When it comes to understanding the social, demographic and economic conditions of a country, the obvious place to turn is its most recent census, a national survey carried out to collect just this kind of information. At least, that’s possible in most developed countries.
In developing countries, census information is much less reliable. In Côte d’Ivoire on the west coast of Africa, for example, the National Statistics Office carried out censuses in 2002 and 2008 but the civil war that broke out during that period makes the results highly unreliable.
Today, Thoralf Gutierrez at the Universite catholique de Louvain in Belgium and a couple of pals say there is a better way to understand the social and economic make-up of a developing country. Given the widespread use of mobile phones in these areas, why not use the datasets that record usage habits, they ask.
These guys say, in particular, that the way individuals buy airtime credit is a good indication of their wealth. And since mobile phone datasets record the buying habits of a significant proportion of the population, they can reveal the distribution and variation of wealth around a country too.
And that’s exactly what they’ve done. Gutierrez and co used a dataset of the mobile phone habits of significant fraction of the population of Côte d’Ivoire which they obtained from one of the country’s large mobile phone operators.
This dataset contains the caller ID and receiver ID for all calls and text messages made in 2012. It identifies the cell tower used and gives each call a timestamp. Crucially, the dataset also contains the timestamp and amount of every airtime credit purchase made by every customer.
This Week’s Top 5 Posts
The analysis is straightforward. Gutierrez and co start by analysing the airtime credit purchases and working out by how much each customer varied the amount they purchased. This revealed several different types of customer: some who made a few big purchases and others who made many small purchases, for example.
“Our hypothesis is that this difference in behavior predicts household income,” they say. “Someone who is poor will have to buy airtime credit in small amounts while someone who is rich can make larger purchases.”
They then mapped the average purchases across the country. This map clearly shows the areas where people tend to spend more on airtime credit and are therefore wealthier. One example is Abidjan, the country’s biggest city and the largest seaport in West Africa.
Another is the border roads to Mali and Burkina Faso in the north and to Ghana in the south-east. These are economic corridors that are likely to generate wealth. The South Coast is also wealthier, probably because of tourism.
The map also throws up some surprising results. “The Liberian border in the South-West is unexpectedly wealthy,” say Gutierrez and co. That’s strange because the population density in this area is low and there is little industry that can account for any extra wealth. Indeed, the area is known for its insecurity and land conflicts.
But Gutierrez and co say there is another explanation. The wealth probably arises from illegal activities on the border, such as drug, arms and human trafficking. Interestingly, that’s not data that an official census would be likely to pick up.
The mobile phone records also reveal areas of inequality which host both rich and poor people. Most urban areas fall into this category however one city, Korhogo, in the north of the country appears to have little inequality for reasons that are not clear.
The Liberian border area mentioned before does not have any diversity either—all the people living there by airtime in large amounts. Exactly why this should be is unclear too.
Gutierrez and co have also studied the social network associated with this dataset. They create a network in which each node is a customer and draw a link between two customers if they communicate at least once per month.
An interesting feature of this network is that people with similar wealth seem to talk to each other. “People tend to be friends with people that have the same purchase average as themselves,” conclude the team.
One problem with this analysis is that there is no ground truth data to compare it against. That’s a shortcoming that Gutierrez and co are only too aware of but, given the unreliability of the official census data, there is little they can do to change that.
What is clear, however, is that the study of airtime credit purchase is a powerful tool for understanding the socio-economic status of countries that do not have the resources to conduct large surveys themselves.
The next steps are many—to try the same technique in other developing countries, to compare the results with reliable ground truth data and to extend the analysis to the developed world, to name just a few.
It’ll be interesting to see where this new science of mobile phone-ology leads next.
Ref: arxiv.org/abs/1309.4496: Evaluating Socio-Economic State Of A Country Analyzing Airtime Credit And Mobile Phone Datasets