All public policies have two things in common. They deal with the future and, as a result, they are based on forecasts or projections. The forecasts or projections may be implicit or based on naive extrapolation or ad hoc assumptions. They may be explicit and based on elaborate extrapolations or on behavioral models. In either case, unfortunately, they are notoriously unreliable. In fact, they almost always are wrong-sometimes just a bit wrong, but often massively wrong. Nonetheless, forecasts are what distinguishes reasoned planning from blind action. Without forecasts, we would be totally at sea. That we have to use forecasts or projections, that we know they will be wrong, and that they usually are wrong raise some difficult questions for policy analysis and policymaking. Regrettably, in my view, they receive too little attention. 2 My purpose today is to urge that they receive more.
My comments are intended to make four points. First, it is important for policymakers to appreciate how error-prone forecasts and projections actually are. Second, it is important not to permit the availability of projections or forecasts to obscure fundamental policy questions that are important in any plausible scenario. Third, uncertainty means that, where possible, it is prudent to design policies with built-in flexibility that respond automatically to diverse possible outcomes. Fourth, where built-in flexibility is impossible, complete analyses should take into account the consequences if forecasts prove wrong, and weigh those consequences against the results of postponing action until information improves or against other policies under the plausible range of possible outcomes.