Loss aversion does disappear and reverse, although estimates of lambda (λ) are not reliable: Reply to André and De Langhe

Abstract

Walasek and Stewart (2015) demonstrated that loss aversion estimated from fitting accept-reject choice data from a set of 50/50 gambles can be made to disappear or even reverse by manipulating the range of gains and losses experienced in different conditions. André and de Langhe (2020) critique this conclusion because in estimating loss aversion on different choice sets, Walasek and Stewart (2015) have violated measurement invariance. They show, and we agree, that when loss aversion is estimated on the choices common to all conditions there is no difference in prospect theory’s λ parameter. But there are two problems here. The first is that while there are no differences in λs across conditions, there are very large differences in the proportion of the common gambles that are accepted, which André and de Langhe chose not to report, reporting instead only the null λ result. These choice proportion differences are consistent with decision by sampling (but are not consistent with any of the alternative mechanisms proposed by André and de Langhe, 2020) and show loss aversion disappearing and reversing. Second, we demonstrate a much more general issue: that λ estimated from the accept-reject choices for 50/50 gambles is extremely unreliable and does not generalise even across random splits within large, balanced choice sets. This point should be of profound interest to anyone estimating model parameters from choice data – it is futile to estimate λ from accept-reject choices.

It’s good and it’s bad: Sampling processes can generate decision-framing effects.

“…However, recent work suggests that sampling processes may also play a role such that people sample gains and losses from memory and compare the relative rank of the present decision (Stewart, Chater, & Brown, 2006; Walasek & Stewart, 2015). However, the existing evidence providing support for this mechanism is debated (André & de Langhe, 2021; Walasek et al, 2020). The current work offers novel and unique evidence that sampling processes can underlie framing effects.…”

“…Indeed, the validity of the experimental evidence testing a sampling mechanism for loss aversion (Walasek & Stewart, 2015) has been called into question as the reported results may stem from measurement invariance created from the way in which loss aversion is computed differently across conditions, rather than being due to the proposed sampling mechanism (André & de Langhe, 2021; cf. Walasek et al, 2020). In the current experiments, we circumvent this issue by exploring situations in which the information is objectively identical across experimental conditions, but merely framed in terms of gains or losses.…”