All partners logo eng

Natural Hazards and Moral Hazards: Understanding the Insurance Coverage Limit

Daniel Felsenstein and Masha Vernick


This research tests the moral hazard hypothesis in the insurance market for natural hazards. This states that insurance coverage does not reflect the distribution of natural hazards due to households tending to under-insure as the liability of risk is likely to be borne by others. We use insurance portfolio data (n=~12,000) from a large private insurance company linked to asset (dwelling unit) data from Tax Authority records for the Haifa. We control for housing attributes including price, size, year built, distance to hazards (natural and anthropogenic) and local socio-econ attributes include income and crime. We use spatial econometrics estimating a SAR (spatial autoregression) model to understand the effect of exposure to hazards on maximum insurance coverage (structure and content). Our estimation strategy accounts for selection bias in data (using Heckman procedure), spurious spatial relationships (residuals testing) and issues of identification (using SUR- seemingly unrelated regression). The findings differentiate between structure and content insurance. The former is directly related to dwelling unit attributes such as size, price, number of floors and other house prices in the vicinity. In terms of hazards we find a positive relationship to local crime rates and distance to industry and an inverse relationship to distance to forests. No relationship is established with distance to the centers of simulated earthquakes of different magnitudes. These findings support the hypothesis with respect to the existence to a moral hazard in relation to earthquakes. Policy implications are suggested.

Final report