Chapter 8: Financial Decision-Making and Heuristics PDF

CHAPTER 8: IMPLICATIONS OF HEURISTICS AND BIASES FOR FINANCIAL DECISION-MAKING 8.1 INTRODUCTION As we saw in Chapter 5, while heuristics are usually excellent time- and effort- saving decision-making mechanisms, they sometimes appear to lead investors in unfortunate directions. In this and later c...

CHAPTER 8: IMPLICATIONS OF HEURISTICS AND BIASES FOR FINANCIAL DECISION-MAKING 8.1 INTRODUCTION As we saw in Chapter 5, while heuristics are usually excellent time- and effort- saving decision-making mechanisms, they sometimes appear to lead investors in unfortunate directions. In this and later chapters, we return to these heuristics when we investigate their potential impact on the behavior of investors, future retirees, analysts, and managers, and how they may potentially impact market out- comes. The focus of this chapter is how heuristics influence investor financial decision-making, with the investment decisions of future retirees reserved for Chapter 17. Section 8.2 deals with financial behaviors stemming from familiarity. One aspect of familiarity is home bias, the tendency to overinvest domestically and locally. While investment close to home can stem from an informational advantage, this is probably not the whole answer. Related to home bias is the tendency to invest in companies you work for or brands you know. In Section 8.3, we turn to behaviors stemming from representativeness and related biases. The tendency to overestimate predictability likely induces investors to believe that good companies are good investments. This, coupled with recency, persuades people to believe that good recent stock market performers are good buys. And the availability bias pushes people into concentrating on investments in securities for which information is freely available. In Section 8.4, we show that anchoring causes people to be excessively influenced by suggested or available cues, instead of relying on their own opinion or expertise. This is demonstrated in the context of expert views of real estate value. 8.2 FINANCIAL BEHAVIORS STEMMING FROM FAMILIARITY HOME BIAS Though preferences are slowly changing in this regard, it continues to be true that domestic investors hold mostly domestic securities---that is, American investors hold mostly U.S. securities; Japanese investors hold mostly Japanese securities; Brit- ish investors hold mostly U.K. securities; and so on. Kenneth French and James Poterba documented this tendency.1 Referring to the first numerical column of Table 8.1, we see displayed the aggregate market values of the six biggest stock markets in the world. The United States, as of 1989, had 47.8% of world market capitalization, Japan 26.5%, the U.K. 13.8%, France 4.3%, Germany 3.8%, and Canada 3.8%.2 Nevertheless, a typical U.S. investor held 93.8% in U.S. stocks; a typical Japanese investor held 98.1% in Japanese stocks; and a typical U.K. inves- tor held 82.0% in U.K. stocks.3 Thus, domestic investors overweight domestic stocks. This behavior is called home bias. Bias toward the home country flies in the face of evidence indicating that diver- sifying internationally allows investors to reduce risk without surrendering return.4 This is particularly true since stock markets in different countries are not highly correlated.5 The average pairwise correlation coefficient for the countries listed in the previous paragraph during 1975--1989 was 0.502, which attests to the gains from diversification. One reason why investors might hold more domestic securities is because they are optimistic about their markets relative to foreign markets. Using an expected utility maximization approach and historical correlations between markets, French and Poterba estimated what expected returns would have to be in order to justify the observed asset allocation, and Table 8.2 reports their results. To justify their overweighted U.S. holdings, American investors would have to believe that their market would beat the second-best market (Canada) by 80 basis points; Japanese investors would have to believe their market would outperform by at least 280 basis points; and in the United Kingdom, the comparable figure was a whopping 430 basis points. Obviously this set of beliefs is contradictory and implies excessive optimism---at least on the part of two of the three sets of investors. The next chap- ter will focus on excessive optimism in financial decision-making. Another behavioral explanation is along the lines of comfort-seeking and familiarity. As we discussed in Chapter 5, people tend to favor that which is famil- iar. U.S. investors are more familiar with U.S. stocks and markets, and so they are more comfortable investing in U.S. securities. The same holds equally for foreign investors.6 As is so often true where behavioral explanations have been advanced to explain apparently anomalous behavior, rational explanations are also put for- ward. International investment may be less attractive because of institutional bar- riers, examples of which are capital movement restrictions, differential trading costs, and differential tax rates. French and Poterba downplay these arguments, however. While at one time there were significant capital movement restrictions, at the time of their work, they were not in effect. As for differential trading costs, if costs in one country are lower than in other countries, this is a reason for all inves- tors to favor the low-cost country, but we do not see this type of behavior. Addi- tionally, especially with the international system of dividend withholding taxes and counterbalancing tax credits, there is little difference between domestic and foreign tax burdens for most investors. [Table]{.smallcaps} 8.1 ------------------------- ------ -- -- ------ 47.8 5.9 26.5 4.8 13.8 82.0 4.3 3.2 3.8 3.5 3.8 0.6 [Table]{.smallcaps} 8.2 ------------------------- -- -- -- DISTANCE, CULTURE, AND LANGUAGE The argument that institutional considerations cause investors to shy away from foreign investments becomes weak if it can be demonstrated that people prefer to invest locally, even within their own country. Gur Huberman reports on a case of such "intra-national" home bias."7 In 1984, AT&T was forced by the court into a divestiture whereby seven "Baby Bells" were created. These companies were cre- ated along regional lines. An example is BellSouth serving the southeastern United States. If people like familiarity, then we would expect a disproportionate number of a Baby Bell's customers to hold a disproportionate number of shares in the same Baby Bell. Indeed, that is exactly what happened after the divestiture. While we often hear that we should buy locally, from a diversification standpoint, if anything, you are wise to underweight (not overweight) local companies. If the economy of your region fares poorly, this will be bad both for the stock market performance of local companies and the employment prospects of local workers (yourself included). If you work and invest locally, technically speaking, your two income sources are highly correlated. Diversification theory says you should look for income streams that are weakly correlated. For this reason, it would have been better for investors to buy stock in Baby Bells outside their region. In a related study, Mark Grinblatt and Matti Keloharju demonstrate that the preference for familiarity extends to language and culture.8 In Finland, there are two official languages, Finnish and Swedish.9 Annual reports are normally pub- lished in Finnish or in both official languages, but in a few cases reports are only published in Swedish. It turns out that, after controlling for other relevant factors, Finnish investors prefer companies whose language of publication is Finnish, and Swedish investors prefer companies whose language is Swedish---with bilingual companies being mid-ranked by both groups of investors. Interestingly, culture matters as well. These authors took note of whether CEOs were Finnish or Swedish. Controlling for the language of the company, Finnish speakers prefer Finnish CEOs, and Swedish speakers prefer Swedish CEOs. The lesson seems clear: familiarity, on all levels, "breeds" investment.10 Moreover, there is evidence that even institutional investors may not be immune from this tendency.11 LOCAL INVESTING AND INFORMATIONAL ADVANTAGES One reason why investors may favor local markets---where local is interpreted as either domestic or close-to-home, but within the same country---is because they may possess, or may feel that they possess, informational advantages. Gains from being geographically close to a company may appear in improved monitoring capa- bility and access to private information. Joshua Coval and Tobias Moskowitz investigated this issue in the context of mutual fund managerial performance.12 They first established that mutual fund managers, consistent with familiarity bias, tend to favor local investments, that is, they tend to buy firms headquartered within a 100-mile (or 161-kilometer) radius of their head office. Specifically, they conclude that the average manager invests in companies that are located about 10% closer to her than the average firm she could have held. Further, local equity preference is related to firm size, leverage and output tradability, with small, levered firms producing goods that are not traded internationally tending to be the ones where local preference comes through strongest. Consider rational motivations for investing locally. One is hedging demand. If you consume local goods at local prices, it can make sense to hedge by investing locally. If locally produced goods are not traded outside the local region, then it is reasonable to talk about local prices. Take haircuts, which are as non-tradable as one gets.13 If you buy the stock of a local haircutting company, your future haircut consumption, which must be local, is well hedged. The finding that local equity preference is more pronounced among companies whose goods are not traded internationally is consistent with hedging demand. Size and leverage, on the other hand, suggest an information differential expla- nation, as smaller, levered firms are likely to be ones for which local informational advantage may be stronger. To test this, Coval and Moskowitz investigate whether local preference can generate a boost to performance. As has been discussed previ- ously, most studies indicate that the average actively managed mutual fund has been unable to consistently outperform its benchmark on a risk-adjusted basis.14 Notably though, Coval and Moskowitz demonstrate a significant payoff to local investing. Fund managers on average earn 2.67% per year more on local invest- ments, while local stocks avoided by managers underperform by 3% per year. Moreover, they find that those better able to select local stocks tend to concentrate their holdings more locally. Are retail investors also able to exploit this? The evi- dence points in this direction as stocks with high levels of local ownership tend to outperform, and this effect lasts for several months, suggesting those with access to such data could earn excess returns. In other research, there is evidence that retail investors take advantage of the opportunity.15 Reminiscent of the money manager finding, based on a dataset of retail investors, local investments outper- form remote investments