Freight forecasting models are data intensive and require many explanatory variables to be accurate. One problem, particularly in the United States, is that public data sources are mostly at highly aggregate geographic levels, while models with more disaggregate geographic levels are required for regional freight transportation planning. Second, supply chain effects are often ignored or modeled with economic input-output models which lack explanatory power. This study addresses these challenges by considering a structural equation modeling approach, which is not confined to a specific spatial structure as spatial regression models would be, and allows for correlations between commodities. A FAF-based structural commodity generation model is specified and estimated and shown to provide a better fit to the data than independent regression models for each commodity. Three features of the model are discussed: indirect effects, supply chain elasticity, and intrazonal supply-demand interactions. A validation of the geographic scalability of the model is conducted using data imputed with a goal programming method.