Understanding how health care utilization responds to cost-sharing incentives is of central importance for providing high quality care and limiting the growth of costs. While there is compelling evidence that patients react to financial incentives, it is less well understood how and why specific aspects of the design of contracts shape the size of this reaction. In this paper, we focus on the question whether the framing of cost-sharing incentives has an effect on health care utilization. To study this, we make use of a policy change that occurred in the Netherlands. Until 2007, patients received a no-claim refund if they consumed little or no health care; from 2008 onward there was a deductible instead. This means that very similar economic incentives were first framed in terms of smaller gains and later as losses. We use claims-level data for a broad sample from the Dutch population to estimate whether the reaction to economic incentives was affected by this. Our empirical approach is to exploit within-year variation using an instrumental variables approach while controlling for differences across years. Our central finding is that patients react to incentives much more strongly when they are framed in terms of losses. Simulations based on our estimates show that the effect on yearly spending is 8.6 percent. This suggests that discussions on the optimal design of cost-sharing incentives should not only involve coinsurance rates and cost-sharing limits, but also how these are presented to patients.