Inter-city trade

We propose and apply a new theory-consistent algorithm, which uses disaggregated inter-city trade data to identify a pyramidic city system with central places and associated hinterlands. Because central places possess more industries than the cities in their hinterlands, and because industries, which are exclusive to central places, are more likely to export to the small, peripheral cities in the central place's hinterland, we find that aggregate exports from central places to their hinterlands are two to five times larger than predicted by gravity forces. Using a simple decomposition approach, we show that this upward bias results from aggregation along the extensive industry margin, which is why the bias is much smaller and only marginally significant if estimation is conducted in a theory-consistent way at the disaggregated industry level.