In the context of loss aversion, which of the following statements is true of the endowment effect?

The Neurobiology of Loss Aversion

Loss aversion is proposed as a likely cause of most of the effects such as framing effect, the status quo bias and the endowment effect. The question then is what neurobiological circuit underpins loss aversion? Tom, Fox, Trepel and Poldrack (2007) addressed this issue using fMRI in a study where subjects were presented with a series of mixed gambles that offered a 50/50 chance to either gain or lose a given amount of money. Potential gains (ranging between $10 to $40, with $2 increments) and potential losses (ranging between –$5 to $20, with increments of $1) were manipulated independently and subjects were required to either accept or reject each proposed gamble. This setup enabled the authors to estimate the individual behavioral loss aversion (λ) computed as the ratio of the (absolute) loss response to the gain response, which yielded a median λ = 1.93, a degree of loss aversion consistent with many previous studies. The values of the potential gains and losses were entered into a regression analysis to identify brain areas showing a parametric change in the response to increasing magnitude of losses or gains. Regions encoding for potential reward responded to the increase in potential gains and this included the striatum, OFC, and dopaminergic midbrain regions. However, potential losses were also coded (deactivation in BOLD signal) by the same network. The observed pattern was suggestive of a signal that computed the net value of the gamble, by averaging the value of potential gains and potential losses. However, the most significant finding was that although a net value signal was monotonic, it was not linear: rather, it reflected the asymmetry between gains and losses shown behaviorally and predicted by prospect theory. In other words, for the majority of participants the decrease in activity in this network was steeper than the increase. Using the parameter estimates from the BOLD signal, the authors constructed a parameter that they called “neural loss aversion” that reflected the ratio between two slopes that correlated with individual behavioral loss aversion (see Figure 5.6).

A key question raised by this study concerns how such asymmetry is generated when the gamble net value encoded by the VMPFC and striatum is computed. A plausible candidate mechanism would be that the potential loss triggers an aversive “breaking” signal that is communicated to VMPFC and/or the striatum. This signal would overweight the loss component of the gamble, thereby affecting the computation of the net value to produce the kink observed in the neural signal and reflected in the value function. Consistent with a number of previous studies on framing and loss processing (some of them presented earlier), the amygdala is a likely candidate to emit such a signal. Indeed, in support of an hypothesis that an emotional response is at the root of the gain-loss asymmetry there is experimental evidence that implementation of standard cognitive appraisal strategies (aimed to down-regulate the emotional response) produces a reduction in autonomic skin conductance response, as well as a marked reduction in loss aversion (Sokol-Hessner, Hsu, Curley et al., 2009).

Quite surprisingly, however, Tom and colleagues did not find any signal in the amygdala. One explanation for this null finding is that the range of potential losses used might not have been large enough to evoke detectable BOLD signal in the amygdala. Another reason might be the peculiar nature of the BOLD signal that is known to reflect local field potential rather than output activity (Logothetis, 2002; Logothetis, Pauls, Augath et al., 2001). If, as proposed here, the amygdala’s contribution to loss aversion is mainly through its inputs to other brain structures like the VMPFC or the striatum, BOLD activity may well fail to reflect this neural signal.

To directly test a hypothesis that the amygdala is part of a computational network that produces loss aversion, we studied individuals with focal damage to the amygdala (De Martino, Camerer & Adolphs, 2010). Two subjects affected by a rare genetic syndrome that produces lesions of the amygdala nuclei were tested on a modified version of the task used by Tom and colleagues (2007). Each subject was compared with a control group closely matched in IQ, age, gender, education and monetary income. Both amygdala-lesioned participants showed a dramatic absence of loss aversion compared with their own control group and in fact one of the lesion subjects showed a mild loss seeking behavior. However, both individuals retained a monotonic sensitivity to reward magnitude (i.e., they preferred larger gains and smaller losses) and showed a marked dislike for increases in variance and risk. Given the well-known role of the amygdala in inhibiting instrumental behavior in response to potential threats (Adolphs, Tranel, Damasio & Damasio, 1995; LeDoux, 2000) it is plausible that the same mechanism is elicited in response to the prospect of a potential monetary loss. Such a “withdrawal” signal elicited by the amygdala would affect the value computation (in the VMPFC or striatum) and generate loss aversion. While this hypothetical model accommodates most of the experimental results currently available it will require direct testing by further studies.

