Effect of Comparison Standard on One's Optimism Bias

In our daily life, we constantly evaluate the likelihood of possible future events in order to take decisions under uncertainty. However, we do not process the obtainable evidence in an objective and strictly realistic manner. Indeed, our reasoning is inherently subjective and contains systematic biases. The optimism bias constitutes one of the most widespread and consistent cognitive biases. People believe to be less at risk of experiencing negative events, while on the other hand overestimating their likelihood to experience positive events (Shepperd, Waters, Weinstein, & Klein, 2015). About 80 % of the population regardless of age, sex, education and occupation are unrealistically optimistic/display optimistic errors (Sharot et al., 2011). Furthermore, the bias is widespread in an extensive range of fields including workspace, finances, well-being, personal relationships or self-evaluation (Dunning, Meyerowitz, & Holzber, 1989).

In the western world, two out of 5 marriages will end up in a divorce (Sharot, 2011). Nevertheless, those about to marry consistently underestimate the longevity of their own marriages (Baker & Emery, 1993). One could argue that they are too naive, having no experience and knowledge about marriages. However, even experts display concerning degrees of optimism bias. Family lawyers underestimate the negative consequences of a divorce even though they are confronted with clients’ divorces on a daily basis (Sharot, Korn, & Dolan, 2011). Moreover, medical practitioners overestimate the success of their therapy and financial specialists predict unusually high profits (Calderon, 1993) to the point that economists have proposed the optimism bias as a root cause for the financial meltdown/crisis of 2008 (Shefrin, 2009).

It has been proposed that cognitive and motivational factors underlie these unrealistic perceptions. From the cognitive perspective, optimistic beliefs may be a natural outgrowth of the process of setting goals. By thinking about ways of achieving a goal, the belief that one will accomplish it is strengthened. Complementary, imagining approaches how to avoid negative events improves the perception that one will be capable of doing so. Additionally, we are highly motivated to be optimistic. By believing that our probability of success is higher than that of failure, a positive sense of oneself, including one’s skills, capabilities, resources and outcomes can be preserved. Moreover, gender seems to influence optimistic perceptions. Lin & Raghubir (2005) show that men display greater optimism bias than women regarding expectations of their marriage. However, the underlying mechanisms are thus far not established.

An important question stemming from these observations is whether costs of irrational and overoptimistic beliefs outweigh potential benefits. On the one hand, constantly expecting positive outcomes can encourage goal persistence, motivation, positive affect and hope (Armor & Taylor, 1998). Furthermore, Taylor and Brown (1988) propose that healthy individuals view their future overly optimistic and that being accurate about one’s personal risk may have negative consequences on mental health. Specifically, it is suggested that depression is associated with to an absence of optimistic beliefs about future life events (Korn, Sharot, Walter, Heekeren & Dolan, 2014). Hence, it is claimed that being unrealistically optimistic about one’s future is natural and beneficial for mental health. At the same time, increasing evidence suggest that underestimating one’s risk is problematic as it can prevent people from taking adequate precautions. Indeed, looking at most health behavior models, a necessary requirement to take precautionary action is to perceive oneself as personally vulnerable to a negative event beforehand. Consistent with this line of reasoning, Kim and Niederdeppe (2013) have found that people who were unrealistically optimistic about evading the H1N1 virus stated lower intentions to wash their hands and use hand sanitizers.

Although the existence of unrealistic optimism has been extensively studied, the bias appears to be remarkably resistant to a variety of manipulations designed to reduce it. Learning theories suggest that people usually adjust their beliefs when confronted with contradicting information (Pearce & Hall, 1980, Sutton & Burton, 1998). Nevertheless, Sharot, Korn & Dolan (2011) have found that when asked to estimate the likelihood of experiencing an aversive life event after receiving the average probability of experiencing those events, participants only changed their estimates significantly when the average frequency was better than their own estimate, but did not so, when it was worse. This belief update method introduced a new way of assessing the optimism bias and suggests that providing people with the actual risk estimates is surprisingly unsuccessful at adjusting individual’s perception of their own vulnerability.

Considering these findings, perceived vulnerability may need to be defined differently by decomposing it into both absolute and comparative terms. The former is linked to the absolute optimism bias, which refers to people’s risk estimates as overly optimistic as pointed out by to a quantitative, objective standard. For example, someone may belief that his risk of getting cancer, is below 10%, while in fact his family background and unhealthy lifestyle cause his risk to be over 50%. The latter is linked to the comparative optimism bias, which states that people evaluate their personal outcomes as more favorable than the outcomes of another specific reference group.

In several experiments, it has been demonstrated that comparative and absolute risk perception influence behavioral intentions. Klein (1997) even postulates that comparative risk perceptions have a stronger effect than the comparison to an objective standard. Especially for self-evaluations, objective criteria are often redundant, demanding the use of social comparison information. It is virtually impossible to evaluate one’s attractiveness, intelligence or achievements without looking at how other perform on these aspects. Accordingly, in many situations, people apply social comparison information as a gauge for judgment, which in most cases, gives rise to a favorable self-assessment (Aspinwall & Taylor, 1993). Indeed, breast cancer patients compare themselves with others in still worse conditions in order to effectively cope with their situation (Wood, Taylor & Lichtman, 1985).

