Paul FAVIER

Abstract

As tourism demand measurement increasingly relies on proprietary and costly data, freely accessible digital traces offer a scalable alternative. We propose mobile Wikipedia pageviews as an interpretable, high-frequency proxy for city-level hotel demand, grounded in a behavioral distinction between platforms: mobile browsing captures on-the-ground information seeking by physically present visitors, while desktop browsing reflects remote planning. Using a daily panel of 704 French cities (2018–2025) matched to proprietary hotel records from Accor, a two-way fixed-effects design controlling for city-level and daily aggregate shocks yields a robust positive elasticity between mobile Wikipedia traffic and same-day hotel room-nights. A COVID-19 natural experiment, exploiting the asymmetric collapse of mobile relative to desktop views during lockdowns directly validates the physical-presence mechanism. Our central contribution is a systematic decomposition of this relationship at the city level, revealing that virtually all dispersion in destination-specific effects reflects true signal differences rather than estimation noise. The indicator performs best for internationally recognized tourist destinations and weakens or reverses sign for business-dominated and transit cities. A typological analysis identifies day-of-week desynchronization between Wikipedia activity and hotel demand as the main mechanism behind negative effects. Taken together, these findings map the boundary conditions under which mobile Wikipedia traffic constitutes a valid, cost-free tourism nowcasting instrument, with direct implications for destination monitoring and tourism policy.