International trade is the outcome of ongoing interaction between individual buyers and sellers. Official statistics have historically revealed only the aggregate trade flows that their activity generates. Quite naturally, trade theory sought to model these aggregates with little attention to the role of individual interactions, which were hidden from view. A limitation is that research was unable to delve into the underlying reasons for such phenomena as the decline of trade with distance or the sluggish response of trade to changes in international prices.
This research exploits newly available data on individual international transactions, from the United States, Colombia, and China, to study the trading relationships among individual exporters and importers. The project has four components. The first is the construction of panel datasets linking domestic buyers and foreign sellers through their international transactions. Each commercial shipment generates a customs record that reveals the identity of both the importing and exporting firm. In addition, the United States, Colombia and China all collect extensive firm- or plant-level panel data on their domestic producers. By merging China-U.S. shipments and Colombia-U.S. shipments with these producer-level surveys, the researchers on this project will create large panel data sets that follow importer-exporter relationships through time, with information on the current state of both parties to each transaction. These data sets will permit very detailed analyses of a wide range of issues concerning trade dynamics and firm-level performance. They will be made available through the Census Bureau to all researchers with the proper security clearance, and they will provide the basis for further analysis within the remainder of this project.
The next component of the project will examine the quality of trade statistics generated from the shipments data. Inconsistency between exports reported by one country and the corresponding imports reported by another has plagued the analysis of international trade. For example, what Colombia reports exporting to the United States in 2007 differs by 8 percent from what the United States reports importing from Colombia. These datasets provide a tool to diagnose the discrepancies. Combined, the shipment records of the exporting and importing country reveal, for example, whether discrepancies arise in the aggregate statistics because they are based on different sets of shipments or different valuations placed on the same shipments. This component of the research should prove useful to the construction of trade statistics.
The third component of the project will describe buyer-seller interactions. These datasets connect firms to each other through their trading relationships. Network theory, which provides various metrics for describing the potentially very complex connections among members of a network, is a natural means of interpreting these connections. The metrics thus provide insight, for example, into the extent to which a single widely-connected buyer or seller contributes to international trade. The matched datasets will allow the observation of the propensity with which different types of firms form links with each other. Copula theory provides a natural means of describing the matching behavior of different types of firms. Preliminary evidence suggests, for example, little tendency for large firms and small firms to differ in the extent to which they match with other large or small firms. In terms of size, in other words, matching is neither positively nor negatively assortative. This component of the research provides insight, for example, into the vulnerability of trade to the potential collapse of a single entity and the extent to which growth in trade is the consequence of new trading relationships or expansion of trade within existing ones.
The final component of the project will use the data to guide the design of economic models of international trade. Recognizing the discrete nature of individual transactions and the behavior of individual firms involved requires a radical reformulation of international trade theory. The dynamics of network formation can be described in terms of a paired set of generation-recombination processes. A generic model that is fairly devoid of specific assumptions about behavior can thus be applied to examine features that a wide class of economic models must satisfy to explain basic features of the data. The research will then turn to specific formulations of firm behavior that connect trading relationships to other decisions, such as hiring and investment. This component of the research should prove useful to the development of a measurement-based dynamic theory of international trade that reflects the interaction of individual firms.