2 Loss aversion

Loss aversion is a cornerstone of prospect theory (Kahneman and Tversky, 1979) which states that, the disutility of a loss is greater than the utility of a comparable gain. Kahneman and Tversky expressed the principle in hedonic terms: “The aggravation that one experiences in losing a sum of money appears to be greater than the pleasure associated with gaining the same amount” (p. 279). This principle has been used to explain many violations of economic theory, including the endowment effect. The endowment effect was first demonstrated by Kahneman et al. (1990, 1991). They randomly distributed mugs to half the students in a classroom. Those who received mugs were sellers. Sellers were asked to state the minimum amount of money they would be willing to accept to give up their mug. Those who did not get mugs were buyers. They reported the maximum amount of money they would be willing to pay to purchase a mug.

Since mugs were distributed randomly, there was no reason to assume that the utility of the mug would differ across groups. Economic theory asserts that prices for buyers and sellers should be approximately equal. Nonetheless, sellers wanted significantly more to give up their mugs than buyers were willing to pay. According to loss aversion, sellers view the exchange as a loss, and buyers perceive it as a gain. Losses loom larger than gains, so selling prices should exceed buying prices.

Hundreds of studies have used this paradigm and demonstrated that selling prices exceed buying prices. However, relatively few have tested the hedonic prediction implied by loss aversion in experimental markets. Mellers and Ritov (2010) asked sellers to imagine the pain of losing their mug. Buyers were asked to imagine the pleasure of getting a mug. If loss aversion described anticipated emotions as well as utilities, the pain of the imagined loss should be greater in magnitude than the pleasure of the imagined gain. Yet the opposite pattern emerged. The anticipated pleasure of the gain exceeded in magnitude the anticipated pain of the loss, (t(110) = 4.57), as shown in Fig. 4.

Mellers and Ritov suggested that this pattern could occur if surprise influenced judged emotions. The absence of “hedonic” loss aversion—and perhaps even the reversal—as shown in Fig. 4, could occur if buyers thought that gains were surprising, and sellers thought that losses were expected. Decision affect theory predicts that, under these circumstances, pleasure would increase and pain would decrease, perhaps even reversing the pattern of loss aversion.

To test this hypothesis, Mellers and Ritov (2010) asked buyers and sellers to rate their surprise with their outcome and the alternative one (i.e., endowment or the absence of endowment). Despite the equal odds, both buyers and sellers said that an endowment was more surprising than the absence of an endowment (5.1 and 3.0, respectively, on a scale of 1 (not at all surprising) to 7 (extremely surprising), (t(54) = 4.15)). This result may have occurred because subjects are typically not given mugs to take home when they participate in experiments. The pattern was consistent with the predictions of decision affect theory. Even if the utilities in decision affect theory were loss averse (see Eq. 1), surprise reversed the relative magnitude of judged pleasure and pain.

Several researchers have examined the hedonic implications of loss aversion in nonmarket contexts, and results are mixed (see Harinck et al., 2007; Kermer et al., 2006; Liberman et al., 2005; Rozin and Royzman, 2001). In a recent paper, McGraw et al. (2010) offered an explanation for the data. They suggested that when judging emotions, people naturally tend to use similar types of outcomes for comparison. Losses are compared to other losses and gains to other gains. Bipolar scales (anchored with “very happy” and “very unhappy” at the ends) have a natural zero point, and because of these natural comparisons, subjects may use the negative and positive sides of the scale differently. Pleasurable and painful ratings might not be comparable if people used different contexts for comparison.

McGraw et al. (2010) offered a method of judging pleasure and pain that encouraged direct comparison of gains and losses. With this procedure, people are asked to consider the pleasure of a gain and the pain of a loss. Then they are asked, “Which feeling is stronger?” McGraw et al. (2010) used this method with fair 50/50 gambles and stakes of $200. The majority of subjects said that the pain of the loss was more intense than the pleasure of the gain. But with bipolar ratings (used by Mellers and Ritov, 2010), McGraw et al. (2010) found that judged pleasure and pain were equal in magnitude.

