Cellphone Data Can Help Track Real-Time Unemployment Levels: Study

Cellphone data can detect unemployment levels in real-time because people's communications patterns change when they are not working, according to a new study co-authored by MIT researchers.

Using a plant closing in Europe as the basis for their study, the researchers found that in the months following layoffs, the total number of calls made by laid-off individuals dropped by 51 percent compared with working residents, and by 41 percent compared with all phone users.

The number of calls made by a newly unemployed worker to someone in the town where they had worked fell by 5 percentage points, and even the number of individual cellphone towers needed to transmit the calls of unemployed workers dropped by around 20 percent.

"Individuals who we believe to have been laid off display fewer phone calls incoming, contact fewer people each month, and the people they are contacting are different," said Jameson Toole, a PhD candidate in Massachusetts Institute of Technology's Engineering Systems Division.

"People's social behaviour diminishes, and that might be one of the ways layoffs have these negative consequences. It hurts the networks that might help people find the next job," said Toole, a co-author of the study.

The paper, published in the Journal of the Royal Society Interface, builds a model of cellphone usage that lets the researchers correlate cellphone usage patterns with aggregate changes in employment.

The researchers believe the phone data closely aligns with standard unemployment measures, and may allow analysts to make unemployment projections two to eight weeks faster than those made using traditional methods.

"Using mobile phone data to project economic change would allow almost real-time tracking of the economy, and at very fine spatial granularities both of which are impossible given current methods of collecting economic statistics," said David Lazer, a professor at Northeastern University and a co-author of the paper.

The study's starting point was an automotive plant in Europe that closed in 2006, leaving about 1,100 workers unemployed in a town of roughly 15,000 people.

Having the information about the layoffs allowed the researchers to build an algorithm that, by analysing phone-use patterns, assigns a probability that someone has become unemployed.

The researchers then extended that usage model to see how well it corresponded with larger-scale unemployment, using eight quarters of unemployment data in 52 provinces of a European country.

The researchers emphasise that they are not proposing the new method as a replacement for time-tested ways of measuring unemployment. Instead, they see it as an additional tool for analysts.

"These methods should not be viewed as substitutes for current methods of collecting data about the economy as much as very powerful complements," Lazer said.