One of the main forces driving behavioral economics is a desire to explain the irrational behavior of individuals. This week I'll be discussing a paper authored by Christopher Olivola and Eldar Shafir which delves into one of the more heroic quirks of human behavior, martyrdom.
Before getting into the research let's establish some premises. First of all, rational individuals should prefer pleasurable activities over painful activities. This premise is a main driver of nearly all social sciences and is taken as a given for the purposes of this posting (if it's disproved there are far greater implications than the inaccuracy of this article). Secondly, rational individuals should be willing to pay more (or at least the same amount) for activities they prefer. Combined, these premises amount to "Individuals should be willing to pay more for pleasurable activities than painful activities all other conditions being equal."
If these premises are correct than why would an average person pay more to participate in a lengthy run than enjoy a picnic in the park? After all, picnics are generally pleasant, provide nutrition, and are far less strenuous than running. Yet Olivola and Shafir found that under certain circumstances study participants were willing to pay more for the strenuous running option than the pleasant picnic.
These researchers conducted four experiments investigating how the difficulty or unpleasantness of a charity event influenced donation amounts. The first experiment was a simple questionnaire provided to a bit over one hundred college students. Each questionnaire described a future charity event, either an outdoor picnic or a five mile charity run. In the case of the picnic participants were told that to attend they would be required to donate any amount of money greater than zero dollars. This voluntary admission fee would be matched by a third party. In the case of the charity run participants were informed that they would have to donate to attend but their donation would only be matched by a third party if they completed the run.
Participants were then asked two simple questions. "Would you attend the fundraiser?" and "How much would you donate to attend the fundraiser?". Recalling our previous premises we would expect that more people would be willing to attend the picnic (as it is pleasurable rather than strenuous), and participants would be willing to donate more to attend the picnic than the run. As expected significantly more participants responded that they would attend the picnic than the charity run (86% versus 76%). However, those who responded that they would participate in the run claimed they would donate nearly twice as much as picnic attendees ($23.87 versus $13.88). It's perplexing why anyone would pay more for an experience which is generally considered inferior (as indicated by participation rates). Luckily the remainder of Olivola's and Shafir's research lends some insight.
The second martyrdom experiment involved participants playing a public goods game. In this game participants were given five dollars and asked if they would contribute any of the gifted funds to a common pool. Any funds they did not contribute to the pool they kept. Contributed funds were doubled and dispersed to the entire group at the end of the experiment. Group sizes ranged from three to five, thus in no case was contributing to the pool beneficial to the individual (a contribution of $1.50 for example would be doubled to $3.00 then divided among three group members resulting in each received a dollar. The contributor is thus left $.50 worse off while the rest of the group gains a dollar each.) Participants were not allowed to speak with each other or in any other way communicate in order to avoid strategic group decisions.
The public goods game is a common experimental tool used by economists. However, this example had an interesting difference. Half of the participants were informed that if they contributed to the group pool they would be required to undergo an unpleasant cold pressor task (cold water immersion of the hands for 60 seconds) or their contribution would not be doubled. These participants could easily avoid the cold pressor task by simply keeping all of their gifted funds if they desired.
Despite the pain and risk involved, individuals in the cold pressor groups contributed significantly larger amounts. In fact the cold pressor group contributed their entire $5 gift 67% of the time as opposed to 28% of the time for the control group. What is it about an unpleasant condition that drives a willingness to contribute? Olivola and Shafir conjectured that perhaps having a difficult or strenuous fundraiser might be a social cue that a cause was worthy of the difficulty, and by association greater contributions. However, in this experiment participants were asked what their expectations were for contributions from others in their group. No significant difference was found between those whom believed everyone was receiving a cold pressor task and those who had no knowledge of the task. If difficulty acted as a social cue of value that drives donation amounts we would have expected to find a difference between treatment groups.
