|
Field
Last Published
|
Before
October 25, 2021 05:34 PM
|
After
October 26, 2021 02:18 AM
|
|
Field
Primary Outcomes (End Points)
|
Before
The data collected in the second data collection wave can be pooled with data from the first
data collection wave. Differences in meal tastes between July and November are controlled for
due to individual by meal fixed effects included in each analysis. A total of eight observations
are made for each individual, as for each of the four meals, the change in WTP taking place both from (1) the first to the second WTP inquiry and (2) the first to the third WTP inquiry is
observed. This approach maximizes power, leading to the following number of observations for
each treatment effect (provided 300 participants can be recruited in the second data collection
wave):
1. Change in demand occurring in reaction to being made aware of emissions (by being asked
to guess emissions caused), without being shown emission labels: 1240 observations (155
participants in Group C observed twice in the first data collection wave)
2. Change in demand occurring in reaction to being made aware of emissions and then shown
emission labels: 1240 observations (155 participants in group T1 and 155 participants in
group T2 in the first data collection wave)
3. Change in demand occurring in reaction to being made aware of emissions and then being
told that emissions will be offset: 1240 observations (155 participants in group T1 and
155 participants in group T2 in the first data collection wave)
4. Change in demand occurring in reaction to being asked for WTP a repeated time, without
being made aware of emissions: 900 observations (75 participants in Group 4 observed
twice and 75 participants in group 5 in the second data collection wave)
5. Change in demand occurring in reaction to being shown emission labels, without previously
being made aware of emissions: 900 observations (150 participants in group 6 and 75
participants in group 5 in the second data collection wave)
6. Change in demand occurring in reaction to being told that emissions will be offset: 600
observations (150 participants in group 6 in the second data collection wave)
In the main analysis, I am interested in whether these treatment effects significantly differ and in
how this interacts with the greenhouse gas emissions caused by the meal in question. Treatment
effects (3) and (6) can be pooled to provide insights to the structural model for decision making
under reduced environmental concerns, while treatment effects (1),(2),(4) and (5) together form
a two by two design varying across the dimensions of awareness (by guessing emissions) and
information (through emission labels).
As an additional analysis, the effects of providing emission labels on WTP can also be analyzed
using only data collected in the second data collection wave, comparing only treatment effects
(4) and (5) in the list above.
|
After
See attached Pre-Analysis Plan
|
|
Field
Randomization Method
|
Before
Participants are randomly sorted into treatment groups.
|
After
See attached Pre-Analysis Plan
|
|
Field
Randomization Unit
|
Before
Randomization at the level of the individual
|
After
See attached Pre-Analysis Plan
|
|
Field
Planned Number of Clusters
|
Before
300 individuals
|
After
See attached Pre-Analysis Plan
|
|
Field
Planned Number of Observations
|
Before
300 individuals
|
After
See attached Pre-Analysis Plan
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
75 participants in group 4,
75 participants in group 5,
150 participants in group 6
|
After
See attached Pre-Analysis Plan
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
Details in attached PDF. The standard error of the effect size is assumed to be 0.0217.
With 300 participants in the experiment, MDE is 0.0583 (9.5%)
With 200 participants in the experiment, MDE is 0.0714 (11.6%)
|
After
See attached Pre-Analysis Plan
|
|
Field
Intervention (Hidden)
|
Before
This document describes additional data collection to Schulze Tilling, Anna. 2021. "Quantifying
the role of greenhouse gas emissions in consumption choice." AEA RCT Registry. June 22.
https://doi.org/10.1257/rct.7858-1.0
The original pre-registration focused on the first wave of data collection, which took place
between the 22nd of June and the 8th of July 2021. I will perform a second wave of data
collection in October and November 2021, which will be further described in this document.
The goal of this second wave of data collection will be to add to the data collected in the first
wave with additional treatment conditions and additional observations (the sample size aimed
for in the first data collection wave was not quite reached due to some potential participants
not physically being in Bonn during the Covid-19 pandemic).
Experiment participants are again recruited via hroot from the participant pool of the BonnEconLab.
