You are here

Forecasting exports with targeted predictors

Francisco Dias
Publication Year 
This article applies factor models to forecast monthly Portuguese exports by resorting to an international dataset covering the country’s main trading partners.We find noteworthy forecasting gains up to 12-month ahead when soft indicators for these countries are pooled and predictors are pre-selected prior to factor estimation. Resorting solely on national data and with no pre-selection of predictors yields greater forecasting accuracy when nowcasting. Hence, data from Portugal’s main trading partners is more informative to produce h-step ahead forecasts. In turn, when hard and soft data are pooled, forecast accuracy is, in general, not enhanced.
Document link 
Journal (repec) 
Economic Studies