USAspending data quality has been audited for more than a decade, by the Government Accountability Office (GAO) and by the inspectors general of federal agencies. Their reports found errors in dollar fields, coded fields and subawards, and found agencies and agreements missing from USAspending.gov [1, 6, 7, 10, 11]. The figures cover different agencies, fields and years, and GAO itself states that its summary of inspector general results "cannot be directly compared with prior GAO or OIG reports" [1]. Each number below therefore carries its own scope and date.
Contract records start with a contract action report. Under section 4.604 of the Federal Acquisition Regulation (FAR), responsibility for the accuracy of each report "resides with the contracting officer who awarded the contract action" [16]. GAO calls the Federal Procurement Data System (FPDS) the government's principal repository for procurement information, "some of which is displayed on USAspending.gov, the official source of federal spending data" [3]. SAM.gov replaced FPDS as that repository on 24 February 2026 [6]. Where each copy of the records comes from, and when each updates, is covered in FPDS, SAM.gov and USAspending: where contract data comes from.
Who audits federal award data
The DATA Act of 2014 required the inspector general of each federal agency to "review a statistically valid sampling of the spending data submitted under this Act by the Federal agency" and to report on its completeness, timeliness, quality and accuracy [14]. GAO summarized the results of the offices of inspector general (OIGs) in two reports used here: 51 OIG reports on first-quarter fiscal year 2019 data, and 57 reports that each tested one quarter of fiscal year 2020 or 2021 data [1, 2].
That requirement "expired in 2021," GAO told a Senate committee in March 2026 [6]. Public Law 119-106, approved on 11 September 2026, restores the reviews for the 24 agencies listed in 31 U.S.C. 901(b), with a first report due within one year and then at least every two years [8, 13]. The last round under the old requirement produced 57 OIG reports [15].
Self-certified accuracy against audit results
According to the Office of Management and Budget (OMB), the 24 Chief Financial Officers (CFO) Act agencies self-certified their FPDS data as 97.68 percent accurate and 99.14 percent complete and timely, on average, for fiscal years 2017 to 2022 [3]. In September 2025 GAO found that none of those agencies' fiscal year 2023 procurement data quality reports met all relevant reporting requirements, and OMB staff confirmed they had no formal procedures to review the reports [3].
Two agencies' own checks fell short of their own bar on some fields [3]. The Department of Health and Human Services (HHS) rated Date Signed 75.9 percent accurate in fiscal year 2022 and Base and Exercised Options Value 78.5 percent accurate in fiscal year 2023 [3]. HHS and the Department of Veterans Affairs both treat rates below 95 percent as needing corrective action, yet "neither agency documented corrective action plans" for elements below that level in consecutive years [3].
GAO's own samples tested whole records, counting a record as consistent only when every tested element matched agency records. In fiscal year 2012, 2 to 7 percent of awards were fully consistent on all 21 elements [4]. In the fourth quarter of fiscal year 2018, 24 to 34 percent of award transactions were fully consistent on all applicable elements [5]. A record-level rate counts a record as inconsistent when any one element differs, so it measures something different from an element-level average.
One inspector general did set certification beside testing. The Small Business Administration certified the accuracy of its fiscal year 2008 contracting data, yet its inspector general found that 92 percent of a random sample of its contract actions had one or more inaccurate or incomplete data elements in FPDS [21].
Errors in dollar fields
In the last round of inspector general audits, 37 of 57 OIGs found accuracy errors in dollar fields [1]. GAO wrote that the Potential Total Value of Award and Current Total Value of Award fields "had the highest absolute dollar amount of errors" [1]. Across the 26 OIGs that reported errors in Potential Total Value of Award, the absolute value of those errors came to $828,464,250,023 [1].
That figure needs its context. It sums absolute errors found in samples, and GAO notes that the inspectors general's audit guide treats such amounts as "not projectable because the statistical sample design focused on estimation of nonfinancial attributes and not on monetary amounts" [1]. Part of the cause sat in FPDS itself: some OIGs traced errors in the two value fields to "known calculation issues within FPDS" [1].
Which dollar field to sum, and at which level, is the subject of how to count federal contract dollars without double counting.
