Modeling Firm Transportation Strategy using Big Text Data

Agent-based freight models are used to simulate a individual agents and their attributes, behaviors, and environments. The population in these models typically comprises business establishments or firms. Due to lack of data, attributes are often limited to readily available traits such as firm size and industry category. However, these basic attributes, while important, yield little insight into the underlying behavioral drivers of firm decisions. This research addresses this data gap by building a data development engine (DDE). The DDE extracts key company data from big, online, text-based information systems and builds a dataset of companies to use in model estimation. The primary, initial focus of the DDE is to develop data regarding strategies that are adopted by companies and used to guide company decisions. Earlier work has shown that including firm strategies in an agent-based freight model is important for estimating freight agent behavior and, ultimately, impacts on energy use and vehicle-miles traveled. Factor Analysis and Structural Equation models are used to identify firm “emphasis areas” and a latent variable that weighs the relative importance of two emphasis areas (service vs. product innovations). The resulting strategic data is used to inform a model of private fleet ownership.