Beyond Greed and Fear

by Hersh Shefrin

  The sentiment index does a good job of predicting the past. It only has difficulty predicting the future. Indeed, bullishness appears to have a backward-looking horizon of about twelve months. The two most recent months exert the strongest impact on bullish sentiment, with the market movements in the other ten months exerting a moderate influence.

  Not surprisingly, market moves from three months ago affect the degree of bullish sentiment today more than market moves of the same magnitude twelve months ago.

  The logic behind treating the Bullish Sentiment Index as a contrarian indicator seems compelling. Most investors are trend followers, or so the logic goes, and believe in technical analysis proverbs such as “the trend is your friend” and “don’t fight the tape.” Hence, we might naturally expect newsletter writers to be trend followers as well. But are they?

  Clarke and Statman (1998) analyze how sentiment responds to moves in the market. Keep in mind that bulls forecast that the recent upward trend in the market will continue, while bears and chickens predict that the recent upward trend will reverse. Suppose that during the preceding four weeks the S&P 500 index increased by 1 percent. Clarke and Statman find that this leads to a 1.23 percentage increase in bulls, as some bears and chickens change their perspective. Specifically, the bears decline by 1.18 percent and the chickens by 0.05 percent.

  Although the behavior described in the preceding paragraph conforms to the idea that newsletter advisers naively extrapolate trends, the situation is more complex. Advisers responded differently to that same 1 percent increase in the S&P 500 had it occurred during the preceding twenty-six weeks, as opposed to the preceding four weeks. Now the chickens register a 0.4 percent increase, which is even larger than the 0.3 percent increase in bulls. Clarke and Statman refer to this phenomenon as nervous bullishness. There is heuristic diversity. Unlike the response to a four-week increase, where the migration in sentiment is from bears to bulls, the response to a twenty-six increase is a migration from bears to an even combination of bulls and chickens.14

  How did the 1987 crash affect the sentiment of advisers? Not surprisingly, the general level of bullishness fell dramatically. However, Clarke and Statman report that once bullishness adjusted to its new lower level, the reaction of sentiment to changes in the S&P 500 was pretty close to what it had been prior to the crash. Before the crash, an increase of 10 percent in the S&P 500 led to a 7 percent increase in the Bullish Sentiment Index. After the crash, the same 10 percent increase led to a 6.4 percent increase.15 In chapter 4, I mentioned that behavioral finance does not have all the answers. Clarke and Statman had hypothesized that the crash would make the sentiment more responsive to changes in the S&P 500. But the evidence refuted their hypothesis.

  Interestingly, newsletter writers find it more difficult to spot market trends when the market is highly volatile. Consequently, volatility serves as a moderating influence, leading these writers to become, relatively speaking, less bullish during up markets and less bearish during down markets.16

  The fact remains that the Bullish Sentiment Index provides no guidance as to where the market is headed next. The index is more of a rearview mirror than a front windshield. Were the past to provide an accurate guide about where the market was headed, then the index, as a reflection of the past, could prove useful. Alas, the past is a poor indicator of the future, and hence so is the Bullish Sentiment Index.

  Whose Sentiment?

  There is no universally accepted sentiment index. The Bullish Sentiment Index is a misnomer in that it only reflects the views of the writers of advisory newsletters. The American Association of Individual Investors (AAII) monitors the sentiments of small investors. Investors Intelligence does the same for the writers of advisory newsletters. Richard Bernstein’s index (see chapter 5) is based on Wall Street strategists’ allocations. Some technical analysts use the Call/ Put ratio to measure sentiment. In chapter 13, I discuss the extent to which the discount on closed-end funds serves as a sentiment index. And many money managers, such as Martin Zweig, have developed sentiment indicators of their own. Kenneth Fisher and Meir Statman (1999b) compare several of these and show that they behave quite differently from one another.17

  Summary

  Technical analysts continue to treat the Bullish Sentiment Index as a contrarian indicator. Why do they succumb to the illusion of validity? Because the logic seems so appealing. Because others believe it. Because the popular financial press continues to perpetuate the myth. Because selective sentimental journeys, especially when presented by successful and entertaining commentators, can be very persuasive. And because in the end, technical analysts have not learned to validate their views. They emphasize evidence that confirms their own point of view, and overlook evidence that disconfirms their point of view. Consequently, they end up with a biased view and an overconfident attitude to boot. These biases are deeply ingrained: Learning is a slow process.

  Chapter 7 Picking Stocks to Beat the Market

  Either market efficiency is an illusion or mispricing is an illusion. In a landmark article, Eugene Fama (1970) argued that financial markets are efficient. Market efficiency holds that unexploited profit opportunities do not lie around for very long, just as a $20 bill does not lie for long on a crowded sidewalk.

  Recall that the third theme of behavioral finance is inefficient markets. In recent years scholars have produced considerable evidence that heuristic-driven bias and frame dependence cause markets to be inefficient. I mentioned in chapter 1 that scholars use the term anomalies to describe specific market inefficiencies. For this reason, Fama characterizes behavioral finance as “anomalies dredging.”

