell when they say neutral, we can safely interpret the recommendation for Unisys to have been a sell.
Behavioral explanations such as the one advanced by De Bondt and Thaler (1985) postulate that investors overreact. When investors are pessimistic, they are overly pessimistic; when they are optimistic, they are overly optimistic. For instance, at the end of June 1997 Dell stock was a winner. During the preceding three years, it had returned 150.5 percent. On the other hand, Unisys was a loser, having returned a negative 2.8 percent.
What were the experiences of Dell and Unisys between July 1997 and June 1998? Dell continued to soar, returning an astonishing 216 percent, prompting Fortune magazine journalist Andy Serwer (1998) to wonder whether Dell would be “crowned stock of the decade, as in the best-performing stock of the S&P 500.” As for Unisys, over the same period it returned 270.5 percent, beating Dell by 54 percent.
Overreaction and Underreaction
The winner-loser effect is puzzling in that if winners and losers are defined in terms of one-year past returns, rather than three-year past returns, an underreaction effect emerges, not an overreaction effect.26 See Navin Chopra, Josef Lakonishok, and Jay Ritter (1993). What we seem to have is overreaction at very short horizons, say less than one month (Lehmann, 1990), momentum possibly due to underreaction for horizons between three and twelve months (Jegadeesh and Titman 1993) and overreaction for periods longer than one year (De Bondt and Thaler 1985, 1987, 1990). This phenomenon is quite complex, and does not lend itself to easy explanations. Chapter 8 will address some of the explanations that have been put forward.
The Three-Factor Model
Proponents of market efficiency have their own theory about value investing. That theory is the Fama-French three-factor model (Fama and French 1992, 1996). Of the three factors, one pertains to book-to-market (the inverse of price-to-book), one to size, and the third to the return on a proxy for the market portfolio. These factors represent sources of systematic risk, meaning risk that is priced.
In the three-factor framework, the average small stock is riskier than the average large stock, and therefore tends to earn a higher return. Consider a security whose return behaves more like a small stock than a large stock. A portion of its return will reflect the premium that small stocks earn over large stocks. The same statement applies to a security whose return behaves like a value stock, where value is measured by book-to-market. Securities that are neutral when it comes to size and book-to-market would earn the market return on average.
According to the efficient market school, size and book-to-market reflect systematic risk, meaning risk that requires compensation in the form of higher expected returns. If this is the case, then what we should find is that investors perceive small-value stocks to be riskier than large-growth stocks.27 And we do, which on its own lends support to market efficiency. But investors consistently expect large-value stocks to outperform small-growth stocks. Likewise, analysts tend to recommend growth stocks more favorably than they do value stocks. In the efficient-market paradigm, expected return and perceived risk should be positively related, not negatively related.
Is beta dead? The existence of three factors, rather than one, implies that beta is at least crippled. But the casualty is traditional beta. At this point, you might be wondering what other beta there might be. Statman and I (Shefrin and Statman 1994) propose the notion of a behavioral beta. We analyze how the concept of beta needs to be adapted to reflect mispricing as well as fundamental risk. The traditional beta is defined relative to a proxy benchmark for the market portfolio, and is a suitable risk measure for a world where prices are efficient. But for a behavioral beta, the benchmark portfolio needs to change so that it tilts in the direction of underpriced securities and away from overpriced securities.
I suggest that tilting is why size and book-to-market give rise to factors that explain realized returns. Consider the book-to-market factor HL, the difference between the returns to high book-to-market stocks minus the returns to low book-to-market stocks. Does this seem like it captures the effect of tilting? How about factor SB, the return to big stocks minus the return to small stocks?—doesn’t this capture the effect of tilting as well? Kent Daniel, David Hirshleifer, and Avanidhar Subrahmanyam (1999) argue that this is the case.
Misinterpreting the Evidence About Overreaction and Underreaction
Fama (1998a, 1998b) claims to be unpersuaded by the evidence coming out of behavioral finance. He raises concerns about both methodology and issues of interpretation. On the methodology side, he argues that behavioral finance needs to impose a clearly defined alternative hypothesis to market efficiency, one that is narrowly constructed. By this, I think he means that a useful theory tells us that only a small set of well specified phenomena are possible. A less useful theory tells us only that many vaguely defined phenomena are possible. That is why Fama suggests that behavioral finance adopt a specific alternative such as overreaction.
