Pulse Lab Jakarta conducted research to explore pseudonymised mobile network data as a source of insights on displacement following the Highlands earthquake in Papua New Guinea (PNG) in February 2018 and compared the findings to the current good practice for displacement tracking, namely IOM’s displacement tracking matrix. Insights from the study were also compared against information on the targeting of humanitarian resources, finding that the targeting of surveys and the allocation of humanitarian assistance are respectably efficient.
The analysis focused on displaced subscribers only, and removed any noise generated by regular commuters. In order to do so, it included mobile subscribers with an average day-to-day travel distance of less than or equal to 15 km, and used this subset to filter the day-to-day subscriber trajectory data set following the earthquake.
The analysis revealed a high-level picture of earthquake impact on communities across Papua New Guinea by modelling the relationship between cell tower activity and cell tower distance relative to the earthquake epicentre. The tower activity ratio, which is defined as a ratio between the number of unique mobile phone subscribers connected to the tower and the number of unique mobile phone subscribers connected to the entire Digicel network in PNG over a week, was first calculated.
In order to observe the earthquake’s impact, the research examined these ratios during two time periods, before and after the earthquake. Findings showed that the tower activity ratio decreases closer to the earthquake epicentre. Following the earthquake, the linear model contained a lower constant compared to the model before the earthquake, which indicates a stronger decrease in tower activity ratio after the earthquake in areas proximate to the epicentre.
The mobile phone penetration rate in Papua New Guinea stands at around 54 per cent and this analysis only examines data from mobile provider Digicel, which holds a majority market share. The absence of data from another mobile network operator, as well as the fact that neither network has coverage of all populated areas introduces some additional gaps in the analysis. Nevertheless, the findings offer new perspectives on the quantification of displacement and underline the potential of mobile network data to offer highly valuable information during what are often chaotic days and weeks following a natural disaster.
