On the first Friday of every month, the U.S. Department of Labor releases its jobs report, showing the number of people employed and unemployed at that moment. The data is based on surveys of employers and households across the country, and there are several reasons to think it isn’t as accurate as sometimes supposed. Economists argue, for example, that it fails to capture all the people now working part-time or flexibly.
David Lazer, a professor at Northeastern University, thinks we could improve the jobs numbers by using a completely different collection-method. Rather than asking people, after-the-fact, if they have a job, you could track their behavior using cell-phone data, he says. That way, we might be able to have jobs data earlier than now, and perhaps improve how we manage the economy.
Lazer’s contention is based on research he recently published with colleagues from Harvard and M.I.T. The researchers took a sample of phone records from a small, undisclosed European town that experienced a factory closure in 2006. They found distinct patterns of calling among those who’d been fired, leading them to believe they could detect layoffs more widely if they had enough real-time information. For example, the out-of-work subscribers used their phones almost half as much as employed people in the sample, moved far less each day and also less predictably (presumably because they were no longer commuting).
“We showed at the population level that, if you saw the same behavioral patterns that we saw in this mass-layoff event, and then picked out individuals in all the provinces, you could get really good predictions of future unemployment reports,” says Lazer in an interview.
He believes the method could give policy-makers as much as four months head-start compared to current survey-based practice. “I hope this kind of approach could help inform real-time measurements of the trajectory of the economy and that could help inform better economic policy-making. At the moment, our methods are really slow and they’re not very spatially granular,” he says.
The paper is published in the Journal of the Royal Society Interface.
The obstacle is probably in getting phone companies to play ball. Like Internet companies, they generally aren’t keen to release even aggregate records, even though, in a sense, it’s our data and we ought to be able to share it if we want.
“The big companies hold data that, in some sense, is in trust for society,” Lazer says. “There needs to be a conversation if that company provides something, at near-zero cost to itself, that benefits to society. If we could have cut off the Great Recession because we knew earlier that the economy was going down the tubes, that could have improved the well-being of millions of people.”