Fama contends that the anomalies literature has not accepted the discipline of an alternative hypothesis, although he does refer to one exception, presented in a paper by Josef Lakonishok, Andrei Shleifer, and Robert Vishny (1994). These authors propose that investors use sales growth and earnings growth as measures of past performance and that ratios involving stock prices, like P/E and book-to-market equity, proxy for future performance. They suggest that many firms having a high P/E or a low book-to-market equity will have experienced strong past earnings and sales growth. Lakonishok, Shleifer, and Vishny call companies with strong past performance and expected future performance glamour stocks. They suggest that investors form expectations by naively extrapolating past performance. Consequently, glamour stocks come to be overpriced, and poor performers eventually come to be underpriced. This leads investors to predict that value stocks will outperform glamour stocks, a prediction that can be tested and refuted if the data do not support it.
Lakonishok, Shleifer, and Vishny test the prediction that value stocks outperform glamour stocks. They do not test the hypothesis that investors naively extrapolate past performance. But because we look at expectations data, Statman and I can test the prediction about investors’
expectations. What we find is that investors attach higher expected returns to stocks that have experienced higher past sales growth (or earnings growth) (Shefrin and Statman 1998).28
Fama makes two important points about a well-specified alternative hypothesis, especially one such as overreaction. First, this particular hypothesis, if formulated to hold regardless of horizon, will be rejected by the data. This we already know from the results on momentum. Second, Fama argues that because of random variation, realized returns will differ from expected returns, and expected returns are the focus of market efficiency. What market efficiency does predict about the spectrum of returns across securities and time is that they will exhibit no systematic deviation from their efficient means. And this leads him to make a novel interpretation of the findings on overreaction and underreaction.
Fama (1998a, 1998b) argues that “apparent overreaction of stock prices to information is about as common as underreaction.” Moreover, this is “[c]onsistent with the market efficiency hypothesis that the anomalies are chance events” (1998a, p. 16).
It seems to me that this argument would make sense if about as many stocks displayed long-term underreaction as long-term overreaction, or if about as many stocks displayed short-term underreaction as short-term overreaction. But that’s not what we see. Rather, what we find is apparent underreaction at short horizons and apparent overreaction at long horizons.29
Regret and Hindsight Bias
The data on return expectations tell us that investors are quite consistent in the mistakes they make. Investors believe that large-growth stocks will outperform small-value stocks. They tend to believe that past winners will continue as winners, past losers will continue as losers, and strong revenue growth in the past will lead to strong revenue growth in the future. A byproduct of this is that investors expect to earn higher returns on stocks they perceive to be less risky.
Efficient market theory predicts that mispricing is eliminated for the same reason that people pick up $20 bills lying on the sidewalk. But these errors are very difficult to override, even when you know about them. To appreciate just how difficult, consider the following excerpt from a January 1988 Wall Street Journal column by Barbara Donnelly, who had interviewed both De Bondt and Thaler about investing in losers.
“It’s scary to invest in these stocks,” says Prof. Thaler. “When a group of us thought of putting money on this strategy last year, people chickened out when they saw the list of losers we picked out. They all looked terrible. …” [A]dds Prof. De Bondt, “The theory says I should buy them, but I don’t know if I could personally stand it. But then again, maybe I’m overreacting.”30
There are two other behavioral phenomena lurking in these quotations: regret and hindsight bias. Some losers will continue to be losers, maybe most of them. The De Bondt-Thaler strategy may not work in any given year. If it doesn’t work, hindsight bias may well set in. It will look obvious that these stocks were going to bomb. And the investors who bought losers will probably feel like fools and experience the pain of wishing they could turn back the clock and do it all over again.
Summary
The weight of the evidence—the success of value investing over the long term, post-earnings-announcement drift, and post-recommendation drift—all go against market efficiency. As I emphasized in chapter 4, this reflects a cause-and-effect relationship. Heuristic-driven bias causes prices to depart from fundamental values.
Fama’s defense of market efficiency would have us believe that representativeness-based errors are nothing more than random chance, or that there are enough investors immune to these errors to exploit them and correct any mispricing in the process. The latter, I believe, is the nub of the issue—whether the market holds enough smart money, enough investors able to learn quickly and overcome biases.
The evidence suggests that not enough smart money exists to eliminate market inefficiency. Representativeness, regret, and hindsight bias are very powerful, and most people are overconfident. The idea that self-interest will induce most people to learn to avoid errors is an empirical proposition, not a universal truth; furthermore, the evidence hardly supports it. In fact, most people are subject to the illusion of validity described in chapter 6. They emphasize evidence that confirms their views and downplay evidence that does not. Even proponents of market efficiency may not be immune from the illusion of validity.
