The social cost of carbon (SCC), a carbon price calculated from cost-benefit based integrated assessment models and used to inform some climate policies, will always be highly disputed, partly because a key model assumption, the centennial climate damage valuation function (CDF), will “always” be highly unknowable. Current disputes are highlighted here by the huge range of SCCs resulting from alternative values of key parameters like discount rates, climate sensitivity and the CDF; by the implausibility to climate scientists of a leading model’s warming projections; and by strong criticisms of mainstream CDFs by many climate economists. The claim that statistical analyses of “weather” impacts on local economies can improve centennial CDFs rests on untestable out-of-sample extrapolation. Compared to astronomy, geology and other earth sciences, prediction testing in climate science is generally harder because of Earth’s uniqueness, and the unprecedented range and speed of likely centennial climate change, but stable underlying laws make modelling based on past observations meaningful. By contrast, the added complexity of human behaviour means there are no reliable laws for modelling centennial CDFs. For this reason alone, SCCs will always be disputed. I suggest instead more use of carbon prices based on marginal abatement costs, computed on costeffective paths that achieve socially agreed, physical climate targets. Downplaying the SCC approach to carbon prices poses challenges to many economists, and a osteffectiveness approach is no panacea, but it avoids the illusion of optimality, and allows more detailed analysis of many current climate policies.