Evaluating the rationality of fixed-event forecasts

A test of forecast rationality based on the weak efficiency of fixed-event forecasts was proposed by Nordhaus. This paper considers the scope for pooling fixed-event forecasts across ‘events’, to deliver more powerful tests of the weak-efficiency hypothesis, when only a small number of fixed-event forecasts are available. In an empirical illustration we demonstrate the usefulness of this approach. We also suggest an interpretation of the rejection of the null hypothesis of weak efficiency in favour of negative autocorrelation in series of revisions to fixed-event forecasts. The relationship between weak efficiency and rationality when loss functions are asymmetric and prediction error variances are time-varying is also considered.