Hyper-altruistic behavior vanishes with high stakes

Using an incentivized experiment with statistical power, this paper explores the role of stakes in charitable giving of lottery prizes, where subjects commit to donate a fraction of the prize before they learn the outcome of the lottery. We study three stake levels: 5€ (n = 177), 100€ (n = 168), and 1,000€ (n = 171). Although the donations increase in absolute terms as the stakes increase, subjects decrease the donated fraction of the pie. However, people still share roughly 20% of 1,000€, an amount as high as the average monthly salary of people at the age of our subjects. The number of people sharing 50% of the pie is remarkably stable across stakes, but donating the the whole pie–the modal behavior in charity-donation experiments–disappears with stakes. Such hyper-altruistic behavior thus seems to be an artifact of the stakes typically employed in economic and psychological experiments. Our findings point out that sharing with others is a prevalent human feature, but stakes are an important determinant of sharing. Policies promoted via prosocial frames (e.g., stressing the effects of mask-wearing or social distancing on others during the Covid-19 pandemic or environmentally-friendly behaviors on future generations) may thus be miscalibrated if they disregard the stakes at play.


S.2 Sharing 5e under certainty
Apart from the main treatments in the main text, we additionally conducted a treatment, in which people donated out of 5e to be received with certainty. That is, people played a standard five-euro Dictator Game. The sample in this treatment was also powered to detect an average effect of 0.33SD with a power of 0.8 and a significance level of 90% while comparing this treatment with the 5A conditions. A total of n = 104 students participated in this treatment under conditions identical to our main treatments (see Methods). We compare the donations in this treatment with the behavior in the 5A treatment from the main text. Below, we label DG5 the treatment under certainty and DG5A the 5A treatment.  Figure 5 suggests that people give somewhat more in DG5A than in DG5. There seems to be more subjects keeping the entire endowment for themselves and fewer people sharing around 50% and 100% of the pie in the latter case. Indeed, subjects donate 50.79% of the pie in DG5A while the fraction drops to 39.32% in DG5. This difference of 11.46 percentage points is statistically significant at conventional 5% (t − test = 2.04, p = 0.043). We thus conclude that subjects do not use symmetric ambiguity as an excuse to give less in our experiment. In fact, they give more. Although we do not isolate the effect of impact uncertainty as proposed by [29], the symmetric ambiguity regarding the payment of both parties seems to eliminate moral wiggling and impact ambiguity seems to prevail, reproducing the result of [29]. Giving is the amount (in euros) and share is the percentage of the pie given in the Dictator Game (DG). Give0 is equal 1 if subject gives 0 in the DG, and 0 otherwise. EnqualSplit, GiveAll, Give < 50 and Give > 50 are defined in the same way. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Share is the percentage of the pie given in the Dictator Game (DG). Give0 = 1 if a subject gives 0 in the DG and 0 otherwise. EqualSplit, GiveAll, Give < 50 and Give > 50 are defined in the same way. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.
In the following sections, we present the instructions that faced subjects in each treatment. The experiment was conducted in Spain. In the following we show the translated instructions.
Note that in the 5e and 1,000e treatments, we let participants choose the charity that would receive their donations by writing it into a space in the instruction right below the selection of the donated fraction. Since the experiment was performed in a jesuist University, the vast majority of subjects chose to donate to the jesuist non-governmental organization Entreculturas (www.entreculturas.org), but other NGOs were also selected despite rarely. All the donations in these treatments were executed following the preferences of the subjects. To simplify the organization of the 100e treatment, we let subjects only choose how much to donate to an unkown charity and we transfered all the donated money to Entreculturas too.

Treatment 5e
If you win the 5e lottery, would you like to donate a fraction to an NGO? Keep in mind that the answer you give us is NOT hypothetical and will determine your final payment. Please mark the percentage of the prize that you would donate on the scale below.

Treatment 100e
If you win the 100e lottery, would you like to donate a fraction to an NGO? Select the amount of money you want to donate from 0 (nothing) to 100% of 100 euros. Please mark only one percentage.

Treatment 1000e
If you win the 1000 euros lottery, would you like to donate a fraction to an NGO (see below)? Keep in mind that the answer you give us is NOT hypothetical and will determine your final payment. Please mark the percentage of the prize that you would donate on the scale below.