Agencies and agreements outside USAspending
GAO found that 49 of the 152 agencies that reported in the U.S. government's consolidated financial statements did not report to USAspending.gov in fiscal year 2022 [7]. Of those, 25 were executive branch agencies that together accounted for more than $5 billion in net outlays for fiscal year 2022 [7]. GAO did not assess whether they were required to report, and it noted that neither Treasury nor OMB has clear responsibility for deciding which agencies must [6, 7].
Records that do arrive may not join up. For fiscal years 2020 to 2022, over 10 million award records with over $1.2 trillion in obligations were unlinked, so users would miss information such as agency accounts and program activities [7].
Other transaction agreements (OTAs) form another gap. USAspending.gov has no identifier for them [7], and Treasury held that they fall outside the Federal Funding Accountability and Transparency Act's (FFATA) definition of a federal award [7]. GAO identified $77.5 billion in OTAs that the Department of Defense reported in SAM.gov for fiscal years 2021 to 2025 and that were not reported to USAspending.gov [6]. Public Law 119-106 now defines a federal award to include OTAs [8]. It requires Treasury, not later than three years after enactment, to see that their data are "automatically transmitted to the website" [8]. Three years after approval is 11 September 2029 (derived from [8]).
Subawards: missing and duplicated
From a generalizable sample of 198 fiscal year 2019 awards in the Small Business Innovation Research and Small Business Technology Transfer programs, GAO estimated that about 70 percent of Phase II awards used subcontractors or consultants, and its review of USAspending.gov found that few of these programs' awards, about 10 percent, display subawards [9]. GAO called the reporting limits "a government-wide issue" [9]. For fiscal year 2020 Small Business Technology Transfer Phase II awards, subawards for 253 of 278 prime awards were not reported [7].
Duplication is the other problem. USAspending states that "Duplication of subaward records is common," because prime recipients often re-report all subawards to date, which can push subaward totals above the prime award [30]. In a November 2023 report on grant data, GAO found 26 percent of non-COVID-19 grant subaward records and 11 percent of COVID-19 ones likely duplicates [10]. Controls added in March 2025 "should prevent duplication for subaward records going forward," USAspending says [30].
Coded fields and identifiers
Coded fields failed when auditors compared them with contract files. In a July 2010 report, GAO found about 18 percent of 107 sampled contracts at three departments coded incorrectly for competition [11]. The Department of Homeland Security's inspector general found that the Federal Emergency Management Agency mislabeled 42 of 128 COVID-19 contract actions awarded from 13 March to 31 May 2020 as "full and open competition after exclusion of sources" [12]. In a September 2020 report on fiscal year 2016 and 2017 procurements, the General Services Administration's inspector general found 10 procurements, totaling $274 million, whose NAICS industry codes in FPDS did not match the award documents [20]. The system copied a base contract's code onto its task orders and did not let contracting officers change it [20].
Identifiers have gaps too. In fiscal year 2023 and 2024 award data, GAO counted 7.4 million of 20.3 million awards with no unique entity identifier (UEI), mostly financial assistance aggregate awards that do not require one [32]. Matching suppliers across records without a shared key is covered in how entity resolution matches records across sources.
USAspending data quality checks before analysis
The audits point to checks for any team using these records in a model, a study or a report:
- Record the pull date and source with every extract, since Department of Defense and Army Corps of Engineers procurement records arrive on a 90-day delay [33].
- Sum obligations, and handle the two total value fields with care [1].
- Compare the agencies present with those expected, using USAspending's list of entities that do not report and its Agency Submission Statistics [23, 27].
- Count unlinked awards before joining award and account data [28].
- Look in SAM.gov for other transaction agreements [6].
- Deduplicate subawards, and test for subaward totals above the prime award [30].
- Spot-check competition and industry codes against award documents where a result depends on them [11, 12, 20].
- Treat self-certified accuracy as the agency's own statement about its data [3].
Rillor's public procurement records carry their version and datasheet with each delivery, so an analysis can name the exact records it used.
Read the full report
The full report sets out each audit's scope and period, GAO's statements on comparability and projection, the inspector general results by dollar field, the new law's deadlines and the complete source list.