Months after they were required to make their financial data public, their reports are rife with errors.
Most federal agencies are required by law to have uploaded their financial data on USASpending.gov, but complying with the federal transparency law has been a little messy, watchdog reports show.
The Digital Accountability and Transparency Act, known as the DATA Act, required agencies make their spending data available to the public online by May 2017. The yearly spending—about $3.7 trillion—is displayed on Beta.USAspending.gov. The act also tasks the Treasury Department and the Office of Management and Budget with overseeing agency reports, and requires agencies’ inspectors general to assess the quality of the data they upload.
Across the board, agencies have uploaded incomplete data or failed to establish a certification process ensuring that the data was accurate.
The Government Accountability Office, in its first review of data quality, also dinged OMB and Treasury, concluding less than 1 percent of agencies’ reported grants, contracts and loans were consistent with information obtained from authoritative sources within the agency. That’s actually a decrease in inconsistency from what GAO found in 2014, when it estimated between 2 and 7 percent of award data was consistent.
Overall, budgetary data, including appropriations, was more consistent, but the grants, contracts and loan errors may have omitted a significant chunk of federal spending, GAO concluded. Information from about 160 financial assistance programs, covering about $80.8 billion, were completely omitted from reports. About 13 agencies, including the Defense and Agriculture departments, submitted files that purported to connect budgetary data to award spending, but didn’t provide any data, GAO concluded.
GAO recommended OMB and Treasury clarify guidance for what to report and also to be up front about any data quality issues. They generally agreed with the recommendations.
Agency IGs were required to file own their assessments of second quarter spending data by Nov. 8. Some highlights:
- The Housing and Urban Development Department didn’t comply by the May 2017 deadline, and underreported “$17.9 billion in incurred obligations, $16.9 billion in outlays, and $4.2 billion in apportionments,” the OIG found. The failure to comply could be due to limited resources for DATA Act implementation and a lack of governance for the various HUD components submitting data to USASpending.gov.
- The Consumer Product Safety Commission uploaded its data in time, but included errors attributable in part to “data supplied by systems outside CPSC control” and data entry problems.
- The Justice Department uploaded timely data, but sometimes was “noncompliant with standards” for the quality of that data, the OIG found. Sometimes financial data was not linked to awards information, and a “legacy accounting system” led to other reporting challenges.
- The Railroad Retirement Board: had “few, if any” procedures in place to validate the data it submitted, and senior officials did not adequately outline that process.
- The U.S. Army Corps of Engineers didn’t comply with the DATA Act because the Treasury’s system couldn’t “identify or separate the USACE procurement award, grant award, awardee and sub-award data from the DoD data,” the report said. The Pentagon component also could not "ensure the completeness, accuracy, and quality of financial data certified and submitted for publication.”
- The State Department-hired auditing firm Kearney couldn’t assess overseas transactions, so it was unable to fully assess compliance. For domestic data, Kearney found that 64.4 percent of the domestic transactions it explained “did not meet the quality requirements outlined by OMB,” and that “[t]hese errors were within the control of the Department.”
- At NASA, there were minor errors with accuracy, despite timely upload. The agency struggled with reporting some "legal entity’s name, address, primary place of performance, or highly compensated officer names,” the OIG found. Some of those errors could be traced to databases containing outdated information and manual errors.