Reconceiving the Relationship Between the Built Environment and Walking Behavior: Examining the Samples, Scales, and Methods in Travel Behavior Research

Abstract

This dissertation aims to formulate a mechanism for the relationship between the urban form and walking choice that can be consistent across contexts. The motivation is the lack of concordant results in the magnitude of the environmental influence on walking choice in urban areas found in the literature. The dissertation identifies a series of limitations in previous research that could cause mixed results in the magnitudes of the association. This research elaborates an approach to overcome these limitations by proposing a mechanism of the activity density over walking modal share by controlling for trip distance distribution. The aim is an approach that can achieve consistent results across applications and help define standard planning policy guidelines. The work is structured as a three-paper dissertation. In the first paper, this research examines how travel choices vary across different urban contexts, challenging the assumptions of spatial homogeneity in the relationship between travel modes and the built environment. The study, using data from Portland, OR, uses an innovative method to spatially segment behavioral models based on the urban environment attributes associated with a spatial grid (cells). Each cell will be related to a combination of different traveler classes. These classes are latent as it is not observable by the analyst. Using the Latent Class Discrete Choice method, the study inferred the class from the travel data. The method assigns a probability of having each class of travelers to each cell (making the final behavior a probability-weighted combination). This probability is based on local accessibility measures (such as the number of people reachable at a certain walking distance) and regional accessibility measures (such as employment reachable by transit or car). Each traveler class is represented with a multinomial choice model based on their socioeconomic attributes and the trip distance. The article's contribution shows that segmentation of the travel choice increases the model's goodness of fit and offers a more realistic representation of the traveler and a more precise relationship. Moreover, local and regional accessibility interaction appears relevant in travel choices and should be included in planning policy. The second paper investigates why previous literature could have found variable magnitudes in the relationship between the built environment and walking. The contribution of the article identifies possible sources of inconsistencies in previous works. Specifically, not controlling for distance in the relationship specification and the modifiable areal unit problem (MAUP) can contribute to inconsistent results across research studies