Government Earns Poor Grades for Spending Data AccuracyTweet
Offices of Inspectors General (OIGs) across the government have been releasing the results of the first coordinated audit campaign of federal spending data, and they have uncovered widespread problems. Federal agencies currently spend approximately $3 trillion in taxpayer money each year. Congress and the public need to be able to rely on their spending data to track the money and hold them accountable.
The Digital Accountability and Transparency Act (DATA Act) passed in 2014 with a goal of significantly improving the quality and scope of government spending data made available to the public. The law tasked Treasury and the Office and Management and Budget with establishing uniform government-wide data standards for the spending information. While it might not sound like much, these data standards were important to address long-standing data quality problems that have persisted in federal spending data. The data standards, properly implemented, would ensure that agencies are reporting complete and reliable spending information and that it would match up across government and allow the data to be merged, broken down and analyzed.
The DATA Act enjoyed strong bipartisan support in Congress. Its supporters had high expectations that it could transform spending transparency in the federal government. However, the supporters had been around long enough to know that expectations are not always easily met. So they included provisions in the DATA Act that required each agency’s Office of Inspector General to audit a statistically valid sample of their agency’s submitted spending data to assess completeness, timeliness, and accuracy of data sampled. The OIGs will conduct follow-up audits in 2019 and 2021
These are the first audits and focus only on the data agencies reported for the second quarter of fiscal year 2017, which was the first data submitted under new data standards. The OIGs established a uniform approach to reviewing the data quality:
- Timeliness: the percentage of transactions reported within 30 days of quarter end.
- Completeness: the data contains all of the transactions during that period, and transactions contain all applicable data elements.
- Accuracy: the percentage of transactions that are complete and match information from source systems.
The Project On Government Oversight (POGO) reviewed 41 OIG audits, including 9 for cabinet level agencies, and discovered a troubling trend of incomplete and inaccurate spending data. There were problems with timeliness, completeness, and accuracy. POGO considered an agency to have missed the standard if the OIG specifically stated the agency submission failed on the issue or if the listed error rate, which some audit reports included for each issue, was above 10 percent.
The agencies had the least problem with timeliness. The OIGs deemed 32 of the agencies to have been timely with their data submissions. The nine agencies that weren’t timely included some big spenders such as the Department of Defense (DoD), General Services Administration, and Department of Veterans Affairs. The DoD widely missed the 30 day deadline with procurement and grant transactions that weren’t submitted until 91 days later. At the other end of the spectrum the Export Import Bank missed being timely by just one day. The audit found that the individual records contained only one untimely error, but the data submission wasn’t certified until the day after the deadline.
|Agencies That Met All Three Data Standards|
|Consumer Financial Protection Bureau|
|Election Assistance Commission|
|Environmental Protection Agency|
|National Credit Union Administration|
|National Endowment for the Arts|
|Securities and Exchange Commission|
|US Agency for International Development|
There was a sharp decline in agency performance when it came to completeness of the data. Only 16 of the 41 agencies were judged to have submitted complete spending records. Only two of the nine cabinet agencies submitted complete spending data – the Departments of Education and Interior.
Many agencies encountered problems both with missing records and missing data within reported records. The Commodity Futures Trading Commission omitted data on purchase card transactions greater than $3,500; however, this only totaled $339,000. The National Science Foundation audit found that 37 percent of the transactions were missing from part of their submission and that 99 percent of the reported records were missing one or more data element. The Department of Agriculture’s data was found to only contain 1 of 670 required Treasury account symbols and 6 of 576 program activities. Agriculture also submitted a completely blank data file that was supposed to contain the financial award data for the quarter.
Agencies struggled most with accuracy. According to the audits only 11 agencies, about one in four, submitted accurate spending information. Most of those agencies were smaller offices like the Federal Labor Relations Authority and the Commodity Futures Trading Commission. None of the nine cabinet level agencies reviewed provided accurate DATA Act information. Of the agencies with satisfactory accuracy, the largest were the Social Security Administration and the Environmental Protection Agency.
