Abstract: Based on theories of regional interactions and competition and the gravity model, this paper first develops a sector-specific spatial statistic model to estimate inter-regional trade flows by employing a geographically weighted regression technique. The model takes into consideration sector-specific input-output relationships. That is, in some sectors there exists strong competition between regions while other sectors may need close inter-regional cooperation in terms of supply linkages. The former case results in less inter-regional trade but the latter witnesses more trade. The model also introduces the spatial lag factor of trade flows between regions. Then, the paper applies the model to estimate inter-provincial trade flows of three sample sectors, i.e., agriculture, chemistry and electronics, with data from the 2007 provincial input-output tables of China. The computing result shows that the sector-specific model can significantly increase the reliability of inter-regional trade flow estimation. It also reveals that the bandwidth of weighting function is a key factor in the sector-specific model; that is, the smaller the bandwidth, the more the trade flows. To a certain degree, the bandwidth reflects the degree of geographical concentration of economic activities while the bandwidth itself is different from sector to sector. Different sectors display different features of inter-regional trade flows. For example, agricultural trade flows are mainly from the inland provinces to the coastal ones and show strong intra-sector competition while trade flows of chemical and electronic sectors take place mainly inside the coastal regions and show an intra-sector cooperation relationship.

Key words: sector-specific spatial statistic model, gravity model, geographically weighted regression, intra-sector trade