**6. Limitations**

Some factors related to the experimental design may have influenced how subjects behaved.

As the link to the survey were distributed at lectures encouraging students to participate, it is unknown when, where and possibly with whom the subjects completed the survey. Hence, there is concerns regarding their anonymity. Considering the relatively high share of correct answers in the CRT, one could expect subjects to have communicated with each other or have accessed the internet to look up the correct answer. Further, the chances of receiving payment for completing the CRT depended on the number of correct answers, which might have further incentivized subjects to look up the correct answer - at least incentivized them to think more carefully about the question, which was unintended. These limitations question whether the categorization of subjects is reliable. A reasonable explanation for the relatively high share of correct answers on the CRT in this study is the test's correlation with math abilities [5]. The vast majority of subjects were students of Economics, Political Science or Psychology. Especially students of Economics are expected to be relatively more capable of math. The survey questions did not elicit from which education the subjects were enrolled.

Only seven of the 295 subjects who completed the study received payment, providing only weak incentives. However, the observations here fit rather well the observations from other studies with stronger economic incentives. In a meta study, the average transfer was found to be 28% of the endowment [18], which is not far from 30.4% observed in this study. In a meta study on the ultimatum game, subjects were found to offer 40% of the endowment on average [17], which is comparable to the 43.5% observed here.

Further, around 20% of the subjects who started completing the survey opted out before the final question. Not being able to control the condition under which the survey was completed increases the probability of subjects sabotaging the experiment by choosing randomly or not reading through the instructions thoroughly. However, including or excluding the "outliers" of the present study did not change results.
