Despite advances in statistical modeling and estimation, policy modeling and infrastructure impact assessments in the transportation area are almost always conducted by assuming linearity. The focus is on the assessment of location premiums associated with accessibility variables that are key policy variables, estimated via hedonic methods. These variables are invaluable for assessing effects of transportation projects on nearby land uses. Most hedonic studies hypothesize a parametric functional relation between accessibility and the dependent variable in estimating premiums. Unfortunately, common parametric approaches cannot control for functional form misspecification of individual variables. That line of research is extended to investigate and explore these issues with a flexible modeling strategy using a cross-section sample of housing sales. The analysis indicates the presence of significant nonlinearities that have implications for impact assessment approaches and for the assessment of accessibility influences on price premiums. The methods suggested can be extended to other areas as well, areas in which theory does not provide any a priori bounds on behavior of variables included in an analysis.
Abstract