Can advisors eliminate the outcome bias in judgements and outcome-based emotions?

Abstract

Purpose An outcome bias occurs when performance is evaluated based upon the outcome of the decision rather than upon the quality of the decision itself. The purpose of this paper is to test experimentally whether advisors eliminating the uncertainty in the quality of decisions as a potential driver of the outcome bias can eliminate this bias in judgements. Additionally, the paper analyses whether such advisors can attenuate the emotional experience after decisions’ outcomes by supporting the cognitive understanding of these outcomes. Design/methodology/approach The paper employs a between-subject experimental setting where decision makers are asked to make investment decisions. The two variables manipulated were advice (receiving advice vs deciding without an advice) and state (loss vs neutral vs gain state). Participants were randomly assigned to each group. One group completed all tasks while receiving advice before making a final decision. Another group completed all tasks without any advice. After completing each investment task, participants were randomly assigned to one of three possible states that determine their payoff. Findings The results reveal that advisors eliminating the uncertainty in the quality of decision can eliminate the outcome bias in the judgements of decision quality, especially after bad outcomes. Nevertheless, after controlling for the perceived quality of the decisions, advised individuals show a greater emotional sensitivity to bad outcomes than non-advised decision makers. These observations suggest that advisors eliminating the uncertainty in the decision quality can improve the understanding that good decisions can lead to bad outcomes just by chance, but they are not able to prevent affective reactions after bad outcomes; on the contrary, they may even reinforce them. Research limitations/implications The observation that, after bad outcomes, advised decision makers are less willing to decide the same way than non-advised decision makers is consistent with empirical findings on the risk-taking behaviour of self-directed and mutual fund investors. Self-directed investors are less likely to revise their decision and sell an investment at a loss than mutual fund investors (Changet al., 2016). While investors might sell funds because they are unable to observe the decision process of the manager and use the outcome to judge the manager’s skills, this study shows that such learning from decision outcomes is not necessary for observing the risk-taking behaviour of the investors. Even if the decision process of the advisor is observable (as in this study), the decision makers’ willingness to decide the same way is influenced by the losses – an effect that goes beyond the assessed quality of advice as the results of this study show. Practical implications The results of this study have important implications for advisors aiming to maintain a positive relationship with their clients. Convincing clients that an advice is optimal supports their understanding that a good advice can have bad outcomes. However, this understanding may not prevent affective reactions after bad outcomes. On the contrary, the affective response after bad outcomes is even stronger with the advice than without it. Hence, advisors should address not only issues related to the quality of the provided advice, but also emotional aspects, which could be related to what clients expect from following the advice. Originality/value This study is one of the few that account for the possibility that the outcome bias may arise because there is uncertainty regarding the optimal choice. In particular, this paper uses a much more powerful criterion to define an optimal choice than the expected value criterion used in previous studies. The criterion represents a minimal requirement for rational behaviour in expected utility theory and many non-expected utility theories.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 5 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.