Answers are grounded in this paper and cite their sources.
Walking dominates mobility in many Global South cities, yet pedestrian route choice behaviour remains insufficiently understood, particularly in data-constrained urban environments with limited behavioural evidence. Existing studies rely merely on either stated preference (SP) surveys or revealed preference (RP) observations, each capturing part of the pedestrian decision-making process. This research examines the complementarity between SP and RP data in modelling pedestrian route choice and assesses whether their joint use improves behavioural realism. Makerere University in Kampala, Uganda, served as a living laboratory representing dense institutional activity hubs in rapidly urbanising cities. SP data were collected through a discrete choice experiment involving 106 respondents. RP data were obtained from GPS tracking of 105 pedestrians generating 309 observed trips. Discrete choice models were estimated separately for SP and RP data and jointly using a combined SP/RP framework with source-specific scale parameters. Model performance was evaluated using likelihood-based statistics and repeated hold-out validation with a 70/30 estimation–validation split. Results reveal systematic differences between stated and observed behaviour. While SP responses emphasise shorter travel times and direct routes, RP data show pedestrians accepting longer routes to access greenery and amenity-rich paths. Greenery and amenities consistently increase route attractiveness, while slope negatively influences route choice, particularly among female pedestrians. The SP/RP model exhibited the best predictive performance (ρ² = 0.458) and the smallest out-of-sample log-likelihood differences compared with SP-only and RP-only models. Hence, integrating stated and revealed preferences better captures pedestrian decision-making, supporting the design of higher-quality walking environments in rapidly growing African cities.
No takes yet. Share an insight, caveat, or question.
Namakula et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: