The Washington PostDemocracy Dies in Darkness

Can big data solve veteran unemployment?

A General Electric representative (C) talks to an applicant at a hiring fair for military veteran job seekers and spouses at the Verizon Center in Washington on April 9, 2014. (REUTERS/Gary Cameron)

The unemployment rate for veterans coming out of America's current wars has consistently lagged several points behind the rest of the labor force. Today, unemployment for veterans who've served since September of 2001 hovers just over 8 percent. For the economy at large: 6.1 percent.

The gap between those two numbers speaks to several challenges: The military trains service members in many jobs with close but imperfect civilian corollaries, leaving veterans with the right job skills but the wrong certifications. Military service demands other skills that civilian jobs don't (managing violence, repairing weapons, defending convoys). But where that unfamiliar experience entails universal qualifications (leadership, judgment, communication), employers don't always know how to recognize them. New veterans, all the while, must navigate the job market — a place that can be hostile to anyone — as they're also re-entering the rest of the civilian world.

"We’ve thrown a lot of policy at this — the G.I. bill, hiring tax credits," says Aneesh Chopra, the former chief technology officer at the White House and now the co-founder of a startup called Hunch Analytics. "But we haven’t solved it."

He is hoping now, though, that big data can help narrow the gap in veteran unemployment. On Thursday, Chopra and a coalition of tech entrepreneurs, open-government advocates and private companies launched a novel experiment mashing together data about job-seeking vets and vet-friendly employers, all in the hopes that we might better understand why so many vets can't find work.

They've pulled together government data on veteran unemployment claims, skills data scraped from Monster resumés and LinkedIn profiles — all of it anonymous — and job requirements from employers who've publicly committed to hiring veterans. The result is a unique dataset for labor economists and workforce planners, as well as a data visualization mapping the many mismatches between veterans and the job market.

The project includes an interactive map, built by Chris Walker at Mic, that can contrast the location of unemployed vets — alongside their demographics and qualifications — with the location of veteran-friendly job openings. It shows, for instance, that Fairfax County is a hub for veteran-friendly jobs in computers and information systems, while Roanoke County is looking for more sales and fast food workers. So far, the tool includes comprehensive data on 18 states.

The project, called Veterans Talent, is meant primarily as a proof-of-concept of what can be built when private companies and government agencies alike open up some of the aggregated data they normally hold dear.

"It's not a job-search tool for veterans, and it’s not a head-hunting tool for recruiters," Walker says. "It’s more a medium- and long-term planning tool for policymakers to help them see where pockets of skills are."

State governments hold much of the information about who's unemployed and where those job-seekers are clustered. Private companies, meanwhile, hold most of the data about the positions employers are looking to fill, or the qualifications would-be workers cite in job-search engines.

It would be difficult, in other words, to answer many questions about the veteran labor market without data from either group.

"What is the art of the possible when you open up data and you mix public and private data sets together?" says Leighanne Levensaler, a vice president at Workday, an HR and finance software firm that contributed to the project. "Could we apply the most modern technology, the best techniques we use to help you find a flight, to help you shop – the machine learning, the big data, the scraping and parsing of data – to actually pop out insights that could be instructive for workforce planners?"

Government agencies could use data like this to determine what kinds of veteran job-training programs to develop in certain communities (there's disproportionate demand in Virginia Beach, for instance, for automotive technicians), or what kinds of companies to recruit there (there are an awful lot of veterans with experience in cargo and freight in Dorchester County, Md.). Companies, likewise, might learn something about where to recruit, or why local veterans aren't applying for jobs.

Before the project's launch, the full data was also shared with several economists and researchers, who've blogged some initial findings on the Veterans Talent Web site.

John Parman at the College of William & Mary, for instance, found that veteran unemployment tends to be lower in those Virginia counties where economic mobility is higher (he compared this latest data to results from a landmark social mobility study released last year). This isn't all that surprising. But Parman found that as economic mobility rises, the unemployment rate falls faster for non-veterans than veterans. This implies that while veterans fare better in places with dynamic economies, they still fall behind non-veterans there.

When Chopra and others started to pull together this data, they weren't sure what questions economists and workforce planners using it could answer. "If you actually take a deeper look at the demographics of job-seeking veterans," Chopra says, "you see some things that are very eye-opening."