Purpose: To strengthen our understanding of the impact of baseline variability in mode choice on the likelihood of travel behaviour change. Methods: Quasi-experimental analyses in a cohort study of 450 commuters exposed to a new guided busway with a path for walking and cycling in Cambridge, UK. Exposure to the intervention was defined using the shortest network distance from each participant's home to the busway. Variability in commuter travel behaviour at baseline was defined using the Herfindahl-Hirschman Index, the number of different modes of transport used over a week, and the proportion of trips made by the main (combination of) mode(s). The outcomes were changes in the share of commute trips (i) involving any active travel, (ii) involving any public transport, and (iii) made entirely by car. Variability and change data were derived from a self-reported seven-day record collected before (2009) and after (2012) the intervention. Separate multinomial regression models were estimated to assess the influence of baseline variability on behaviour change, both independently and as an interaction effect with exposure to the intervention. Results: All three measures of variability predicted changes in mode share in most models. The effect size for the intervention was slightly strengthened after including variability. Commuters with higher baseline variability were more likely to increase their active mode share (e.g. for HHI: relative risk ratio [RRR] for interaction 3.34, 95% CI 1.41, 7.89) and decrease their car mode share in response to the intervention (e.g. for HHI: RRR 7.50, 95% CI 2.52, 22.34). Conclusions: People reporting a higher level of variability in mode choice were more likely to change their travel behaviour following an intervention. Future research should consider such variability as a potential predictor and effect modifier of travel and physical activity behaviour change, and its significance for the design and targeting of interventions.@en

Despite a large body of research suggesting that the built environment influences individual travel behavior, uncertainty remains about the true nature, size, and strength of any causal relationships between the built environment and travel behavior. Residential self-selection, the phenomenon whereby individuals or households select a residential area based on their transport attitudes, is a frequently proposed alternative explanation for the reported associations. To resolve the issue of residential self-selection, longitudinal studies are often recommended. In this paper, we argue that intervention study designs are insufficient to fully resolve the problem and that intervention studies on the built environment and travel behavior may still be biased by residential self-selection. The aim of this paper is to extend existing conceptualizations of the relationships between the built environment, travel behavior, and attitudes and to provide suggestions for how a causal relationship between the built environment and travel behavior may be ascertained with more accurate estimates of effect sizes. We discuss the complexities of determining causal effects in intervention studies with participants who relocate, and the biases that may occur. We illustrate the complexities by presenting extended conceptualizations. Based on these conceptualizations, we provide considerations for future research. We suggest repeating analyses with and without individuals who relocated during the study, and with and without statistical controls for residential relocation. Additional quantitative and qualitative analyses will be necessary to obtain more accurate effect size estimates and a better understanding of the causal relationships.@en