By this account, the pattern of judged pleasure and pain found by Mellers and Ritov (2010) was due to the use of a bipolar response scale that did not force participants to directly compare gains to losses. To find out whether this method would reverse the pattern of loss aversion found by Mellers and Ritov (2010), Mellers and Berman (2012) asked buyers and sellers about their feelings using direct comparisons. Buyers and sellers were told, “We would like you to consider the emotional impact of two situations, A and B. In situation A: You did not get a mug. How much pleasure would you feel if you got one? In situation B: You got a mug, but had to give it up. How much pain would you feel if you had to give it up? In which situation would your feelings be stronger (not better or worse, but rather, more intense)? Situation A, Equal, or Situation B?” Participants who answered “Situation A” or “Situation B” were then asked to rate the intensity of the difference on a 5-point scale ranging from 1 = “very little” to 5 = “extremely.” Results were still inconsistent with loss aversion. Gains and losses were no different in their intensity. This leaves decision affect theory as the remaining account of why gains loomed larger than losses in the experimental markets.

To find out whether the direct comparison method suggested by McGraw et al. (2010) was sensitive to surprise effects, Mellers and Berman (2012) conducted another experiment in which people anticipated their feelings about the monetary outcomes of gambles. Outcomes were gains and losses of either $10 or $100, and the probabilities of winning were 10%, 50%, or 90%. Participants compared the pleasure of the gain to the pain of the loss and indicated which feeling was stronger. A follow-up question asked, “By how much?” Responses ranged from 1 = no difference to 5 = extremely different. Figure 5 shows the results.

The relative intensity of pleasure and pain is plotted on the y axis for the six gambles with light gray bars for $10 and dark gray bars for $100 gambles. According to McGraw et al. (2010), direct comparisons should result in “hedonic” loss aversion; all bars should fall below the zero point, regardless of the probability of outcomes. With fair 50/50 gambles, Mellers and Berman (2012) were able to replicate the results of McGraw et al (2010). Losses loomed larger than gains. But when the probabilities of winning were small, the relative magnitudes of pleasure and pain reversed. When gains were surprising (i.e., a 10% chance of winning), pleasure exceeded pain for the $10 gamble and pleasure was identical to pain for the $100 gamble. Figure 6 shows judged surprise for the outcomes of the gambles. Differences in surprise ratings (surprise of a gain − surprise of a loss) indicated that, when the probability of winning was 10%, gains were more surprising than losses for $10 and $100 gambles, and when the probability of losing was 10%, losses were more surprising than gains.

To summarize, surprise effects may have made the pleasure of a surprising gain of $10 exceed the pain of an expected $10 loss and the pleasure of a surprising gain of $100 equal in magnitude to the pain of an expected $100 loss, even when judgments were placed on a common continuum. Our results in Fig. 5 show that the relative magnitude of pleasure and pain is not fixed; it depends on the probabilities of occurrence. Surprise amplifies emotional experiences. The pleasure of a surprising gain can exceed the pain of an expected loss, and the pain of a surprising loss can be greater in magnitude than the pleasure of an expected gain.

3 Loss Aversion and the Status Quo Bias

A fundamental fact regarding people's reaction to outcomes is loss aversion: The loss of utility associated with giving up a good is greater than the utility associated with obtaining it (Tversky and Kahneman 1991). An immediate implication of loss aversion is that people will not accept an even chance to win or lose $X, because the loss of $X is more aversive than the gain of $X is attractive. Indeed, people are generally willing to accept an even-chance prospect only when the gain is substantially greater than (about twice as large as) the loss. Loss aversion entails that the loss of utility associated with giving up a good that is in our possession is generally greater than the utility gain associated with obtaining that good. This yields ‘endowment effects,’ wherein the mere possession of a good (thus viewing it as a potential loss) can lead to higher valuation of it than if it were not in one's possession (Kahneman et al. 1990).

