Behavioural finance is an effort to build new and different models of stock price behavior that better fit the observable data. The element of psychology in such models proceeds from the assumption that important forms of human behavior are unlikely to be ‘washed out’ in the financial markets as conventional economists have long assumed. Large numbers of so-called noise traders and investors buy and sell stock. If their cognitive biases are systematic enough, they will have an impact on prices that others do not arbitrage away. Thus, the models are constructed by reference to the many sorts of biases that have been identified in the explosion of work on investor judgment and decision-making that has occurred over the last thirty years or so.28
As with the literature about stock price anomalies, the research connecting psychology and finance has become too voluminous to catalogue here.29 Commentators have considered virtually every well-recognized bias some way or another, as well as some less obvious ones. Finance scholars, for example, have found evidence that moods triggered by good or bad weather can affect stock prices on a given day.30 However, I want to concentrate on four biases on which finance scholars have built the most visible and sustained research efforts, with special emphasis on the last bias—the phenomenon of investor over- confidence.
The first of these biases, and the one with the most distinguished pedigree in the cognitive bias literature, is loss aversion. People appear to approach risk-taking differently depending on the framing of the choice before them. When evaluating a potential gain, people exhibit a strong degree of risk aversion. But if prompted to see the choice as one of trying to avoid a loss of that which is currently possessed, people tend to be more risk-seeking. Thus, there could be differences in investors’
risk-taking approaches in buying versus selling stocks. We might expect
28See Kahneman and Riepe (1998). In addition to the finance studies on which we will focus, there has been an explosion in laboratory studies that seek to replicate features of the financial markets. See, e.g., Gangulyet al.(2000).
29See Hirshleifer (2001).
30See Saunders (1993); Hirshleifer (2001: 1560).
people to hold on to their losing stocks too long, and sell their winners too readily.31
A prominent behavioral model incorporates loss aversion with an interesting twist. Drawing on prior work sometimes described as the
‘house money effect,’ Barberis, Huang, and Santos argue that one’s degree of loss aversion will vary depending on recent prior performance (Barberis et al. 2001).32 If one has recently enjoyed gains, ‘possession’
effects do not operate as strongly; people are willing to take considerable risks with ‘found money.’ On the other hand, when one has suffered recent losses, people are reluctant to gamble much unless it is necessary to preserve what they have left. This suggests that after a run up in prices, people become more aggressive—one reason why we might observe greater volatility than traditional models might suggest.
The next two biases seem to present opposing tendencies (Barberiset al. 2001).33 Cognitive conservatism is an extremely robust behavioral construct showing that people change their views slowly even in the face of persuasive evidence. In other words, people cling as long as possible to what they previously believed. This bias could be the basis for the underreaction phenomenon described earlier. However, under some circumstances, this tendency is reversed, as new information has an excessive effect on judgment, prompting overreaction. This could be explained by the ‘representativeness’ effect, under which people’s attentions are distracted from the baseline. Much work in psychology and finance tries to reconcile these conflicting tendencies.34 One possible reconciliation relates the new information to the pattern of prior news events. Another centres on the salience of the new information.35When the investor receives the new information in a particularly dramatic way, for example, it might be overweighted; when it is presented normally, it is not, so that cognitive conservatism controls the process of inference.
Salience would explain the EntreMed story. The ever-increasing volume of media coverage of investment information—on the internet, cable TV, and in the financial press—means that some stories will gain substantial saliency, while others will be buried under a heavy load of other information.36
While media attention is a well-recognized influence, many IMH scholars contend that the most underexplored aspect of behavioral
31See Odean (1998a); Shefrin and Statman (1985).
32See also Barberis and Huang (2001).
33See Shleifer (2000: 128–30).
34See Hong and Stein (1999); Nicholas Barberiset al.(1998); Michaelyet al.(1995). For work in the options market, see Poteshman (2001).
35See Klibanoffet al.(1999).
36In addition, increases in demand—for whatever reason—may themselves start bandwagon effects, even if no other information is conveyed. See Shiller (1999: 60–2).
finance is informal social contact among investors.37It is very likely that investors affect each other not simply by trading, but through conversations, including internet-based talk,38 and other forms of social influence.39 Hence, it is possible that further research will be able to document an ‘epidemiology’ of investor behavior—tracking the contagion of excitement or panic within embedded communities of traders. That research eventually may help us understand better, if not predict, why information becomes overweighted in some circumstances while similar information is underweighted in others.
