The attention–aversion gap: how allocation of attention relates to loss aversion
Introduction
Loss aversion—the tendency to weight losses more heavily than equivalent gains—has been described as “one of the basic phenomena of choice under both risk and uncertainty” (Tversky & Kahneman, 1992, p. 298, emphasis added].1 The idea that “losses loom larger than gains” (Kahneman & Tversky, 1979) when people make decisions—that the pain of losing is much more significant than the pleasure of gaining—has been central to the understanding of how people make decisions under risk (and beyond; Tversky & Kahneman, 1991). It has had a profound impact in psychology, economics, and finance, and has also influenced fields such as political science and law. Loss aversion has been invoked to explain a variety of empirical phenomena, including the endowment effect (Kahneman, Knetsch, & Thaler, 1990; Thaler, 1980), the disposition effect (Weber & Camerer, 1998), effort in sports (Riedl, Heuer, & Strauss, 2015), and underinvestment in the stock market (Benartzi & Thaler, 1995).
Loss aversion was initially demonstrated in decisions under risk, where people select between options whose possible outcomes and corresponding probabilities are known. For example, someone whose preferences are described by the original parameters of prospect theory (Tversky & Kahneman, 1992) would reject a gamble that offers a 50% chance to win $1000 and a 50% chance to lose $1000, because the potential loss of $1000 outweighs the possible gain of the same size. Although both options—the gamble and the safe option of staying with $0—have the same expected value, this overweighting of losses renders the gamble relatively unattractive.2 To be attractive, a gamble with a 50% chance of losing $1000 would need to offer a possible gain (again with a 50% chance) of roughly $2750 (Tversky & Kahneman, 1992). Loss aversion has been viewed as “one of the most fundamental and well-documented biases” (Rozin & Royzman, 2001, p. 306).
Although the concept of loss aversion appears to be well established, there is also evidence challenging its generalized nature. Ert and Erev (2013) were among the first to note the empirical inconsistency of findings on loss aversion in decisions under risk. Specifically, they identified a set of situations in which loss aversion is unlikely to emerge, including situations with feedback and situations with a clear and safe status quo.3 Moreover, Walasek and Stewart (2015) showed that loss aversion can be made to appear, disappear, and even reverse by manipulating the ranges of possible outcomes that people could expect to obtain. As predicted by decisions-by-sampling theory (Walasek & Stewart, 2015), they showed that loss aversion emerged in a context where gains ranged between $0 and $40 and losses between $0 and $20, but disappeared when gains and losses both ranged between $0 and $20. According to decisions-by-sampling theory, it is the asymmetry in the magnitudes of the possible gains and losses (specifically, that the former are usually larger than the latter) that leads to the occurrence of loss aversion.
Systematic reviews and comprehensive analyses have confirmed the empirical fragility of loss aversion. Studies using large and diverse sets of choice problems have found only small, though reliable, degrees of loss aversion (Glöckner & Pachur, 2012; Kellen, Pachur, & Hertwig, 2016; Pachur, Mata, & Hertwig, 2017). Yechiam and Hochman (2013b) reviewed articles that examined loss aversion using symmetric gambles (i.e., gambles offering an equal chance of winning or losing an equal number of points or amount of money), and found that only four out of 24 studies reported evidence for loss-averse choices when averaged across participants. The four studies in which there was evidence for loss aversion used decisions from description, as did the original studies of Tversky and Kahneman (1992). In decisions from description (also known as decisions under risk), decision makers are informed about all possible outcomes of the options and their corresponding probabilities. In decisions from experience, in contrast, they learn about the outcomes and their probabilities from previous decisions or by sampling options (Hertwig, Barron, Weber, & Erev, 2004; Hertwig & Erev, 2009). In the review conducted by Yechiam and Hochman (2013b), none of the 13 studies examining decisions from experience found evidence for loss aversion. As evidence against loss aversion accumulated, Yechiam (2018) went back to the early studies of loss aversion (Fishburn & Kochenberger, 1979; Galanter & Pliner, 1974) that Kahneman and Tversky (1979) had drawn on to substantiate the assumption of loss aversion as a stylized fact in their formulation of prospect theory. Yechiam concluded that this early evidence had been “over-interpreted” (p. 1) by Kahneman and Tversky.
