Preregistration Skeleton — Green Ads × Source × Style × VBN

1. Title

Effects of Message Style and Source on Environmental Concern: The Moderating Role of Environmental Values and the Mediating Role of Ad Effectiveness

2. Hypotheses

• H1 (total/direct): Green ads increase environmental concern (EC). Expect partial mediation by ad effectiveness (AE); direct effect remains after controlling for AE.
• H2 (moderation): Environmental values (VBN) moderate the effects of Style (informational vs emotional) and Source (NGO vs government vs company).
  • High VBN → informational > emotional on AE and EC.
  • Low VBN → emotional ≥ informational.
• H3 (mediation): AE mediates the effect of ads on EC.
• H3b (moderated mediation): The indirect effect via AE varies with VBN (index of moderated mediation ≠ 0).

3. Design

• Between-subjects 3×2 factorial: Source (NGO/Gov/Company) × Style (Informational/Emotional).
• Each participant views 2–3 short ads from their assigned cell (distinct stimuli per cell).
• Random effects: Participants and Stimuli.
• Control arm (optional): neutral video (no environmental content).

4. Participants

• Target N: XXX (from power analysis; see attached R script).
• Inclusion: adult, fluent in language, passed attention checks.
• Exclusion (pre-registered): failed ≥1 manipulation check, failed ≥1 attention check, completion time < p1 cutoff, prior familiarity with exact ads (self-report), duplicate IPs (for online).

5. Materials & Manipulations

• Stimuli: Matched sets of environmental ads varying only in source and style (duration, topic, call-to-action held constant).
• Manipulation checks: perceived informativeness, perceived emotionality, perceived source credibility/identification.

6. Measures

• Moderator (pre): VBN values (biospheric, altruistic, egoistic), standardized.
• Pre DV: Environmental Concern (EC-T1).
• Mediator (post): Ad Effectiveness composite (believability, attitude-toward-ad, recall/comprehension, credibility).
• Outcome (post): EC-T2 (same scale as T1).
• Behavioral DV (optional): incentivized donation split / eco-choice task.
• Covariates (recorded): political ideology, environmental skepticism, brand familiarity, trust in NGOs/gov/companies.

7. Procedure

1. Consent → Pre-measures (VBN, EC-T1).
2. Randomize to Source × Style (Qualtrics/experiment platform).
3. Watch ads (order randomized). Attention checks embedded.
4. Post measures: manipulation checks, AE, EC-T2, behavioral DV.
5. Demographics and debrief.
6. Optional follow-up at 1–2 weeks (EC + behavioral DV).

8. Analysis Plan

• Manipulation checks: Style → informativeness/emotionality; Source → credibility (mixed models with random intercepts for Participant and Stimulus).
• Primary models (trial level):
  • Mediator: AE ~ VBN × Style × Source + (1|Participant) + (1|Stimulus).
  • Outcome: EC-T2 ~ EC-T1 + VBN × Style × Source + AE + (1|Participant) + (1|Stimulus).
• Probing moderation: simple slopes at VBN = -1, 0, +1; pairwise comparisons of Style within Source across VBN levels (Holm-corrected).
• Mediation: Report indirect effect a×b and its 95% CI. Moderated mediation via conditional indirect effects at VBN = -1, 0, +1 and the index of moderated mediation.
• Missing data: listwise for primary outcomes; sensitivity analyses with multiple imputation if >5% missing.
• Outlier handling: winsorize extreme response times; robust check excluding ±3 SD on DVs.
• Alpha: .05, two-tailed. Familywise control for multiple simple-slope tests (Holm).

9. Power

• Determined by simulation with mixed-effects structure (see power_simulation_template.R). We target ≥ .80 for the Style×VBN interaction and ≥ .70 for the three-way, assuming effect sizes in the small-to-moderate range.

10. Data & Code Availability

• We will share anonymized data, stimuli thumbnails/links, and analysis code on OSF upon publication.

11. Deviations

• Any deviations from the above will be transparently noted in the final report.