A closely related manifestation of loss aversion is a general reluctance to trade, which is illustrated in a study (Knetsch 1989) in which subjects were divided into two groups: Half of the subjects were given a decorated mug, and the others were given a large bar of Swiss chocolate. Later, each subject was shown the alternative gift, and offered the opportunity to trade their gift for the other. Because the initial allocation of gifts was arbitrary and transaction costs minimal, economic theory predicts that about half the subjects should exchange their gifts. On the other hand, if losses loom larger than gains, then most participants will be reluctant to give up the gift in their possession (a loss) in order to obtain the other (a gain). Indeed, only 10 percent of the participants chose to trade their gifts. This contrasts sharply with the 50 percent predicted by standard economic analysis in which the value of a good does not change when it becomes part of one's endowment.

Loss aversion entails a strong tendency to maintain the status quo, because the disadvantages of departing from it loom larger than the advantages of its alternative (Samuelson and Zeckhauser 1988). A striking framing effect which relies on people's tendency to maintain the status quo has been observed in the context of insurance decisions, when New Jersey and Pennsylvania both introduced the option of a limited right to sue, which entitles automobile drivers to lower insurance rates. The two states differed, however, in what they offered consumers as the default option. New Jersey motorists had to acquire the full right to sue (transaction costs were minimal: a signature), whereas in Pennsylvania the full right to sue was the default. Presented with the choice, only about 20 percent of New Jersey drivers chose to acquire the full right to sue, while approximately 75 percent of Pennsylvania drivers chose to retain it. The difference in adoption rates due to the different frames had financial repercussions estimated at around $200 million dollars (Johnson et al. 1993).

Loss aversion promotes stability rather than change by inducing people to maintain their current position. Among other things, the reluctance to change induced by loss aversion can hinder the negotiated resolution of disputes. If each side to a dispute evaluates the opponent's concessions as gains and its own concessions as losses, then agreement will be hard to reach because each side will perceive itself as relinquishing more than it stands to gain.

Loss Aversion

Recent research (see Chapter 3 and the Appendix) has begun to investigate the neurobiological mechanism underlying loss aversion. In one fMRI neuroimaging study, subjects were presented with a series of mixed gambles that offered a 50/50 chance to either gain or lose a given amount of money (Tom et al., 2007). Potential gains (ranging between $10 to $40, in $2 increments) and potential losses (ranging between −$20 to −$5, in $1 increments) were presented independently and subjects were required to either accept or reject each proposed gamble.

Individual behavioral loss aversion λ was computed as the ratio of the (absolute) loss response to the gain response, which yielded a median λ=1.93 across all subjects, a degree of loss aversion consistent with many previous studies. The values of the potential gains and losses were entered into a regression analysis to identify brain areas showing a parametric response to increasing magnitude of either losses or gains. Activity encoding potential reward, increasing with the magnitude of potential gains, were found in regions including the striatum, OFC, and dopaminergic midbrain regions. Consistent with the computation of the net gamble value, potential losses were coded as decreased signal by the same network. Critically, while the net value signal was monotonic, it was asymmetric between the parametric estimates for gains and losses. This asymmetry in the neural estimates in striatum for gains and losses (called “neural loss aversion”) correlates with well-documented asymmetry in behavioral loss aversion. An important question is how this neural asymmetry in net gamble value emerges, a point not yet resolved (DeMartino et al., 2010).

Loss aversion and the endowment effect

Two key principles deriving from Prospect Theory, and used as evidence for reference-dependent preferences, are loss aversion and the endowment effect (Kahneman et al., 1991). Loss aversion reflects a person’s preference to prefer avoiding losses to acquiring gains. The endowment effect is a manifestation of loss aversion, wherein people place extra value on goods they own compared to identical goods they do not own. In other words, the value of a good increases once a person establishes his or her property right over it. In the original endowment effect experiment (Kahneman et al., 1990), students demanded a higher price for a mug that had been given to them but put a lower price on a mug they did not yet own – when the actual price of each mug was identical. The endowment effect has been described as an anomaly in neoclassical theory, which predicts that a person’s willingness to pay (WTP) for a good should be equivalent to their willingness to accept (WTA) payment to be deprived of the same good. In other words, valuation should not be affected by ownership. In reality, as the endowment effect demonstrates, references points (as predicted by the Prospect Theory value function) do influence valuations and decisions and can result in WTA being greater than WTP.