A recent review by Gal and Rucker (2018b) went even further. These authors questioned the evidence for loss aversion on conceptual grounds, attributing the idea that it is a ubiquitous regularity in human choice to its intuitive appeal among behavioral scientists. This led to the existence of loss aversion being overgeneralized—often inappropriately—to everyday situations. Gal and colleagues did not mince their words in concluding that loss aversion, “the most important idea in behavioral decision-making,” is “a fallacy” (Gal, 2018) and suggested “moving beyond loss aversion as a generalized principle” (Gal & Rucker, 2018b, p. 513). Irrespective of whether loss aversion is indeed the most important idea in behavioral decision-making research, the key unresolved issue is this: Why is it that the empirical phenomenon of loss aversion is so fragile and even elusive? In search of an answer, let us first turn to the conceptual question of what can reasonably be considered as a manifestation of loss aversion.
Section snippets
What is loss aversion—and what is it not?
Although Edgeworth observed that “minus pain is sweeter than plus pleasure” as early as 1877 (Edgeworth, 1877, p. 75), Kahneman and Tversky (1979) were the first to test this notion empirically and to give it a name: loss aversion. Their empirical test required people to make choices between gambles. Accordingly, prospect theory's concept of loss aversion has long been closely aligned with choice behavior. In fact, Tversky and Kahneman (1992) described loss aversion as one of the “basic
Current evidence on the relationship of loss attention and loss aversion
Two recent empirical studies have examined the relationship between attention and aversion to losses. Pachur et al. (2018) tested whether attention patterns in decisions between gambles reflect the psychological constructs assumed in prospect theory, such as nonlinear probability weighting, risk aversion, and loss aversion. The authors had participants choose between gambles, and then estimated each participant's prospect theory parameters from those choices. One of these parameters is λ, the
Three unresolved issues
First, as outlined above, loss aversion has sometimes been conceptualized as a multidimensional construct, which can manifest across various neurophysiological and cognitive processes (including hormones and neural activation). Implicit in this view is the assumption that these various manifestations of loss aversion operate in tandem, such that, for example, neural signatures and hormonal markers necessarily correlate with the degree of loss aversion in choice (because they reflect a common
Choices between gambles: the limits of a convenient paradigm
Experimental tasks involving choices between gambles are archetypal in the study of risky choice, and are the home turf of prospect theory (Kahneman & Tversky, 1979). For example, a participant is given a choice between two options: a sure gain of EUR 3 and a gamble offering an 80% chance of winning EUR 4 and a 20% chance of winning nothing. Participants make several decisions of this type, with varying outcomes, probabilities, and the possibility of winning or losing money. A common way to
Are people more attentive to losses than to gains?
We analyzed data from two experiments studying decisions under risk between gambles in a description-based setup. Both experiments used MouselabWEB (Willemsen & Johnson, 2019)—a browser-based version of Mouselab—in a repeated-measures design. The only difference between the two experiments was that different acquisition modes were used: mouseover and click. In mouseover mode, the person moves the mouse pointer over the box on the screen and the box opens automatically to reveal the information
Do losses prompt more attention than gains?
We addressed this question by examining the amount of time that participants kept the individual boxes open while making decisions in problems in the loss domain versus the gain domain. We first computed the average opening time for each participant across problems, and then computed the average across participants for each domain. Fig. 2 shows the mean opening times in milliseconds across participants.6
Discussion
The evidence for loss aversion has recently been questioned. Using the standard experimental paradigm for the study of risky choice, we observed a robust pattern of increased attention to losses relative to gains, even though the large majority of participants showed no loss aversion in their choices.
The consistency of results across various neurophysiological and cognitive measures suggests that organisms devote more attention to losses than to gains of equivalent magnitude. Loss attention
Acknowledgements
We thank David Pietraszewski and Jan K. Woike for helpful discussions and Susannah Goss for editing the manuscript.
References (68)
- et al.
Cognitive models of risky choice: Parameter stability and predictive accuracy of prospect theory
Cognition
(2012) - et al.
The description–experience gap in risky choice
Trends in Cognitive Sciences
(2009) - et al.
Clarifying the relationship between prospect theory and risk-sensitive foraging theory
Evolution and Human Behavior
(2014) - et al.
The evolution of error: Error management, cognitive constraints, and adaptive decision-making biases
Trends in Ecology & Evolution
(2013) - et al.
How (in)variant are subjective representations of described and experienced risk and rewards?
Cognition
(2016) - et al.
How choice ecology influences search in decisions from experience
Cognition
(2012) - et al.
Bad riddance or good rubbish? Ownership and not loss aversion causes the endowment effect
Journal of Experimental Social Psychology
(2009) - et al.
Hierarchical Bayesian parameter estimation for cumulative prospect theory
Journal of Mathematical Psychology
(2011) The logic of risk-sensitive foraging preferences
Animal Behaviour
(1981)Toward a positive theory of consumer choice
Journal of Economic Behavior & Organization
(1980)