by 3.2% per year. INVESTING IN YOUR EMPLOYER OR BRANDS THAT YOU KNOW There is also abundant evidence that investors overweight the stocks of companies whose brands are familiar or that they work for. As for the first, Laura Frieder and Avanidhar Subrahmanyam looked at survey data on perceived brand quality and brand familiarity (recognition) and asked whether these attributes impacted inves- tor preferences.16 To answer this question, they correlated institutional holdings with these factors. Note that high institutional holding in a stock implies low retail holding in that same stock. These researchers found that institutional holdings are significantly and negatively related to brand recognition, but no discernible impact was present for brand quality. The former implies that retail investors have a higher demand for firms with brand recognition, which is consistent with comfort- seeking and familiarity. Still, Frieder and Subrahmanyam argue that recognizable brands are associated with companies with more readily accessible information for average investors. They provide a model that shows that investors will, ceteris paribus, demand more of a stock when they have more precise information about the stock. Therefore, in this context as in others, a natural informational advantage may stem from familiarity. As for overweighting companies that one works for, while the same sort of familiarity versus informational advantage debate is possible, the extent to which some investors invest in these companies seems to transcend an informational explanation. Many "employee-investors" put a very high percentage of their inves- tible wealth in their employer's stock, thus foregoing a significant amount of possi- ble diversification.17 This will be discussed in more detail in Chapter 17. 8.3 FINANCIAL BEHAVIORS STEMMING FROM REPRESENTATIVENESS There is evidence that representativeness and related biases induce inappropriate investment decisions. To casual observers it seems obvious that if a company has high-quality management, a strong image, and consistent growth in earnings, it must be a good investment. Students of finance, of course, know better. In valuation, future cash flows are forecasted and discounted back to the present using an appropriate risk-adjusted discount rate. All the aforementioned attributes that make a company a good company should theoretically be reflected in these estimates of future cash flows (including the growth in cash flows) and the risk- adjusted discount rate---that is, they should already be impounded in price. Loosely speaking, good companies will sell at high prices, and bad companies will sell at low prices. But, once the market has adjusted, there is no reason to favor a good company over a bad company, or, for that matter, a bad company over a good company. Quite simply, it is a mistake to think that a good company is representa- tive of a good investment, and yet, that is exactly what people often seem to believe. Further, according to market efficiency, excess returns should be unpredict- able. Nevertheless, as we have noted, there is a tendency to overestimate predict- ability. In this context then, there may be a tendency to associate past success (which led to high past returns) with likely future returns. GOOD COMPANIES VS. GOOD INVESTMENTS Hersh Shefrin and Meir Statman provide some very revealing evidence.18 As they report, Fortune magazine has been surveying senior executives on company attri- butes for a number of years.19 Executives are asked to assign values between "0" (poor) and "10" (excellent) to each company in their industry for the following items: quality of management; quality of products/services; innovativeness; long- term investment value; financial soundness; ability to attract, develop, and keep tal- ented people; responsibility to the community and environment; and wise use of corporate assets. While Fortune reports average scores on all attributes as a proxy for company quality, because 82% of respondents consider quality of management as the most important attribute of a company's quality, these researchers use it as their proxy for company quality. In Table 8.3 we report some regressions from Shefrin and Statman. From the first panel, we see that management quality (i.e., good company measure) and value as a long-term investment (i.e., good stock measure) are very highly corre- lated: the R2 value from the first regression suggests a correlation (take the square root of R2) of 0.93---that is, executives believe that good companies are good stocks. As discussed before, it is important to understand that no company attri- bute should be associated with investment value: all information on company qual- ity should already be embedded in stock prices so that all companies (good ones and bad ones) are equally good investments (on an ex ante basis). The bottom three regressions (i.e., those in the lower panel) reveal that two firm characteristics, size and the book-to-market ratio, are strongly associated with perceived management quality. Specifically, big companies and those that have low book-to-market ratios (where the latter are considered growth companies) are seen to be good companies. This is not overly surprising. Big companies have often be- come big because they are good (i.e., well managed), and growth should come from quality. Turn to the last regression in the upper panel. In this regression, value as a long- term investment is regressed on size, book-to-market, and management quality. As before, the latter strongly impacts perceived investment value. Additionally, however, size and book-to-market, even after accounting for their impact on man- agement quality, independently influence investment value. Big firms are viewed as good investments, and growth companies are viewed as good investments. In other words, big high-growth firms are representative of good investments. Interestingly, as was discussed in Chapter 4, the empirical evidence points in the exact opposite di- rection. It is small-cap value firms that have historically outperformed. Indeed, the tendency for individuals to use representativeness in this context may have contrib- uted to the small-firm and value anomalies.20 In related research, there is evidence that firm image impacts the perception of investment attractiveness. As argued previously, while a positive firm image can only be seen as a good attribute, its ability to generate cash flows and growth should have already been capitalized in the price of the stock. In one experiment, disclosures related to image that are not value-relevant are released to partici- pants.21 Subjects are more likely to invest in firms with a positive image than those with a negative image, even controlling for such value-relevant attributes as indus- try membership and financial data.22 If firm image has such impact, one might expect the same would be true for perception of brand quality. Nevertheless, in one previously discussed study, there was no evidence that perception of brand quality, once brand recognition was controlled for, led to retail investors increasing their demand for a stock.23 CHASING WINNERS Research has also shown that investors choose securities and investment funds based on past performance. To those with this view, investment performance in the recent past is representative of future investment performance. This form of representativeness is often called recency. Such trend-following, or momentum- chasing, has long been a popular strategy, and, coupled with detecting turning points, is at the heart of technical analysis.24 A survey of individuals from the American Association of Individual Investors reports that more people become bullish if the market has recently turned up.25 In the context of mutual funds, strong past performance leads to abnormally high inflows of investor money.26 Trend-following is an international phenomenon. From Japan, the evidence is that stocks that experience increases in individual ownership were past winners.27 In Canada, a survey of workers managing their own retirement money indicates they are momentum-chasers, rather than contrarians.28 More specifically, respondents were asked to start their pensions from scratch and allocate money between two stocks, one with an "average return over the last 5 years of 5%," and a second with an "average return over the last 5 years of 15%." Further, they were told that "analysts forecast that both stocks should earn about 10% per year over the next 5 years." Those neutral on future direction would go 50/50 in order to maximize diversification. Momentum-chasers would put more than 50% in the second stock, while contrarians would put more than 50% of their money in the first stock. Figure 8.1 shows the frequency distribution of the percentage difference between investment in the "loser" stock and the "winner" stock. A high percentage of respon- dents (63.8%) were momentum-chasers, while far fewer (11.6%) were contrarians. Shlomo Benartzi evaluates investment in company stock in 401(k)s in relation to momentum-chasing.29 When he divides plans into quintiles based on company stock performance over the previous 10 years, he finds that employees of the top- performing companies contribute 40% of their discretionary money into company stock versus 10% for the bottom-performing quintile. Did momentum-chasing work for these investors? Unfortunately not, as in the year after portfolio forma- tion employees who allocated the most to company stock earned 6.77% less than did those who allocated the least. So is there any evidence in favor of the popular notion that momentum- chasing is profitable? The answer is both yes and no. There is evidence that risk- adjusted returns are positively serially correlated for 3- to 12-month return inter- vals.30 For longer periods of three years or more (as in both the Benartzi paper and in the Canadian survey), the evidence favors reversals or negative serial cor- relation.31 Later, in Chapter 13, we will present behavioral models that seek to account for this pattern of intermediate-term momentum followed by long-term reversal. AVAILABILITY AND ATTENTION-GRABBING In Chapter 5 we saw that when information on certain types of events is freely available, people often get the impression that such events are more likely. For ex- ample, news reports of violent crime may induce people to revise upward their sub- jective probabilities of such violent attacks. Brad Barber and Terrance Odean investigated whether information availability impacts the trading behavior of inves- tors.32 They argue that since attention is a scarce resource and there is a plethora of possible investment opportunities, the transactions of retail investors are likely to be concentrated in stocks where information is freely available. "Attention- grabbing" is proxied in three ways: news reports on a stock, unusually high trading volume, and extreme returns. The latter two factors control for impact since some- times news might be neutral. While news can be of a positive or negative nature, since individual investors rarely short-sell and normally own only a small subset of stocks, negative news is likely to be ignored, while positive news may attract pur- chases. On this basis, these researchers suggest that news is likely to lead to net purchases for retail investors. On the other hand, institutional investors are much less likely to be so affected, because in their work they typically consider all the securities in their universe, without requiring any external prompt. Indeed, the empirical evidence is in line with the expectations of Barber and Odean. 