4 From Basics to States and Traits: Assessing Approach, Avoidance, and Goal Conflict

Our analysis of the basics of approach, avoidance, and goal conflict shows that care must be exercised when using complex combinations of motivational stimuli and complex paradigms. Variations in valuation, such as loss aversion, differential effects of approach and avoidance gradients, direct interactions between approach and avoidance systems, and the asymmetric impact of goal conflict on avoidance relative to approach, must all be taken into account when interpreting many of the paradigms currently used. However, in principle, state analysis of these systems is straightforward.

One simplifying step is to use money as the source of motivation. Organizations that find work for students and other casual workers can supply participants with a hunger for money sufficient to make them willing to work for the local minimum wage. Importantly, loss of money from an existing store can then be used as a motivator, with the knowledge that its external value is the same as the gain of the same amount of money used as a positive motivator. As shown in Figure 3, gain and loss can be presented or omitted to generate approach or avoidance. The amounts of gain and loss can then be varied parametrically to allow mathematical extraction, separately, of the contribution of gain/loss sensitivity differences and of approach/avoidance sensitivity differences. Using these methods, loss aversion and approach preference have been demonstrated.188

For neuroimaging, it is also important to use designs that allow the calculation of appropriate contrasts. If one wishes to image goal conflict activation, one must accept that gain, loss, approach, avoidance, and other systems will all necessarily be activated when approach-avoidance conflict is being generated. To deal with this requires the use of at least three conditions. For example, with conditions that deliver two alternatives with a 50% probability on any trial, one could have: (1) net gain (−10c, +20c); (2) conflict (−15c, +15c); and (3) net loss (−20c, +10c). A contrast of neuroimaging activation in condition 2 against the average of condition 1 and condition 3 would assess goal conflict-specific activation while eliminating the effects of external value (15c = (10c+20c)/2) and controlling for effects of factors such as risk. In practice, because of loss aversion, to statistically eliminate the effects of gain, loss, approach, and avoidance, when assessing conflict, one would need the ratio of gain/loss amounts tailored to each individual’s degree of loss aversion. Additional conditions would allow the separation of the effects of gain from the effects of loss and effects of approach from the effects of avoidance.188

For those interested in goal gradients (Figures 2 and 4), existing virtual reality maze paradigms (see Section 2.2) or even simpler runway analogues could be used. These have already demonstrated effects related to distance from a “predator,” as well as differences between simple anticipation of shock and the response to actual shock delivery. Combined with the presentation of money (to selected money-hungry participants), these virtual reality paradigms allow manipulation of the full gamut of parameters that have previously been used in animal behavior tests.

It is tempting, in the imaging of personality, to select questionnaires that have been designed, in theory, to tap into specific neurobiological functions (e.g., scales purporting to measure Gray’s Behavioral Inhibition System) but that have not in fact been neurobiologically validated. However, as we noted earlier, the nascent neuroscience of personality should not assume the very hypotheses that need to be tested. Psychologists’ presuppositions about which neural systems are responsible for any given trait, as measured by a questionnaire, may well be wrong. With approach, avoidance, and conflict, we are dealing with primordial biological systems whose elements have evolved to fulfill system-specific purposes. The state activation of these systems can be, and has been, assessed directly, with specific components extractable through appropriate contrasts. These specific components of neural state activation provide, we would argue, the best basis both for assessing personality-related variation in activation and for deriving questionnaire scales or other measures of approach, avoidance, and goal conflict traits, using the criterion approach described in Section 3. How the sensitivities of the approach, avoidance, behavioral inhibition, and other neural systems give rise to variation in traits is the key question that the field must strive to solve. A genuinely neuroscientific approach will provide a solid basis for future attempts to understand the contribution of these fundamental neural systems to traits such as extraversion, neuroticism, impulsivity, and others.

This article describes current trends and controversies in behavioral decision research. Much of the empirical research in this field has focused on violations of rational choice theory. Some violations are utility-based, such as loss aversion, framing effects, and contextual effects. Others are belief-based, such as base rate neglect and conjunction errors. This abundance of evidence has led behavioral decision researchers to develop descriptive theories based on more realistic assumptions about human nature. Issues such as fairness, self-control, emotional satisfaction, social pressure, and cultural norms are directly incorporated into the choice process. This approach to theorizing is gradually gaining popularity in other areas of social science, such as behavioral economics, behavioral finance, behavioral game theory, and behavioral accounts of the law.

