Abstract There are many explanations why the more structured and formalised techniques of forecasting have not yet provided major input to government policy except in specialised areas. This article gives an assessment of the present state of the art in our ability to predict the consequences of current actions in the long-term future. The relevance of this to the ongoing debate about the place of formal methods in policy analysis is considered. In many instances it seems that the methods used run counter to the ideal of scientific liberalism to which the forecasters and officials involved often subscribe. The article indicates where institutional arrangements can be adjusted to ensure that the forecasting ability available is better employed, and points to areas in which forecasting methodologists should increase their attention if forecasting methods are to support more open and more flexible institutional arrangements.