AbstractNeighborhood walkability is important to planners and policy makers in the public health, land use planning, and transportation fields. This research contributes to knowledge of walking behaviour by aggregating GPS-tracked walking trips as neighborhood walking densities, and investigating local characteristics affecting those densities. The study maps walking trips in urban and suburban neighborhoods of Halifax, Canada, using data from the Space-Time Activity Research (STAR) survey conducted in 2007–2008. Respondents completed a two-day time-diary, and their movements were tracked using a GPS data logger. The 1971 primary respondents recorded 5005 geo-referenced walking trips.From mapped walking tracks, walking distances were aggregated to 87 census tracts, and expressed as walking densities (per resident, per meter of road, and per developed area). Multivariate regression was used to examine which neighborhood variables and socio-demographic controls are most useful as estimators of walking densities. Contrary to much of the walkability literature, built-environment measures of road connectivity and dwelling density were found to have little estimating power. Rather, retail lot coverage ratio was the single most useful estimator, acting as a proxy to identify traditional retail shopping streets. Office and institutional land uses were also important contributory estimators (highlighting areas of dense employment), as were measures of residents' income and age.

Highlights•Employs GPS tracking to map over 5000 urban walking trips•Creates three neighborhood walking densities for mapping and analysis•The best density estimator is retail lot coverage ratio.•Other useful estimators are office and institutional land use, and income.•Dwelling density and road connectivity are much less important.