The effect of cycling infrastructure on commuting mode choice: a quasi-experimental evidence

This paper examines the causal impact of cycling infrastructure on commuting behavior using a quasi-experimental research design. Drawing on four waves of the Canadian Census of Population and National Household Survey data from 2001 to 2016, we assess changes in cycling to work in Toronto and Vancouver, two major metropolitan areas that differ in climate, demographics, cycling culture, policy and investment toward cycling infrastructure. We construct a panel of individual commuters with origin and destination (O-D) information at the census tract (CT) level and develop five measures of cycling infrastructure exposure that incorporate commuters' home to work activity space into infrastructure usage. Our difference-in-differences design leverages variation in infrastructure expansion across O-D pairs to estimate the effect of infrastructure construction on cycling mode choice. In a pooled sample of both cities our results show that a 1 km increase in cycling infrastructure raises the odds of cycling to work by 1.6 to 4.1%. An analysis of heterogeneity of the effects across socioeconomic groups and trip characteristics reveals that individuals aged 15 to 35, those living within 5 km of their workplace, individuals with no formal education, those earning below the median income, and recent immigrants (i.e., those living outside Canada prior to the census wave) are the most responsive to improvements in cycling infrastructure.