We investigate whether a profit-maximizing insurer with the opportunity to modify the loss probability will engage in loss prevention or instead spend effort to increase the loss probability. First we study this question within a traditional expected utility framework; then we apply Koszegi and Rabin's (2006, 2007) loss aversion model to account for reference-dependence in consumer preferences. Largely independent of the adopted framework, we find that the profit-maximizing loss probability for many commonly used parameterizations is close to 1/2. So in cases where the initial loss probability is low, insurers will have an incentive to increase it. This qualifies appeals to grant insurers market power to incentivize them to engage in loss prevention.