The underlying mechanism which operates to purposely elect inferior target has been appointed to downward comparison. In a study by Perloff & Fetzer (1986) participants rated their own risk of experiencing several negative events and provided ratings for the average student, their closest friend, or “one of your friends”. Participants exhibited less optimism bias when comparing themselves with a best friend than with “one of your friends”, or the average person. These findings indicate that vague targets encourage downward comparisons, while specific targets make this kind of comparison more difficult (Perloff & Fetzer, 1986).

Several researchers have offered motivational and cognitive mechanisms as an explanatory basis for this phenomenon. Weinstein (1980) explains/claims that a basic cognitive short-cut can be attributed to the main cause of comparative optimism. The representativeness heurististic describes the tendency to base one’s risk estimation merely with regard to how closely an event corresponds to a person’s prototype (Tversky, 1977). Thus, a vague and general comparison target may facilitate the selection of a general prototype of the risk category. This leads to lower risk judgements to the extent that people feel they are dissimilar to the prototypical target. Others argue that similarity to the comparison target is associated with less comparative optimism bias. It was found that optimism bias for smoking related health problems was greatest among those smokers who considered themselves least similar to the typical smoker (McCoy, Gibbons, Reis, Gerrard & Sufka, 1992). Furthermore, in a number of studies, Alicke, Klotz, Breitenbecher, Yurak & Vredenburg (1995) enhanced individuation of the comparison target by having participants look at a videotape or photo of the person with whom they were comparing themselves. Participants showed less optimism bias than a control group that compared itself with the average person. According to the person-positivity bias, vague and generalized targets are seen as less human and thus less favorable (Sears, 1983).

An interesting question stemming from these observations is whether the type of comparison standard differentially affects optimistic belief updating. Therefore, the current study was set up. Although optimism bias is characterized both as the overestimation of positive future events and the underestimation of future negative events, we concentrated on the latter, as this is strongly associated to a concern that people do not take precautionary action to protect themselves against hazards.

In order to provide additional confirmatory data on the phenomenon of comparative optimism bias we examine how the comparison standard affects participant’s optimism bias. Half of the participants were asked to estimate their likelihood of experiencing various adverse life events by comparing themselves to a fellow peer (vague target). The other half was instructed to compare themselves to their closest friend (specific target). It is hypothesized that participants who compare themselves to a specific target will show less optimism bias than participants who compare themselves to a vague target. Thereafter, using a belief update task (Sharot et al., 2011), participants were asked to reevaluate their likelihood compared to either a fellow peer or their closest friend, while being presented to the actual population risk estimates. Of critical importance and contrary to the disappointing results of using the belief update task alone (Sharot et al., 2011), we expect that participants who compare themselves to a specific target will show less optimism bias at the second evaluation and update their beliefs more in the direction of the population risk estimates than participants who compare themselves to a vague target.

Replicating past studies and combining differential levels of comparison target with the belief update method could shed light on new methods to reduce the optimism bias. The optimism bias has large societal implications since it has strong effect on decision-making. Therefore, knowing about ways to diminish the bias.

Next to replicating previously reported findings that the comparison standard affects participant’s optimism bias, the current study was primarily set up to examine whether this type of comparison standard differentially affects optimistic belief updating. Previous findings showed that a specific comparison target, compared to a vague target, is associated with less optimism bias (Perloff & Fetzer, 1986). However, the current study was not able to replicate this.

Nonetheless, beyond the goal of replicating previous findings, it was hypothesized that manipulating the level of the comparison standard differentially affects optimistic belief updating. Sharot et al. (2011) demonstrated that using the belief update task alone was not successful in altering participant’s optimism bias. Therefore, solely providing people with actual risk estimates is remarkably unsuccessful at altering participant’s optimism bias. We, therefore, hypothesized that combining the belief update task with a specific comparison target would lead to a reduction in the optimism bias score and to an increase in belief updates in the direction of the population risk estimate. This was tested by employing a belief update task in which participants were presented with various negative life events and asked to evaluate their individual likelihood of experiencing those events, before they were presented with the same events again along with the average probability of that event occurring in the population. Upon reading the base rate, participants were asked to re-evaluate their likelihood estimation compared to a fellow peer or their closest friend (depending on condition). It was expected that participants who compared themselves to their closest friend would show less optimism bias at the second evaluation and will update their estimation more in the direction of the provided.

The results show that in both conditions the update scores were positive, indicating that, on average, participants updated their beliefs in the direction of the population base rates. However, the level of comparison standard had no significant effect on optimistic belief updating. The data suggest effects in expected direction, with people who compared themselves to their closest friend (specific target) having higher positive update scores than people who compared themselves to a vague target. One could infer from these results that combining specific comparison target with the presentation of actual population base rates could result in estimations which are updated more in the direction of the population base rate.