8.4 ANCHORING TO AVAILABLE ECONOMIC CUES In Chapter 5 we provided an example where experimental subjects, when asked to estimate an uncertain magnitude, anchored their estimates on obviously meaning- less red herrings. Anchoring is even more likely to occur when the potential anchor appears prima facie to have economic content. AN EXPERIMENTAL STUDY OF REAL ESTATE APPRAISALS Gregory Northcraft and Margaret Neale investigated whether anchoring might oc- cur in the context of real estate appraisals.33 Two randomly selected groups of real estate agents were taken to a house and asked to appraise it. They were given the same tour and identical packages of information, which included the house's (pur- ported) list price. The only difference between the two groups was that the first group was given a list price of \$65,900, while the second group was given a list price of \$83,900---\$18,000 more. Put yourself in the place of the agents. There is always some uncertainty in an appraisal. While you can exclusively use your own expertise and totally ignore the list price, perhaps it should not be surprising that agents were influenced by the list price. Yet, list prices are quite variable and often have a strategic component. The average appraisal price of the first group came in at \$67,811, and that of the second group was \$75,190. These dollar figures are summarized in Table 8.4. If we take the mid-point of these values (\$71,500.50) as our best estimate of the true appraisal value, the gaps between the two appraisal averages was a full 10%. Clearly, the real estate agents were anchored on the list prices that they were ex- posed to---despite the fact that only 25% mentioned the list price as one of the fac- tors that they considered. One can think of the agents as using the following appraisal estimate mechanism: 8.1 Appraisal estimate = a \* Personal appraisal estimate + (1 − a) \* List price Only those ignoring the list price would set a =1. For the first group of real estate agents, it turned out that a =.34, suggesting that the list price was very influential; for the second group, a =.70. We can also calculate the appraisal price that sets a equal for the two groups. One can show that this is \$69,136.43, at a =.59, still suggesting significant influence of the list price on the real estate agents' appraisals. There is no reason to think that the tendency to anchor is not present in other economic and financial situations. The reality is that anchors in such con- texts are likely to be common. All of us anchor on market prices. There is a rational side to this, though, because market prices are consensus estimates of value. But unfortunately this implies some circularity---if everyone is anchored on market price. Any initial value, however "off," would have an influence on the eventual market price. Consider the high valuations of Internet stocks in 1999. Quite a few observers had misgivings about their levels, but many were clever in their ability to justify them. Were they heavily influenced by the anchor of the current market price? Was this anchor "dropped" by the irrationality of a subset of traders who had little idea of fundamental value? In retrospect, this seems to be a valid view. ANCHORING VS. HERDING AND ANALYSTS Since anchoring and herding are closely related, it makes sense at this point to say a few words about herding. There is a social component to herding behavior, so we will mostly leave it to Chapter 12. In the real estate appraisal experiment, if an agent had been told that a second agent had come up with a certain appraisal, and the first agent's appraisal was pulled toward this value (even taking into account the influence of the list price), this would be an example of herding or fol- lowing the crowd. Professional financial analysts who publicly estimate value, forecast earnings, and make buy/sell recommendations, are often said to anchor or herd. Let us briefly consider whether analysts exhibit anchoring and/or herding behavior. One way in which anchoring can be exhibited by analysts is if they are slow to change their initial opinion. In Chapter 13 we argue that this behavior may be the source of certain anomalies. Analysts may herd if some analysts are influenced by the recommendations or earnings estimates of other analysts. There is research indicat- ing that analysts go with the crowd when it comes to recommendation revisions.34 The evidence for earnings estimates is more mixed, with some of it pointing in the direction of herding and other research suggesting "anti-herding" (i.e., running contrary to the crowd).35 For example, a recent study using U.K. data on earnings forecasts is consistent with herding behavior, while another, using German data, is consistent with anti-herding behavior.36 Note that while herding makes sense because going with the crowd is easy and safe, anti-herding can make sense if you believe you have private information and you want to make yourself visible for the purpose of career advancement. CHAPTER HIGHLIGHTS 1\. There is a preference for investing close to home. This manifests itself in home-country bias, investing locally within the domestic market, and prefer- ring one's own language and culture. 2\. One explanation for home bias is the comfort-seeking associated with familiarity. 3\. Another explanation for home bias is informational advantage, a view rein- forced by evidence on the efficacy of local investment on the part of money managers and retail investors. 4\. Representativeness causes investors to think that good companies are good investments, whereas known positive characteristics should already be im- pounded in the price of a stock. 5\. Because of recency, investors are prone to chasing winning stocks and funds. While there is some evidence of medium-term (3--12 months) momentum, in the longer-term (3--5 years), reversal is the order of the day. 6\. Availability bias is evidenced when investors tend to buy stocks that are in the news. 7\. Anchoring appears in research showing that real estate appraisals are an- chored to list prices. CHAPTER 9: IMPLICATIONS OF OVERCONFIDENCE FOR FINANCIAL DECISION-MAKING 9.1 INTRODUCTION As we saw in Chapter 5, overconfidence is pandemic in society. In this chapter, we address the extent to which this behavioral tendency impacts financial decision- making. As in the previous chapter, our focus will be on investors and other mar- ket practitioners. Later, in Chapter 16, we will consider how entrepreneurs and corporate managers might be affected. In Chapter 13, overconfidence will be seen to play a central role in models that seek to explain various market anomalies. In Section 9.2, the various manifestations of overconfidence and excessive trad- ing are related. We begin with a simple model illustrating the relationship between overconfidence and trading, and then move to evidence from naturally occurring markets, surveys, and experiments. We turn to the demographics and dynamics of overconfidence in a financial setting in Section 9.3. Some groups (e.g., men) tend to display greater overconfidence. Moreover, we investigate whether overconfi- dence can be "learned" by past experience in markets. In Section 9.4, evidence that relates overconfidence to underdiversification and excessive risk taking is ex- plored. Finally, in Section 9.5, we briefly present evidence that analysts exhibit ex- cessive optimism. This is likely due to more than psychology, as will be discussed in Chapter 12 when we revisit the financial behavior of this group of practitioners. 9.2 OVERCONFIDENCE AND EXCESSIVE TRADING There is evidence that the overconfidence of investors leads to excessive trading. Theoretical models have been constructed that yield this result. To illustrate the in- sights provided by these models, we begin with a simple illustrative model that re- lates overconfidence and trading activity. OVERCONFIDENT TRADERS: A SIMPLE MODEL Consider the demand for a particular security. At the level of the individual, de- mand will be a function of the investor's estimate of the security's (intrinsic) value. If the investor believes that the value exceeds the market price, he will wish to hold more of the security than if the security was perceived to be fairly priced. Let qn equal the (neutral) number of shares that an investor would hold if price and value were equivalent.1 If the value exceeds the price, the investor will want to hold more than qn shares. On the other hand, if value falls short of price, the investor will want to hold less than qn shares. The difference between investors is that they respond differently to prices that deviate from their value estimates. In order to understand how different prices af- fect desired holdings, we begin with a mechanism for value estimation. First assume that since there are many investors, all are price-takers.2 Further, we will assume that when estimating value, an investor uses two items of information, his own opinion (prior value) and the market price (which is the weighted average of all in- vestors' opinions), as follows: vi = aiv\* + (1− ai) p, 0 ≤ ai ≤ 1 where vi is the (posterior) estimate of value of investor i; vi\* is the same investor's prior estimate of value; p is the market price; and ai is the weight investor i puts on his prior relative to the market price. The higher ai is, the higher is the weight an investor puts on his own opinion. Since there is a very large number of investor views determining p, any value of ai more than slightly above zero suggests some overconfidence, with higher values suggesting more overconfidence than lower values. Here, by overconfident we primarily mean miscalibrated, which implies an inflated view of the precision of one's information (or opinion). The better- than-average effect, here the feeling that one is better at estimating value than other market participants, also likely plays a role. Consider how 9.1 feeds into demand for the stock. Suppose that the demand curve can be written as: qi = qn + θ(vi − p), θ \> 0 where qi is investor i's demand and θ is the sensitivity of demand to a divergence between the posterior value estimate and price.3 Substitute 9.1 into 9.2 and sim- plify to arrive at: qi = qn + θai(v\* − p) Next take the partial derivative of qi with respect to p: ∂qi /∂p = − θai The higher the investor's level of overconfidence (ai), the more responsive demand is to changes in price. As ai approaches one, which means market price has no in- fluence, the closer ∂qi/∂p is to --θ. On the other hand, as ai moves toward zero, the demand changes little when the price changes. It is conventional to write demand curves with p on the y-axis and q on the x-axis. Using this approach, the higher the investor's level of overconfidence (ai) is, the flatter is the demand curve. And as ai moves toward zero, the demand curve becomes close to vertical. Figure 9.1 illustrates graphically the situation for three investors. On this graph, their demand curves for a given security are depicted. These are labeled D1PC, D2LOC, and D3HOC, where "PC" refers to "proper cali- bration," "LOC" refers to "low overconfidence," and "HOC" refers to "high overconfidence." As has been discussed, a more overconfident investor in this con- text is one who more strongly believes in his ability to appropriately value the secu- rity. The three investors are similar in some respects. They all analyze the security in question and arrive at the same prior value estimate, which is designated as in the v0 graph. For this reason the equilibrium price (and all posterior value estimates) is also equal to v0. This is why the three individual demand curves intersect at (qn, v0). One investor (the one whose demand curve is D1PC) has a vertical demand curve. For him, a1 = 0. The other two investors have negatively sloped demand curves, implying that lower prices increase demand, and higher prices decrease de- mand. For both investors, ai is positive, but note that a3 \> a2. While the second in- vestor pays some attention to her own opinion, the third investor pays the most attention to his own opinion. Since Investor 3 is more influenced by prior value- price discrepancies than is Investor 2, Investor 3 is relatively more overconfident. Investor 3 puts less weight on the market price and more credence in his own prior estimate. Thus, when the market price increases, he responds by adjusting de- mand further down relative to Investor 2. Similarly, when the market price adjusts down, Investor 3 responds more strongly than Investor 2 by demanding relatively more shares. Let us use this framework to elucidate the role of overconfidence on trading and volatility. To do so, we will assume that there are 300 shares outstanding (Q = 300). For this illustration, we will assume that the demand curves of the three investors are as follows: 5. q1 = 100 (Investor 1) 6. p = 20 − 0.1 \* q2 (Investor 2) 7. p = 15 − 0.05 \* q3 (Investor 3) Notice that the more overconfident trader has a flatter, more price-responsive de- mand curve (i.e., the slope for 9.6 is less than the slope for 9.7). Figure 9.2 shows the aggregate supply and aggregate demand curves on a single graph. The aggre- gate supply for shares is 300. The aggregate demand curve is a horizontal summa- tion of all individual investors' demand curves. At \$20 or more, Investor 1 demands 100 shares and no other investors express interest; between \$20 and \$15, Investor 1 continues to demand 100 shares and Investor 2 now demands a positive amount that declines with price; at lower prices, all investors have positive demands. Aggregate supply and aggregate demand intersect at \$10, which is where vo = p and qn = q1 = q2 = q3 =100. Periodically, investors reassess their prior value estimates. Many will do so when material news arrives. To keep this example simple, let's suppose that one investor alters her value estimate after a thorough (second) analysis of the stock. Specifically, suppose Investor 2 believes that the security has become more valu- able. We operationalize this by a \$5 parallel shift in the demand curve of Investor 2\. The new demand curve for this investor is: 9.8 p = 25 − 0.1 \* q2 (Investor 2 − Scenario 1) Note that we call the environment as specified Scenario 1. Figure 9.3 shows how the aggregate demand curve looks after this revision. Not surprisingly, the aggre- gate demand curve has shifted up. The new equilibrium price is \$11.67. This illus- trates that price is a weighted average of the three value estimates---while the other two investors still believe the stock should sell for \$10, the third thinks \$15 is right. To investigate the role of overconfidence, we will alter the situation again by increasing the overconfidence level of one of the traders (while returning to the ini- tial situation where for all traders v0 = \$10), thus generating Scenario 2. We do so as follows: 9.9 p = 15 − 0.05 \* q2 (Investor 2 − Scenario 2) Investor 2 now has the same demand curve as Investor 3, which implies that they both now have the same (high) level of overconfidence. Figure 9.4 illustrates that the initial equilibrium price, \$10, is the same as before, since all investors still ini- tially believe that the value of the security is \$10. The difference, though, is apparent if we consider Scenario 3, which combines Scenarios 1 and 2 in the sense that Investor 2 is both more overconfident than be- fore and she also increases her estimate of value by \$5. Her new demand curve is: 9.10 p = 20 − 0.05 \* q2 (Investor 2 − Scenario 3) Figure 9.5 shows that, once again, not surprisingly, the price rises, this time to \$12.50. Note that the price rise is higher than before since the investor with the ex- treme view, being more overconfident than before, is more willing to trust her opinion and transact on this basis. The price is still a weighted average of the three value estimates, but the investor with the extreme view exerts a greater influence on it because of her willingness to trade more. It is straightforward to show that a value revision in the negative direction will work the same but in reverse. The first lesson is that volatility increases with over- confidence. The same value revision led to a greater price change when one of the traders was more overconfident. The second lesson is that overconfidence induces greater trading activity---as well as higher levels of volume at the level of the market. Assuming that all inves- tors begin with 100 shares (the initial situation), in Scenario 1, Investor 2, who has become more optimistic, increases her holding to 133.33 shares. This is accom- modated by a 33.33 sale by Investor 3. Contrast that to Scenario 3. In this case, In- vestor 2 increases her holding to 150 shares, for a net purchase of 50 shares. This is accommodated by a 50 share sale by Investor 3. Thus higher overconfidence is associated with more trading. While this example is merely suggestive, it is consistent with rigorous theoreti- cal models. For example, Terrance Odean formulates a model where investors receive noisy signals on the future value of a stock.4 While investors realize their in- formation and opinions are imperfect, they believe them to be more precise than they really are. In other words, they are overconfident in the sense of being miscali- brated. Several predictions are derived from Odean's model. Consistent with the overconfidence example presented here, it is demonstrated that: 1) expected trading volume increases as overconfidence increases; and 2) price volatility increases with overconfidence. Several other notable results emerge as well: 3) overconfidence worsens the quality of prices, which means they are less likely to be accurate esti- mates of value; and 4) overconfident traders have lower expected utility than do those who are properly calibrated. The third prediction follows from the fact that divergent views sometimes receive a lot of weight if the trader in question is well- capitalized and egregiously overconfident. The fourth follows from the fact that in- vestors take on excess risk relative to those who are well calibrated. EVIDENCE FROM THE FIELD Are these predictions corroborated by evidence from the field? Brad Barber and Terrance Odean investigated the performance of individual investors by examining the trading histories of more than 60,000 U.S. discount brokerage investors be- tween 1991 and 1996.5 Their goal was to see if the trades of these investors were justified in the sense that they led to improvements in portfolio performance. Think about why a market transaction would make sense. Suppose, for example, you sell one stock and use the proceeds to buy another, and in doing so incur \$200 in transaction costs. This transaction is only logical if you expect to generate a higher portfolio return---high enough to at least offset the transaction cost. To be sure, in- dividual investors do a lot of trading. In their study, Barber and Odean found that, on average, investors turn over 75% of their portfolios annually. This means that, for a typical investor who holds a \$100,000 portfolio, in a given year she trades \$75,000 worth of stock. Barber and Odean divided their sample of individual investors into five equal groups (quintiles), where the groups were formed on the basis of portfolio turn- over. Specifically, the 20% of investors who traded the least were assigned to the lowest turnover quintile (no. 1), the 20% of investors who traded the next least were assigned to quintile 2, and so on---all the way to quintile 5, which was re- served for those investors trading the most. To put all this into perspective, those trading the least only turned over 0.19% of their portfolio per month---less than 3% per year. Those trading the most turned over 21.49% of their portfolio per month---more than 300% per year. Referring to Figure 9.6, we see for each quin- tile the gross average monthly return and the net (after transaction costs) average monthly return. The returns for all quintiles (both gross and net) were fairly high during this period (even for those trading excessively) because the overall stock market was performing quite well. Was all this trading worthwhile? Was it based on superior information, or was it based on the perception of superior information (i.e., misinformation)? An inspection of the figure reveals that while the additional trading did lead to a very slight im- provement in gross performance, net performance suffered. In other words, most of the trading was not helpful. The evidence reported by Barber and Odean suggests that the trades were not based on superior information, but rather were often con- ducted because of misinformation. While it is impossible to prove without a doubt that overconfidence was the culprit, this view appears to be a reasonable one. While Figure 9.6 is in terms of raw returns, sometimes returns are high because greater risk is taken and investors are merely being properly rewarded for the risk borne. If an investor earns high average returns only because high risk has been borne, this does not imply any sort of stock-picking skill. After risk-adjusting re- turns, Barber and Odean found that their results were quite similar to those dis- played in Figure 9.6. For all investors, the net risk-adjusted annual return (after taking into account transaction costs, bid-ask spreads, and differential risk) was be- low the market return by well over 3.00%. The 20% of investors who traded the most underperformed the market (again on a net risk-adjusted basis) by about 10% per year. EVIDENCE FROM SURVEYS AND THE LAB While the previous study by Barber and Odean is an important one, it does have one unavoidable drawback. It is difficult to unambiguously explore the potential nexus between overconfidence and trading activity using market data, since nor- mally no psychometric data on individuals (or markets) exist. New research seeks to overcome this problem. For example, Markus Glaser and Martin Weber combined naturally occurring data with information elicited from a survey.6 Using trading data from online brokerage accounts and psychomet- ric data obtained from the same group of investors who responded to an online questionnaire, they correlate various measures of trading activity with a number of metrics of overconfidence. While there was solid evidence that those who are most subject to the better-than-average effect trade more, there was little such corre- sponding evidence for those who were most overconfident based on calibration tests.7 In a