Experimental Design Details
Experimental sessions are planned for the end of October and beginning of November 2021.
In each session, participants first fill out a survey online (45 minutes duration) and then pick
up their payment in person on the same day and no later than 2 pm. In addition to their
payment, participants receive a meal for immediate consumption when picking up their payment.
Participants can only sign up for one of the sessions.
The timing of the survey and pick-up of payment was chosen so as to mimic the usual process of
choosing a lunch meal as far as possible. Due to COVID-19 regulations, the survey is conducted
online instead of in-person. In some parts of the survey, participants make incentivized guesses:
of the greenhouse gas emissions and calories attributable to each meal and of entirely unrelated
issues (as a time-filling task to proxy for the greenhouse gas emissions guessing task). These
guessing tasks are restricted to 60 seconds so that participants are not able to search for solutions
online.
Participants are randomly sorted into either group 4 (total of 75 participants), group 5 (total
of 75 participants) or group 6 (total of 150 participants). These treatment conditions add to
those observed in the first data collection wave (groups T1, T2 and C). Depending on the
treatment group participants are assigned to, the information conditions under which they make
consumption decisions in the core part of the survey differ.
In the course of the survey, participants’ willingness to pay (WTP) for various restaurant meals
is elicited under different information conditions. In one information condition, participants are
shown the greenhouse gas emissions caused by each meal. These emission values are calculated
using the Eaternity Institute (2020) database. I purchased an Eaternity personal license, and
Eaternity has confirmed that I may use this license to calculate values for the experiment.
The experiment procedure is:
1. Questions on demographic information, allergies, eating preferences, current hunger level.
2. WTP elicitation for meals A,B, C and D.
3. A time filling task in which participants make incentivized guesses on completely unrelated
issues. This task is to proxy for the greenhouse gas emissions guessing task experienced by
groups T1, T2 and C in the first data collection wave, without drawing any attention to
the issue of greenhouse gas emissions. For example, participants are asked to estimate how
much world population increased over the past 20 years, how long a certain running route
in Bonn is and how many yellow cards were shown in the last global football tournaments.
For incentivization and to keep the same protocol as in the emissions guessing task,
additional e0,10 are added to participant’s payment for every guess within 30% of the
true value. Each guess is restricted to 60 seconds.
4. Repeated WTP elicitation for meals A, B, C and D.
• Groups 4 and 5 repeat the previous baseline WTP elicitation.
2
• Group 6 is now shown emission labels for each meal.
5. Repeated WTP elicitation for A, B, C and D.
• Group 4 repeats the previous baseline WTP elicitation.
• Group 5 is now shown emission labels for each meal.
• Group 6 is told that the emissions attributable to the meal chosen will be offset.
6. Group 4 (who has not yet seen any emission labels) guesses the greenhouse gas emissions
caused by eleven different meals, analogous to the emissions guessing task performed by
groups T1, T2 and C in the first data collection wave. However, one additional meal was
added to the guessing task to provide further insights. For incentivization, additional
Euro 0,10 are added to participants’ payment for every guess within 30% of the true value.
Each guess is restricted to 60 seconds.
7. Incentivized guess of the calories attributable to meals A, B, C, D and the cheese sandwich.
Groups 4 and 6 are shown emission labels in this procedure, while group 4 is not.
8. WTP elicitation for receiving emissions information for meals E, F and G.
9. WTP elicitation for meals E, F and G, with information conditions depending on the
previous decision.
10. Participants answer questions on attitudes towards the environment and psychological
traits such as self control in eating. Further, they are asked how much they would support
the introduction of (1) carbon labels or (2) a carbon tax in the student restaurant.
In steps 2, 4, 5 and 9 of the survey, participants make a total of 15 consumption decisions. Each
decision is a choice between receiving a cheese sandwich or a warm meal. This warm meal is a
typical student restaurant meal, and the meals which are handed out to experiment participants
after completing the experiment are in fact prepared by Bonn’s student restaurant. The cheese
sandwich is also prepared by the student restaurant and is a typical cheese sandwich (bread roll,
slices of cheese and some lettuce garnish).
Regardless of the decisions participants make in the survey, they always receive one meal at
pay-out (i.e. cheese sandwich or warm meal). This mimics usual meal choice: the alternative to
not eating a certain meal is not "not eating", but eating something else. The WTP captured
for a certain meal is thus relative to the participants’ WTP for a cheese sandwich, as it is the
participant’s WTP to receive the meal instead of the cheese sandwich. If a participant prefers
the cheese sandwich, this is interpreted as negative relative WTP for receiving the meal. As the
main object of interest in this study is the change in WTP for meals which is induced by the
treatments, it is secondary whether absolute or relative WTP values are captured and analyzed.
In each of the 15 decisions, participants first state whether they prefer receiving the cheese
sandwich or the warm meal at payout, and then state the maximum amount they are willing to
pay to exchange the two options if they are handed their less-preferred option. Participants are
incentivized to respond truthfully, since one of these decisions is in fact implemented. For this
decision, with 50% probability, a participant is handed their preferred option for free. With
50% probability, she is first allocated the less-preferred option, and receives her preferred option only if her WTP lies above a price which is randomly drawn from the interval (0,3), where each
value in 5-cent steps is equally likely. If her WTP lies above the price drawn, the drawn price is
automatically deducted from the participant payment. If her WTP lies below the price drawn,
she receives her less-preferred meal and no amount is deducted.
For each step, the order in which meals are shown to participants is randomized, i.e. there
is randomization across meals A, B, C and D, there is randomization across the incentivized
emission guesses and there is randomization across meals E, F and G. Further, one aspect of
the layout of the design decision - whether the warm meal or the sandwich is shown on the left
or right part of the screen - will differ across experimental sessions to ensure that results are not
driven by this feature.
Which decision is relevant for pay-out is partly pre-determined for logistic reasons, but not
known to the participants. Great care was taken to ensure that participants are not able to guess
which of the decisions is relevant for pay-out. For each participant, there are a total of seven
meals playing a role in her 15 payout decisions. These seven meals differ depending on whether
the participant is vegetarian or not. On each day, the meal which is relevant for payout is the
same across non-vegetarian participants and the same across vegetarian participants. However,
the relevant meal differs across days. It is thus not possible for participants to potentially learn
from experiment participants from previous days which of the meals is relevant. Further, all
meals asked for in the experiment are typical student restaurant meals and are regularly offered
by the student restaurant in Bonn, so that participants should not be inferring that one of the
meals is unlikely to be relevant.
The meals for which WTP is elicited in the second data collection wave are identical to those of
the first data collection wave. Correspondingly, the same meals will be relevant for pay-out as
in the first data collection wave, of course again split across days in such a way that it is not
possible to predict which meal will be relevant for payout on a given day.
For the WTP elicitation for meals E, F and G in step 8 of the survey, participants have the
opportunity to purchase emissions information for these meals. Participants decide whether
they prefer the information to be shown, and indicate a WTP for their preferred display option.
With 50% probability, a participant’s preferred display option is implemented for free. With
50% probability, she is first allocated the less-preferred display option, and receives her preferred
option only if her WTP lies above a price which is randomly drawn. The price drawn for this
information is only deducted from participants’ payment if one of the final three decisions is the
decision relevant for payout. Under these information conditions, the WTP for meals E, F and
G is elicited.