However, considering the optimism bias scores, the results do not appear to support our hypothesis. Contrary to our expectations, there is no significant effect of comparison standard on optimism bias score at baseline, nor at optimism bias score at the second evaluation. Moreover, the second optimism bias score, in which participants were provided with the actual population base rates, is significantly larger than the optimism bias score without information. This result indicates that participants surprisingly displayed more optimism bias when faced with information about the population base rate. Furthermore, manipulation of the comparison standard seems to be no effective mean by which one can modulate participant’s optimism bias.

At first sight, these findings seem to be at odds. How can one have more optimism bias while at the same time having high update scores that indicate an updating of individual estimations towards the provided population information? The reason for this apparent contradiction constitutes a severe methodological limitation of this study. Previous studies (e.g. Sharot et al., 2011) discriminated whether participants received desirable information, e.g. were presented with a population base rate of 20% while their initial estimate was 50%, or undesirable information, e.g. the presented population base rate was 20% while their initial estimate was 10%. The present study was not able to differentiate between these two situations which results in interpretational limitations. Thus, a positive update score could mean either that the participant received desirable information and consequently adjusted his/her estimate toward the population base rate, hence the participant became more “optimistic”. At the same time, it could mean that the participant received undesirable information and consequently adjusted his/her estimate toward the population base rate, hence the participant became more “pessimistic”. Sharot et al. (2011) demonstrated that participants only changed their estimates when the average frequency was better than their own estimate, thus received desirable information, but did not so, when it was worse. Accordingly, our finding of positive update scores could be primarily due to the fact that participants mainly received desirable information. However, this remains speculative.

The present study is also not in line with the results of the study conducted by Perloff & Fetzer (1986). In this study, participants rated their own risk of experiencing several negative events and provided ratings for the average student, their closest friend, or “one of your friends”. The finding that participants exhibited less optimism when comparing themselves with a best friend (in our case closest friend) than with “one of your friends”, or the average person (in our case fellow peer) could not be replicated.

Important to note, is that in the current study no manipulation check of comparison standard was conducted. Thus, it is unclear if the manipulation was strong enough to elicit an observable effect on optimism bias scores. In the light of previous studies, the formulation of “Thinking of a fellow peer/your closest friend, how likely do you think that this event will happen to you?” probably elicited weaker effects than the manipulation used in Perloff & Fetzer who separately asked participants to estimate their own risk and thereafter estimate the likelihood of the comparison standard.

Furthermore, it remains ambiguous whether participants stayed with the same closest friend as a comparison standard throughout the entire experiment. It is possible that participants selected a different friend for each event, one that is highly at risk for each event. In particular, it is proposed that any negative event induces an image of a prototypical individual who displays certain features to be especially high risk of experiencing the event. Consecutively, the prototype acts as a primer to browse through one’s collection of friends and select the one that is similar to the prototype. Participants then estimate their risk in comparison to this friend. Future studies should control for this, e.g. by emphasizing it in the instruction and by asking participants to e.g. write down the initials of their closest friend they had in mind. To enhance the strength of the manipulation, individuation and specificity of the comparison target should be increased. Future research could work with photos and videos as has been successfully done by Alicke et al. (1995). These difference in study design as well as the mentioned limitation might be the reason why the findings of Perlicke and Fetzer (1986) could not be replicated.

Moreover, one could argue that prior knowledge about the optimism bias could affect participant’s optimism bias score at baseline as well as their updating behavior. People who are familiar with the optimism bias might be influenced by this knowledge and hence show less optimism bias and update their beliefs more in the direction of the population base rates. However, the results show that prior knowledge does not seem to affect optimism bias or optimistic belief updating. It is important to mention that the question whether participants have heard about the optimism bias before, was very vaguely formulated. As a result, it is unclear to what extent participants are familiar with the bias and its underlying mechanisms. Future research should take prior knowledge into account by better screening and controlling for it.

As shown by Lin and Raghubir (2005) both women and men show levels of optimism bias regarding expectations of their marriage. However, males show greater optimism bias than do females. Considering these results, post-hoc analysis revealed no significant effect of gender on levels of optimism bias or on the magnitude and direction of the update. Nonetheless, the results are in the expected direction with males having more optimism bias score than females before being presented to population base rates, as well as after being presented to population base rates.

It should be taken into consideration that the gender distribution was not equal across our sample since there were more female participants than male participants. Consequently, results of gender analyses might be flawed and lack external validity.

Further, this study was not able to adequately control for confounding factors. Participants were asked to rate the events regarding vividness, familiarity, extent of negativity and arousal. However, this was assessed once over all scenarios. Hence, these scores were not truly informative since they are certainly dependent on the exact situation presented. Moreover, future studies should employ a screening procedure prior to the experiment asking whether participants experienced any of the events before as this can have an enormous bias diminishing effect (Weinstein, 1980).

Although suffering from several limitations, the present study makes a valuable contribution to the comprehensive spectrum of optimism bias. To our knowledge, it is the first study that attempts to combine the belief update task with differential levels of comparison targets. Moreover, the study presents contradicting results to a several unanimous findings. Disconfirming results are an enriching and essential part of research. They essentially inspire future research and help to remain critical towards prior findings. Finally, from a societal perspective, understanding underlying mechanisms of the optimism bias and ways to reduce the bias, have a tremendous relevance.