Another Social Divide: Internet Usage
French study examines the link between socioeconomics and mobile data usage
By Irving Wladawsky-Berger
A few weeks ago I attended an MIT seminar by Esteban Moro, professor at Universidad Carlos III in Madrid and visiting professor at MIT. Moro’s talk discussed the socio-economic digital divide based on how we use the internet, focused on two key questions: how profound is the digital usage gap in our society? and can we use mobile phone traffic to infer socioeconomic status? His seminar was based on a recently published article in the Journal of the Royal Society Interface co-authored with four collaborators, News or social media? Socio-economic divide of mobile service consumption.
“Inequality is a central societal problem, especially within rapidly expanding urban areas,” wrote the authors. “While it is a crucial driver for economic growth, the progressive ‘clusterization’ of workers, industries, companies, and services in cities has a tremendous cost in terms of segregation and discrimination. This cost is not only economic: in the same city, different areas can have a 10- to 15-year imbalance in life expectancy and highly divergent education levels, with little chances of social mobility.”
“The design and successful implementation of policies to alleviate these problems require fine-grained, frequently updated information about income, education or inequality across metropolitan areas,” they added. “However, most data sources employed today, such as population censuses or surveys, suffer from sparsity in population coverage or infrequent updating, hence they do not allow the swift evolution that urban societies experience… Thus, the traditional ways of understanding cities tend to explain what happened five years earlier rather than ‘nowcasting’ or even predicting urban transformations.”
The First Divide: Access
In his seminar, Moro referred to the socio-economic divide based on how we use the internet as the second digital divide, in contrast to the original digital divide which was based on access to the internet.
Getting online in the 1990s required a personal computer and an account with a service provider, and e-commerce transactions required a credit card and bank account. As our economy became increasingly digital, major new inequalities were arising because so many around the world could neither afford a PC nor an internet account and had no bank relationship or credit card. The reach and connectivity we were so excited about in this initial phase of the internet era was in reality not so inclusive. While the internet was truly empowering for those with the means to use it, it led to a growing digital divide both within countries and across the world. The internet was ushering a global digital revolution, but it was disconcerting to have a global digital revolution that left out the majority of the world’s population.
This picture started to change in the 2000s. Continuing technology advances were now bringing the empowerment benefits of the digital revolution to a majority of the planet’s population. Mobile phones and wireless internet access went from a luxury to a necessity that most everyone could now afford, initially in advanced economies, and later in most of the rest of the world. We were transitioning from the connected economy of PCs, browsers and web servers to an increasingly hyperconnected digital economy of ubiquitous, powerful and inexpensive mobile devices, cloud-based apps, and broadband wireless networks.
While the original internet access gap is now minimal in developed economies, Moro and his collaborators found that a digital usage gap has now emerged, representing the distinct uses of the internet by different socio-economic groups based primarily on their income and educational status.
The Second Gap: Usage
To measure this digital usage gap, their article argued that the specific consumption of mobile services by each group was a suitable indicator for analyzing the economic and social inequality in advanced economies, since consumption patterns were a direct reflection of the users’ personal interests and digital skills. Consequently, their study is based on the analysis of 3.7 billion records of mobile traffic collected by France’s leading mobile operator in the 45 days or so between mid-May and June of 2017. The records were time-stamped and linked to the specific network base stations from which the internet access originated.
The records captured the mobile behavior of about 15 million individual across the whole country as they accessed different applications — e.g., YouTube, Facebook, Netflix — from their Apple or Android devices. To investigate the differences in the relative usage of specific mobile applications by income and education groups, the analysis only considered internet usage recorded between 8 p.m. and 7 a.m. during weekdays — the hours when individuals were more likely to be at or near their homes. Their mobile phone usage was then linked to the demographic data of their home area based on France’s 2014–2015 census report. The results are based on a fully privacy-preserving analysis of mobile phone data. Let me summarize the study’s key findings.
The study found quantitative evidence of a significant digital divide in internet usage between two socio-economic groups, each with different income and educational attainment.
In principle, all individuals had access to the same internet. But, the study found that each group generally accessed its own distinct version of the internet, and socio-economic behavior was therefore influenced by the fairly different services and information they were exposed to. By analyzing mobile traffic flows, the study identified the key services that each group accessed:
· Higher income and education demographics: Information-seeking traffic predominates, e.g., news, mail, search. Instagram, WhatsApp, and Twitter are the dominant social media apps; games like Clash of Clans are the most widely used.
· Lower income and education demographics: Entertainment traffic predominates, e.g., video-streaming, gaming, adult services. Facebook and Snapchat are the dominant social media apps; games like Candy Crush are the most widely used.
“The digital usage gap is so profound between low- and high-income, or low- or high-education areas, that it can be used to clearly distinguish between them or even identify the relative composition of these groups in a given area,” wrote the authors. “High-income areas or those with higher education attainability show a more pronounced utilization of mobile devices to consume news, exchange e-mails, search for information or listen to music. At the same time, they display a reduced use of some social media platforms or video-streaming services. These results hold even when we control for age composition and other census variables such as an immigrant population.” In addition, the findings highlight “a pronounced spatial uniqueness in the consumption of each application, when relative usage is compared across different geographical units at a national scale.”
“The digital usage gap is so large that we can accurately infer the socio-economic status of a small area or even its Gini coefficient [a measure of income inequality] only from aggregated data traffic.
Our results make the case for an inexpensive, privacy-preserving, real-time and scalable way to understand the digital usage divide and in turn, better understand poverty, unemployment, or economic growth in our societies through mobile phone data.”
This blog first appeared June 3, here.