similar vein, Mark Grinblatt and Matti Keloharju investigate whether trading activity, based on a comprehensive dataset of equity trading data in Finland, is related to overconfidence and sensation seeking.8 Sensation seeking is a personality trait whose four dimensions are thrill and adventure seeking (i.e., a de- sire to engage in thrilling and even dangerous activities); experience seeking (i.e., the desire to have new and exciting experiences, even if illegal); disinhibition (i.e., behaviors associated with a loss of social inhibitions); and boredom susceptibility (i.e., dislike of repetition of experience).9 We might reasonably expect those with a high degree of sensation seeking to be prone to excessive trading because of the novelty derived from the experience of the trade and new stock in their portfolio. In the Grinblatt-Keloharju paper sensation seeking is proxied by the number of speeding tickets obtained by an individual. Arguably, this only captures the thrill and adventure-seeking dimension of sensation seeking. A measure of overconfi- dence is obtained from a mandatory psychological profile given to all Finnish males upon entry into military service. While precise details are not publicly available, this overconfidence measure appears to be closest to the better-than-average effect. The authors conclude that trading activity is positively related to both overconfi- dence and sensation seeking in their sample. In an experimental setting, Bruno Biais, Denis Hilton, Karine Mazurier and Sébastien Pouget considered the impact of two psychological traits, overconfidence, based on calibration tests, and self-monitoring, namely the disposition to attend to social cues and appropriately adjust behavior.10 Both measures were taken prior to the participation of students in a series of trading sessions. They found that while overconfidence did not lead to a significant increase in trading intensity, it did serve to significantly reduce profits. High self-monitors, on the other hand, earned rela- tively greater trading profits. In another experiment, researchers explored the relationship between trading activity (number of transactions) and their miscalibration-based proxy for overcon- fidence, the better-than-average effect, and illusion of control.11 Their approach was novel in that their experimental design induced overconfident traders to believe that their signals were more informative than those of others. In previous experi- mental work, when private information was provided there was either no differ- ence in signal quality or, when differences in quality existed, signals were randomly assigned. They took their cue from naturally occurring markets where many, through some form of analysis, habitually generate their own information. Referring to Table 9.1, Specification (1), they found that both miscalibration and the better-than-average effect led to more trading. No significant effect was found for illusion of control. Specification (2) explored additional determinants of trading ac- tivity beyond overconfidence measures. Older subjects (p-value = 0.016) with more financial education (p-value = 0.003) traded less. On the other hand, those with real- world trading experience felt more comfortable "pulling the trigger" and hence traded more (p-value = 0.023).12 9.3 DEMOGRAPHICS AND DYNAMICS GENDER AND OVERCONFIDENCE IN THE FINANCIAL REALM Recall from Chapter 5 that men tend to be more overconfident than women. Does this translate into the financial realm? The answer appears to be yes. Barber and Odean, using the dataset discussed previously in this chapter, explored the role of gender in the context of investment decision-making.13 They reported that, on av- erage, men traded 45% more than did women, thus incurring higher trading costs. While both genders reduce their net returns by trading, men do so by 0.94% more than women. The difference between single men and single women is starker, with single men trading 67% more, thus reducing their returns by 1.44% more than women. Other studies comparing the activity of male and female portfolio managers and male and female business students find little difference between the genders and trading activity and overconfidence.14 One possible reason is that the finance and business professions, being often viewed as male activities, attract women who are relatively more overconfident.15 DYNAMICS OF OVERCONFIDENCE AMONG MARKET PRACTITIONERS While most of the preceding discussion has focused on retail investors, there is no reason to think that sophisticated investors, even those whose success relies on hav- ing a good sense of the limits of their knowledge, are immune. In the realm of fi- nancial markets, market practitioners are often called on to generate forecasts. Analysts and portfolio managers, for example, forecast revenues and earnings, and economists forecast GDP and the level of the stock market. In the case of analysts who suffer from conflicts of interest, as discussed later in the chapter, there are rea- sons to take their prognostications with a grain of salt. This argument, however, does not hold for money managers who are seeking to capture alpha (excess re- turns) and who have no such conflicts of interest. Nor does it obviously hold for macroeconomic or market forecasters. One advantage to looking at professionals is that they often make public fore- casts. This begs the question: Do professionals learn by their mistakes and over time develop a good sense of their knowledge? The dynamics of overconfidence is clearly an important issue. It seems logical to think that if people recall their suc- cesses and failures equally clearly, they should move toward an accurate view over time. Experience should engender wisdom. On the other hand, the prevalence and persistence of overconfidence suggest that forces able to eliminate it are weak. Cog- nitive dissonance sometimes induces us to forget what is unpleasant or did not go our way.16 Moreover, as discussed earlier, self-attribution bias leads us to remem- ber our successes with great clarity, if not embellishment; hindsight bias induces us to idealize our memory of what we believe or forecasted in the past; and confirmation bias, the tendency to search out evidence consistent with one's prior beliefs and to ignore conflicting data, also contribute.17 A strict efficient markets view of the world would seem to argue that those fooling themselves in this way will be driven from the marketplace, but some have called this into question.18 The dynamic nature of overconfidence is stressed in a number of theoretical models. In the multi-period of setting of Simon Gervais and Odean, past successes, through the mechanism of self-attribution bias, exacerbate overconfidence, while past failures tend to be downplayed.19 The inference is that those who have had the good fortune of being successful in their fields might for a time be more over- confident than those who have just entered the market. Eventually, however, expe- rience should reveal to people their true knowledge level. The evidence on whether professionals are overconfident is mixed. One study examined a dataset of futures market traders and was unable to find any costs as- sociated with their trading activity.20 From this the authors inferred that these tra- ders were not overconfident. On the other hand, they found that less-disciplined traders were less successful than other traders, arguing that a lack of discipline can stem from overconfidently ignoring new public information. In another study, the forecasts of a group of German market practitioners were examined.21 These individuals were asked to provide both forecasts for the future level of the DAX (the German counterpart to the Dow) and 90% confidence bounds. This respondent group was egregiously overconfident. Their dynamic be- havior, however, seemed more in line with rational learning than self-attribution bias because respondents narrowed their intervals after successes as much as they widened them after failures. At the same time, this research found that market ex- perience made overconfidence worse, which is more consistent with a "learning to be overconfident" view and self-attribution. A likely reason for this is that experi- ence is a double-edged sword. While we learn about our abilities (or lack thereof) from experience, those surviving in financial markets often have done so because of a run of success (good luck?), which has reinforced overconfidence through self- attribution bias. The latter research also provided evidence that overconfidence can increase even at the level of the entire market, which was apparent from the correlation of past returns and changes in overconfidence. This is in accord with what would be expected since high past returns are likely to make many in the market feel success- ful. Previously, this tendency had been shown indirectly, in that lagged market re- turns were correlated with increases in trading activity (which proxied for increases in overconfidence).22 9.4 UNDERDIVERSIFICATION AND EXCESSIVE RISK TAKING Another investor error likely related to overconfidence is the tendency to be under- diversified. This is suggested by the illustrative model previously presented--- underdiversified people are too quick to overweight/underweight securities when they receive a positive/negative signal, and insufficient diversification results. An- other factor is that most retail investors, lacking the time to analyze a large set of securities, will stop after several. As long as they believe they have identified a few "winners" in this group, they are content. After all, if they are so sure that certain stocks are good buys, why dilute their portfolios with stocks that they have not studied? In one study, the portfolio composition of more than 3,000 U.S. individuals was examined.23 Most held no stocks at all. Of those households that did hold stocks (more than 600), he found that the median number of stocks in their portfo- lios was only one. And only about 5% of stock-holding households held 10 or more stocks. Most evidence says that to achieve a reasonable level of diversifica- tion, one has to hold more than 10 different stocks (preferably in different sectors of the economy). Thus it seems clear that many individual investors are quite underdiversified. William Goetzmann and Alok Kumar sought to ascertain who were most prone to being underdiversified.24 Not surprisingly, they found that underdiversifi- cation was less severe among people who were financially sophisticated. Moreover, diversification increased with income, wealth, and age. Those who traded the most also tended to be the least diversified. This is likely because overconfidence is the driving force behind both excessive trading and underdiversification. Also less di- versified were those people who were sensitive to price trends and those who were influenced by home bias. Once again, these are likely markers of a lack of sophistication. Related to underdiversification is excessive risk taking. This is actually tauto- logical, in that underdiversification is tantamount to taking on risk for which there is no apparent reward. It is done, of course, in the hope of finding undervalued securities. The disposition effect, the tendency to hold on to losers too long with deleteri- ous consequences for performance, while often linked to