1. Groups that view the experimental conditions (i.e., Psychological Distance, Mechanism Framing, Social Framing, Economic Framing (personal & social), and Health Framing) will report increased intentions to adopt wildfire risk reduction measures relative to the Active Control group.
2. Groups that view the experimental conditions will report intending to adopt wildfire risk reduction measures sooner than the Active Control group.
3. Groups that view the experimental conditions will show higher levels of support for wildfire risk reduction policies than the Active Control group.

For all 3 main hypotheses we will conduct ANOVAs (alpha = .05) with condition as the independent variable and intention or policy support measures as dependent variables as appropriate. These will be followed by post-hoc tests of differences between conditions using either Tukey's HSD test (if sample sizes are roughly equal across groups, as expected) or Scheffe's test (in the event sample sizes are not equal across groups).

Primary:
1. Intentions to adopt risk reduction measures within the next 2 years: Raw responses are on a 5-point Likert scale (from not at all likely to extremely likely) with 2 additional N/A options (wildfire is not a risk in my area, other reasons) that will be excluded from analysis here. Ratings were made for 15 risk reduction actions. We expect to reduce these ratings into 3 categories as follows and conduct 3 separate ANOVAs accordingly (but this may change if e.g. Principal Components Analysis suggests a more appropriate grouping):
i. Miscellaneous actions (comprising 5 items, rated by everyone)
ii. Home features (comprising 6 items, rated by homeowners only)
iii. Yard/vegetation management (comprising 4 items, rated by everyone).

2. Planned timeline for adoption of intended risk reduction measures: Raw responses are on a 5-point ordinal scale (Within the next year, in 1-2 years, in 2-5 years, in more than 5 years, No specific timeline) with “I don’t know” and “prefer not to say” options that will be excluded from the analyses here. Respondents were asked to rate this for any home features and yard/vegetation management items to which they responded “no [I haven’t done this], but I am planning to make this change” earlier in the survey. We expect to conduct 2 separate ANOVAs, combining ratings across the home features and yard/vegetation management items (but as above this may change on the basis of other analyses).

3. Intentions to engage in recurring/maintenance behaviours in the 2024 wildfire season: Raw responses are on the same Likert scale as in #1 with an N/A option that will be excluded from analysis here, and respondents gave ratings for 4 maintenance behaviours and 4 types of actions related to wildfire smoke protection. We expect to conduct 2 separate ANOVAs, combining ratings across the maintenance and smoke protection items (but as above this may change on the basis of other analyses).

4. Policy support: Respondents rated their support for 9 policies related to wildfire risk reduction on a 5-pt Likert scale (from strongly oppose to strongly support), with a “not familiar” option that will be excluded from the ANOVA. Responses will be combined into a single score for the ANOVA.

Secondary:
Link clicking: As a proxy for information-seeking behaviour, we have 4 binary variables indicating whether respondents opened 4 different links related to wildfire risk reduction shown at the end of the study.

See above for primary analyses. For analysis of our secondary outcome (link clicking), we will conduct logistic regressions to explore the effects of condition.

Planned N = 3185 (n = 3000 approximately representative of the Canadian general population with respect to age, gender, and region, + additional oversample of n = 185 from the Atlantic provinces).

With 6 experimental conditions and a control group, achieving 80% power to detect between-groups differences of 5 percentage points at alpha = 0.05 requires N = 2981. The oversample n of 185 additional participants from the Atlantic provinces was the maximum number of respondents the survey provider was able to commit to recruiting from that region.

Data cleaning steps undertaken by the survey provider include:
1. Removing data from respondents who do not complete the survey or are screened out for not meeting eligibility criteria, declining to respond to eligibility questions, or failing an early attention check
2. Removing data from respondents who complete the survey in less than 10 minutes (based on soft launch data showing a median completion time of 19 minutes)
3. A data scrub to identify and remove poor quality responses (e.g., suspected duplicate participants, nonsensical responses to open-ended questions, suspicious response patterns like selecting the same response for all items in matrix questions, etc.)

Further exclusions will be decided on a case-by-case basis.