This paper separates the moral hazard and selection effects of a voluntary deductible scheme using Dutch individual claims data of all inhabitants in the Netherlands for the period 2008-2013. In the Netherlands,(basic) health insurance is mandatory and features a mandatory deductible. On top of the mandatory deductible, people can choose a voluntary deductible. Healthcare expenditure for people with a voluntary deductible is lower than for people with (only) a mandatory deductible. This could be due to two effects. First, a selection effect: lowrisk individuals with low expected healthcare expenditures are more likely to opt for voluntary cost-sharing than high-risk individuals with high expected expenditures (Cutler and Zeckhauser, 2000). Second, a moral hazard effect: higher cost-sharing increases the price of healthcare, and leads to lower healthcare demand (Zweifel and Manning, 2000). With a panel regression discontinuity design we separate selection and moral hazard effects. The motivation for our paper is the public debate in the Netherlands on whether the voluntary deductible should be abolished or not. Currently, the Dutch have a mandatory deductible of 385 euros for all individuals over 18 years old. On top of this mandatory deductible, each individual above 18 years old can opt for a voluntary deductible of either 100, 200, 300, 400 or 500 euros. In return, they receive a discount on their health insurance premium. Below 18, people face neither a mandatory nor a voluntary deductible. Three questions play a central role in the debate. First, to what extent are persons with and without a voluntary deductible different? We answer this question by disentangling moral hazard and selection effects. Second, how much will healthcare spending increase if the voluntary deductible is abolished? This boils down to the extent to which a voluntary deductible reduces healthcare expenditures. If this moral hazard effect is substantial, abolishing voluntary deductibles may lead to a large increase in healthcare expenditures. Third, if there is (adverse) selection resulting from the voluntary deductible: does it distort health insurance prices? Dutch policymakers worry that high-risk individuals, such as the chronically ill, pay a higher price for their insurance because of this selection effect of the voluntary deductible. The argument goes that high-risk individuals, for whom a voluntary deductible is unprofitable, partly finance the discount in premiums that low risk individuals with a voluntary deductible receive. It is empirically challenging to disentangle selection and moral hazard effects (Bajari et al., 2014; Trottmann et al., 2012; Geruso and Layton, 2017). Randomized control experiments, such as the RAND and Oregon experiment (Newhouse, 1993; Finkelstein et al., 2012; Chiappori et al., 1998) are less suited as the idea of randomization is to remove endogenous choice, and thus selection, problems. That is, in such an experimental set-up there is no selection effect to