Extracting the population mobility information contained in Call Detail Records (CDRs) is of critical importance in data poor contexts such as in low- and middle-income countries (LMICs), where it can support humanitarian and human development efforts. Such contexts however present additional challenges compared to high-income countries (HICs) for mobility analysis from mobile operator data: often only CDRs are available and they are sparser over time and space, mobile networks are more unstable, particular in crises, which are often more frequent, and the geographic coordinates of cells are sometimes missing and erroneous.
Further, the proportion of the general population using mobile phones is significantly lower in LMICs (e.g. 47% of households on average in 7 provinces of the DRC, down to 35% in the more rural provinces) and therefore differences in the mobility of phone users and non users have a larger impact on the representativity and applicability of CDR-derived statistics.
At Flowminder, we have specialised in addressing such challenges and we present here an overview of our live systems, from ingestion and automated quality assurance (QA) checks of pseudonymised CDR data and cell data, to the extraction of mobility information from CDRs and bias correction using survey data, resulting in the semi-automated production of a set of standard indicators, ready to be disseminated to decision makers in LMICs through dashboards, standard reports or as data sheets. We present the standards we've produced, and recommend using, when producing population mobility and distribution estimates from CDR data, to reduce biases inherent to the data, and ensure robustness of estimates.
Read the full abstract
Flowminder Foundation, Lefebvre V., et al. (2023). Flowminder standards in producing mobility and population estimates from call details records in low- and middle-income countries. https://doi.org/10.5281/zenodo.8414430