Natural disasters may increase in frequency and severity in the future as a result of climate change, which is likely to have an impact on the demand for natural disaster insurance. Insights about individual risk beliefs and behavioural responses to changing risks are relevant for insurers, as it allows them, for example, to estimate the demand for new insurance products that cover weather-related damage. This study elicits individual risk beliefs and the demand for low-probability, high-impact flood insurance using the contingent valuation survey method among approximately 1000 homeowners in the Dutch river delta. This study is of practical relevance since currently flood insurance is not available in the Netherlands, while insurers have been considering to provide such insurance. Individuals generally do not behave in accordance with the expected utility model since a significant proportion of homeowners neglect the low-probability flood risk. The willingness-to-pay (WTP) of those individuals who demand flood insurance is on average considerably higher than the expected value of the flood risk they face. Moreover, the WTP for flood insurance is less than proportionally related to increased flood probabilities that were presented to respondents in the questionnaire. Individuals follow a process of Bayesian updating of flood probabilities, since perceptions of flood risk are an important determinant of the WTP, while objective risks derived from geographical characteristics influence the WTP to a lesser extent. Communication of baseline probabilities and changes in flood probabilities using risk ladders facilitate the comprehension of risk by respondents, and has a considerable effect on the level of the WTP and its sensitivity to probability changes. The results indicate that the current ex post public compensation scheme of flood damage lowers demand for private insurance. © 2012 Elsevier B.V.