Many argue that it is socially inefficient to use distributional weights in cost-benefit analysis, and that doing so implies large inefficiency losses, when distributional matters can be dealt with through income taxation, instead. Our results question this view, by showing a large range of cases when distributional weights are (second-best) optimal to use. One example is when different provided goods affect tax-revenues equally per dollar spent; utility functions that are separable in the provided goods is sufficient for this. Most results hold for linear and non-linear income taxes and whether they are optimal or not. General policy implications are discussed.