Increasing the Efficacy & Diffusion of Covid-19 Messaging for Vaccination

We have 3 main treatment groups + control.

Treatment 1: Direct Messaging
Facebook users in the area receive ads which include videos of health professionals telling them to get vaccinated. The health care professionals answer common questions about Covid-19 vaccines (e.g. are they safe / are they free).

Treatment 2: Friends Messaging
Facebook users in the area receive ads which include videos of health professionals encouraging them to help their friends to get vaccinated.

Treatment 3: Gossips Messaging
Facebook users in the area receive ads which include videos of health professionals encouraging them to nominate their most influential friends to help their friends get vaccinated.

Moreover, when individuals arrive at the study webpage, they will be randomized into different versions of the same videos:

Content Variation Treatments: "Fact" vs "Myth" Frame
Some of the health professionals' videos will be framed as responding to a question about the vaccines, while others will be framed as responding to a myth about the vaccine.

2021-12-22

2022-01-27

1. Changes in county-level Covid-19 vaccination
Measured by publicly available county- or zip code-level vaccination data. The study will examine total vaccinations as well as vaccinations by age group. The study will also examine people 1) getting their first dose of vaccine, 2) completing their vaccine series, and 3) getting booster shots.

2. Social shares of vaccine-related content
Number of social media shares of vaccine-related videos and messaging that the study team has created, disaggregated at the county or zip code level.

3. Number of views of vaccine-related content
Number of views of the website and vaccine-related videos, disaggregated at the county or zip code level.

1. COVID-19 cases
Measured by publicly available county- or zip code-level Covid-19 cases, in some areas also split by vaccinated vs unvaccinated status.

2. COVID-19 symptoms prevalence
Measured by publicly available county- or zip code-level Covid-19 symptoms, in some areas also split by vaccinated vs unvaccinated status.

3. COVID-19 hospitalization and mortality
Measured by data on patients admitted to hospitals or who have been recorded to have died from Covid-19, , in some areas also split by vaccinated vs unvaccinated status.

4. Vaccine hesitancy
Measured by publicly available county- or zip code-level surveys

5. Movement and Travel
Measured using mobile phone location data of Facebook users. Vaccinated people may be more likely to take risks, including travel.

6. School closures
Increased cases could lead to more school closures. The investigators do not expect the study to be well-powered on this outcome.

7. Number of vaccine ambassador volunteers
This will be compared between the "friends" and "gossips" treatment groups. This will include people who start to sign up to be vaccine ambassadors as well people who complete the vaccine ambassador consent process.

The main study will have three treatment arms and a control group. Each of the treatments will be randomized at the county level and will be initiated through a Facebook ad campaign.

Experimental Sample: The experimental sample includes where vaccinated rates remained low in fall 2021.

The following procedures will be used:

Treatment Group 1 ("Direct" messaging): The investigators will conduct a Facebook ad campaign to show content about COVID-19 vaccination to a large number of Facebook users. Facebook will allocate ad credits across users in treatment areas. Like for any Facebook ad, individuals can choose to whether or not to watch the video and can close the ad at any time. If individuals wish, they can also share any of the content with others.

Treatment Group 2 ("Friends" messaging): For this group, the Facebook ad campaign will recruit individuals to become "vaccine ambassadors" to help spread the word about vaccination to their friends, families and communities. Recruitment ads will be disseminated in a similar manner to the content in T1. Individuals will be able to share the ad with others. Individuals interested in learning more can click through a link in the ad to the vaccine ambassador webpage. Participants can consume or share any of the information on the webpage. Interested individuals can also enroll formally as vaccine ambassadors. They can share their contact information with the study to a) be entered into prize lotteries, b) receive reminders to share content with others, c) be contacted in the future for possible follow-up surveys. They will be asked to give their informed consent at this time and will be screened based on whether they live in a treated county and whether they are 18 years or older.

Treatment Group 3 ("Gossips" messaging): This treatment looks the same as Treatment Group 2, except the recruitment ads will differ slightly. If the individual clicks through to learn more about vaccine ambassadors, they will also be reminded to share the invitation with people they know who are often the source of important news or information.

Control: individuals in these counties will receive no intervention and will never be contacted by members of the study team.
Treatment and control status at the county level will be merged with aggregated and de-identified datasets to measure COVID vaccinations, COVID symptoms, and cases, and engagement with the vaccine ambassador content.

The study team will create a series of websites with materials that the vaccine ambassadors can share with friends, family and others in their communities. This same content will be disseminated through sponsored Facebook ads in T1. On the website, the framing of the vaccine-related videos with either a "fact" or "myth" frame will be randomized at the individual level. The purpose is to study which framing is more effective at engaging users and encouraging sharing.

Randomization was done using Stata on a computer

U.S. county for the main experiment
Individual for the variation in content framing

Yes

1,397 counties

1,397 observations of county-level data

468 counties in the control group, 310 counties in T1 ("direct" messaging) treatment, 309 counties in T2 ("friends" messaging), and 310 counties in T3 ("gossips" messaging)

We do not know how many individuals will visit the project website. Randomization into content framing will be done by the website itself with 50% seeing the myth framing and 50% seeing the fact framing.