The purpose of this paper is to compare the cost efficiency of private and public property insurance providers in Switzerland. The most commonly used measure for this kind of exercise is the claims-premium ratio. We argue that this measure may give strongly biased results. We develop a simple model to test whether the elasticity of premiums with respect to claims is less than unity. We address the fact that premium income is relatively stable across time, while claims are not, using estimation techniques that correct for measurement error. We develop tools to cope with heteroskedasticity in such measurement errors and apply the model to a data set on 19 firms in housing insurance markets in Switzerland. We show that the public insurance providers are about 20% more cost efficient than their private counterparts.