If learning-by-doing is important for macro-forecasting, newcomers might be different to regular, established participants. Stayers may also differ from the soon-to-leave. We test these conjectures for macro-forecasters point predictions of output growth and inflation, and for their histogram forecasts. A bootstrap approach is used to overcome the problems associated with the relatively small numbers of joiners and leavers. Controlling for the numbers of forecasters with the bootstrap approach is required to correctly determine whether there are systematic differences between experienced forecasters and newcomers, and between stayers and leavers.
Forecast accuracy, experience, learning-by-doing, probability forecasts