Although loss aversion, cognitive conservatism, and representa- tiveness are mainstream subjects in behavioral finance, the last of our four has gained a particularly high level of prominence in recent years:
the phenomenon of investor overconfidence. In an oft-repeated quotation in the finance literature, De Bondt and Thaler state that ‘perhaps the most robust finding in the psychology of judgment is that people are overcon- fident’ (De Bondt and Thaler 1995: 385-6). People have a strong tendency to have greater faith in their intuitions and judgments than the evidence warrants.40 They put too much weight on their privately-acquired information or inference, and calibrate poorly even when they realize the presence of some uncertainty. This bias has a comparative dimension to it: people are overconfident in their skills vis-a-vis others. Indeed, far more than fifty per cent of a sampling of active investors will rate themselves as above average as compared to their peers at the task of investing.41Notably, there is an interesting gender element at work here:
overconfidence is dominantly a male trait.42
This bias of investor overconfidence is a popular subject of study among economists, including some conventional ones,43 for a few reasons. First, much observable economic behavior seems hard to explain except by reference to a hubris hypothesis. The volume of corporate takeover activity is an example, as winning bidders consistently pay high premiums for what often turn out to be unprofitable acquisitions.44 Secondly, there is an interesting evolutionary story behind the bias, which
37See Shiller (1999: 153–62); Hirshleifer (2001: 1552–53, 1577); Shiller and Pound (1989). For some early efforts in this direction, see Baker (1984); Klausner (1984). See generally Ellison and Fundenberg (1995).
38See below Part III.
39Of substantial relevance here is work on rumours among investors. See DiFazio and Bordia (1997).
40See Odean (1998b). For a review of the psychology literature, see Griffin and Varey (1999). For a linking of overconfidence with corporate disclosure policy, see Langevoort (1997).
41See Mooreet al.(1999: 97). Interestingly, the authors demonstrate that people overrate their past performance, as well as their future prospects.
42See Barber and Odean (2000a).
43See Rubenstein (2001: 17–18).
44See Roll (1986); see also Camerer and Lovallo (1999).
appeals to economists.45 Illusions of control and overoptimism are associated with a variety of positive outcomes: greater willingness to take risk, more persistence in the face of adversity, etc. One can readily see why being unrealistically confident, within moderation, can lead to greater success, even if it also leads to more mistakes as well. Those who bear greater risk are compensated for it, on average. Indeed, when they are also beneficiaries of a streak of good luck, we might expect that highly successful people—an important group in the world of investing—might be particularly infused with hubris.46
Finally, and most importantly, there is an increasing body of empirical evidence that directly supports investor overconfidence as an important trait. In what became a widely-reported study, Barber and Odean (2000) studied the investment performance of a large number of online brokerage accounts, which are held by those who think they can make their own trading decisions without the assistance of a stockbroker as adviser and have been the fastest growing segment of the brokerage industry over the last few years. What the researchers found is that the rate of trading increased once the accounts were established, especially after an initial spurt of good performance (or good luck). Notwith- standing this increasing volume of trading, overall average performance lagged behind what a more passive, well-diversified trading strategy would generate. Not surprisingly, almost all of the lag could be traced to the costs (e.g., commissions) associated with active trading. The authors state their conclusion simply: ‘Overconfident investors will overestimate the value of their private information, causing them to trade too actively and, consequently, to earn below-average returns’ (ibid.: 800).
Overconfidence is notably dynamic in character. Another long- recognized trait in human behavior, ‘biased self-attribution,’ is the tendency to take credit for positive results, but externalise blame for bad results. This tendency is one reason that people learn poorly from experience, as people are not willing to recognize that their failures stem from a lack of competence or skill. As a result, people will attribute a streak of good luck as skill and will attribute a run of losses to bad luck or someone else’s fault. Thus, when prices rise and investors gain from that alone, their investment decisions are readily self-characterizable as talent, which in turn will promote even more aggressive trading. Downswings will not have a comparable cautionary influence.47 In recent years, academics have developed several different behavioral finance models on overconfidence and biased self-attribution. Perhaps the best known is by
45For a formalization of this idea, see Benabou and Tirole (2002). For applications in the financial markets, see Kyle and Wang (1997); Benos (1998).
46See Gervais and Odean (2001).
47Obviously, there are many parallels here with gambling behavior. See Statman (2002). For a legal discussion linking these two phenomena, see Gabaldon (2001).
Daniel, Hirshleifer, and Subrahmanyam (1998), who explicitly use a dynamic model that assumes that overconfident traders overreact to private informational signals but underreact to those which are public.
This allows them to reconcile the observed underreaction and overreaction phenomena.48