When looking at all three data quality characteristics, only 7 agencies met all of them. Once again, smaller agencies like the Consumer Financial Protection Bureau and the Election Assistance Commission made up the majority of this group. These agencies likely have considerably less spending data to report and confirm. A few medium-sized agencies such as the EPA and Securities and Exchange Commission notably met all three data standards.
GAO Government Wide review
In addition to the OIG audits, the Government Accountability Office (GAO) conducted a broader data quality review that evaluated the performance of multiple agencies. The GAO sampled 402 records from across 24 agencies. The number of records reviewed seems a bit low considering the large scope of data they are attempting to evaluate, but the report claims that the sample is large enough to be statistically sound with a 95 percent degree of confidence.
The assessment found results similar to the individual OIG audits. Most of the records were timely, but completeness and accuracy were problematic.
The GAO review found that agencies failed to report any data on financial awards for 160 different assistance programs with a combined estimated annual spending total of almost $81 billion. The GAO found that 13 agencies including the Departments of Defense and Agriculture submitted blank files that were supposed to link budgetary data and award transactions.
The GAO estimated that 56 to 75 percent of the new budgetary information agencies are reporting under the DATA Act was consistent with source data. However, the assessment found that only between 0 and 1 percent of the award transaction records were consistent. That is a 99 percent error rate. The GAO noted this was down from a review it conducted in 2014 that projected between 2 and 7 percent of the award records to be consistent.
The OIG audits for several agencies are not currently available to the public including those for several cabinet agencies such as the Departments of Commerce, Health and Human Services, Homeland Security, Labor, and Transportation. POGO could not locate the reports either on the agencies’ websites or on the central OIG website Oversight.gov. Given the legal requirement, the OIGs for these agencies have almost certainly conducted the audits and have simply not released them yet. These agencies spend significant amounts of taxpayer money, so their audits are incredibly important. Hopefully these audits will soon be made available.
While it is hard to put a good face on such widespread data quality problems, there are a few brighter points worth noting.
First, the OIG reports examined the very first quarter of new data that agencies were submitting. So one could reasonably expect that errors and problems would be at their highest point as there was no time to identify and iron out the anticipated kinks. The timing just worked out this way. Oddly, the DATA Act itself instructed OIGs to perform the first audit in 2016, before any new data was to be reported. The OIGs came together and decided to push the audits back one year so there would be something to review. And the OIGs knew the earliest data would be more error-riddled than usual. In guidance for the audit, a 50 percent error rate was proposed as the baseline for judging if the data met expected standards.
|Agency||Agency Error Rate||Total Error Rate|
|Consumer Products Safety Commission||30%||100%|
|Railroad Retirement Board||81%||91%|
|Department of Energy||28%||76%|
|Broadcasting Board of Governors||70%||86%|
|Government Accountability Office||7%||67%|
|National Science Foundation||43%||62%|
Second, there were government-wide reporting problems associated with the system the Treasury Department set up to pull data directly from other sources, called the Broker System. Essentially, the source systems have correct information but the Broker System seems to misalign with the records and pulls the wrong data, creating new errors in the data that the public sees on USASpending.gov.
A number of agencies noted that the Broker System errors significantly contributed to accuracy problems. Some of the OIG reports were even able to break down the errors attributable to the agency and the Broker System and show just how much the glitch exacerbated their data quality problems. For instance, the Department of Energy OIG reported that the agency had an accuracy error rate of 28 percent, but that problems with the Broker System brought that error rate up to 76 percent.
The good news is that, with a single fix, error rates at multiple agencies should drop significantly. Treasury is well aware of the problem, and several agencies’ audits noted that fixes are in the works.
The third bit of good news is the OIG reports themselves. The data quality problems the audits uncovered were exactly the reason Congress included audit requirements in the DATA Act. Now that a bright light is shining on these data problems, agencies will be under pressure to improve. The bipartisan support for the DATA Act will turn into broad political anger toward the agencies if they don’t improve quickly. And these aren’t one-time-only audits. There are two more rounds coming in 2019 and 2021 that will check to see just how much progress the agencies are making.
See the full results below.