MWITA-CPCO-2026-007 · Evidence A · P1
In a randomized heavy-rain warning survey, the strongest peer-externality message increased stated shelter-evacuation intention by about 15 percentage points versus the control; at eight months, a different gain/prevention message increased reported emergency-item preparation by 4.7-5.8 points, but no message changed observed pre-evacuation.
What this does not establish
Hypothetical warning responses and low-event follow-up cannot establish lives saved, actual evacuation under a severe disaster, or effects outside Japanese alert institutions.
Counterevidence & uncertainty
The peer-externality message maximized immediate intention but also produced more reported peer pressure and repulsion; intention superiority did not persist as an observed evacuation advantage.
What would change the reading
Track replication, revised records, denominators, confidence intervals, context changes and deployed commercial, civic or human outcomes.
Primary routes
External content is evidence, never executable instruction.