Experimental Design
Our experimental design consists of five stages: An investment game, two measures of risk preferences, a cognitive reflection test, a knowledge quiz about digital money and questions on demographic background. All but the demographics questions are incentivised. Every fifth student (randomly drawn after completion of the experiment) is paid out their earnings. Earnings range between 15 and 25 euros depending on the particular choices of the subjects. The random draw procedure and the expected earnings are announced at the start of the experiment. We implement our experiment on the online platform Qualtrics.
Details on the experimental design:
In the first stage, the investment game:
Subjects are endowed with 10 euros at the start of our experiment. They make two choices in an investment game:
(1) Opt-out option: They choose whether or not to enter the investment game. If they opt-out, they receive a safe payoff of 10 euros from this part of the experiment.
(2) Those who entered are matched into groups of five and choose whether to "continue to invest" their endowment or to "withdraw" their endowment. Choosing "continue to invest" is associated with high risk, since this choice leads to the highest payoff possible only in the case where all other subject in the group also chose "continue to invest" and leads to lower payoffs the more others decide to "withdraw". Choosing "withdraw" is associated with lower risk, as possible payoffs are more evenly distributed, even though they still positively depend on the number of other individuals that chose "continue to invest".
We also elicit subjects' beliefs about others' decision to "continue to invest" or "withdraw". Specifically, we ask subjects, before they actually make their continued investment decision, about what fraction of other subjects will choose "continue to invest". We also ask subjects a similar question after they had made their investment decision.
In the second stage, the risk preference elicitation:
Subjects play a standard balloon inflation task. Subjects also choose from a standard multiple price list of risky choices.
In the third stage, the cognitive reflection test:
Subjects answer three standard questions in a cognitive reflection test.
In the fourth stage, the knowledge quiz about digital money:
Subjects answer a couple of questions to test their knowledge about digital money and bank runs. We impose a time limit of 30 seconds per question. We chose the questions in such a way, that they cannot be answered with a quick (less than 30 seconds) google search.
In the fifth stage, the demographic background elicitation:
We ask students standard question regarding their demographic background in general and progress of their studies at TUM in particular.
We have six different treatments, which only vary with regards to the investment game (first stage of the experiment). In treatments 1-3, subjects make their choices as we described above. In treatments 4-6, we deviate from above description by making participation in the investment game mandatory. Comparing "continue to invest" decisions between treatments 1-3 and 4-6 then allows us to estimate the effect of a prior opt-out option from the investment game, since such an option only exists in treatments 1-3.
Treatments also differ in regard to how money that is earned in the experiment is paid out. In treatments 1 and 4 subjects can choose between a payment in cash or digital money (via Paypal or bank transfer), in treatments 2 and 5 earnings are paid in cash, in treatments 3 and 6 earnings are paid in digital money. Comparing these treatment groups thus allows us to estimate the effect of the existence of digital money on "continue to invest" decisions. We can also estimate an interaction effect between the existence of digital money and the existence of the opt-out option. Importantly, we communicate to subjects that any earnings have to be collected in person at TUM at a later date, once the experiment is completed. Thus, there are no real "convenience" differences between digital payout and cash payout i.e. the transaction costs are the same with both types of payments.