Surely a starting point for assessing the accuracy of such forecasts as these would be the predictor's forecasts for the last, say, 20 years? Accepting (relatively) short term variances, eg 2002/3, have they predicted the overall trend correctly or not?