The importance of taking into account multidimensionality in poverty measurement has been recently emphasized. The poverty alleviation literature has not, however, yet addressed the important issue of policy design for efficient multidimensional poverty reduction. From a positive perspective, it is regularly observed that different poverty dimensions are often correlated and mutually reinforced, especially over time. From a normative perspective, it can be argued that, in addition to being concerned with impacts on multiple dimensions of poverty, policy should also consider impacts on their joint distribution. The paper integrates these two perspectives into a consistent policy evaluation framework. Targeting dominance techniques are also proposed to assess the normative robustness of targeting strategies. The analytical results are applied to data from Vietnam and South Africa and illustrate the role of both normative and positive perspectives in designing efficient multidimensional poverty targeting policies.