In recent years there has been a growing recognition that forecasts of the most likely outcome of the future value of a variable may often be of limited value. Michael P. Clements describes methods of evaluating forecasts which provide a fuller description of the range of possible future outcomes.