Any economic criteria for an efficient allocation of resources is based on marginal “thinking”. Such criteria can equally be applied to the evaluation of the public allocation of R&D funds. Differently from the usual evaluation schemes - mainly dichotomous - this study implements the continuous treat- ment matching approach to investigate the optimality of the modulation of public funding. With this method, the marginal treatment effects can be identified and sub-optimal amounts of public funding determined. Although we can distinguish cases of input additionality, the substitutability outcome seems to prevail also when unobserved heterogeneity is accounted for.