The third martyrdom experiment was extremely simple conceptually and in implementation. Subjects were presented with a questionnaire very similar to the one discussed in experiment one. The primary difference was the length of the charity run varied between questionnaires. The goal of this experiment was to determine whether the degree of adversity had a significant impact on contribution rates or if the phenomenon was more binary in nature. Results indicated that length of the charity run had little effect on contribution rates. The researchers interpreted this mean that the degree of adversity is of minimal importance as compared to presence or lack of an unpleasant condition.
The final experiment sought to establish if perhaps the martyrdom effect was partially due to a sort of empathy with victims of a disaster. In other words, would arduous tasks still increase donations if the charity cause was not meant to relieve suffering, but instead promote pleasure. In this study participants were again given a simple survey which represented one of four treatment groups. Each survey informed participants that a charity, either a public park construction project or a feeding starving children effort, was to host a charity event. The event was represented as either a public picnic or a thirty hour fast. Thus the four possible surveys given to participants were park project hosting picnic, park projecting hosting fast, starving children relief hosting picnic and starving children relief hosting fast. With the exception of the change in questions the procedures remained the same as in experiment one.
If the martyrdom effect is in fact based upon an empathetic link with victims of disasters then we would expect to see that in the case of fasting for child hunger, donations would be much higher than in the case of fasting for public park construction. In fact, this experiment demonstrated exactly that. For the public park construction project the picnic option resulted in a larger donation amount on average than the fasting option. Contrarily, for the hunger relief project the fasting option yielded higher donation amounts. Between projects the picnic option showed no statistical difference in donation amounts but the fasting option nearly tripled. This may indicate (though evidence is weak) that a pleasurable experience such as a picnic is minimally influenced by charity cause while a strenuous task such as a fast or run will only yield large donation amounts when combined with a charity which seeks to relieve suffering.
Curiously it seems as though the results are experiments two (ice pressor public goods game) and four (picnic verus fast survey) are somewhat contradictory. Experiment four seems to show that the martyrdom effect is most profound when individuals believe their efforts go towards relieving the suffering of others. However, in the ice pressor experiment it's hard to imagine that participants believed their contribution to the community pool resulted in any relief of suffering. One possible interpretation of this curiosity is that charity cause impacts the magnitude of the martyrdom effect and that fasting has a negative real value. In experiment four a picnic was valued at approximately $15 regardless of cause. This value represents a certain value of the picnic plus a value attributed to a charity relief effect. In other words, if you give $15 you likely value the picnic at some lesser amount and give the remainder as an act of generosity. Viewing the fasting option from a similar perspective but starting with a negative value results in a model which agrees with both experiments two and four. Assume that normally someone would have to pay you $20 in order to induce you to participate in a thirty hour fast. Then when asked how much you would pay to participate in the same fast for the benefit of a public park you answer $10. This is a $30 difference from the implied value of the fast alone but still a lesser amount than the $15 donation for a picnic. In other words we still see a larger net change in value due to the martyrdom effect, but a smaller overall donation. When the charity cause is changed to one which alleviates suffering we then find the martyrdom effect increases in magnitude resulting in a higher overall donation while the picnic option donations remain unchanged. This theory is consistent with the data presented by Olivola and Shafir but requires experimental verification to be confirmed. Interestingly, it does suggest that the martyrdom effect is greater than shown by the majority of these experiments as the value change would be calculated from the negative value associated with the unpleasant experience rather than a zero value.
It's important to note that the martyrdom effect did not seem to have much effect on whether people donated or not, only how much they donated. In fact, unsurprisingly in most experiments less people chose to participate in the unpleasant tasks. However, it is a curious quirk of human nature that under some circumstances we are willing to pay more to suffer than we would pay for enjoyment. Olivola and Shafir offer some interesting data that give hints to what might cause this particular irrational behavior but stop short of offering a suggestion of mechanism or cause. Hopefully their continued research will result in substantial answers to the questions raised by their recent work.
More economics next week, until then stay safe and rationale.