The Washington PostDemocracy Dies in Darkness

Here’s how the Recovery Act became a test case for open data

Earl Devaney in the Recovery Operations Center. (Sarah L. Voisin/The Washington Post)

Making sure that government money is spent efficiently and without fraud can be difficult. You need to collect the right data, get the information to the right people, and deal with the sheer volume of projects that need tracking. Open data make the job easier to draw comparisons across programs and agencies. And when data are released to the public, everyone can help be a government watchdog.

When President Obama was first elected in 2008, he promised transparency. Almost immediately after he was sworn into office, he had an opportunity to test that promise with the implementation of the Recovery Act. And it worked.

In fact, according to Earl Devaney, the former chair Recovery Accountability and Transparency Board, transparency efforts around the bill became a "proof of concept of whether government data, particularly gov spending data, could be standardized and become machine readable and usable by all American citizens."

Before retiring in 2011, Devaney led a team of 50-some auditors and investigators that sifted through data from those who received a sliver of the Recovery Act's $840 billion pie and sniffed out fraud. He spoke about their efforts during a panel at the Data Transparency 2013 conference last week. The conference was hosted by the Data Transparency Coalition, of which Devaney is a board member.

According to Devaney, the board first met to discuss their "jaw-dropping assignment" around 10 days after the passage of the Recovery Act in February, 2009. Their goal was to do accomplish two major things: Provide some sort of umbrella fraud prevention, and "bring an unprecedented level of transparency to government spending."

They quickly decided a standard data format was essential. "The fraud piece was made easier by the fact that the data was standardized," explained Devaney. But just how that data should be collected, and just how much of it, was more of a debate.

"We made a decision early on to collect data directly from recipients," says Devaney -- a choice that was not particularly popular at the time because there were concerns that it wouldn't be as accurate as having government investigators collect the data themselves. Recipients plugged in their data at a private portal at But the key to keeping them honest was actually the public according to Devaney. used geospatial technology to "allow Americans to drill down to their zip codes exactly where government money was being spent in their neighborhood." It's this micro-level of attention that increased accountability, according to Devaney.

"The degree of transparency forced them to get it right because they didn't want to be embarrassed by their neighbors who they knew were going to these Web sites and could see what they were doing with the money."

As to the second question of what data to collect: "I finally put my foot down and said no more than 100 pieces of data," Devaney recalls, "So naturally, we came up to 99." Of course, even with limiting themselves to that number of data points, transparency and fraud prevention was a daunting task, with the 300,000 some grantees to keep tabs on.

But having those data points in an open format was what allowed investigators to use "sophisticated cyber-technology and software to review and analyze Recovery-related data and information for any possible concerns or issues." And they were remarkably successful on that end. A status report in October, 2010 showed "less than 0.2 percent of all reported awards currently have active fraud investigations." Indeed, for Devaney's  tenure leading the board he says the level of fraud hovered somewhere below half of one percent of all awards.

There are some concerns that the tight level of oversight resulted in a somewhat slower than expected roll-out of funds, thus compromising the major goal of the Recovery Act: Spurring the economy. But as a test case for deploying government data on a large scale, it appears to have been a success.

But looking forward, Devaney worries that the lessons he learned leading the board will take a long time to make their way into the government at large. The president issued an Executive Order calling on agencies to make open and machine-readable data the default for new government data sets. But Devaney thinks legislative measures like the DATA Act is more likely to actually spur agencies to action: "A law is the way to ensure the government moves quickly to standardize and make data machine readable." After 41 years in the government, he thinks it's "a little naive" to the hope that government agencies will move there themselves.