regret, is also sometimes associated with overconfidence. An overconfident trader, overly wedded to prior beliefs, may discount negative public information that pushes down prices, thus holding on to losers and taking on excessive risk. Indeed, there is evidence that fu- tures traders exhibit this behavior. Traders with mid-day losses increase their risk and perform poorly subsequently.25 9.5 EXCESSIVE OPTIMISM AND ANALYSTS Abundant research has established that analysts tend to be excessively optimistic about the prospects of the companies that they are following.26 This is true both in the United States and internationally. While this issue will be revisited in Chap- ter 12, in order to set the stage, consider Table 9.2 that shows the distribution of analyst recommendations among strong buy, buy, hold, sell, and strong sell for G7 countries.27 It is clear that analysts are much more likely to recommend a purchase than a sale. In the United States, where this tendency was most pro- nounced, buys/sells were observed 52%/3% of the time. In Germany, where this tendency was least pronounced, the buy/sell ratio was 39%/20%. As will be made clear in our later discussion, while excessive optimism is one interpretation, another is a conflict of interest induced by a perceived need to keep prospective issuers happy. CHAPTER 10 INDIVIDUAL INVESTORS AND THE FORCE OF EMOTION 10.1 INTRODUCTION Market movements are commonly attributed to the emotions of investors. Yet it is not obvious how to separate the role of emotions from that of fundamentals in producing market outcomes. In Chapter 7 we considered the foundations of emo- tion. We learned that emotion includes cognitive, physiological, and evolutionary aspects. It was argued that emotions, when in balance, can facilitate decision- making, rather than hinder it. In this chapter, we will consider the extent to which the various aspects of emotion influence observed individual behavior in the finan- cial realm. The chapter begins, in Section 10.2, with a discussion of how mood impacts the decisions of individual investors. We will see that it is not easy to characterize the interaction between an investor's mood and risk attitude. Next, Section 10.3 considers two emotions that have received a lot of attention: pride and regret. Re- searchers have shown that these two emotions have very important effects on in- vestor behavior. Section 10.4 focuses on the disposition effect, one investor behavior that can be explained by emotion. The empirical evidence indicates that people tend to sell stocks that have performed well too soon, while holding on to poorly performing stocks too long. Though traditionally this behavior has been ra- tionalized using prospect theory, theoretical and experimental evidence suggest that emotions may provide a better explanation. Next, Section 10.5 discusses the house money effect, so-named from the observation that gamblers take increased risks af- ter winning because they feel they are betting with the house's money. A house money effect has been documented even for very large gambles, as research of game show contestant behavior shows. Finally, Section 10.6 considers how a per- son's assessment of a situation or impression of another, referred to as affect, shapes financial decision-making. 10.2 IS THE MOOD OF THE INVESTOR THE MOOD OF THE MARKET? In his best-selling book Irrational Exuberance, economist Robert Shiller argues that "the emotional state of investors when they decide on their investments is no doubt one of the most important factors causing the bull market" experienced around the world in the 1990s.1 Do traders' emotional dispositions translate into a market mood that, in turn, moves the market? This is a very interesting question. Some re- cent research concludes that what appears to be anomalous financial behavior can be explained by emotion. Here are some examples of this work. One study using data from 26 interna- tional stock exchanges argues that good moods resulting from morning sunshine lead to higher stock returns.2 A sunny day might make people more optimistic so that, in turn, they are more likely to buy stocks. Other researchers report that stock markets fall when traders' sleep patterns are disrupted due to clock changes with daylight savings time.3 A third recent study suggests that the outcomes of soccer games are strongly correlated with the mood of investors.4 After a loss in a World Cup elimination game, significant market declines are reported in the losing coun- try's market. Whether these aggregate studies of the effect of mood on stock market pricing provide clear evidence on how individual behavior translates into market outcomes is debatable. For example, even if people were irrationally optimistic on a sunny day, does it necessarily mean that they run out and buy stocks? Would you? Even if some people do rush to buy stocks on sunny days, market behavior can be con- sistent with rational pricing when individual behavior is characterized as irrational, as theoretical and experimental evidence suggests.5 At a more fundamental level, though, it is not clear that there is a simple way to characterize the relationship between mood and risk attitude. As we discussed in Chapter 1, risk attitude is important because it affects how a person values an as- set. If risk aversion changes in response to changes in mood, how much a person is willing to pay for a stock will change. When someone is in a poor mood, does he take more risks or fewer? The answer probably depends on the context and the individual's personality. For example, one person who is in a very sour mood may engage in risky behavior like driving recklessly or drinking too much alcohol. Another person who is not having a good day may shy away from risk more than usual and simply withdraw from others. The evidence does not provide compelling evidence that a buoyant mood consistently leads to lower risk aversion or that a poor mood consistently leads to increased risk aversion, particularly in a financial context. Some research suggests that happier people are more optimistic and assign higher probabilities to positive events.6 But at the same time, other decision- making research indicates that even though people may be more optimistic about their likelihood of winning a gamble when they are happy, the same people are much less willing to actually take the gamble.7 In other words, they are more risk averse when they are happy. When you are in a good mood you are less likely to gamble because you do not want to jeopardize the good mood. Thus, taken to- gether, it is unclear how positive and negative emotional states translate into changes in risk attitude and, in turn, market pricing. In addition to the studies that tie market movements to changes in mood, some researchers link depression induced by reduced daylight to stock market cycles.8 As with the evidence on the effect of mood on risk choices, evidence on the relation- ship between risk attitude and depression does not provide a clear picture. Clinical depression is clearly different from a simple bad mood---depression has a biochem- ical basis and can occur with no cognitive appraisals. The current view of depres- sion by psychologists recognizes that it may involve altered brain circuitry.9 A person with no chemical imbalances will naturally experience anxiety in some si- tuations (e.g., a job interview) but a depressed person can feel chronically anxious. Some researchers question the importance of anxiety or depression in explaining choices across risky alternatives.10 Others conclude that risk aversion is correlated with depressive tendencies, but the correlation between depressive symptoms and risk aversion may arise from the correlation between anxiety and depression.11 Thus, the fundamental issue of how depression and risk attitude are linked re- mains unresolved.12 While a depressed person who shies away from risk with no apparent basis may seem to be irrational, an anxious person may be completely rational when he decides to move toward safer alternatives. Further research is needed before we can move toward definitive conclusions. Neuroscience research, as will be discussed in Chapter 20, is making inroads into the workings of the hu- man brain. 10.3 PRIDE AND REGRET While it may be premature to assert that we understand every factor that affects decision-making, some emotions have proven to be useful in understanding the fi- nancial choices people make, perhaps most notably, pride and regret. Regret is ob- viously a negative emotion. You might regret a bad investment decision and wish you had made a different choice. Your negative feelings are only amplified if you have to report a loss to your spouse, friends, or colleagues. Pride is the flip side of regret. You probably would not mind too much if it just slipped out in conversa- tion that you made a good profit on a trade. Psychologists and economists recognize the important impact regret and pride have on financial decision-making. Researchers believe that people are strongly mo- tivated to avoid the feeling of regret.13 Importantly, the effects of pride and regret are asymmetric. It seems that the negative emotion, regret, is felt more strongly by people. Researchers found that a number of the implications of expected utility theory are not corroborated by experimental evidence. This led to the development of al- ternative models of decision-making under uncertainty, prospect theory being the most popular of these. As was discussed in Chapter 3, central to prospect theory is that people are sometimes risk seeking. This occurs in the domain of losses and in the domain of gains for lottery-type prospects. Is it possible that regret and pride are behind these two tendencies to be risk seeking? In the case of risk seeking in the domain of losses, it may be that people want to avoid the negative feeling of regret that would occur if they had to recognize a loss, and so they gravitate away from their natural tendency to be risk averse. As for the lottery effect, a big low-probability gain, whether from picking a long shot at the track or from undertaking some research to find a "diamond in the rough" stock that you think is about to take off against all odds, may lead to anticipated pride and even risk seeking as you can just see yourself telling your friends about your acumen. Whatever the reality, it is clear that pride and regret are powerful emotions that impact the decisions people make. Now we will consider a specific financial behavior and investigate whether emotion explains observed choices. 