Gender-Reflective Framing and HPV Vaccine Intentions

Parents of children under age 9 in Germany read a short informational message about HPV vaccination. The message emphasizes either female-specific cancers (cervical cancer) or male-specific cancers (penile cancer).

Each participant reads one of two brief informational messages (~140 words) about HPV vaccination, identical in structure, length, and tone. The female-specific version names cervical cancer as the primary example; the male-specific version names penile cancer. Both messages note that HPV causes cancers in both sexes and reference the STIKO vaccination recommendation. Each message is accompanied by a gender-congruent illustrative infographic. The message page enforces a minimum viewing time of ~20 seconds. Participants are not informed that an alternative version exists. All respondents receive a balanced factual debrief on HPV vaccination for girls and boys after completing the survey.

2026-04-13

2026-05-15

Vaccination intention (0–10 scale): How likely the parent would be to vaccinate their oldest child against HPV once the child reaches the recommended age.

Measured immediately after treatment exposure via a single item on an 11-point scale (0 = "definitely will not vaccinate" to 10 = "definitely will vaccinate").

(1) Intention to discuss HPV vaccination at the next pediatric visit (0–10 scale). (2) Binary opt-in for additional HPV vaccination information. (3) Video watch behavior and duration (conditional on opt-in).

The pediatrician intention uses the same 0–10 scale as the primary outcome. The information opt-in is a behavioral engagement proxy (yes/no). Video watch duration is recorded in seconds for respondents who opt in and serves as a revealed-preference measure of engagement. Secondary outcomes are analyzed using the same regression specifications as the primary outcome.

2 × 2 × 2 between-subject factorial design. Factor 1 (randomized): message emphasis (female-specific vs. male-specific HPV cancers). Factor 2 (observed): parent gender (mother vs. father). Factor 3 (observed): oldest child's gender (daughter vs. son). Target population: parents in Germany whose oldest child is under 9 years of age. Target sample: N ≈ 2,160 (≈270 per cell). Six pre-specified two-sided hypotheses tested on the primary outcome.

The target population is mothers and fathers residing in Germany whose oldest child is below age 9 — i.e., below the age at which STIKO recommends initiating HPV vaccination (9–14 years) — so that the vaccination decision is prospective at the time of the survey. Respondents must cohabit with their oldest child. Respondents reporting non-binary gender are screened out because the confirmatory analysis requires the binary factorial structure (mother/father × daughter/son). Recruitment is via Dynata, an international online panel provider. Stratified quota sampling targets equal cell sizes across the four parent–child subgroups; within each subgroup, randomization to the message condition is 1:1 at the individual level.
Survey flow: informed consent → screening (gender, number/age/gender of children, cohabitation) → attention check (incorrect responses screened out) → pre-treatment measures (HPV awareness, vaccine attitudes, baseline beliefs on infection risk, vaccine safety, disease severity, side effects) → randomized treatment exposure → manipulation check → primary outcome → secondary outcomes → post-treatment beliefs (identical to baseline items) → moderator/mediator items (identity salience, parental concern, severity perceptions, compassion, responsibility norms) → perceived HPV statistics (vaccination coverage, infection rates, cancer incidence by gender) → willingness to pay for HPV vaccination → sociodemographics → optional KBV video → debrief.
The first ~60 observations serve as a soft launch to verify technical functionality (randomization, piping, filter logic, redirect parameters). Provided no issues are detected, these are retained in the final sample. All specifications use OLS with robust standard errors. The reference specification controls for P, C, and P × C; a covariate-adjusted robustness specification additionally includes pre-treatment covariates. Balance is assessed via a joint F-test across all eight cells.

Computer-based individual-level randomization via the Tivian (EFS) survey platform's quota system, stratified by parent gender × child gender (1:1 within each subgroup).

Individual respondent (no clustering).

No

2,160 individuals (not clustered; same as number of observations).

2,160 respondents (8 cells × ~270 each).

~1,080 per message condition. Within each of the four parent–child subgroups (mother-daughter, mother-son, father-daughter, father-son), ~270 receive the female-specific and ~270 the male-specific emphasis.

Monte Carlo simulation (10,000,000 replications, N = 2,100, 262 per cell, simulated Likert-scale outcome). MDEs at 80% power, 5% significance (two-sided): (i) H0/H1/H2 (four-vs.-four cell comparisons): ≈ 0.275 scale points; (ii) H1a/H2a (difference-in-differences): ≈ 0.553; (iii) H3 (triple interaction): ≈ 1.122. These are conservative upper bounds that do not incorporate precision gains from covariate adjustment.