The attitude towards flood insurance purchase when respondents' preferences are uncertain: a fuzzy approach

Abstract

Individuals may have difficulty in determining whether or not to buy insurance against low‐probability, high‐loss events. This ambivalence would cause preference uncertainty and decrease homeowners' interest in voluntarily buying insurance. This paper incorporated fuzzy set theory into contingent valuation analysis to examine the determinants of attitude towards buying flood insurance under preference uncertainty. The results show that the perceived levels of flood risk, experience with flood, disposable income, as well as house conditions, are influential factors in the decision‐making process for insurance purchase. However, both the estimated price and income elasticities are low for flood insurance purchase. It is worthy to note that governments' artificial structures provide a disincentive for buying insurance, although respondents perceived or/and are exposed to a high level of flood risk. The findings also show that the spread of fuzzy willingness to pay regions is wide, resulting from respondents' high uncertainty on their value judgment to insurance. This indicates that preference uncertainty and conservatism rule are the key factors that cause respondents to tend to reject buying insurance.

Keywords:

• flood insurance
• fuzzy contingent valuation
• fuzzy willingness to pay
• preference uncertainty
• flood risk

Notes

1. Ehrlich and Becker (1972) defined self‐insurance as the actions to reduce the size of potential loss.

2. The elasticity associated with each of the continuous variable x_j is calculated as: (∂P/∂x_j )×(x_j /p), where P is the probability for the yes answer, and x_j and P are measured at their means.

3. A high value of h could cause the very wide spread of the estimated output ˜Y_i . For instance, the outputs obtained at h = 0.5 are twice larger than the estimates obtained at h = 0.

4. In Model 1, the mean WTP was estimated at the mean of respondent's WTP. In Model 2, a non‐negative mean WTP was calculated by: WTP_mean = ln(1+expα)/β, where α is the intercept and β is bid coefficient in the binominal logit analysis (Hanemann 1989).