Sociodemographic matters: Analyzing interactions of individuals' characteristics with walkability when modelling walking behavior

The concept of walkability encompasses a multitude of features of the built and social environments that impact walking behavior, leading to the creation of a wide range of walkability indices. While past studies have compared the research outputs from multiple walkability indices, little research has looked at how relevant each index is across sociodemographic groups and how the relationship between sociodemographic characteristics and behavior vary from one index to another. Using trip data from Montréal, Canada, this paper evaluates two commonly used walkability indices at different scales – Walk Score© (meso-scale) and the MAPS-Mini audit tool (micro-scale) – and their interactions with sociodemographic characteristics. Weighted binary logistic regressions are used to model the probability of adults taking a homebased utilitarian walking trip. A total of ten models are generated, including eight models in which one of four socio-demographic variables – gender, age, household income, and presence of children below 13 years old in the household – is interacted with one of the two walkability indices. Sensitivity analyses are then conducted using the interaction variables' outputs by varying the values of the interacted sociodemographic variable. Results show significant interactions for all variables for both indices except from Walk Score©’s interaction with gender. Opposite effects are observed between the two indices in the sensitivity analysis for household income. The differential results observed between the indices and between sociodemographic groups underscore the need to properly test the equity implication of using certain walkability measures in research.