Summary Protected areas may impose local welfare costs through the enforcement of use restrictions. Predicting their welfare impacts before their establishment could help with the design of compensation schemes. Discrete choice experiments (DCEs) are increasingly used for ex ante evaluations but their validity is largely untested in low-income settings. Using a case study of a new REDD+ (Reducing Emissions from Deforestation and forest Degradation) project in eastern Madagascar, we explore the validity of DCEs in two ways: (i) whether the estimates of welfare costs derived from DCE are affected by respondents’ prior experience of conservation (ii) whether DCE results have high theoretical and content validity. We surveyed households who have varying degrees of experience of restrictions to swidden agriculture. We also qualitatively debriefed a sub-sample of respondents to better understand their thought processes. Latent class analysis shows that DCE outcomes vary with conservation experience. Households more experienced with forest protection are less willing to trade-off rights to clear forest for swidden agriculture with any compensatory interventions whereas less experienced households highly favor support for alternative agricultural techniques and a secure right to clear one hectare of forest. Although the results show apparent non-attendance to some attributes (e.g., cash payments), qualitative debriefings suggest that respondents infact do expect relatively low or no utility from the given attributes and hence have theoretically valid preferences. Similarly, the DCE has generally high content validity. Although DCE can elicit current preferences in this context, using ex ante DCE to estimate the welfare costs of such a long-term intervention requires caution. We conclude that it is difficult to robustly estimate compensation in advance of an intervention, there is therefore a need to rethink conservation approaches, and the feasibility of achieving fair compensations for conservation-imposed restrictions.