Natural experiments provide explicit and robust identifying assumptions for the estimation of treatment effects. Yet their use for policy design is often limited by the difficulty in extrapolating on the basis of reduced-form estimates of policy effects. On the contrary, structural models allow us to conduct ex ante analysis of alternative policy situations. However, their internal validity is often questioned. In this paper, we suggest combining the two approaches by putting structure on a regression discontinuity (RD) design. The RD estimation exploits the fact that childless single individuals under 25 years of age are not eligible for social assistance in France. The behavioral model is identified by the discontinuity and by an additional exclusion restriction on the form of financial incentives to work. We investigate the performance of the behavioral model for predictions further away from the threshold, check external validity and use the model to predict important counterfactual policies, including the extension of social assistance to young people and the role of in-work benefit components.