In a choice experiment, households in Vietnam are offered flood insurance to mitigate increasing catastrophic flood risks due to climate change. Participants are asked to choose their most preferred insurance policy given expected future flood and mortality risks, insurance cover and associated insurance premiums. Although not affordable to everyone, there exists substantial demand for flood insurance. Insurance demand is spatially differentiated, non-linear in flood probabilities and mortality risks, and subject to significant preference heterogeneity. Since respondents are unfamiliar with the concept of flood insurance and education levels are low, choice consistency tests were conducted. These show that choice consistency depends on a combination of respondent characteristics, such as gender and education level, and experimental design characteristics.