‘Mind the Gap’—The impact of discrepancies between Google Maps API and reported travel data in the Global South

Over the past decade, online navigation services have been adopted increasingly as a source of ‘ground truth’ in estimating choice alternatives during travel behaviour. These services, including Google Maps, Bing Map, and Waze, which are designed to provide real time traffic information and navigation guidance to the users, are believed to offer comprehensive and precise information regarding travel attributes. Nevertheless, discrepancies between the travel attributes collected from those services and the travel data that is reported by the travelers may introduce a systematic bias into travel behaviour analysis and modelling. This paper attempts to explore this challenge by investigating the discrepancy between the reported travel times and costs and the corresponding values derived from the Google Maps API. The comparison is conducted in the context of a developing country, through the use of travel diary survey data from Greater Jakarta, where there is a greater variety of transport modes and individuals may have varying capacities to gauge travel attributes due to the unpredictability of traffic conditions. Results show that even minor adjustments to which observations are included and which specific attribute treatments are used can completely change values of travel time savings (VTTS) estimates. Further, the characteristics of the observations excluded in the process of pre-processing are investigated to provide insight into preventing loss of data in future mobility surveys. Recommendations to address both of these issues are discussed along with policy implications. • Information on travel attributes provided by navigation map providers like Google Map are widely used in transport modelling. • Comparison of travel attributes obtained via the Google Maps API and data reported in surveys in the context of developing countries indicate significant discrepancies. • An investigation into the characteristics of excluded observations can help to ensure data quality in future RP surveys. • Data pre-processing guidance and travel attribute treatments are proposed to overcome the impact of such discrepancies. • Careful selection of appropriate treatments is essential to obtain reasonable mode-specific VTTS measures.