Bottom Time Series Research Articles

A Hierarchical Forecasting Model for China’s Foreign Trade

Analyzing the underlying characteristics of trade values movements has attracted much attention in the domestic research. However, the proposed understanding of these characteristics is limited by the intrinsic complexity of the imports/exports. Since economic systems are naturally organized by hierarchies, a novel hierarchical model is proposed in this paper to forecast China’s foreign trade. First, the foreign trade data are disaggregated from perspectives of trading partners and trading products, forming two independent hierarchical models with total exports and imports as target variables. Second, a bottom-up strategy is applied. All bottom time series are modelled by corresponding control variables according to trading theories. Forecasts for bottom time series are then combined to generate initial forecasts for total exports and imports. Finally, forecasts for total imports and exports from the two hierarchical models, plus a single VECM model are combined to generate final forecasts. Empirical experiments show that this proposed forecasting model approach significantly outperforms benchmark models and produces consistent forecasts for both total imports and exports or detailed items, which helps a lot for analyzing future trading structure evolution and making foreign trade policies.