Assessing the representativeness of mobile phone data

Abstract

New forms of mobile phone data offer enormous potential for the advancement of our understanding of human activity and mobility. However, uncertainty in the representativeness of data creates ethical risks that underlying biases could impact results and mislead policy recommendations, limiting their usefulness. This paper assesses the spatial and socio-demographic representativeness of two extensive independent mobile phone application datasets over three years, covering a large and diverse city-region (Glasgow), home to over 1.8 million people. Towards this end, we advance methods in home location detection by utilising high resolution residential land use data as part of the process. Results show a very high level of representativeness across both datasets, comparable to survey-based techniques commonly used for mobility analysis in Scotland. The techniques employed to advance home location detection improve the fit substantially over a conventional technique. Our findings provide further confidence for the use of mobile phone data as an alternative data source to understand human activity, furthermore, our methods offer an improved benchmark upon which to assess the quality of similar data sources in the future.