Mitigating Wildfire-Urban Spread Through Risk and Strategy Communication

Participants are randomized into two treatment groups or a control: one group receives information from the B.C. Wildfire Dashboard about wildfire incidence in their region over the past year; the other receives personalized FireSmart recommendations based on their self-assessment results. The project applies the behavioural and communication insights to a new domain, exploring how tailored risk information can shift protective behaviours in high-stakes public policy contexts.

2026-05-08

2026-10-01

Adoption of recommended BC FireSmart mitigation actions; reduction in risky behaviour; insurance coverage (if possible); perceived risks to property and community

We are broadly interested in what respondents have actually completed in terms of fire-mitigation maintenance around their properties and what socio-economic and behavioral factors are associated with mitigation efforts.

In our Wave 1 survey, we measure the resilience of respondents' homes to wildfires and elicit the reasons behind their adoption — or non-adoption — of recommended mitigation actions, linking these behaviours to individual characteristics. We then randomize respondents into one of two information interventions or a control group. A follow-up survey will assess longer-term changes in behavior and expectations.

Not available

Randomization is done through Qualtrics platform

We randomize the treatments (control, B.C. wildfire dashboard, BC FireSmart Home Assessment) by region.

No

We plan to collect in six different regions of the province of B.C.

2160 respondents

We are aiming for 360 respondents per region, with 120 per treatment. Some regions may be more challenging to collect data from.

The study enrolls 720 respondents per arm across three treatment conditions, stratified by six geographic regions (120 respondents per region–treatment cell). MDE calculations assume 80% power, a pilot-estimated within-outcome standard deviation of SD = 0.50 scale points, a baseline–endline correlation of ρ = 0.90 (conservative bound: ρ = 0.50), and individual-level randomization within strata (ICC = 0). Holm–Bonferroni correction is applied jointly across eight primary hypothesis tests (α ≈ 0.00625). For anticipation of fire damage to home or community (7-point scale, usable range = 6 points): the MDE is 0.030 scale points (SD = 0.50; 0.5% of the usable scale range) under the preferred assumption (ρ = 0.90), and 0.067 scale points (SD = 0.50; 1.1% of range) under the conservative assumption (ρ = 0.50). For likelihood of fire-mitigation activities (5-point scale, usable range = 4 points): the MDE is 0.030 scale points (SD = 0.50; 0.8% of the usable scale range) under ρ = 0.90, and 0.067 scale points (SD = 0.50; 1.7% of range) under ρ = 0.50.