Recovery.gov is supposed to be a transparency clearing house for information on the federal stimulus spending appropriated in the $787 billion American Recovery and Reinvestment Act passed earlier this year. Unfortunately, the reports on spending and jobs saved or created are showing errors across the board.

Clay Johnson at Sunlight Labs looked at the “dirty secret” that is FederalReporting.gov, the site where agencies and stimulus fund recipients file their reports before that data is pulled by Recovery.gov:

Looking into FederalReporting.gov is a lot tougher than Recovery.gov. Not a lot of light has shone upon this website. In terms of costs– the only thing I can find on usaspending.gov is that the EPA has set up a $4,000,000 helpdesk for the operation. It looks like right now there are three ways to send data into FederalReporting– via an Excel Spreadsheet, a Web Form, and via an XML API.

The question on my mind is– what kind of validation is being done on the data before it goes into federalreporting.gov? For instance, how is data getting being accepted by FederalReporting.gov saying that jobs are being created in Arizona’s 15th District when Arizona’s 15th district doesn’t exist? Shouldn’t FederalReporting.gov be validating that? It seems from the documentation that all three methods of submission have a validation process. Is the validation so lax that obviously wrong data can get through?

My initial reaction upon seeing the Arizona 15th District story was that this could have been a state-level agency or contractor reporting that jobs were created in the 15th District of the Arizona Legislature (Arizona elects one state senator and two state representatives from each of their 30 legislative districts). That was until I saw that jobs and spending were being reported from the 86th District and other states were seeing reporting coming from the 99th District and other non-existent legislative boundaries. This problem, which is huge for a project that is relying on transparency for legitimacy, stems from a patchwork reporting structure that, as Clay reported, is not being overseen properly. It looks like some of the state and local agencies and private contractors and subcontractors are simply putting a number into a box where they decided not to figure out the correct answer. Subsequently, the reporting site that they submit to is apparently not checking for errors.

In response to a letter sent by House Oversight and Government Reform Committee ranking member Darrell Issa, Recovery Accountability and Transparency Board chairman Earl Devaney answered questions about the accuracy of Recovery.gov reporting by stating, “Your letter specifically asks if I am able to certify that the number of jobs reported as created/saved on Recovery.gov is accurate and auditable. No, I am not able to make this certification.” The accurate part is obvious from the many examples pointed out by ABC, Sunlight Labs and others, but the auditable seems a bit shocking. Why isn’t the data able to be audited? Is it really that bad? Or is the Recovery board’s staff that over-stretched. While Devaney promises “increasingly higher levels of accuracy in the future,” this problem of accuracy and auditability should have been tackled before issuing press releases claiming the positive effects of stimulus spending.