  This chapter discusses the following:

  • the evidence that recommended stocks have consistently beaten the market

  • the riskiness attached to recommended stocks

  • the role of risk in the market efficiency debate

  • the implications of heuristic-driven bias for value investing and momentum-based strategies

  • the roles played by regret and hindsight bias

  Some managers have beaten the market consistently. Two examples come to mind.1 The first pertains to the television program Wall $treet Week with Louis Rukeyser, which first aired in 1970. Stocks recommended on that program have beaten the market by 4 percent per year.

  The second example is the “Pros vs. Darts” contest that the Wall Street Journal has run since January 1990. In this contest, four security analysts (pros) each select a single stock. Four Wall Street Journal staffers throw a dart at a sheet of stocks. The contest winner, pros or darts, is determined by the performance of the respective selections over the subsequent six-month period. Not only have the pros’ recommendations trounced the darts, but they beat the S&P 500 and Dow Jones Industrial Average as well, by 5 percent a year.

  According to the efficient market school, investors cannot rationally expect to beat the market, except by taking on more systematic risk than the market. Active money managers disagree. Some, such as David Dreman, argue that these errors and biases create profitable opportunities for those who recognize them. In his book Contrarian Investment Strategies: The Next Generation, Dreman (1998) states: “Nobody beats the market, they say. Except for those of us who do.”

  There are two sides to the market efficiency debate, and one side is wrong. Either money managers such as Dreman are subject to the illusion of mispricing, or the efficient market school is subject to the illusion of market efficiency.

  I argue that the evidence goes against market efficiency. Much of this evidence pertains to the long-term success of value investing, momentum investing, and the reaction to changes in security analysts’ recommendations. To be sure, one of the main contributions of behavioral finance has been to explain the long-run success of value investing in terms of heuristic-driven bias, rather than risk.

  At the same time, there are subtle issues. The meaning of market efficiency is that prices reflect fundamental values, not “you wo
n’t find $20 bills lying along crowded sidewalks.” This is important. As I stressed in chapter 4, the “smart money” traders typically stop short of exploiting all of the mispricing they identify. Why? They conclude that it’s not worth the risk. There are limits to arbitrage.2 For example, investors who are convinced that value investing almost always outperforms a growth-based strategy, may experience a rude shock.

  One other note. Knowing that prices are inefficient and exploiting that inefficiency are two different things. A lot of people seem to think that the message of behavioral finance is that beating the market is a no-brainer because errors cause mispricing. Well, it’s not easy money; just the opposite, in fact. One of the main messages of behavioral finance is that heuristic-driven bias and frame dependence get in the way. There was a lot of California gold waiting to be discovered in 1849, but how many prospectors actually got rich? Precious few.

  Do Brokerage House Recommendations Beat the Market? Yes, But …

  In his survey of market efficiency, Robert Merton (1987b) describes many ways of looking for evidence of inefficient prices. One way is to see whether professionals beat the market. Stock-picking advice is one of the major services investors buy when they deal with a full-service brokerage firm. If mispricing is an illusion, then stock-picking advice should be worthless. Have brokerage house recommendations beaten the market? Yes, but yes, they have, but with a lot of noise.

  In June 1986, the Wall Street Journal and Zacks Investment Research began a joint study to investigate the value of this advice. In their study they track the performance of stocks recommended by brokerage firm analysts and compare that performance against the Dow and the S&P 500. Results were initially issued on a quarterly basis.

  Some of the brokerage houses in the Wall Street Journal/Zacks study issue monthly “buy” lists. Other firms do not issue such lists but communicate their opinions by rating all the stocks they followed. Typically, these ratings range from “1” for a strong buy to “5” for a strong sell.3

  The study tracks the recommended stocks from about fifteen brokerage houses at any one time. The firms selected include local, regional, and national brokerages.4 Over the course of time, the identity of firms in the study has changed as brokerage houses have merged or gone out of business.

  A 1990 Wall Street Journal article describes the procedure used to evaluate the recommended stocks.

  The study assumes that an investor buys every strongly recommended stock and sells all others, even if they are designated as holds or weak buys. Equal dollar amounts are put into each stock; portfolios are rebalanced monthly to keep the amounts equal. All buying and selling is on the last trading day of each month. Commissions and taxes are disregarded, mostly to make it easier to compare the results with market indexes. Dividends are included.5

  The Wall Street Journal/Zacks study offers some intriguing lessons. Thirty months into the study, journalist John Dorfman wrote: “Stocks recommended by major brokerage houses have done, on average, no better than the market as a whole. That means investors could have done as well plunking their money into a mutual fund that mimics Standard & Poor’s 500-stock index.”6 As figure 7-1 illustrates, this pattern continued through 1989.

  The period 1990 through 1992 was an entirely different matter. During these three years, recommended stocks returned over 72 percent, more than double the 35.9 percent return of the S&P 500. Now John Dorfman was communicating an entirely different message. In a 1993 Wall Street Journal article he wrote:

  Taken your broker to lunch lately?

  Maybe you should. Major brokerage houses tore up the track in stock-picking last year.

