PathShift: Relational trip reconstruction from passive mobility data for official-statistics-like mobility products

Abstract

Population mobility is a core domain of Official Statistics, yet standard sources such as commuting matrices and travel surveys are infrequent, costly and only partially observe behaviour. Passive mobility data can improve timeliness and coverage, but they do not contain trips directly: trip boundaries must be inferred from noisy, purpose-blind trajectories. This paper presents PathShift, a trip-reconstruction methodology that treats stop significance relationally rather than through a single global dwell-time threshold. Candidate trips are defined between daily anchors and are recursively split only when the realised path between the same origin and destination is unusually long relative to the population distribution. The approach is assessed on mobile-network-derived data from four Italian provinces selected to represent contrasting mobility contexts. Results show that the recursive procedure stabilises after a limited number of iterations and produces an aggregation-ready trip layer that preserves behavioural structure while reducing the brittleness of threshold-based segmentation. We then show how PathShift outputs can be translated into products analogous to an official commuting matrix and a travel-diary observatory, thereby clarifying both the opportunities and the conditions under which passive-data trip layers can support mobility measurement in the orbit of Official Statistics.