Population Dynamics from Mobile Network Data

How can we understand where people actually live, work, and move — not just where they're officially registered? We explored the potential of mobile network location data to complement traditional census methods, creating a near real-time picture of population distribution and mobility across Slovakia.

Challenge

Census data provides a snapshot once per decade, while policy decisions happen daily. Slovakia needed better tools to understand dynamic population movements — commuting patterns, seasonal shifts, urban-rural flows — without waiting years for official statistics. With over 7 million active SIM cards and nearly 89% LTE coverage, mobile networks offered an untapped data source.

Mobile location data comes with inherent complexity. One person may carry multiple SIM cards; children and seniors are underrepresented; rural areas have lower network density than cities. Translating raw network signals into meaningful population insights required navigating technical limitations, privacy requirements, and statistical methodology simultaneously.

Approach

Working with Comenius University researchers, we systematically analyzed the limitations and opportunities of mobile positioning data. We examined how network infrastructure affects localization accuracy (hundreds of meters in cities, kilometers in rural areas), developed methods to filter duplicate SIM cards, and created frameworks for interpreting movement patterns while respecting strict anonymization thresholds.

Mobile network data cannot replace census data — but it fills critical gaps between census periods with unprecedented timeliness and geographic detail. When properly processed and interpreted with awareness of its limitations, it provides robust insights into spatial distribution and population dynamics that no survey could match in scale or frequency.

Outcome

The project delivered a comprehensive methodology for extracting population insights from mobile data, including identification of regular day/night locations, commuting flow estimation, and seasonal population variations. The approach achieved robust results while maintaining privacy protection through minimum threshold requirements of 3 users per geographic unit.

Beyond methodology, the project produced concrete recommendations for future applications: using raw signaling data instead of pre-aggregated records, incorporating statistical address points to capture marginalized communities, and calibrating models for different socio-economic contexts. These insights pave the way for mobile data to become a standard complement to official statistics.

Explore the full case

If you like to explore the full scope of this case study, read the publication with our contributions to experimental statistics using big data available for download.

Download full version of case study as .pdf