At the same time, it is harder to beat the market than most people think. That is an important reason why the moral of this chapter is not that investors can use behavioral finance to make a killing. I think most investors would be better off holding a well-diversified set of securities, mainly in index funds, than they would be trying to beat the market. In other words, they would be better off acting as if Fama were right, that markets are efficient.
In truth, real-world performance is complicated. Stocks recommended by the major brokerage houses and those endorsed on the television program Wall Street Week with Louis Rukeyser did beat the market—and not by investing exclusively in small-cap value stocks. In fact, just the opposite: recommendations emphasized momentum investing, large cap, and growth.31
The moral of the story, for most investors, is not to be overconfident. Markets may fail to be efficient, but that doesn’t mean it’s easy to beat the market—either by oneself or by relying on the advice of some guru. This statement applies even to the recommended stocks tracked by the Wall Street Journal/Zacks study. If an investor picked just one brokerage firm in 1986 and stayed with it for the duration, the odds of beating the market were no better than even. Why? Because only half the brokerage firms recommended stocks that beat the market.
What about investing based on learning some behavioral finance? Well, understanding the relevance of representativeness means having a little knowledge, and you know what they say about a little knowledge: It’s a dangerous thing.
Chapter 8 Biased Reactions to Earnings Announcements
Mispricing is complicated. Sometimes mispricing results in reversals, while at other times it results in momentum.
Momentum and reversals coexist, despite lying at diametrically opposite ends of the spectrum. The academic scholars participating in the market efficiency debate have been grappling with what this coexistence means. Proponents of market efficiency view these phenomena as nothing more than random deviations from efficient prices. On the other hand, proponents of behavioral finance view them as systematic departures from efficient prices.
To shed additional light on the coexistence issue, I devote this chapter to post-earnings-announcement drift, a phenomenon that involves both reversals and momentum. What underlies reversals and momentum associated with earnings announcements? Are they nothing more than random disturbances? Or are they systematic, an effect caused by heuristic-driven bias? I believe that the evidence supports heuristic-driven bias stemming from conservatism—anchoring-and-adjustment—overconfidence, and salience.
This chapter discusses the following:
• a case that enables us to get a closer look at both momentum and price reversal
• the academic literature dealing with post-earnings-announcement drift
• the experience of one particular money management firm whose trading strategy is based on post-earnings-announcement drift
• recent theoretical work addressing momentum and overreaction
Case Study: Plexus Corporation
Plexus Corporation, located in Neenah, Wisconsin, is a contract provider of design, manufacturing, and testing services. Plexus develops, assembles, and tests a variety of electronic component and subsystem products for major corporations in industries such as computer, medical, automotive, and telecommunications.
At the beginning of 1997 Plexus was an $87 million company being followed by a single analyst, Robert W. Baird and Co. A year later, its analyst coverage had expanded from one to five, owing to a series of major earnings surprises. The surprises gave rise to price momentum, which was then followed by a shar
p reversal.
Brief history: For the quarter ending December 31, 1995, Plexus’s earnings per share (EPS) were 11¢. The EPS forecast for December 31, 1996, was for slightly less than double that amount. During the first week of January 1997, a report from Robert W. Baird & Co. said the company was on track to meet, or possibly exceed, its first-quarter earnings projection, which would reflect earnings growth of 26 percent from the prior year.
The first surprise: On January 16, 1997, Plexus registered a significant surprise when it reported earnings of 40¢. How did the market react that day? Volume was certainly dramatic. The day before the announcement, volume stood at 19,200 shares. On the day of the announcement, volume jumped to 420,500, and then on the subsequent three days volume was 718,400, 391,000, and 142,600 respectively.
But on the day of the announcement, the closing price only jumped $2.00 a share, from $25.50 to $27.50. Over the following three days, it climbed only a little more, to $28.75. Unlike the change in volume, the price change could hardly be called dramatic. Perhaps this was because the increase had already been anticipated. On Friday, December 20, the stock closed at $17.75, a little less than its all-time high of$19.00. Yet when the market reopened the following Monday, the stock soared to $26.00 on a volume of 54,800 shares. (See figure 4-2.)
Expanded coverage: As the market value of Plexus grew, so too did its analyst coverage. A month after the January surprise, on February 10, Stephens Inc. initiated coverage of Plexus Corp. with a “buy” rating. A report on Dow Jones News Service stated: “In a research note, the firm said Plexus should see rapid earnings growth in 1997 from its focus on increasing operating efficiency, reducing fixed costs and improving cash management. Stephens estimated Plexus will earn $1.60 a share in fiscal 1997 and $2.00 a share in fiscal 1998. The company’s fiscal year ends in September. The firm set a 12-month price target of 40.”1
Beyond Greed and Fear Page 14