Anticipated Peer Effects: An Application to COVID-19 Vaccinations

We run an online survey experiment in Germany. We recruit participants through a professional survey company in German states that at the moment of the intervention allow for an online pre-registration for COVID-19 vaccination appointments. Our experimental intervention groups participants into pairs and varies whether information on willingness to get vaccinated is conveyed to the partner.

2021-04-08

2021-12-31

(1) Stated willingness to get vaccinated against COVID-19
(2) Stated willingness to register for a vaccination appointment
(3) Revealed measure of actually registering for a vaccination appointment
(4) Revealed measure of actually getting vaccinated in follow-up survey

(1) Stated willingness to get vaccinated against COVID-19
Survey participants are asked whether they are willing to get vaccinated against COVID-19

(2) Stated willingness to register for a vaccination appointment
Survey participants are asked whether they are willing to register for a vaccination appointment via the online registration platform of their state of residence right now with the option of answering yes or no.

(3) Revealed measure of actually registering for vaccination appointment
We will give participants the option to register for the vaccination online via the online registration platform of their state of residence which we see as a revealed preference measure of vaccination willingness. To this end, subjects will have the option to click on a link redirecting them to the official vaccination registration portal. Participants willing to do this fill out the registration form outside of our survey and receive an automatically generated registration confirmation mail. Once participants return to the survey, we ask them to type in the subject line and email address from which they received the confirmation email. We do so to obtain verifiable proof of their registration while upholding anonymity.

(4) Revealed measure of actually getting vaccinated in follow-up survey
In a follow-up survey several weeks later to which all subjects that have participated the main experiment are invited, we will ask respondents to state whether they have been vaccinated and, in case they do, to receive verifiable information, ask them to report the name and batch number of their COVID-19 vaccine (or the last vaccine they received in case they report not having been vaccinated against COVID-19) as stated in their official vaccination certificate.

Same decision of both partners in a pair
Belief about same decision of both partners in a pair (only Senders)
Belief about own decision influencing decision of partner (only Senders)

We group subjects into pairs where they take different roles: individuals either act as Senders or Receivers. The decision of Senders whether to get vaccinated is (potentially) made observable to Receivers but not vice versa. We assign pairs of subjects into different experimental conditions: In condition (A), Senders’ decisions are not observable, while in condition (B) Senders’ decisions are observable to their partner. We hypothesize that the observability of Senders’ decisions may simultaneously give rise to anticipated peer effects and social image effects. In order to isolate anticipated peer effects from social image effects, we hold the latter constant across Senders whose behavior can be observed by their Receiver and, at the same time, induce random variation in Senders’ beliefs about the strength of the potential peer effect they can exert on their Receivers. To induce variation in this belief among Senders, we further randomize the timing when the Receiver is informed about the Sender’s decision in condition B. In condition (B1), Senders are informed that their decision is only reported to Receivers after he has already decided. In condition (B2), Senders are informed that their Receivers learn about their decision before the latter get to decide for themselves. For further details on the experimental design, please refer to our pre-analysis plan.

We group subjects into pairs where they take different roles: individuals either act as Senders or Receivers. The decision of Senders whether to get vaccinated is (potentially) made observable to Receivers but not vice versa. We assign pairs of subjects into different experimental conditions: In condition (A), Senders’ decisions are not observable, while in condition (B) Senders’ decisions are observable to their partner. We hypothesize that the observability of Senders’ decisions may simultaneously give rise to anticipated peer effects and social image effects. In order to isolate anticipated peer effects from social image effects, we hold the latter constant across Senders whose behavior can be observed by their Receiver and, at the same time, induce random variation in Senders’ beliefs about the strength of the potential peer effect they can exert on their Receivers. To induce variation in this belief among Senders, we further randomize the timing when the Receiver is informed about the Sender’s decision in condition B. In condition (B1), Senders are informed that their decision is only reported to Receivers after he has already decided. In condition (B2), Senders are informed that their Receivers learn about their decision before the latter get to decide for themselves. For further details on the experimental design, please refer to our pre-analysis plan.

Randomization done in office by a computer

Randomization takes place at the individual level.

No

4,000 individuals

4,000 individuals

Expected treatment group sizes:
(A): 667
(B1): 1,667
(B2): 1,667