The requirement for participation in the experiment is that the participant does not
follow a very restrictive diet (e. g. vegan, lactose-free, gluten-free or halal). Vegetarians are
permitted to participate. The reason for this restriction is that people following these restrictive
diets only make up a sub-part of the population and I consider them negligible in determining
the effect a CO2 label has on the population. Vegetarians, in contrast, make up a larger part of
the population. In the pre-survey, 20% of participants were vegetarian. Two of the meals shown
to participants in the main decision scenarios are the same across all participants (vegetarian
meals), while the other two differ. This way, half of the meals shown to non-vegetarians contain
meat, while vegetarians are only shown vegetarian meals.
Due to the fact that the experiment procedure differs from usual procedures at the BonnEcon-
Lab and that some potential participants may not physically be in Bonn due to the Covid-19
pandemic, it is again difficult to predict how successful recruitment will be. I plan for 200-300
participants.
The sample will be restricted as previously for the main analysis:
• The fastest 3% of participants are excluded from the main analysis.
• There are four comprehension questions to check the participants’ understanding on the
incentivization of WTP. If participants’ response to at least one of these questions is
incorrect, participants receive an error message and this counts as one error. I expect the
average participant to make one to two mistakes as the questions are designed to make
the participants further think about the mechanism. Participants who make more than
five mistakes are excluded from the main analysis.
|
After
This document describes additional data collection to Schulze Tilling, Anna. 2021. "Quantifying
the role of greenhouse gas emissions in consumption choice." AEA RCT Registry. June 22.
https://doi.org/10.1257/rct.7858-1.0
The original pre-registration focused on the first wave of data collection, which took place
between the 22nd of June and the 8th of July 2021. I will perform a second wave of data
collection in October and November 2021, which will be further described in the attached Pre-Analysis Plan.
The goal of this second wave of data collection will be to add to the data collected in the first
wave with additional treatment conditions and additional observations (the sample size aimed
for in the first data collection wave was not quite reached due to some potential participants
not physically being in Bonn during the Covid-19 pandemic).
|
|
Field
Secondary Outcomes (End Points)
|
Before
WTP, guess and survey data can be pooled with data from the first data collection wave to add
to the following secondary analyses (already described in the original PAP):
• WTP for carbon labels without and with having experienced carbon labels previously (as
elicited in step 8) suggests the (expected) effect on welfare of being provided with these
labels. The carbon labels tested in the experiment were designed together with Bonn’s
student restaurant and the student restaurant is considering implementing these labels on
a large scale in the future.
• Participant’s guesses for the emissions attributable to meals can be tested for their accuracy.
Camilleri et al. (2019) found that people are insufficiently sensitive to the magnitude
of differences in emissions between food items. I expect consumers to overestimate the
emissions of low- and underestimate the emissions of high-carbon meals.
• The data gathered in step 2 can be used to construct a demand curve for each meal,
allowing to evaluate the effect a carbon tax would have. Thus, one can compare the
effectiveness of carbon labels versus carbon tax as policy instruments.
• In step 10 of the experiment, participants are asked for their approval of (1) the introduction
of carbon labels and (2) the introduction of a carbon tax in the student restaurant. I will
examine whether approval differs between treatment groups. Further, these answers can
be used as a check on the WTP which participants indicate for being shown emissions
information.
• Suggestive within-subject estimates of treatment effects can be constructed by comparing
step 4 and step 5 WTP for a given meal of a given subject with her baseline WTP for
the meal. This allows for some heterogeneity analysis. The effectiveness of the label
might differ depending on (1) subjects’ education, (2) subjects’ income, (3) subjects’
environmental attitude, (4) subjects’ degree of self control in eating. The same factors
might influence subjects’ WTP for being shown the label.
• One might argue that participants shown emissions labels use these labels to infer nutritional
characteristics of the meal. To check whether this is the case, I have participants
guess the calories attributable to meals in step 6 of the first data collection wave/ step 7
of the second data collection wave. One group of participants is not shown emission labels
for this guess, while other participants are shown the emission labels. If it is the case that
participants infer nutritional information from emissions labels, the guesses made by the
two groups should systematically differ.
Similarly, data from the second data collection wave can also be included in the estimation of
the structural model.
|
After
See attached Pre-Analysis Plan
|