Quantifying the Self-Selection Effects in Residential Location Choice with a Structural Equation Model

Land use policy is viewed as a way to mitigate congestion and alleviate greenhouse gas emissions. Many studies have confirmed the reduction effect for vehicle miles traveled (VMT) of compact and mixed land use development. However, debates on self-selection effects have arisen in recent years. Researchers argue that the correlation between land use pattern and VMT may be caused by self-selection of residential location based on a person's attitude toward traveling. This paper develops an urban form indicator for describing local land use patterns and then establishes a structural equation model (SEM) with VMT, vehicle ownership, and the urban form of residential location estimated simultaneously. Residential self-selection was controlled in two ways in the model: implicitly through the correlated error terms in multiple equations and explicitly through the incorporation of expected VMT in the equation of residential location choice. The model was estimated with household travel survey data collected in 2007 in the Washington, D.C., area. The results showed that land use itself could influence travel behavior after self-selection effects were removed. A comparison of the results of the full SEM results with that of a reference SEM confirmed the existence of self-selection effects. The comparison verified that the VMT reduction effect of land use would be exaggerated without consideration of self-selection, and self-selection effect accounted for a larger part of the total effect in more compact and better mixed development areas. However, the self-selection effect was small compared with the effect of land use itself in the analyzed case.