8.2 More risk seeking in losses compared to gain domains

A common effect widely discussed in decisions from description implies that the subjective enjoyment from gaining a certain amount tends to be less than the subjective pain from losing the same amount (Kahneman and Tversky, 1979). Some researchers have demonstrated that loss aversion does not hold in decisions from experience, where decision makers seem indifferent between an equal chance of gaining or losing the same amount (Erev et al., 2008; Ert and Erev, 2011). In decisions from description, decision makers are risk averse in the gain domain and risk seeking in the loss domain (Kahneman and Tversky, 1979), and this pattern may reverse or disappear in decisions from experience (Erev and Barron, 2005).

Although much work needs to be done in regards to the differences between gains and losses in decisions from experience, our initial analyses of decisions from experience in the sampling paradigm of the TPT indicate no difference in risky behavior between gains and losses (χ2 = 0.308, p = 0.580). The IBL model, however, predicts a difference between gains and losses, which although small, it is significant (χ2 = 12.462, p < 0.001). These effects are illustrated in Fig. 9. Interestingly, human behavior as well as the IBL model prediction are in disagreement with the predictions from prospect theory: Humans do not show higher risk-seeking tendency in problems involving losses than gains and the IBL model, shows a higher tendency for risky choices in problems involving gains than losses. Both, human data and the IBL model data illustrate opposite effects than those expected in prospect theory.

Loss Aversion

The axiom that “losses loom larger than gains” is one of the best-known effects in behavioral economics. If people are offered a gamble on a coin toss, on which they would win $10 if a tail is thrown, but lose $10 if a head is thrown, most participants decline this gamble [16]. As the expected value of the coin toss is zero (it is risk neutral), we might expect participants to be indifferent regarding these two options, but they are not. The win amount can be gradually increased in a series of steps while keeping the loss constant. Participants typically require the win to be approximately twice as large as the loss (i.e., win $20, lose $10) before reliably accepting the gamble. This is parameterized by λ values of around 2, and these values are prone to substantial individual differences [17]. Loss aversion is explained within Prospect Theory by the relative steepness of the value function in the loss domain relative to the gain domain [18].

Prima facie, this phenomenon is obviously relevant to gambling behavior. On one hand, loss aversion may be a deterrent for many people to engage in gambling in the first place; the risk of losing one's bet overrides any potential gain. From this, we might expect problem gamblers to display reduced loss aversion relative to healthy participants; that is to say, their value functions should be similarly steep in the gain and loss domains. Such effects have been described in some neurological studies of patients with brain injury. In a coin-toss task similar to the aforementioned scenario, a mixed group of brain injury patients with damage to the ventromedial prefrontal cortex, amygdala, or insula were more likely to accept a positive expected value coin toss (win $2.50, lose $1) than healthy controls, with the healthy participants declining a substantial proportion of such bets owing to loss aversion [19]. These findings were later confirmed in two cases with selective amygdala damage as a result of Urbach–Wiethe disease [20].

However, reduced loss aversion is not the only prediction for real-world gamblers. Loss aversion is conceptually related to loss chasing, a key symptom of gambling disorder in which the gambler persists with betting in a desperate attempt to recoup mounting debts. Loss chasing is regarded by some as the defining feature of the pathological gambler [21], and empirically, it is the most frequently endorsed item on the DSM checklist [13]. In a compelling field demonstration in online gamblers who closed their accounts because of gambling-related problems, Xuan and Shaffer [22] observed that these gamblers (relative to a control group who did not close their accounts) responded to mounting losses by increasing their bet size, rather than placing more bets or taking bets at longer odds.

Could loss chasing arise from increased loss aversion? Such an explanation requires an additional feature of Prospect Theory: the reference point [18]. Most decisions are made from the origin of the value function, such that between any two choices, people “re-reference” so that the options are evaluated independently. In a gambling episode, players may fail to re-reference between successive bets, so that a player on a losing streak progressively shifts leftward on the value function, getting steadily farther from the origin. As the loss curve plateaus, the subjective value of a win becomes dramatic, whereas the subjective value of sustaining a further loss becomes negligible.

