Observability in food choices

2022-03-15

2022-07-01

The primary outcome measure is the food choice of participants signing up for conference lunches. They can select between a vegan, vegetarian or meat-or-fish dish for each of the three conference lunches. This multinomial choice variable 'food choice' with three levels serves as the major variable of interest. From this choice, one binary choice variable, two personal food choice scores and three count variables will be derived as described below.

The study hypothesis concerns the role of knowing that one's choice will be publicly revealed through a colour-coding on the conference attendance badge. We hypothesize that the participants who will be informed of the visibility of their choice are more inclined to choose vegetarian or vegan meals for fear of being judged as behaving inappropriately or not conforming to the social norm of choosing environmentally friendly food options that likely prevails at an environmental economics conference.

From the multinomial 'food choice' variable, a binary variable named 'meatless food choice' will be derived. This variable takes the value 1 if participants choose a vegan or vegetarian meal and 0 if they choose the meat/fish option. This captures the fact that a meal including meat or fish is seen as the norm in many cultures and allows for an investigation of whether participants do or do not deviate from this option.

Two additional derived outcome variables serve to account for the fact that the three choices made by one and the same subject are likely not independent of one another, but reflect some common unobservable preferences. The first score variable, termed 'personal food score', is derived by assigning 1 point for a meat/fish choice, 0.5 points for a vegetarian choice and 0 points for a vegan choice. This way, each subject's personal food score will represent a value between 0 and 3 that captures the 'animal-intensity' of their choices, with higher values representing higher animal-intensity. The second score, termed 'personal binary food score', is derived from the 'meatless food choice' by assigning 1 point to meat/fish-choices and 0 points to meatless choices, which also results in a score ranging from 0 to 3.

While the scores serve to make the three choices of each participant comparable, they necessarily involve some weighting of the different options against one another. Therefore, three additional count variables are derived for each subject that capture the amount of vegan, vegetarian and meat/fish choices an individual makes (and therefore also range between 0 and 3). This way, both the interdependence of choices made by a single person as well as the difficulty of comparing different options to one another (e.g., are two vegetarian choices more animal-intensive than one vegan and one meat choice?) are accounted for.

Deriving these different variables allows us to analyze the robustness of identified effects to different specifications of the outcome variable and enables a more fine-grained analysis of the effects that may drive whether and how often a specific option is chosen by each subject.

The experiment involves the participants at an international conference. During the registration for the conference, participants must select which type of lunch (vegan, vegetarian or meat/fish) they would like to receive. Decisions over the lunch options are studied under different communication treatments.

The participants of the experiment will be the participants who register online for the Annual Conference of the European Association of Environmental and Resource Economists 2022 in Rimini. During the registration phase, they are asked to choose which lunch they would like to have during the conference. As lunch is provided on three subsequent conference days, participants are asked to choose three times, with different options for each day being possible.

Half of the participants will be informed that a symbol will be inserted in the conference badge indicating their three choices by colour. In contrast, the remaining half receives this information upon arrival.

At the end of the registration form, participants can decide whether they allow the transfer and analysis of an anonymized part of their data (including the meal choice) to a research team, consisting of the persons indicated within this form.

At the conference, all participants will receive a badge with the colour representing the choice made during registration.
- The recruitment will therefore be carried out automatically during the voluntary registration phase for the conference.
- The treatment to which each person will be assigned will be selected at random.
- Data on the identity of the participants will not be disclosed to the researchers.
- A post-conference questionnaire will elicit some information about participants' perception of the meal choice, including a manipulation check.

Randomization is automatically done by a computer in the registration phase.

Randomization is done at the individual level.

No

We expect to have 700 individuals participating at the event giving their informed consent to analyze the registration data.

The planned number of observation of the outcome variable is an estimated 2.100, as each participant is making three decisions.

The treatment groups are allocated and random and will therefore likely contain an estimated 350 participants and 1.050 observations each.