10.4 THE DISPOSITION EFFECT Researchers have recognized the tendency of investors to sell superior-performing stocks too early while holding on to losing stocks too long.14 Perhaps you have ob- served this behavior in others, or even experienced it yourself. Have you ever heard someone express a sentiment such as, "This stock has really shot up so I better sell now and realize the gain?" Or, can you imagine yourself thinking, "I have lost a lot of money on this stock already, but I can't sell it now because it has to turn around some day?" The tendency to sell winners and hold losers is called the disposition effect. EMPIRICAL EVIDENCE We begin with some recent empirical evidence documenting the existence of the disposition effect. For example, Terrance Odean, using a database that included trading records for 10,000 discount brokerage accounts with almost 100,000 transactions during 1987--1993, carefully documented the tendency of individual investors to sell winners and hold on to losers.15 To distinguish between winners and losers we need a reference point. Consistent with prospect theory, Odean used the purchase price of each security (or average purchase price in the case of multi- ple transactions). One issue that had to be confronted is that in an up market many stocks will be winners, so it is natural that more winners than losers will be sold. Odean dealt with this by focusing on the frequency of winner/loser sales relative to the opportunities for winner/loser sales. Specifically, he calculated the proportion of gains realized (PGR) as: For example, when the sale of a winner occurs in an account, you compare this to all winners that could have been sold. Paper gains include any sales that could have been made at a gain. Similarly, the proportion of losses realized (PLR) was calcu- lated as follows: To provide insight into the tendency of these individual investors to sell winners while holding losers, Odean tested the hypothesis that the proportion of gains real- ized exceeded the proportion of losses realized. From Table 10.1, which aggregates over all investor accounts, there is a clear tendency to sell winners over losers (PGR \> PLR) over the entire year. It is impor- tant to note that for tax reasons investors should prefer to sell losers, not winners. An investor with a positive tax rate should put off realizing gains on winners be- cause of the tax liability generated, but should recognize losses sooner in order to reduce current tax liability. The second numerical column in the table shows that the disposition effect operates despite the fact that some investors understand this tax issue and act accordingly. In the month of December, when investors are most likely to transact for tax reasons, there is actually a greater tendency to sell losers rather than winners. It is in the other 11 months (the third numerical column) where the disposition effect dominates. To explain these observations, Odean considers several possibilities related to rationality. First, portfolio rebalancing suggests that losers, whose aggregate value is now lower than winners, need to have their positions increased relative to win- ners in order restore desired portfolio allocations. Odean investigated this and found it did not matter appreciably. Second, perhaps investors anticipate that losers will outperform winners looking forward. This is symptomatic of the tendency for long-term reversal discussed in Chapter 4. Unfortunately, investors have their tim- ing wrong, as they are selling medium-term (not long-term) winners and holding on to medium-term (not long-term) losers. This is exactly the opposite of what they should do. Indeed, looking ahead over the next year, Odean finds that win- ners sold outperform losers held by 3.41% on a risk-adjusted basis. It is for this reason that researchers sometimes speak of the disposition effect as selling winners too soon and holding on to losers too long. PROSPECT THEORY AS AN EXPLANATION FOR THE DISPOSITION EFFECT Hersh Shefrin and Meir Statman were the first to try to explain why the disposition effect is observed.16 Their explanations fall into two categories: prospect theory (coupled with mental accounting) and regret aversion (coupled with self-control problems). While nothing precludes the possibility of a role for both behavioral ex- planations, Shefrin and Statman emphasize prospect theory over the emotion of re- gret, and many commentators since then have followed this cue. Based on recent research described next, however, emotion may be the more important factor. First we begin with the prospect theory explanation. Consider Figure 10.1, which shows how gains and losses appear according to prospect theory, pro- vided that prior outcomes are integrated. Stocks A and B have suffered losses, while C and D have experienced gains. How would these gains and losses affect your behavior as an investor? After a large gain (D), you have moved to the risk- averse segment of the value function. Only major reversals of fortune are likely to move you back to the origin. On the other hand, after a large loss (A) you have moved to the risk-seeking segment of the value function and, again, you are unlikely to move quickly back to your reference point. The implication is that since you are less risk averse for losers than winners, you are more likely to hold on to them. Still, why not engage in a tax swap (the simultaneous purchase and sale of two similar securities for tax reasons) in order to reduce tax payments? With a tax swap, an investor sells a losing stock and buys another stock with similar risk in order to realize a loss for tax purposes without changing the risk exposure in her portfolio. Though this strategy seems to make sense, if the investor uses mental ac- counting and evaluates the stocks separately, a tax swap would entail closing one account at a loss. As we have seen, many have difficulty doing so. Closing an account at a loss is difficult because of regret aversion. Shefrin and Statman argue that the fear of triggering regret leads an investor to postpone losses, whereas on the other side, the desire for pride (and/or rejoicing) leads to the reali- zation of gains. An investor feels regretful when closing a position with a loss be- cause of the (ex post) poor investment decision that was made, but feels pride when closing a position with a gain because the financial decision resulted in a profit. As for self-control, it is argued that even though investors often know they are doing the wrong thing, they have difficulty controlling the impulse to hold on to losers. ANOTHER POSSIBLE EXPLANATION Nicholas Barberis and Wei Xiong have recently revisited the prospect theory expla- nation of the disposition effect.17 Noting that previous justifications have been in- formal at best, they adopt a rigorous theoretical perspective. These researchers conclude that depending on the assumed parameters of the model, the implied be- havior of investors can easily be the very opposite of what prospect theory would suggest. They argue that the problem with prospect theory is that it does not take account of the initial decision to purchase the stock. In the two-period version of the Barberis-Xiong model, the parameterization of prospect theory preferences put forth by Kahneman and Tversky always predicts behavior opposite of what prospect theory calls for. The following simple example is provided. Given loss aversion, the expected return on the stock must be high. Otherwise, investors would not hold it to start with. Say a stock is priced at \$50 and can go up or down next period with equal probability. If it goes up, it rises to \$60 for a \$10 gain. A loss aversion coefficient of two (λ = 2), which is in the neigh- borhood of the Tversky-Kahneman value, implies that an investor would have only been willing to acquire the stock in the first place if the possible loss next period were \$5 or less. Let's assume \$5. Now consider what happens after the stock is ac- quired and either a loss or gain has been experienced. In their model, if a stock has initially done well, an investor will take a position in a stock that leads to breaking even if, in the worst-case scenario, the stock falls the following period. On the other hand, if a stock has initially done poorly, the position taken will be one that will lead to their breaking even in the best-case sce- nario. Suppose the stock initially did well and increased in value to \$60. Since after a gain of \$10 a subsequent loss will only take away half of this, the investor dou- bles the number of shares. This is because the value function is mildly concave in the gain domain so the investor is close to risk neutral. Conversely, suppose the stock fell to \$45 the first period. With a loss of \$5, only a half a share is required to get the investor back to square one if the stock increases in value. This suggests a partial liquidation, with the investor selling half a share. In short, the exact oppo- site of the disposition effect is implied. The investor with prospect theory prefer- ences buys after a gain and sells after a loss. EXPERIMENTAL EVIDENCE A recent experiment by Barbara Summers and Darren Duxbury also favors emo- tions over prospect theory in explaining the disposition effect.18 Their experimental design is predicated on whether or not individuals have chosen their investments. Suppose, contrary to what generally occurs in reality, there is no choice and you merely have to sit back and observe how your stocks are performing. When a stock you own fares poorly, you experience disappointment, and when your stock per- forms well you experience elation. If you actually selected these stocks yourself, ar- guably you will experience emotions with higher valence---in the face of a loss, you will experience regret (which is stronger than disappointment), and in the face of a gain, you will experience rejoicing (which is stronger than elation). Summers and Duxbury hypothesize that anticipated regret and rejoicing are necessary to generate behavior that is consistent with the disposition effect. In order to separate the emotional responses, Summers and Duxbury manipu- lated choice and responsibility regarding participants' current stock positions. Each participant was shown a graph of a single stock, with some groups given a winning stock and others a losing stock. Participants were then allowed to sell some or all of their stock. In the first treatment, there was no initial choice---sub- jects were told that they had inherited the stock from a relative. In the second treat- ment, there was an earlier stage where subjects could freely decide how much (if any) of the stock to hold. If prospect theory without emotion explains the disposi- tion effect, the mere experience of a gain or loss without personal responsibility for the choice of investments (first treatment) should induce the disposition effect. It did not. The second treatment where choice was given did, however, reveal a dis- position effect, with the proportion of gains realized greater than the proportion of losses realized (with statistical significance at less than 1%). Summers and Dux- bury concluded that responsibility for an outcome is a prerequisite for the disposi- tion effect, which highlights the importance of emotions in understanding the choices of investors. The disposition effect has been documented in another experimental study by Martin Weber and Colin Camerer.19 Interestingly, in one condition, participants' stock holdings were sold at the end of each period, regardless of their preferences. When shares were automatically sold, the disposition effect was moderated. This finding is consistent with a role for emotion in traders' choices, because when they begin anew each period, the negative feelings of regret and rejoicing are suppressed. 