This article estimates the demand for forest insurance of French private forest owners against fire risk. We combine experimental and real world data and use a Heckman selection model to separately analyse insurance participation and coverage-level decisions. Our results showthat some determinants explain either participation to insurance (e.g., ambiguity about the probability of occurrence of natural hazards) or coverage-level decision (e.g., the past perception of a public help in real life), some of them do not allow to explain the owner’s insurance behaviour (e.g., having ever suffered from a fire in the past) while the other explain both the two behaviours (e.g., a contingent fixed help program).