Development of a Freight Generation Model through Linear Regression: An Application to California

The growth in demand for freight transportation prompts the need to develop better tools to evaluate and analyze the goods movement system. Tools such as a freight generation model can improve decision-making through the analysis of the types and quantity of goods moved from one point to another. This thesis' model building process involves the disaggregation of the Federal Highway Administration's Freight Analysis Framework database on freight origin-destination data and the development of linear regression equations to describe the relationships between commodity outputs (productions/attractions) to specific economic variables. Several freight generation models are presented and "validated" for California with actual 2007-year data and applied to predict 2015 commodity outputs. The models are distinguished from one another by the different groupings of commodities that are evaluated. Instead of generating a production/attraction equation for each commodity, the grouping of commodities can simplify model development and application. Commodity equations with high R2 values were more likely to generate outputs closer to the actual 2007 data in the calibration process. Because model output results are highly dependent on grouping configurations, this study does not provide a recommendation on which model is "better". Due to economic and geographic differences among the nation's 50 states, one set of production/attraction equations cannot accurately predict tonnage outputs for every region. This study can be used as a guideline for city, county, metropolitan and state level planning agencies to develop their own customized freight generation model.