Pierre Deville of Belgium's University of Louvain in Belgium and Northeastern University's Center for Complex Network Research and Catherine Linard of Belgium's Université Libre de Bruxelles are the study's lead authors, and their newly released research appears in the Proceedings of the National Academy of Sciences. What Deville and Linard set out to discover was whether the rapid global adoption of mobile phones can provide more finely-grained insight into population distributions than traditional, census-driven demographic methods do. Is cellphone data the answer for countries that can least afford other counts of their people? Can that information provide a usefully nuanced reading of human geography? And can it be done without violating the privacy and security of the cellphone toting population?
Existing ways of understanding population distribution have tended to be high-level. In the 1990s, there was as push among demographers and geographers to scale down census data, and there were advances made in figuring out how to use land cover (Pavement? Grass?), street grids, and even climate data to divide those population counts down to the hundred-meter square. But that approach has its own failings. Downscaling uses national censuses as its baseline. In many spots of the globe, though, those counts are rarely done, often because of the expense or logistic complexities. That presents a paradox: the populations most at risk during disasters and other epidemics often live where census data is the least available or least reliable.
But the rapid adoption of mobile phones across the planet, Deville and Linard found, offers a new way to fill in where there are gaps in the data. The researchers acquired from telephone companies a data set of one billion phone calls made by two million users in Portugal and 17 million users in France. Then, in a first experiment, they mapped Portugal's calls using both long-standing census downscaling methods and the mobile phone data.
To plot the data, Deville and Linard drew around existing cellphone towers what's known as Voronoi polygons — or cells drawn around a set of points so that any spot within each cell is closer to its own contained point than to any other — and then plotted the number of calls connecting to that tower. Then, using the telephone providers' data on service adoption rates, the researchers extrapolated per-tower populations.
One benefit that revealed itself about that approach, says Linard on the phone from Belgium, is that, "we can produce maps for day and night, for weekdays and weekends, for holy days and working periods."
Of course, there's a worry that using mobile phone data to plot population is a personal privacy and security risk. The study points out, though, that user identities weren't contained in the data set; in fact, they were working with simple call data aggregated by cell tower, without even a record of when calls were passed from one cell tower to the next.
In their Portugal mapping, the researchers found that the mobile phone data matched the accuracy of the census-based data on a high level. Drill down to local levels, though, and things got even more interesting: around the capital city of Lisbon, the mobile-phone data captured details that the other method missed. The digital information, concluded the authors, can provide "a near real-time understanding of patterns and processes in human geography."
That said, concluded the authors, the value of the method increases as the number of represented cell towers increases. "Its accuracy depends directly upon the network structure," the study reads. "The higher the density of towers, the higher the precision of the [mobile phone] communication geo-localization." In other words, the research technique they're advancing will only improve with age, as long as more people keep using more cellphones.
Moreover, found the researchers, the mobile-phone data was far more responsive to changes in populations over time than traditional methods. Applying the approach to France as well revealed that the population of that country's main cities decreased dramatically during the summer and the country's other holiday periods, with people pouring out of Paris — save for large pockets of people at Disneyland Paris and Charles de Gaulle Airport — and toward France's mountains and coastlines.
"In operational and governmental decisions," Deville and Linard write, "these data may also be valuable for supporting rapid responses to disruptive events or longer-term planning purposes." Mobile-phone mapping could make it easier to provide targeted responses to everything from where to send earthquake relief supplies to where to direct medical treatment to address Ebola.
But one challenge is simply getting the data. Linard says the team has on hand phone data from Senegal and the Ivory Coast, but "we don't have phone data from Liberia at the moment."
Deville and Linard's study fits in a broader debate over the power of mobile phone data, and into whose hands that power should rest. If one of the benefits of a traditional census is that people can stand up and be counted, and thus count, we could seen underrepresented communities demanding their own mobile-phone data from phone companies to show where they're actually living, day in and day out, and what resources they need there.