  Fourteen out of 15 major brokerage houses beat the overall stock market with their “recommended lists” of stocks to buy, according to a quarterly study by The Wall Street Journal and Zacks Investment Research of Chicago.7

  Figure 7-1 Recommended Stocks versus S&P 500, June 1986–December 1989

  From the inception of the study through December 1989, stocks recommended by major brokerage houses did no better than the S&P 500.

  Where does that leave us? It leaves us in need of more data. Fortunately, we have some, and it’s better data too. Starting in 1993, some improvements were made to the study. First, performance was tracked monthly, rather than quarterly. Second, a theoretical 1 percent commission was added for trades associated with changes to the recommended stock list. Third, only recommendations made by noon Eastern time are priced at that day’s close. Recommendations made in the afternoon are priced at the following day’s close.

  Consider what the sixty months of monthly data from January 1993 through December 1997 indicate, as shown in figure 7-2. On a cumulative basis, the recommended stocks slightly, but consistently outperformed the S&P 500. For the five-year period as a whole, they returned 165 percent, somewhat larger than the 151 percent returned by the S&P 500.8 In other words, the recommended stocks beat the market by 106 basis points per year.9

  Figure 7-2 Recommended Stocks versus S&P 500, January 1993–December 1997

  Over the five-year period January 1993–December 1997, the recommended stocks slightly, but consistently, outperformed the S&P 500 on a cumulative basis.

  What Happens When an Analyst Changes a Recommendation?

  Interesting things result. Kent Womack (1996) finds that not only does the market price immediately react to the announcement that an analyst has changed his or her recommendation on a stock, but the adjustment continues for a substantial period thereafter. This phenomenon is hardly in keeping with market efficiency.

  Womack studied the period 1989–1991.10 Here is a short illustrative example taken from his database.11 Tom Kurlak is a Merrill Lynch analyst who follows Intel. Back on February 1, 1989, Kurlak upgraded his intermediate-term rating on Intel from the second highest recommendation of “accumulate” to the highest, “buy.” The price of Intel shares jumped by $1.25 to $27.25, on volume of 3.4 million shares, compared with average daily volume of about 2 million shares. Intel also increased relative to the market: The Dow Jones Industrial Average closed down 4.11 that day, at 2338.21.

  Is this event consistent with market efficiency? Possibly. It may be that Kurlak was providing new information to the market, and the price jump represented the appropriate adjustment to fundamental value. According to Dow Jones News Service,12 Kurlak’s upgrade stemmed from a changing inventory picture. But Intel’s stock displayed significant post-recommendation drift. It outperformed the S&P 500 in each of the next five months: Intel rose 11.5 percent, while the S&P 500 rose 8.8 percent.

  The preceding scenario is typical of Womack’s findings for what takes place on average. On the day of the recommendation change, and for two days thereafter, there is a large jump in the price of the stock. Moreover, the change is in the direction forecast by the analyst who altered his or her recommendation. But the surprising thing is the existence of significant post-recommendation drift. On average, the price of a stock that is upgraded to “buy” goes up by 5 percent, relative to a comparison benchmark group. Analogously, the price of a stock that is downgraded to “sell” drops by 11 percent.

  Market efficiency holds that price adjusts virtually immediately to new information. Post-recommendation drift is not a property of efficient prices. So, where does that leave us?

  Risk

  There are still many questions left to answer, the chief one centering on risk. Did the recommended stocks beat the market because they were riskier? Does a change in recommendation affect the riskiness of a stock? In particular, do stocks that have been upgraded to a “buy” become more risky? They would have to in order for the post-recommendation drift identified by Womack to be consistent with market efficiency.

  A lot depends on what we mean by risk. If by risk we mean beta, the recommended stocks were actually less risky than the S&P 500. Suppose that some investor holds the recommended stocks from a particular brokerage firm, say, Merrill Lynch. Imagine that in January 1993, this investor formed his
portfolio by purchasing the stocks recommended by Merrill Lynch, and then updated his portfolio monthly, as Merrill changed its recommendations. What would the beta of this portfolio have been over the five-year period January 1993 through December 1997? The answer is 0.90. For the seventeen brokerage firms in the Wall Street Journal/Zacks study, the average beta was 0.94. It thus appears that recommended stocks both outperform the S&P 500 and have lower risk.

  However, this does not close the issue of whether or not markets are efficient. In recent years a belief has sprung up that as a risk measure, beta is dead. Does this mean that the practitioners and academics who for all those years thought beta to be the correct risk measure were committing a cognitive error? Let me add that the recommended stocks were more volatile than the S&P 500 during the period 1993–1997. The volatility of the S&P 500 was 3.05 percent per month, lower than for any of the seventeen recommended stock portfolios. The volatility of the average recommended stock portfolio was 3.77 percent.

  Recently other measures of risk have been proposed. These alternative risk measures are based on firm characteristics such as market capitalization (size) and price-to-book. Price-to-book is routinely used to label stocks as “growth” or “value.” Growth stocks have high price-to-book ratios, while value stocks are the opposite. Related measures are used in this connection as well, such as price-to-earnings and dividend yield.