Schedule - Parallel Session 2 - Limits to Loss AversionIMC Room 246 - 14:00 - 15:30
Financial Loss Aversion Illusion
Loss aversion has been frequently documented in psychology and economics, with the conclusion that losses loom larger than gains by a magnitude of about two. In finance, loss aversion is suggested to explain, for instance, the equity premium puzzle and stock market participation. In the evaluation of gains and losses, one has to distinguish between anticipated and experienced outcomes. Most experiments on gambles or lotteries focus on the trade-off between anticipated gains and losses. However, this implies that people are able to forecast the hedonic impact of gains and losses. In contrast, recent evidence suggests that people’s ability to cope with losses is much better than they predict (Kermer et al., 2006). Using a unique dataset, we test this proposition in the financial domain. In a panel survey with UK investors, participants state their subjective ratings of anticipated and experienced returns. In a time period of frequent losses and gains in investors’ portfolios, we examine how the subjective ratings behave relative to expected and experienced portfolio returns. To this end, we define several potential reference points investors might use. Loss aversion is strong for anticipated outcomes. From regressions of subjective ratings on expected returns, we infer a loss aversion coefficient of about 2.2 for a reference point of zero. This means that investors react more than twice as sensitive to negative expected returns as to positive expected returns. However, when evaluating experienced returns, the loss aversion coefficient decreases to about 1.2 and is statistically indistinguishable from one (loss neutrality). Investors do not react more sensitive to losses than to gains when confronted with realized portfolio performance. The loss aversion they show ex ante seems to be a projection bias. As a second property of reference-based utility, we test for diminishing sensitivity with respect to outcomes more distant from the reference point. We indeed find that investors’ reaction is strongest for returns close to the reference point. An improvement from 2% to 4% in portfolio return has a greater impact on subjective ratings than moving from 12% to 14%. This is true for expected as well as for experienced outcomes. But while for expected returns the sensitivities in each interval are far greater for losses than for gains, this is not the case for experiences. Our findings have implications for individual investing. While loss aversion can be a legitimate part of people’s preferences, the financial loss aversion illusion we document clearly is not. If investors systematically overestimate their personal loss aversion when thinking about financial outcomes, their investment decisions will differ from what is justified by their experience of these outcomes. In particular, they will invest less riskily than they probably should and will avoid potential losses unless they receive a substantial compensation.
Certainty Preferences, Random Choice, and Loss Aversion
I revisit recent evidence uncovering a preference for certainty in violation of dominant normative and descriptive theories of decision making under risk. I explore two alternative explanations of the preference patterns found: i) systematic noise; and (ii) reference dependence activated by salient outcomes. I develop choice lists that allow to disentangle these different explanations, and test them on rural subjects in southern India. The results reject explanations based on a preference for certainty in favor of explanations based on random choice. The estimates are further distorted by response mode effects, with loss aversion leading to an over-estimation of risk aversion.
Canonical Riskless Choice Over Bundles: Ain't No Reflection Effects Here
Hui-Kuan Chung; Paul Glimcher; Agnieszka Tymula
Since Kahneman and Tversky proposed prospect theory, numerous studies have found that the evaluation of changes in wealth by individuals is best modeled (in a positive sense) as gains and losses with respect to a reference point rather than as being computed over total wealth. Prospect Theory presumes that people have diminishing sensitivity from the reference point and this feature of the utility function is consistent with the commonly discussed reflection effect: a switch from risk aversion in gains to risk seeking in losses. Paradoxically, although it is widely assumed that such gain-loss asymmetry also describes riskless choices, diminishing sensitivity (with most reasonable parameterizations) would suggest normally convex indifference curves over gains, but concave indifference curves over losses. This predicts that people prefer mixed bundles in the gain domain, but retaining all of one type of good to any mixed bundles in the loss domain. There is little empirical evidence supporting the existence of asymmetries in indifference curves around the reference point for choices made under conditions of certainty. We, therefore, used two incentive-compatible (randomly interleaved) riskless and risky choice tasks conducted with the same consumer goods and the same experimental participants in the gain and loss domains. Our novel design allows us to estimate utility curvature for gains and for losses in risky choices and indifference curves for gains and losses in riskless choice with minimal assumptions. More parametrically, we formalize the relationship between the curvature of the utility function and indifference curves under three currently dominant theories of choice: expected utility, prospect theory, and Koszegi-Rabin reference point. For each theory, we derive our predictions assuming that the same utility function describes choice in risky and riskless conditions as in these theories. Surprisingly, while we find reflection effects in risky choice as expected “we see no evidence of such a phenomena in randomly interleaved, nearly identical, riskless choices. These findings are not consistent with any of the three standard models “expected utility, prospect theory, and Koszegi-Rabin preferences. In the domain of losses, in riskless choice over bundles, our participants have a consistently concave utility function, rather than convex ones, as Thaler and others have interpreted prospect theory to predict. This happens even though the same participants in the same experimental session show the convex utility in losses in a risky choice task. These results imply that one of the key tenants of reference-dependent theories, diminishing sensitivity relative to a reference point, might not be present under some conditions of riskless choice. Parametrically, we do not find any relationship between parameterized utility functions elicited in risky and riskless conditions, even when we restrict ourselves to gain or loss domain only.
Meta Analysis of Loss Aversion in Risky Context
Lukasz Walasek; Neil Stewart; Tim L. Mullett
Recent research has shown that the magnitude of loss aversion is highly dependent upon the environment in which choices and judgements are made (Walasek & Stewart, 2014). People exhibit large loss aversion when the typical values on offer are skewed such that the average gains are twice as large as the average losses. However, when there is no such skew people are loss neutral and when the skew is reversed people actually exhibit loss seeking behaviour. This has important implications for existing findings. We are presenting a meta-analysis of loss aversion in risky context. We collected data on choices and valuations of mixed gambles in an effort to estimate the loss aversion parameter using the same functional form of the Prospect Theory. Our findings indicate that the majority of papers examining loss aversion used values with a significant skew in the distribution of gains and losses. In addition, there is a significant correlation between the size of this skew and the degree of loss aversion reported. By taking this design characteristic into account, we were able to determine what is the average level of loss aversion reported in the literature. A critical finding is that the magnitude of loss aversion is partially driven by the distributions of gains and losses used in the elicitation task. Consistently with our earlier experimental work, studies that used asymmetric distributions of gains and losses (wide range of losses and a narrow range of gains) found a higher level of loss aversion. Our work offers an important review of the loss aversion literature. Numerous researchers simply assume that losses loom twice as large as gains, and use loss aversion to explain a range of behavioural phenomena. We point out to the limited evidence for this assertion. We find that the asymmetric weighting of gains and losses in risky context may be much weaker and that loss aversion can be partially a product of the experimental design.