Direct vs Indirect Forecasts of Foreign Trade Unit Value Indices

This paper examines the forecasting approach of foreign trade unit value indices followed in the compilation of quarterly national accounts of Italy. Total imports and exports indices are indirectly obtained from the aggregation of ARIMA forecasts of disaggregated components, derived from the program TRAMO with automatic identification options. An out-of-sample forecasting exercise is performed to validate the automatic choices made by TRAMO and to evaluate the relative performance of a direct forecasting approach of imports and exports aggregates. Also, we show how the use of international raw commodity prices can improve the forecasting accuracy of aggregate unit value indices.