Vaccine information provision

kept private to avoid biasing the study

Four conditions, mutually exclusive.
1. Control: participants are told about a flu vaccine clinical trial and told the placebo arm severe adverse event (SAE) rate.
2. Industry info: Participants are additionally told that the SAE rate was not significantly higher in the vaccine arm.
3. Academic info: As control, plus people are told about an academic study of flu vaccine side effects, and informed that encouragement to vaccinate did not raise the SAE rate.
4. Personal info: Identical to the "academic info" arm, except participants are additionally told "the study participants were Prolific users who were not sure they wanted to get the flu vaccine."

2025-12-02

2025-12-08

Our primary outcome variable is the believed, personal side effect rate of vaccination. We ask people to imagine they do not vaccinate tomorrow - what is the percentage point likelihood that they experience one or more severe adverse events over the next few days. This is the "posterior_no_vacc." Next we ask people the same question but imagining that they DO vaccinate. This is the "posterior_vacc" Our key dependent variable, the believed side effect rate, is posterior_vacc minus posterior_no_vacc.

various measures of vaccine intentions, plus persistence in beliefs and reliability of the provided information

kept private to avoid biasing the study

sample consists of prolific participants who indicate vaccine hesitancy in a baseline survey.

vaccine hesitancy is defined as answering "do not intend to vaccinate" or "may or may not vaccinate" whe nasked about flu vaccination status for the 2025-26 season.

Participants are recruited in a baseline survey that runs for 7 days or until screening 8000 participants. All vaccine hesitant participants invited to the main survey, with a target sample size of 4000. Main survey will run for 7 days or until we hit the target.

In the main survey, participants are randomized to one of the four arms. We will analyze the experiment with OLS regression of outcome on treatment arm indicators and covariates.

Covariates:
- indicators for prior belief levels (which are elicited in multiple choice questions)
- indicators for vaccine intent (may or may not vs. do not intend), measured in screening survey
- indicators (separate) for demographics: age, gender, education, income, race, ethnicity, political views
- indicators for prior vaccine experience: prior flu vacc, prior covid vacc, reporting experiencing prior adverse event from covid vaccine, reporting experiencing prior adverse event from flu vaccine
- indicators for reported trust in government sources (current) and flu vaccine indicators for health conditions.

qualtrics randomizer

individual

No

4000

4000

4000

Simulations on pilot data indicate an MDE of just over 2.9 percentage points, with 80% power. The SD of the unadjusted outcome metric is 31.6 and the mean is 19.6, so the MDE is 15% of the mean.