A 2014 experiment sought to test this hypothesis, quantifying loss aversion in patients with gambling disorder [23] using a straightforward behavioral economics procedure to titrate a λ coefficient for each participant. Notably, the gamblers did not differ in the overall group comparison of loss aversion values. Nevertheless, individual differences in loss aversion were related to the length of time in treatment, such that a subgroup tested later in the course of treatment was more loss averse. It would be worthwhile to explore further any bimodality in the loss aversion scores, whereby gamblers may polarize toward extreme high or low values, with the former group hypothetically disposed to loss chasing. It is also worth noting that loss chasing could be driven by several distinct characteristics of the value function, such as the relative steepness of the gain and loss domains, the distance to plateau, or the tendency to re-reference between gambles.

Characterizing loss aversion in gambling disorder may also benefit from a greater emphasis on the underlying neuroscience. Psychophysiological measures like skin conductance provide objective markers of the relative impact of losses and gains: the skin conductance response to a losing outcome is greater than that to an equivalently sized gain [17]. As participants select bets during a gambling task, skin conductance levels scale closely with bet size and are further correlated with the sensitivity to losing outcomes [24,25] (see Fig. 27.1). Skin conductance levels also change following near losses on a “wheel of fortune” task, and these near losses similarly “loom larger” than equivalent near wins [24]. Using fMRI, a seminal experiment in neuroeconomics by Tom et al. [26] showed widespread neural sensitivity to losses over gains in brain regions that included the striatum and medial prefrontal cortex. Converging with neuropsychological studies, the amygdala response to losing outcomes also scales with behavioral estimates of loss aversion [27].

Behavioral Decision Theory

Behavioral Decision Theory refers to the research paradigm that explores consumers' decision tendencies. For example, research on the framing effect shows that the manner in which choice alternatives are framed or described has a profound effect on choice, even though consumers typically believe that their preferences are stable. This occurs because consumers think differently about outcomes described in terms of gains (above a reference point) versus losses (below a reference point). Loss aversion, or the tendency to weigh losses more heavily than equivalent gains, can explain the framing effect, the compromise effect, the endowment effect, the status quo bias, and the sunk cost fallacy. For example, when outcomes are framed in terms of gains, consumers are risk-averse. When outcomes are framed in terms of losses, consumers are risk seeking. Consequently, when asked to choose between a program that would save 200 out of 600 lives for sure and a program that would save 600 with p = 0.33 or 0 with p = 0.67, most people prefer the former safe option. When asked to choose between a program in which 400 out of 600 would die for sure and a program in which 0 would die with p = 0.33 or 600 would die with p = 0.67, most people prefer the latter risky option. Of course, the two safe options are identical and the two risky options are identical too.

Consumers often prefer a compromise brand that is average on two equally important attributes over brands that are excellent on one attribute but poor on the other. Poor performance on an attribute is often interpreted as a big loss and loss aversion leads consumers to avoid extreme brands.

Research on the endowment effect shows that owning an object leads consumers to value that object more. This occurs because buying a new object is interpreted as a gain and selling an object one currently owns is interpreted as a loss. Loss aversion makes consumers reluctant to sell objects they currently own. Consequently, selling prices are often much higher than buying prices.

People often prefer the status quo to change (the status quo bias). This occurs because change often involves gains on some dimensions and losses on others, and loss aversion can make people reluctant to accept change. Consequently, when consumers are told that a pharmaceutical product is currently on the market, they recommend keeping it on the market. When they are told that the same product is currently not on the market, they recommend keeping it off the market. People also prefer to keep their current jobs, automobile color, financial investments, and medical insurance policies.

As the amount of time, money, or effort invested in a project increases, the reluctance to abandon the project increases even if people would be better off doing so (the sunk cost fallacy). After spending millions of dollars on a dam, building, or construction project, people are reluctant to abandon the project even when it would be less expensive to do so. After spending millions of dollars on old mixing technologies, many steel companies went out of business because they refused to invest in new superior technologies. Loss aversion makes people reluctant to give up what they have already invested, even when the investment is unlikely to ever pay off.