The diversity of papers submitted to this special issue attest to the importance that researchers in forecasting attach to the various roles for human judgement in both the process of forecasting and the forecasts themselves.
Edmondson, focussing on the process, decomposes a single time series into its trend and seasonal pattern, presents this information graphically and asks the user to exercise judgement about each aspect. Recombining the judgements provides forecasts that can be at least as good as the best statistical methods, and are often better by a modest amount.
Forecasts of the UK economy made by two macroeconomic models, one from the London Business School, and the other from the National Institute for Economics and Social Research, are examined by Turner for their sensitivity to judgements made in various parts of the models. He finds substantial influence of judgement in several ways that are not obvious in the reports of the forecasts.
Two papers report studies showing how judgement can improve forecasts. Andreassen and Kraus demonstrate that focussing the forecaster's attention on changes from one time period to the next in an exponential time series can substantially improve judgement extrapolation. Wolfe and Flores show that purely statistical forecasts of the quarterly earnings per share of companies can be improved by explicitly making judgemental adjustments that take account of general economic conditions, financial data of the firm and the firm's products. Descriptive