Spatial interaction modeling of interregional commodity flows

Understanding the determinants of interregional commodity flows is critical for both transportation infrastructure planning (highways, railroad tracks, river/port facilities) and regional development policies (location of activities, reducing regional disparities). Unfortunately, limited data availability has, in the past, hindered empirical research in this area. Drawing from both the spatial price equilibrium theoretical framework and the empirical literature on spatial interaction modeling and international trade, this paper expands past spatial interaction models of commodity flows by incorporating new variables into the model, using a flexible Box-Cox functional form, and applying the analysis to all manufacturing commodities. The recently released 1993 U.S. Commodity Flows Survey provides the empirical basis for estimating state-to-state flow models for 16 commodity groups over the 48 continental U.S. states. Based on input-output considerations and in order to differentiate intermediate from final commodity demands, the new variables include more detailed descriptions of the economies of the origin and destination states, such as employment and value added for the commodity sector at the origin state, wholesale employment at both ends, manufacturing employment at the destination state, and population and per-capita income at both ends. In addition, the average establishment size for the commodity at the origin is intended to measure scale or diversification effects. The competitive or agglomerative effects of the economic spatial structure are captured with competing destination and intervening opportunities variables. In addition to the average hauling distance between states, the model includes dummy variables measuring whether (1) having a common physical border, and (2) the origin or destination states being custom districts, have an effect on flows. Overall, the results show that the selected variables and functional form are very successful in explaining flow variations. The optimized Box-Cox specification proves to be superior to the log-log one in all cases. The results include the following findings: (1) the distance effect is negative and highly significant, with bulkier products hauled over shorter distances; (2) the adjacency effect is significant, with neighboring states trading more with one another, even after accounting for distance; (3) the impact of the spatial structure is of the competitive type in most cases; (4) the effects of imports and exports are significant for specific commodities; (5) wholesale activities at both origins and destinations are important facilitators of commodity flows; (6) except in one case (furniture), flows increase with product diversification; and (7) the role of intermediate and final demands for the commodities are clearly reflected by the selected employment, population and per-capita income variables. Various areas for further research are outlined.