10.5 HOUSE MONEY Next, we turn to another example of path-dependent behavior. Path-dependent be- havior means that people's decisions are influenced by what has previously tran- spired. Richard Thaler and Eric Johnson provide evidence regarding how individual behavior is affected by prior gains and losses.20 After a prior gain, peo- ple become more open to assuming risk. This observed behavior is referred to as the house money effect, alluding to casino gamblers who are more willing to risk money that was recently won. After a prior loss, matters are not so clear-cut. On the one hand, people seem to value breaking even, so a person with a prior loss may take a risky gamble in order to try to break even. This observed behavior is referred to as the break even effect. On the other hand, an initial loss can cause an increase in risk aversion in what has been called the snake-bit effect. EVIDENCE OF A HOUSE MONEY EFFECT ON A LARGE SCALE The first evidence of path-dependence in decision-making came from hypothetical sur- veys or experiments conducted using student subjects, so whether the findings would carry over to high-stakes financial decisions was always open to challenge. To obviate this concern, some researchers turned to consideration of the decisions of game show contestants to provide insight into behavior when the stakes are large. One recent study by Thierry Post, Martijn J. van den Assem, Guido Baltussen, and Richard Thaler examined the choices made on the popular game show "Deal or No Deal?" This show first aired in the Netherlands in 2002 and has since been broadcast in numerous countries including Germany, Mexico, Spain, and the United States. Indeed, the stakes are large, with possible payouts in the Netherlands ranging from 0.01 to 5,000,000 euros. Though the rules of the game vary across countries, here is the basic setup. A contestant is presented with 26 suitcases each containing a hidden payout. The contestant selects one of the 26 as her own. This suitcase remains closed as she selects six others and views their contents. Next, a "bank offer" is made to the contestant, and, if she accepts it, she walks away with the offer with certainty. Otherwise, there is "no deal" between the contestant and the bank. She holds on to her suitcase, selects five more, and views their con- tents. The bank offers another deal, and the game continues until a deal is accepted or the contestant walks away with the contents of her suitcase. While the bank of- fers are not perfectly predictable, they typically begin low, rise over time, and in- crease (or decrease) when low- (or high-) value suitcases are opened. The researchers find that contestants' decisions are strongly influenced by what has happened before. When suitcases with low values are opened, contestants take on more risk. This is consistent with a house money effect because when low pay- offs are eliminated, expected winnings are higher and a contestant experiences a gain. On the other hand, when high-value suitcases are opened, a contestant ex- periences a loss in terms of expected winnings. Consistent with a break-even effect, contestants' decisions reflect decreased risk aversion, and they take risky gambles that give them the opportunity to break even. Importantly, the bottom line is that significant changes in expected wealth regardless of the sign lead to more risk taking. PROSPECT THEORY AND SEQUENTIAL DECISIONS Some of the findings on behavior following gains and losses appear to contradict prospect theory. The house money effect suggests reduced risk aversion after an ini- tial gain, whereas prospect theory makes no such prediction. It is notable, though, that a house money effect is not inconsistent with prospect theory because prospect theory was developed to describe one-shot gambles. Recall our discussion of inte- gration versus segregation in Chapter 3. Under integration, an investor combines the results of successive gambles, whereas, under segregation, each gamble is viewed separately. Instead of presenting a challenge to prospect theory, the house money effect is best seen as evidence that sequential gambles are sometimes inte- grated rather than segregated. If one integrates after a large gain, one has moved safely away from the value function loss aversion kink, serving to lessen risk aver- sion. Thinking in terms of emotions, how emotions like pride and regret are felt de- pends on how experiences are classified, as incremental or grouped together. The evidence provided by Thaler and Johnson provides important insight into how individuals make sequential decisions. People do not necessarily combine the outcomes of different gambles. Other researchers also document a house money ef- fect on individual behavior.21 Financial theory is increasingly incorporating insights on individual behavior provided by psychology and decision-making research. For example, in the model of Nicholas Barberis, Ming Huang, and Tano Santos, inves- tors receive utility from consumption and changes in wealth. In traditional models, people value only consumption.22 In this extension, investors are loss averse so that they are more sensitive to decreases than to increases in wealth, and, thus, prior outcomes affect subsequent behavior. After a stock price increase, people are less risk averse because prior gains cushion subsequent losses, whereas after a de- cline in stock prices, people are concerned about further losses and risk aversion in- creases. Therefore, Barberis, Huang, and Santos's model predicts that the existence of the house money effect in financial markets leads to greater volatility in stock prices. After prices rise, investors have a cushion of gains and are less averse to the risks involved in owning stock. Indeed, as in this model, aspects of prospect theory are increasingly being embedded in financial models. Despite progress, it does not seem that our understanding of sequential behav- ior in a market setting is complete. How does individual behavior translate to a market setting? A recent experimental study that includes a market with sequential decision-making provides some insight.23 Traders who are given a greater windfall of income before trading begins bid higher to acquire the asset, and, thus, the mar- ket prices are significantly higher. In fact, prices remain higher over the entirety of the three-period markets. As the house money effect would predict, people seem to be less risk averse after a windfall of money, as if the earlier gain cushions subse- quent losses. Observed behavior does not always suggest that traders will pay more to acquire stock after further increases in wealth. There is no evidence that traders become more risk taking if additional profits are generated by good trades when the market is open. The results indicate that the absolute level of wealth has a dominating influence on subsequent behavior so that changes in wealth are less important. This observed behavior among traders could be because professional traders are trained to act in a more normative (i.e., less prospect theory-like, less emotional) fashion. Indeed, more work is required to allow us to better understand the dynamics of markets and whether individual behavior adapts to or influences market outcomes.24 10.6 AFFECT Thus far we have argued that emotions, particularly regret, can impact financial decision-making. Emotional responses are also caused by the many stimuli we ex- perience continuously every day. A person's affective assessment is the sentiment that arises from a stimulus. For instance, imagine yourself negotiating a contract for your firm. Then imagine you had an immediate dislike for the other negotiator. Would you guess that the outcome is probably affected by your sentiment? Affect refers to the quality of a stimulus and reflects a person's impression or assessment. Cognitively, a person's perception includes affective reactions and, thus, judgment and decision-making are tied to the particular reactions the person has. Some psychologists have argued that peoples' thoughts are made up of images that include perceptual and symbolic representations.25 The images are marked by positive or negative feelings that are linked to somatic (or body) states. At the neu- ral level, somatic markers arising from experience establish a connection between an experience and a body state (such as pleasant or unpleasant). In effect, affective reactions are cognitive representations of distinct body states, and the brain uses an emotion to interpret a situation. People are attracted to a stimulus linked with a positive somatic marker and avoid those associated with negative somatic markers. Affective reactions that are easy for a person to access provide convenient and effi- cient means for decision making because the reactions allow a far easier way to evaluate the plusses and minuses of a stimulus.26 Some research has examined the role of affect in financial decision-making. In Chapter 16, we discuss how managers might be influenced. Affect also plays a role in markets. For example, some argue that a relationship exists between the im- age of a market and what has occurred in the market.27 This conclusion is based on the observation that experimental participants' willingness to invest in a firm is influenced by the subjects' affective reaction to the firm's industry membership. Other experiments also indicate that firm image has a significant effect on the port- folio allocation decisions of participants.28 In the future, we will likely see more research on the role of affect in financial decisions. Psychologists believe that affective reactions influence judgment and deci- sion making, even without cognitive evaluations, but we do not have a full under- standing of how the influences mesh into outcomes.29 In addition, psychologists suggest that when affective reactions and cognitive evaluations suggest different courses of action, the emotional aspects can be the dominating influence on behav- ior.30 But again, we have a lot to learn if we are to understand when a particular force is likely to dominate. CHAPTER HIGHLIGHTS 1\. Some researchers suggest that the mood of the investor translates into the mood of the market and, in turn, impacts market outcomes. These conclu- sions should be interpreted with caution because we do not fully understand the relationship between emotion and risk attitude. 2\. Much evidence suggests that two emotions, pride and regret, have significant effects on individual financial decision-making. 3\. According to the disposition effect, people sell winners too soon and hold on to losers too long. Empirical evidence documents this tendency. 4\. The disposition effect has traditionally been explained by prospect theory. Because of the shape of the value function, investors are less risk averse for losers, so they are more likely to hold on to them. 5\. Recent theoretical arguments and experimental evidence suggest that loss aversion resulting from a fear of regret may provide a better account of the disposition effect. 6\. According to the house money effect, after a prior gain, investors become less risk averse. 7\. After losses, the snake-bit effect (whereby people are less likely to take on risk), and the break-even effect (whereby people are more likely to take on risk) operate in opposite directions. The latter seems to usually dominate. 8\. Path-dependence in decisions, which suggests that people sometimes integrate sequential gambles, is corroborated for large-scale gambles by considering the choices made by game show contestants. 9\. Affect reflects a person's impression or assessment of a stimulus. Because a per- son's perception is tied to the affective reaction, decisions are impacted by affect.