A method to integrate strategic alignment in freight transportation behavioral models

Companies use high-level strategies to guide their decision-making and maintain strategic alignment in their actions. For example, companies may adopt a strategy of providing excellent customer service and own a private truck fleet, giving the company complete control over delivery. Despite its relevance, the concept of strategic alignment is a major omission in existing freight transportation models. In this study, we develop a methodology to integrate strategic alignment into agent-based, freight transportation models. We first identify a suitable modification to the typical agent-based structure, then outline a conceptual model relating strategy to strategic decisions. We develop a mathematical formulation to operationalize the conceptual model by introducing latent variables, which represent strategies, into the Seemingly Unrelated Regression (SUR) formulation, permitting a mix of continuous and Tobit equations. The new method is named SURTLV (Seemingly Unrelated Regression of Tobit Equations with Latent Variables). Our methodology offers many powerful features for forecasting. Binary, continuous, and contingent decisions are modeled. Choice set generation parameters are modeled as strategic decisions. Strategic decisions are modeled jointly, which acknowledges their interrelationships. Bayesian estimation with Gibbs sampling supports rich model specifications. In an empirical demonstration, we apply SURTLV to simulate a nationwide network of distribution centers and private fleets using real-world data of Fortune 500 companies. Our latent strategy measurement data come from parallel work, featuring the first real-world implementation of a novel, Natural Language Processing-based measurement generation method.