“The intuition was that people's lives shrink when they lose their jobs,” explains David Lazer, a professor at Northeastern University and one of the authors of the paper. “If you're employed, you usually have two full sides to your life—your home context and your work context and you go back and forth between the two.”
The researchers first looked at cell-phone data in a town where an auto-parts manufacturing plant had closed. To determine who was laid off, they looked at individuals who made calls near the plant before the layoff date, then dropped off from that cell-phone tower when the plant closed. They found that the number of calls made by those laid-off individuals dropped 54 percent, and the number of incoming calls dropped 41 percent. The laid-off individuals also reduced how much they moved around geographically, which the researchers tracked by their call logs and established a pattern. “Indeed, we saw that people made fewer outgoing calls and fewer incoming calls. Their mobility reduced and their lives became a little less predictable,” explains Lazer.
They then turned their attention to a dataset of over 10 million subscribers in a European country (they won’t say which), and found that adding mobile-phone data significantly improved the accuracy of existing methods of predicting employment numbers. Currently, labor statistics in the U.S. are based on the Current Population Survey which derives from a sample of 60,000 households.
“I think there are three problems with current government statistics. The first problem is that they are somewhat slow, because they're based on survey methods as well as administrative data and it takes a long time to get all the information together. In this day and age, when we have so much data streaming it would seem like we should be able to do better—and that's in part our objective: to show a way that we can speed up our insights about the current state of the economy,” says Lazer.
Lazer also believes that the current government statistics on employment lack accuracy, and that they have a larger-than-ideal margin of error. They also don’t reveal geographical specificity at a level that statistics derived from cell-phone data would. He believes that using cell-phone data has the potential to enable more precise and real-time picture of the economy, which would aid in projections and policy decisions.
“There's a lot of money and a lot resources that go into figuring out the state of the economy, this would seem to be a stream of data that no one has really utilized for these purposes before,” says Lazer. “If we could have some type of method that would allow more zooming in, down even to the state level or the town level that could really change our paradigm of economic intervention.”
Of course, the possibility that researchers or government officials could use cell-phone data for any purpose might be off-putting to many Americans in these post-Snowden times. But the possibility of non-conventional data streams to improve current methods is certainly a tempting one that doesn’t necessarily have to violate the public’s privacy. For example, moving away from personal cell-phone data, aggregated data from the retail sector could turn out to be a good predictor of GDP. As the Freedom Act would still allow the government to acquire data on a large scale, perhaps the question is whether collecting data in an effort to improve the economy is more palatable a motivation than counterterrorism.