Uncertainty in econometrics: evaluating policy counterfactuals

There is without question a great deal of uncertainty in planning and policy formation. Our starting point in approaching this uncertainty is that ‘Counterfactuals are the very guide of life’.1 For rational planning we need counterfactuals because we need to evaluate what would happen were each of the actions under consideration implemented. For this kind of deliberation there is always a vast amount we do not know – facts, laws, probabilities, people’s reactions etc. This essay will focus on a more abstract source of ignorance that contributes to uncertainty in planning: We do not know how in principle to evaluate counterfactuals. There is a considerable literature in philosophy on the semantics of counterfactuals and recently contributions in economics itself. The bulk of this literature, we shall argue, puts the problems back to front. It attempts to use counterfactuals to evaluate causal claims; we argue instead that one must use causal claims to assess counterfactuals. That said, there is as yet no general account of how to do so.