The aim of this study is to investigate temporal and spatial dynamics of falls among older pedestrians in Sweden. These events involve self-inflicted falls, as well as individuals' collisions with other pedestrians or other road users, such as bicycles and vehicles. We first assess the nature and the distribution of falls over time and by type of environments between 2011 and 2016. Then, using Geographical Information Systems (GIS), we visualize the concentrations of falls by street segment. Finally, we assess whether temporal and spatial factors affect the occurrence of falls among older pedestrians, controlling for contextual factors using a multinomial logit model for all of Sweden, and then for rural municipalities only. Falls take place in more densely populated areas and are associated with specific characteristics of the built environment (e.g. mixed roads, bus stops but less frequently in inclined streets). Given the concentrated pattern of falls (80% of falls happen in 5% of the total street segments), a systematic maintenance service across all municipalities should be tailored according to the local geography of falls to ensure elderly safety. Both models confirm that elderly pedestrians' falls in connection with collisions with car, motorcycle and bicycle happen more often in the afternoons, weekends and in autumn than do self-inflicted falls. These findings call for the need for considering elderly routine activities when defining preventative measures.